Datasets:

bzantium
/

LongPerplexity

Dataset card Viewer Files Files and versions Community

Subset (3)

arxiv · 1.64k rows

Split (1)

test · 1.64k rows

text stringlengths 27.9k 483k
high - redshift quasars provide a powerful tool to study the early universe . the first @xmath0 quasars ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) were found by the sloan digital sky survey ( sdss ; * ? ? ? most of these quasars are very luminous ( @xmath9 mag ) . deep near - ir spectroscopy has shown that these quasars harbor billion - solar - mass black holes and emit near the eddington limit , suggesting the rapid growth of central black holes in the early epoch ( e.g. * ? ? ? * ; * ? ? ? their broad emission lines exhibit solar or supersolar metallicity , indicating that vigorous star formation and element enrichment have occurred in the host galaxies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? their optical absorption spectra show that the state of the intergalactic medium ( igm ) at @xmath0 is close to the reionization epoch ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? furthermore , the mid / far - infrared , mm / sub - mm , and radio observations of these quasars have provided rich information about dust emission and star formation in their host galaxies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . therefore , @xmath10 quasars are essential to understanding black hole accretion , galaxy evolution , and the igm state in the first billion years of cosmic time . in recent years , more than 70 quasars at @xmath0 have been discovered . the sdss pioneered quasar studies at these redshifts , followed by the canada - france high - redshift quasar survey ( cfhqs ; * ? ? ? * ) and the ukirt infrared deep sky survey ( ukidss ; * ? ? ? * ) . to date @xmath1130 quasars have been discovered in the main imaging survey ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and the stripe 82 deep survey @xcite . the ukidss has found several quasars @xcite , including the most distant quasar known at @xmath12 @xcite . the cfhqs produced 20 quasars down to a fainter luminosity limit @xcite . most recently , two large surveys , the panoramic survey telescope & rapid response system 1 ( pan - starrs1 ) and the visible and infrared survey telescope for astronomy ( vista ) kilo - degree infrared galaxy ( viking ) survey have started to yield high - redshift quasars . pan - starrs1 is sensitive mainly to brighter quasars than sdss but over a much larger area , and @xmath1310 new quasars at @xmath0 have been published from this survey @xcite , including three quasars at @xmath14 @xcite . viking data have also been used to discover three quasars at @xmath15 @xcite . the number of high - redshift known quasars is increasing steadily . in this paper we introduce a new method for finding @xmath0 quasars by using regions with overlapping imaging in the sdss . the sdss imaging survey was conducted in drift scan mode along great circles , and the imaging runs usually overlap with each other due to the survey geometry and strategy . this results in multiple observations of individual sources within the overlap regions . about one - quarter of the sky area covered by the sdss has multiple imaging coverage . these overlap regions allow us to select quasars that are fainter than those found in the sdss single - epoch images . they ensure that quasar candidates are free of spurious detections ( such as cosmic rays ) , which is particularly useful for objects with single - band ( @xmath3-band in this paper ) detections . here we report eight new quasars found in our first search of @xmath0 quasars in the overlap regions . these quasars are @xmath110.5 mag fainter than the quasars selected from sdss single - epoch data . the layout of the paper is as follows . in section 2 we briefly describe the sdss overlap regions . in section 3 we introduce our quasar selection procedure and follow - up observations . we present the new quasars in section 4 , and summarize the paper in section 5 . throughout the paper sdss magnitudes are reported on the ab system instead of the asinh system @xcite , and all near - ir magnitudes are on the vega system . we use a @xmath16-dominated flat cosmology with @xmath17 km s@xmath18 mpc@xmath18 , @xmath19 , and @xmath20 . in this section we describe the sdss overlap regions , the regions that were scanned by two or more sdss imaging runs . the sdss imaging survey was carried out in drift - scan mode using a 142 mega - pixel camera @xcite . an sdss run ( strip ) consists of 6 parallel scanlines ( camera columns ) for each of five @xmath21 bands @xcite . the scanlines are @xmath22 wide with gaps of roughly the same width , so two interleaving strips make a stripe . sdss scanlines are divided into fields , and a field is the union of five @xmath21 frames covering the same region of sky . processing of individual runs includes assessments of the photometric quality @xcite . the sdss imaging runs generally overlap each other , due to the survey geometry and strategy @xcite . the imaging survey in drift - scan mode was along great circles , and had two common poles at [ r.a . = [ @xmath23 , @xmath24 and [ @xmath25 , @xmath24 . the fields overlap more substantially when they approach the survey poles ( figure 1 ) . in addition , the two interleaving strips that make any stripe slightly overlap , leading to duplicate observations in a small area . furthermore , if the quality of a run , or part of a run , did not meet the sdss standard criteria , the relevant region of this run was re - observed , yielding duplicate observations in this region . obviously , a combination of the above reasons may result in more than two observations in some regions . the upper panel in figure 1 shows the coverage of the sdss overlap regions at high galactic latitude @xmath26 . the sdss deep stripe , stripe 82 , has been excluded , because it has much deeper co - added data @xcite . the data points represent bright ( @xmath27 ) point sources with at least two observations . in the lower panel of the figure , the coverage map is zoomed in to an area of r.a . = @xmath28@xmath29 and decl . = @xmath30@xmath31 , where the data points are point sources with @xmath32 . we estimated the total area of the overlap regions in the upper panel by calculating the area occupied by the actual data points . the sources we used are point sources with @xmath33 . we divided the r.a.decl . plane into small grid elements , where each element measures @xmath34 at a side . if a grid element contains more than one object , it was assumed to be covered by the overlap regions . the total area by summing the area of all selected grid elements is 3700 square degrees . this rough estimate is accurate to roughly 10% , and could be improved by using more sources and by refining grid elements . we leave this for a future work . the total area of all sdss overlap regions , including low galactic latitude @xmath35 , is more than 4000 square degrees . for a given source in the overlap regions , the sdss pipeline selects one detection as a ` primary ' detection from a ` primary ' run / field . additional detections are classified as ` secondary ' . the primary runs / fields are generally deeper due to better image quality in terms of seeing , sky background , atmospheric transparency , etc . figure 2 illustrates the difference between primary and secondary detections in the @xmath3 band . the sources used in the figure are point sources with @xmath36 mag , selected from the overlap regions displayed in the lower panel of figure 1 ( r.a . = @xmath28@xmath29 and decl . = @xmath30@xmath31 ) . the upper panel of figure 2 shows photometric errors versus @xmath3-band magnitude . the shaded regions include central 68% of the sources at any magnitude . they indicate that primary detections have slightly better data quality than secondary detections . this is also suggested by the three lower panels , which show the distribution of photometric errors in three different magnitude bins . compared to primary detections , the distribution of errors for the secondary detections has a longer tail towards greater errors . therefore , a secondary detection does not always significantly improve the photometry of a source . however , it generally rules out the possibility that its primary detection is a false detection . this is particularly important for single - band detected sources , such as @xmath10 quasars . in this section we present the details of our candidate selection and spectroscopic identification of @xmath0 quasars in the sdss overlap regions . the selection procedure is similar to the procedures that were used to select sdss @xmath0 quasars in previous studies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the procedure consists of several steps as shown in the following subsections . we start with sdss data release 9 ( dr9 ) ` primary ' sources at high galactic latitude @xmath26 , by using the ` photoprimary ' table with the sdss query / casjobs online server , excluding stripe 82 . the ` photoprimary ' table includes both point and extended sources . although distant quasars are point - like objects in ground - based images , faint point - like objects could be mis - classified as extended objects ( and vice versa ) ; thus we include extended objects . we then remove sources with critical sdss processing flags such as bright , edge and satur . for each source , we searched for ` neighbors ' ( ` secondary ' detections from different runs ) within a @xmath37 radius . all matched pairs are considered to be multiple observations of the same object . from these dual observations we select @xmath4-band dropout objects using the following color cuts , @xmath38 these selection criteria are applied to the individual observations , i.e. , each of the dual observations must meet the criteria except criterion 4 . we require that the primary detections meet criterion 4 ( at least @xmath39 detections ) , and the secondary detections should be detected at a significance of @xmath40 . in this first step , we obtain a total of @xmath11550 @xmath4-band dropout objects . in the previous search of luminous @xmath0 quasars with the sdss single - epoch images ( see fan et al . papers ) , the magnitude limit was roughly @xmath41 ( @xmath42 detections ) . below @xmath42 , the number of contaminants in @xmath4-dropout selection increases dramatically . from single - epoch images , one would get tens of thousands of @xmath4-dropout candidates down to a @xmath39 threshold , where the vast majority of them are spurious objects . using duplicate observations ensures that most of the @xmath4-dropout objects we select are physical sources . we visually inspect the dropouts in the sdss @xmath21 images , and remove obvious contaminants , such as suspicious detections close to very bright stars and objects that are apparently extended . we reject about one - third of the 550 objects in this step . we then remove previously known objects , such as l / t dwarfs and @xmath0 quasars . we match our objects to the list of known l / t dwarfs from dwarfarchives.org , and find 51 known dwarfs . we also match the objects to the list of @xmath4-dropouts that have been spectroscopically identified as non - quasars during our previous searches for @xmath0 quasars . in addition , we find that 8 objects are previously discovered quasars at @xmath0 . these quasars are summarized in table 1 . they are generally brighter than @xmath41 mag . after removing these objects , about 250 objects remain in our candidate list . we match the remaining objects to the 2mass catalog @xcite and to the catalog of the ukirt infrared deep sky survey ( ukidss ; * ? ? ? we obtain 10 matches from the 2mass , and 53 matches from the ukidss , providing near - ir photometry for 63 objects . the majority are then rejected as quasar candidates based on their location in the @xmath43 versus @xmath44 color - color diagram ( see section 3.3 ) . this leaves a total of 190 @xmath4-dropout candidates selected for additional observations . ccccl j000239.40 + 255034.8 & 5.80 & 21.49@xmath450.12 & 18.94@xmath450.05 & @xcite + & & 21.39@xmath450.13 & 18.88@xmath450.05 & + j084119.52 + 290504.4 & 5.96 & 23.30@xmath450.66 & 19.78@xmath450.13 & @xcite + & & 22.47@xmath450.28 & 19.78@xmath450.09 & + j103027.09 + 052455.0 & 6.28 & 22.93@xmath450.41 & 19.59@xmath450.08 & @xcite + & & 23.14@xmath450.42 & 19.91@xmath450.11 & + j104433.04012502.1 & 5.80 & 21.58@xmath450.20 & 18.99@xmath450.07 & @xcite + & & 21.62@xmath450.16 & 19.09@xmath450.07 & + j131911.29 + 095051.3 & 6.13 & 21.79@xmath450.33 & 19.35@xmath450.14 & @xcite + & & 22.69@xmath450.28 & 20.07@xmath450.10 & + j141111.28 + 121737.3 & 5.93 & 22.91@xmath450.35 & 19.55@xmath450.08 & @xcite + & & 23.49@xmath450.42 & 19.56@xmath450.07 & + j162331.81 + 311200.6 & 6.22 & @xmath46 & 19.64@xmath450.10 & @xcite + & & 24.47@xmath450.85 & 20.06@xmath450.15 & + j163033.90 + 401209.7 & 6.05 & 23.42@xmath450.39 & 20.33@xmath450.12 & @xcite + & & 22.69@xmath450.27 & 20.14@xmath450.11 & + the next step is to improve the @xmath4 and @xmath3-band photometry . the majority of the objects have @xmath3-band detections weaker than @xmath42 . for these objects we obtained deeper @xmath4 and @xmath3 band images using the wide - field optical imager 90prime on the 2.3 m bok telescope . the typical integration time was 5 min per object ( single exposure ) in the @xmath4 band and 3 min per object ( single exposure ) in the @xmath3 band . the observing conditions for many objects were poor - to - moderate with significant moonlight , poor seeing , or non - photometric transparency , because most of them were observed as backup targets for other programs . to date we have observed roughly 110 objects . the 90prime images were reduced in a standard fashion using our own idl routines . first we made master bias and flat images from bias and flat images taken in the same night . a bad pixel mask was created from the flat image . then science images were overscan and bias corrected and flat - fielded . saturated pixels and bleeding trails were identified , and incorporated ( along with the bad mask ) into weight images . next the first - round sky subtraction was performed by fitting a low - order 2d polynomial function to the background . the 90prime ccds are thin chips , so they produce strong fringing in the @xmath4 and @xmath3 bands . we removed fringes in two iterations . then another round of sky subtraction was performed . we detected objects using sextractor @xcite and calculated astrometric solutions using scamp @xcite by matching objects to the sdss . photometry was measured within an aperture ( diameter ) size of 8 pixels ( @xmath47 ) and calibrated by matching a large number of nearby bright point sources to the sdss . with the improved @xmath4 and @xmath3 band photometry , we rejected objects with relatively blue @xmath44 colors . the 90prime @xmath4 and @xmath3 filters are slightly redder than the sdss @xmath4 and @xmath3 filters . depending the strength of ly@xmath48 emission , quasars at @xmath49 could have much bluer @xmath44 colors in the 90prime system . we rejected objects with @xmath50 ( in the 90prime system ) , instead of @xmath51 ( in the sdss filter system ) that was used in our previous studies . this is similar to the color cut used by the cfhqs ( e.g. , * ? ? ? the next step is to obtain near - ir ( @xmath52 and/or @xmath53 band ) photometry . we excluded objects that already have @xmath52-band photometry from the 2mass or the ukidss ( section 3.1 ) . near - ir observations for the other objects were obtained with the mmt swirc @xcite and the kpno 4 m newfirm . the objects were observed as backup targets for other programs , so the observing conditions were usually poor . the exposure time per object was between 3 and 10 min . we have observed roughly 40 objects to date . the near - ir images were reduced using standard iraf routines . the near - ir photometry is used to separate quasars from late - type dwarf stars , which are the primary contaminants in @xmath4-dropout searches , as mentioned earlier . stars can be rejected based on the @xmath43 versus @xmath44 color - color diagram ; a more detailed description can be found in @xcite . in short , we use the following criterion to further select quasar candidates , @xmath54 in the final step we conducted optical spectroscopy to identify quasar candidates using the mmt red channel spectrograph . we have observed about 10 candidates , preferentially choosing candidates with ukidss photometry in the @xmath53 and @xmath52 bands . the @xmath55 color is one of the best criteria for separating @xmath0 quasars l and t dwarfs . the mmt observations were made in long - slit mode with a spectral resolution of @xmath1110 . the exposure times for each target were 510 min , sufficient to classify our candidates as quasars under normal weather conditions . if a target was identified as a quasar immediately after its first spectrum was obtained , additional 5 or 10 min exposures were taken to improve the spectral quality . two quasars were identified with poor observing conditions , and their additional spectra were taken one year later . the spectra of one quasar ( sdss j0842 + 1218 ) was obtained with keck / esi . the spectra were reduced using standard iraf routines . in total we identified eight new quasars from the optical spectroscopic observations . ccccccc j000825.77062604.6 & 5.929@xmath450.003 & 23.91@xmath450.65 & 20.04@xmath450.15 & 53997 & & 19.43@xmath450.13 + & & 24.04@xmath451.03 & 19.55@xmath450.13 & 55120 & & + & & 22.85@xmath450.25 & 20.35@xmath450.09 & 56561 & & + j002806.57 + 045725.3 & 6.04@xmath450.03 & 22.73@xmath450.27 & 20.30@xmath450.14 & 54742 & 19.59@xmath450.09 & 19.16@xmath450.12 + & & @xmath124 & 20.42@xmath450.22 & 54764 & & + & & 24.00@xmath450.50 & 20.49@xmath450.10 & 56560 & & + j014837.64 + 060020.0 & 5.923@xmath450.003 & 22.25@xmath450.18 & 19.31@xmath450.06 & 53655 & 18.91@xmath450.06 & 18.37@xmath450.07 + & & 22.09@xmath450.22 & 19.35@xmath450.08 & 53654 & & + j084229.43 + 121850.5 & 6.055@xmath450.003 & 23.31@xmath450.39 & 19.56@xmath450.08 & 53441 & & 18.84@xmath450.03 + & & 23.43@xmath450.72 & 19.46@xmath450.12 & 53684 & & + j085048.25 + 324647.9 & 5.867@xmath450.007 & 24.25@xmath450.78 & 20.34@xmath450.16 & 52664 & & 18.90@xmath450.10 + & & 23.66@xmath450.52 & 20.39@xmath450.15 & 52664 & & + & & & 19.95@xmath450.08 & 56369 & & + j120737.43 + 063010.1 & 6.040@xmath450.003 & 22.70@xmath450.29 & 20.16@xmath450.12 & 52338 & 19.51@xmath450.09 & 19.35@xmath450.14 + & & 23.49@xmath450.61 & 20.17@xmath450.16 & 52435 & & + & & @xmath123.5 & 20.39@xmath450.11 & 56370 & & + j125757.47 + 634937.2 & 6.02@xmath450.03 & 24.91@xmath450.84 & 20.20@xmath450.15 & 51987 & 20.39@xmath450.16 & 19.78@xmath450.08 + & & 23.55@xmath450.54 & 20.29@xmath450.14 & 51633 & & + & & 23.50@xmath450.40 & 20.60@xmath450.12 & 56394 & & + j140319.13 + 090250.9 & 5.86@xmath450.03 & 22.53@xmath450.23 & 20.17@xmath450.14 & 52757 & 19.70@xmath450.12 & 19.17@xmath450.10 + & & 23.01@xmath450.32 & 20.26@xmath450.13 & 52345 & & + & & 22.73@xmath450.22 & 20.48@xmath450.11 & 56394 & & + in this section we present the discovery of eight new quasars at @xmath0 . table 2 lists the coordinates , redshifts , and optical and near - ir photometry of the new quasars . the naming convention for sdss sources is sdss jhhmmss.ss@xmath45ddmmss.s , and the positions are expressed in j2000.0 coordinates . we use sdss jhhmm@xmath45ddmm ( or jhhmm in figures ) for brevity . the new quasars span a relatively narrow redshift range of @xmath6 . the redshifts were mostly measured from several emission lines including ly@xmath48 , oi @xmath56 ( hereafter oi ) , cii @xmath57 ( hereafter cii ) , and siiv @xmath58 ( hereafter siiv ) , by fitting a gaussian profile to the top @xmath59% of these lines . three quasars do not show apparent line emission in their optical spectra , so their redshifts were measured from the wavelength where the sharp flux decline occurs . they are discussed in more detail in the next subsection . the redshift errors quoted in table 2 include the uncertainties from our fitting process and wavelength calibration . they do not include any possible velocity shifts between these lines and systemic redshifts . for the quasars without apparent line emission , we adopted a redshift error of 0.03 , corresponding to the scatter in the relation between ly@xmath48 redshifts and systemic redshifts ( e.g. * ? ? ? * ) . two of the quasars ( sdss j0148 + 0600 and sdss j1207 + 0630 ) were independently discovered by s. warren et al . ( in preparation ) . in table 2 , there are multiple entries for the @xmath4 and @xmath3 band photometry . the first two entries for each quasar are the duplicate observations from the sdss , and the third one ( if present ) is the improved photometry from bok 90prime . figure 3 shows two examples of the multi - epoch observations of the quasars sdss j00080626 and sdss j0028 + 0457 . in the sdss observations ( epochs 1 and 2 ) , the two quasars were not ( or barely ) detected in the @xmath4 band , and were very weak ( @xmath60 ) in the @xmath3 band . in the bok 90prime observations ( epoch 3 ) they are clearly detected in the @xmath4 band , and have strong detections ( @xmath61 ) in the @xmath3 band . note that quasars are variable objects with a typical variability amplitude of a few tenths of a magnitude ( e.g. * ? ? ? * ) , so the difference of the photometry in different entries for the same quasar may partially reflect this variability . sdss j0148 + 0600 and sdss j0842 + 1218 do not have the third entry of the @xmath4 and @xmath3 band photometry . they are more than @xmath42 detections in the sdss observations , so we did not take more images for them . the other six quasars are weaker than @xmath42 detections , which was the limit in the previous survey of @xmath0 quasars with the sdss single - epoch images . the last two columns of table 2 show the near - ir ( @xmath53 and @xmath52 ) photometry . the @xmath53 and @xmath52 band photometry for four quasars was obtained from the ukidss . as we mentioned in section 3.3 , in this first study we preferentially observed candidates with ukidss @xmath53 and @xmath52-band photometry . the near - ir photometry of the other four quasars was obtained with the mmt swirc . figure 4 shows the optical spectra of the quasars . as addressed in section 3.3 , one quasar ( sdss j0842 + 1218 ) was observed with keck / esi , and its total integration time was 20 min . the other seven quasars were observed with the mmt red channel spectrograph . the total integration time per object was from 10 min to 40 min , depending on the quasar brightness and observing conditions . each spectrum in figure 4 has been scaled to match the best corresponding @xmath3-band magnitude in table 1 , thereby placed on an absolute flux scale . these quasars show a diversity of ly@xmath48 emission . for example , sdss j00080626 and sdss j0842 + 1218 have strong ly@xmath48 emission lines . sdss j0148 + 0600 and sdss j1207 + 0630 are likely broad absorption line ( bal ) quasars . sdss j1257 + 6349 and sdss j1403 + 0902 do not show any prominent ly@xmath48 emission . quasars with weak emission lines like these quasars are not rare at high redshift . we have found a significant fraction of lineless quasars at @xmath0 ( e.g. * ? ? ? * ; * ? ? ? table 3 shows the properties of the quasar continua and ly@xmath48 emission lines . columns 3 and 4 list @xmath62 and @xmath63 , the apparent and absolute ab magnitudes of the continuum at rest - frame 1450 . they were calculated by fitting a power - law continuum @xmath64 to the spectrum of each quasar . our optical spectra only cover a short wavelength range in the rest frame ( @xmath65 ) , and this range contains several uv emission lines , so we were not able to reliably measure the slopes of the continua . we assumed that the slope is the average quasar uv continuum slope @xmath66 ( e.g. * ? ? ? * ) . the power - law continuum was then normalized to match visually identified continuum windows with little contribution from line emission . after the power - law continuum was subtracted from the spectrum , we measured the rest - frame equivalent width ( ew ) of the ly@xmath48 emission line , including the nv @xmath67 ( hereafter nv ) line that is usually blended with ly@xmath48 . this was done by fitting a half gaussian profile ( ly@xmath48 ) and a full gaussian profile ( nv ) simultaneously to the spectrum redward of the ly@xmath48 line center . we ignored the weak siii @xmath68 emission line on the red side of nv . for the quasars without prominent ly@xmath48 emission , we simply integrated flux ( instead of gaussian fitting ) above the continuum . we did not correct for the flux absorbed by the ly@xmath48 forest . the results are shown in column 5 in table 3 . these measurements are crude due to uncertainties from continuum measurement , redshift determination , and ly@xmath48 forest absorption . cccccc j00080626 & 5.929 & 20.63@xmath450.09 & 26.04@xmath450.09 & 78 + j0028 + 0457 & 6.040 & 20.34@xmath450.10 & 26.38@xmath450.10 & 17 + j0148 + 0600 & 5.923 & 19.60@xmath450.06 & 27.08@xmath450.06 & @xmath69 + j0842 + 1218 & 6.055 & 19.85@xmath450.09 & 26.85@xmath450.09 & 44 + j0850 + 3246 & 5.867 & 19.92@xmath450.08 & 26.74@xmath450.08 & 10 + j1207 + 0630 & 6.040 & 20.10@xmath450.11 & 26.60@xmath450.11 & 31 + j1257 + 6349 & 6.02 & 20.56@xmath450.12 & 26.14@xmath450.12 & 18 + j1403 + 0902 & 5.86 & 20.39@xmath450.11 & 26.27@xmath450.11 & 8 + _ sdss j00080626 ( z=5.929 ) . _ sdss j00080626 has a strong ly@xmath48 emission line . it also shows prominent nv , oi , and siiv emission lines . its redshift was measured from the oi emission line . _ sdss j0028 + 0457 ( z=6.04 ) . _ sdss j0028 + 0457 was independently discovered by the pan - starrs1 survey @xcite . it does not have prominent emission lines in the spectrum shown in figure 4 . the redshift measured from the weak ly@xmath48 line peak is 6.09 , and the redshift measured from the wavelength where the sharp flux drop occurs is 6.04 . here we adopt 6.04 . this is slightly higher than the redshift reported in @xcite ( @xmath705.99 ) . an interesting feature is a strong narrow emission line at @xmath71 . it exists in all three individual mmt spectra , as well as the @xcite spectrum , so it is a real line feature . it could be a luminous ly@xmath48 emitter at @xmath72 , or an ionized igm bubble at @xmath72 . higher resolution imaging and spectroscopy are needed to explore the nature of this line . _ sdss j0148 + 0600 ( z=5.923 ) . _ sdss j0148 + 0600 is the brightest quasar in this sample . it was well detected in the sdss single - epoch images and the ukidss near - ir images . it is likely a low - ionization bal ( lobal ) quasar , and its ly@xmath48 emission line has been mostly absorbed . the ew listed in table 3 is a lower limit . it does not have obvious emission lines other than cii redward of nv in its optical spectrum , so we used cii to determine the redshift . this quasar was independently discovered by s. warren et al . ( in preparation ) . @xcite found that it showed a very long gunn - peterson trough in a deep vlt / x - shooter spectrum . _ sdss j0842 + 1218 ( z=6.055 ) . _ sdss j0842 + 1218 was discovered in 2007 . its optical spectrum was obtained with keck / esi ( 20 min exposure in total ) . it has been observed in the near - ir and mid - ir wavelength ranges , and its physical properties such as metallicity and black hole mass have been well measured ( e.g. * ? ? ? * ; * ? ? ? this quasar was initially selected from the sdss single - epoch images , but also satisfies our overlap region selection criteria . we include this quasar because it is in the overlap regions and its optical spectrum was never published . the redshift of this quasar measured from the ly@xmath48 line is 6.055 . this is consistent with the redshift 6.08 measured from its mgii emission line @xcite , given the large scatter ( @xmath73 ) between ly@xmath48 redshifts and systemic redshifts . _ sdss j0850 + 3246 ( z=5.867 ) . _ sdss j0850 + 3246 shows a very weak ly@xmath48 emission line with an observed rest - frame ew about 10 . it does not show any other line emission in its optical spectrum in figure 4 . _ sdss j1207 + 0630 ( z=6.040 ) . _ sdss j1207 + 0630 is a high - ionization bal ( hibal ) quasar , as seen from the strong nv absorption line in the spectrum . it was also covered and detected in the near - ir by the ukidss . _ sdss j1257 + 6349 ( z=6.02 ) . _ sdss j1257 + 6349 does not show any significant line emission in its optical spectrum in figure 4 . its redshift was measured from the sharp flux drop at the ly@xmath48 line . _ sdss j1403 + 0902 ( z=5.860 ) . _ sdss j1403 + 0902 is another weak line quasar , with an observed rest - frame ly@xmath48ew of only 8 . its redshift was measured from the sharp flux drop at the ly@xmath48 line . it was also covered and detected in the near - ir by the ukidss . the currently known sdss quasars at @xmath0 were selected either from single - epoch imaging data encompassing more than 8000 deg@xmath2 with a limiting depth of @xmath74 mag , or from the stripe 82 multi - epoch co - added images over 300 deg@xmath2 that are two magnitudes deeper . the sdss overlap regions provide another unique dataset that is in between the sdss single - epoch data and the stripe 82 data in terms of areal coverage and image d epth . these regions cover roughly 3700 deg@xmath2 area at @xmath26 , which is nearly one third of the total sdss sky coverage , and is an order of magnitude greater than the area of stripe 82 . we have demonstrated that the overlap regions can be used to efficiently select quasar candidates that are @xmath130.5 mag fainter than those found in single - epoch images . duplicate observations allow us to select fainter @xmath0 candidates primarily because the initial candidate selection is largely free of spurious detections . the survey limit for @xmath0 quasars in single - epoch images was roughly @xmath8 mag ( or @xmath42 detections ) . beyond this limit , the number of spurious detections increases dramatically , making follow - up observations infeasible . the multi - epoch observations in the overlap regions ensure that the majority of single - band ( @xmath3-band ) detections are real . in our first quasar search we have found eight quasars in the overlap regions . we also recovered several known quasars that are located in the overlap regions . as we mentioned earlier , the follow - up observations of our candidates have not been systematic . they were usually observed as backup targets in other programs , when weather conditions were poor . hence these quasars do not form a statistically complete sample . however , they provide guidance for a future systematic quasar survey over the whole overlap regions . in addition , we will use the wise mid - ir imaging data to improve our selection efficiency ( e.g. * ? ? ? * ; * ? ? ? in this paper we have presented the discovery of eight new quasars at @xmath0 . they were found in our first high - redshift quasar search in the sdss overlap regions . these regions were scanned twice or more by the sdss imaging survey , allowing us to select quasars fainter than those found in sdss single - epoch images . we have introduced the details of our quasar selection procedure , including the selection of @xmath4-band dropouts , the follow - up optical and near - ir imaging observations , and the final spectroscopic identifications . the eight quasars span a redshift range of @xmath6 . five of them are fainter than @xmath8 mag , and the faintest one is @xmath75 mag , @xmath110.5 mag fainter than those found from sdss single - epoch images . these quasars show a diversity of emission line properties in their optical spectra , from strong emission lines to lineless features . in addition , we recovered eight previously known @xmath0 quasars that are also located in these overlap regions . these results indicate that the sdss overlap regions can be used to efficiently select high - redshift quasars . we plan to carry out a systematic survey of @xmath0 quasars in the whole sdss overlap regions at high galactic latitude . we acknowledge the support from a 985 project at peking university . l.j . and z.f . acknowledge supports from the national natural science foundation of china ( nsfc ) under grants 11003021 and 11373003 . x.f . and i.d.m . acknowledge support from u.s . nsf grant ast 11 - 07682 . observations reported here were obtained in part at the mmt observatory , a joint facility of the university of arizona and the smithsonian institution . this work is based in part on observations at kpno , noao ( prop . i d : 2014b-0099 ; pi : linhua jiang ) , which is operated by the aura under cooperative agreement with the nsf . this work is based in part on data obtained as part of the ukirt infrared deep sky survey . this research has benefited from the m , l , t , and y dwarf compendium housed at dwarfarchives.org . funding for the sdss and sdss - ii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the u.s . department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england . the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium for the participating institutions . the participating institutions are the american museum of natural history , astrophysical institute potsdam , university of basel , university of cambridge , case western reserve university , university of chicago , drexel university , fermilab , the institute for advanced study , the japan participation group , johns hopkins university , the joint institute for nuclear astrophysics , the kavli institute for particle astrophysics and cosmology , the korean scientist group , the chinese academy of sciences ( lamost ) , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , ohio state university , university of pittsburgh , university of portsmouth , princeton university , the united states naval observatory , and the university of washington . annis , j. , soares - santos , m. , strauss , m. a. , et al . 2014 , , 794 , 120 baados , e. , venemans , b. p. , morganson , e. , et al . 2014 , , 148 , 14 becker , r. h. , fan , x. , white , r. l. , et al . 2001 , , 122 , 2850 becker , g. d. , bolton , j. s. , madau , p. , et al . 2015 , , 447 , 3402 bertin , e. , & arnouts , s. 1996 , , 117 , 393 bertin , e. 2006 , in asp conf . 351 , astronomical data analysis software and systems xv , ed . c. gabriel , c. arviset , d. ponz , & e. solano ( san francisco , ca : asp ) , 112 brown , w. r. , mcleod , b. a. , geary , j. c. , & bowsher , e. c. 2008 , , 7014 , 90 carilli , c. l. , wang , r. , fan , x. , et al . 2010 , , 714 , 834 carnall , a. c. , shanks , t. , chehade , b. , et al . 2015 , arxiv:1502.07748 de rosa , g. , decarli , r. , walter , f. , et al . 2011 , , 739 , 56 fan , x. , white , r. l. , davis , m. , et al . 2000 , , 120 , 1167 fan , x. , narayanan , v. k. , lupton , r. h. , et al . 2001 , , 122 , 2833 fan , x. , strauss , m. a. , schneider , d. p. , et al . 2003 , , 125 , 1649 fan , x. , hennawi , j. f. , richards , g. t. , et al . 2004 , , 128 , 515 fan , x. , strauss , m. a. , richards , g. t. , et al . 2006 , , 131 , 1203 fan , x. , strauss , m. a. , becker , r. h. , et al . 2006 , , 132 , 117 fukugita , m. , ichikawa , t. , gunn , j. e. , et al . 1996 , , 111 , 1748 goto , t. 2006 , , 371 , 769 gunn , j. e. , carr , m. , rockosi , c. , et al . 1998 , , 116 , 3040 gunn , j. e. , siegmund , w. a. , mannery , e. j. , et al . 2006 , , 131 , 2332 hogg , d. w. , finkbeiner , d. p. , schlegel , d. j. , & gunn , j. e. 2001 , , 122 , 2129 ivezi , . , lupton , r. h. , schlegel , d. , et al . 2004 , astronomische nachrichten , 325 , 583 jiang , l. , fan , x. , vestergaard , m. , et al . 2007 , , 134 , 1150 jiang , l. , fan , x. , annis , j. , et al . 2008 , , 135 , 1057 jiang , l. , fan , x. , bian , f. , et al . 2009 , , 138 , 305 jiang , l. , fan , x. , brandt , w. n. , et al . 2010 , , 464 , 380 jiang , l. , fan , x. , bian , f. , et al . 2014 , , 213 , 12 kurk , j. d. , walter , f. , fan , x. , et al . 2007 , , 669 , 32 lawrence , a. , warren , s. j. , almaini , o. , et al . 2007 , , 379 , 1599 lupton , r. h. , gunn , j. e. , & szalay , a. s. 1999 , , 118 , 1406 macleod , c. l. , ivezi , . , sesar , b. , et al . 2012 , , 753 , 106 mcgreer , i. d. , mesinger , a. , & fan , x. 2011 , , 415 , 3237 morganson , e. , de rosa , g. , decarli , r. , et al . 2012 , , 143 , 142 mortlock , d. j. , patel , m. , warren , s. j. , et al . 2009 , , 505 , 97 mortlock , d. j. , warren , s. j. , venemans , b. p. , et al . 2011 , , 474 , 616 shen , y. , strauss , m. a. , oguri , m. , et al . 2007 , , 133 , 2222 skrutskie , m. f. , cutri , r. m. , stiening , r. , et al . 2006 , , 131 , 1163 stoughton , c. , lupton , r. h. , bernardi , m. , et al . 2002 , , 123 , 485 vanden berk , d. e. , et al . 2001 , , 122 , 549 venemans , b. p. , mcmahon , r. g. , warren , s. j. , et al . 2007 , , 376 , l76 venemans , b. p. , findlay , j. r. , sutherland , w. j. , et al . 2013 , , 779 , 24 venemans , b. p. , baados , e. , decarli , r. , et al . 2015 , , 801 , 11 walter , f. , riechers , d. , cox , p. , et al . 2009 , , 457 , 699 wang , r. , wagg , j. , carilli , c. l. , et al . 2011 , , 142 , 101 wang , r. , wagg , j. , carilli , c. l. , et al . 2013 , , 773 , 44 warren , s. j. , hambly , n. c. , dye , s. , et al . 2007 , , 375 , 213 willott , c. j. , delfosse , x. , forveille , t. , delorme , p. , & gwyn , s. d. j. 2005 , , 633 , 630 willott , c. j. , delorme , p. , omont , a. , et al . 2007 , , 134 , 2435 willott , c. j. , delorme , p. , reyl , c. , et al . 2009 , , 137 , 3541 willott , c. j. , delorme , p. , reyl , c. , et al . 2010 , , 139 , 906 wu , x .- b . , hao , g. , jia , z. , zhang , y. , & peng , n. 2012 , , 144 , 49 york , d. g. , adelman , j. , anderson , j. e. , jr . , et al . 2000 , , 120 , 1579
the carina dwarf spheroidal ( dsph ) galaxy has an important place in the history of our understanding of stellar populations in external galaxies . at the time of its discovery ( cannon et al . 1977 ) , there were six known dsph satellites of the milky way ( for an up - to - date census of these systems in the local group see da costa 1998 and mateo 1998a , b ) . all were believed to be essentially old stellar systems similar in stellar population to globular clusters . some evidence to the contrary did exist : baade ( 1963 ) and later zinn ( 1980 ) emphasized that the variable stars of dsph galaxies in particular the rr lyr stars and anomalous cepheids suggested that their underlying stellar populations were fundamentally different from those of globular clusters . zinn postulated that the difference was age , such that the dsph systems were on average younger than globulars . subsequent deep photometric observations of the main sequence turnoffs ( msto ) in these galaxies confirmed this conjecture so spectacularly that it is now difficult to identify _ any _ dsph galaxy whose population is as uniformly old as that of a typical globular cluster ( see da costa 1998 for an excellent recent review ) ! carina was the first dsph galaxy so observed in the early days of modern ccd detectors ( mould and aaronson 1983 ) and remains one of the more spectacular examples of a dsph galaxy with a significant intermediate - age population . mighell ( 1990 , 1997 ) , smecker - hane et al . ( 1994a ) and hurley - keller et al . ( 1998 ; hereafter paper i ) have obtained additional photometry of the carina main - sequence that confirms the bulk of its stars have ages ranging from about 4 - 7 gyr , though older ( saha et al . 1986 ) and younger stars are also clearly present . one consequence of the relative youth of carina is that its main sequence extends far above the canonical msto luminosity of globular clusters . many main - sequence stars are found in carina within the instability strip where short - period pulsating stars known as dwarf cepheids , sx phe stars , @xmath7 sct stars , or ultra - short - period cepheids are found ( see mateo 1993 for a review of the properties and confused nomenclature of these stars ) . in this paper we shall refer to these variables as ` dwarf cepheids ' ( dc ) to emphasize that they are main - sequence stars pulsating in basically the same manner as classical cepheids . nemec et al . ( 1994 ; hereafter nnl ) provide a comprehensive survey of the then - known population ii dwarf cepheids , referring to these stars as ` sx phe ' variables . many dcs also happen to be ` blue stragglers ' in their parent clusters , located above and to the blue of the msto . these blue - straggler sequences sometimes extend to the dwarf cepheid instability strip , a region normally forbidden to stars older than about 8 - 10 gyr ( see figure 1 ) . carina is a logical place to search for dcs as its main sequence extends up to and through the region where dwarf cepheids are found in other stellar systems as a result of its prominent intermediate - age population ( figure 1 ) . in this paper we describe a search for dwarf cepheids in the carina dsph galaxy . the data used here are the same as those reported in paper i , except that now we have analyzed the frames individually to perform time - series photometry , rather than combining the images to go as deep as possible . two challenges face us in this search . first , dcs have very short periods ranging from about 1 hour to 3 hours . any exposures longer than 10 - 15 minutes can ` smear out ' these rapid light variations . second , because dcs have modest luminosities ( @xmath8 to @xmath9 ) , they are expected to have @xmath10 magnitudes of @xmath022 - 24 in carina , rather faint for such short individual exposures . nonetheless , as we describe here , we have discovered 20 dcs in carina . the observations used here are the same as reported in paper i. to review , the data were obtained using the ctio 4 m telescope with the prime - focus ccd camera during two runs in 1989 and 1990 . at the time , this camera contained a thinned , backside - illuminated ti @xmath11 ccd . the typical exposure time was 500 s in both the @xmath10 and @xmath12 filters . a log of the individual exposures is given in table 1 . three fields were observed , and a chart showing their relative locations is given in paper i. field 1 was observed only in 1989 , and fields 2 and 3 were observed in 1990 . a known flat - fielding defect afflicts the first 100 columns of the ccd used in both runs ; no data from this section of the ccd were used in the analyses here or in paper i. further details of the observations can be found in paper i. the individual processed frames were measured using the dophot photometry program ( schechter et al . 1993 ) in the same manner as in paper i. some of the frames with slightly poorer seeing than average that were excluded from the analysis of paper i , but they are used here because we could reject individual poor measurements at later stages as required . details of the calibration of the resulting light curves are described below , but these are based on the transformations described in paper i. once the individual ccd images were reduced with dophot , we registered the coordinates to a common system . the typical rms of these transformations was 0.07 pixels , or about 0.04 arcsec . the data for a given field and filter were then combined into databases in which the individual photometric results were tabulated for each object found on every frame . approximately 800 - 1000 stars per field were found in each exposure compared to the 2000 - 3000 stars found in each of the deep images of each field in paper i. there are fewer total epochs for fields 2 and 3 compared to field 1 , but all three fields were observed on at least two separate nights within their respective runs . the photometric results for each image were then shifted to a common system defined by an arbitrarily - chosen template image . after some experimentation , we found that a simple additive shift to the instrumental @xmath10 and @xmath12 magnitudes was sufficient to bring all the images of a given filter and field to a common zero - point . we then determined the photometric transformation of the template images using the calibrated data from paper i for each field . this transformation included the mean color terms found for the three fields , as well as a frame - specific offset used to convert to the standard system . for reasons we describe later , these transformations were not applied to the database photometry at this point . instead , we first identified candidate variables from the pre - calibrated but photometrically - shifted results . a first cut was made by calculating @xmath13 under the hypothesis that the stars are non - variable . although pulsating variables tend to have larger amplitudes in @xmath12 than in @xmath10 , we determined @xmath13 from the @xmath10-band data because we obtained many more epochs in that filter . stars with ` large ' values of @xmath13 were then extracted from the databases . we chose to be rather conservative at this stage , and selected a value of @xmath13 that allowed us to extract about 10% of the total number of stars in the databases for each field . our motivation was to try to ensure complete coverage of all variables at the cost of having to sort through a significant number of non - variables in the process . at this point we also required that any candidate variable must have good photometric results from at least 90% of the individual images . this was particularly important for fields 2 and 3 which have relatively few observations ; any fewer would have made it impossible to identify variables reliably . about 50 - 80 variable - star candidates were identified in each field with this procedure . we then took the photometry of the candidates and did a period search using the string algorithm described by dworetsky ( 1983 ) . the nine best periods were then used to fold the data and the resulting phased light curves were plotted together and inspected . the initial period search was rather broad , ranging from 0.05 to 2.0 days . in this manner we isolated 25 candidate dcs in the three carina fields . we then turned to each individual star and determined a ` best ' period using the same algorithm , but this time searching only in the range 0.05 - 0.10 days . because of our limited time coverage , the correct period is not always obvious ; we estimate an uncertainty of up to 10% in the worst cases . for most of the stars , the periods we derived are the only ones consistent with the photometry and are probably accurate to the stated precision . when there was some ambiguity about the period , we selected the shortest period consistent with the data in the final period range noted above . this procedure probably biases our most uncertain periods to small values . because every field was observed on multiple nights , we deleted some promising candidates that did not repeat consistently even when their phased light curves looked quite convincing . the phased @xmath10-band light curves were used to determine the ephemerides for each candidate ; these are listed in table 2 in the form of the period ( in days ) and the approximate time of minimum light . because of the long interval between the time the data were taken and the present , these quantities are likely to be of little use to phase newer data . we have included these ephemerides in case some other data more contemporaneous to those reported here were obtained and are eventually analyzed . the @xmath12-band light curves were phased using these ephemerides . in a few remaining cases these were found to be inconsistent with the @xmath10-band data , either in the phasing or because the @xmath12-band data did not confirm variability . such stars were rejected as candidate variables . we then calibrated the light curves of the remaining candidates as follows . because the transformations to the standard system involve modest color terms , we could not simply transform the @xmath10 and @xmath12 photometry independently . also , because dcs vary in color during their pulsation cycle , we could not adopt a mean color . to deal with this we took one of the phased light curves ( say , @xmath10 ) for a given star , and for each point calculated the corresponding instrumental magnitude at the same phase from the phased light curve in the other filter ( in this case @xmath12 ) . this was done by linearly interpolating the second light curve about the the desired phase from the first light curve . this procedure was then repeated for the other light curve ( in this case @xmath12 ) . at the end of this process we identified 20 dwarf cepheids in our three fields ; 12 in field 1 , and four each in fields 2 and 3 . a summary of the properties of these 20 variables is provided in table 2 , and their light curves are shown in figure 2 . we have named the stars v1 - 20 . the individual photometry results for every epoch and every faint variable is listed in table 3 . note that the mean @xmath12 and @xmath10 magnitudes and mean @xmath14 colors for the variables are taken from paper i. particularly for the @xmath12 band , we felt it was more reliable to derive the averaged photometry from the co - added images than from the light curves . only two stars , v11 and v12 , were not found in the photometry list from our deep images from paper i. in these cases the mean photometry values in table 2 are obtained from intensity averaged means of the @xmath12 and @xmath10 light curves . note that the colors listed in table 2 are strictly defined as @xmath15 , and this is true of v11 and v12 also . the locations of the carina dcs is shown in figure 3 . the coordinates of the centers of the three fields are listed in the caption to figure 3 . further details can be found in paper i. given the quality of the light curves of the faintest variables ( v7 and v8 ) , we estimate that out procedure was capable of finding variables in this data set to @xmath16 , or about @xmath17 in carina . this limit is about 2 magnitudes brighter than the detection limit of our deep combined data , and may serve as a figure of merit for any future surveys that obtain sufficient exposures ( @xmath18 ) to identify such faint , short - period variables . we have not attempted to determine the completeness of our data as a function of magnitude , location , or photometric amplitude ; our results represent a lower limit to the number of dcs in carina . the locations of the variables in the color - magnitude diagram of carina is shown in figure 4 using the mean photometry from table 2 . two features are apparent from this diagram . first , the dcs do indeed cluster near the upper ms ; we shall discuss this further below . second , the approximate completeness limit for the detection of dcs in these data is well below the mean brightness of the variables we did identify . _ many _ stars above this limit and in the same general location as the dcs are definitely not variable with amplitudes larger than 0.02 - 0.04 mag in the @xmath10 band . we shall also return to the implications of this point below . we have plotted in figure 5 the period - apparent magnitude data for the dwarf cepheids in carina . plotted as solid lines are the fundamental and first - overtone dwarf cepheids from nnl ( they refer to them as sx phe stars in this paper ) for an apparent @xmath10-band distance modulus of 20.17 and a mean metallicity of @xmath19 using their @xmath10-band period - luminosity - metallicity relations for these stars . this modulus is taken from smecker - hane et al . ( 1994b ) who estimate a mean reddening of e@xmath3 = 0.025 . we have also adopted the metallicity proposed by these authors that they claim applies to the younger stars in carina . this value also agrees with the latest spectroscopic estimate of da costa ( 1994 ) . we need to know the abundance for this analysis because nnl explicitly include a metallicity term in their period - luminosity relations for dcs . figure 5 shows that the carina dcs split into two groups that obey distinct period - luminosity ( pl ) relations . the brighter group consists of first - overtone pulsators , while the fainter group contains fundamental pulsators . the principal uncertainties in the data plotted in figure 5 are uncharacteristically in the horizontal axis , i.e. in the the periods . however , it is important to note that no stars can reasonably be switched from the overtone to fundamental group ( or _ vice versa _ ) due to period uncertainties , apart from v2 . from figure 5 it is also clear that the period range of the fundamental pulsators overlaps that of the overtone variables . unlike rr lyr stars , short - period dcs are _ not _ predominately overtone pulsators . inspection of figure 2 also reveals that the light curve shape is not uniquely correlated with the pulsation mode ( which is noted for each variable in the figure ) . thus , one can not simply assume that sinusoidal light curves correspond to overtone pulsators ( see v5 ) , nor that a ` sawtooth ' light curve identifies a fundamental - mode pulsator ( see v12 and v18 ) . because mass determinations for dwarf cepheids require knowledge of the pulsation mode ( e.g. see da costa et al . 1986 ) , some caution is needed to assign the mode . figure 5 also reveals very good agreement between the carina dcs and the nnl period - luminosity relation when we adopt the smecker - hane et al . ( 1994b ) distance , metallicity and reddening noted above . for these parameters the data agree well with the slope of the pl relation from nnl , particularly for the fundamental - mode pulsators which exhibit a wider period range than the overtone pulsators . the data are also consistent with the relative separation of the overtone and fundamental pl relations . all of these points suggest strongly that the carina dcs are ` normal ' for their class , despite their possibly different evolutionary origin ( mateo 1993 ) . mcnamara ( 1995 ) derived a pl relation for dcs and used preliminary results for the carina dcs ( nemec and mateo 1990 ) to derive a distance modulus of @xmath20 and [ fe / h ] = @xmath21 , for e(b@xmath4v ) = 0.025 . for [ fe / h ] = @xmath22 , the mcnamara ( 1995 ) pl relation gives a true modulus of @xmath23 . we can repeat this analysis using the most recent ( and reliable ) data on the carina dcs , and incorporating some other constraints on the metallicity and reddening of the galaxy . for example , if we fix the slope of the pl relations and set the metallicity dependence to the values suggested by nnl , we find a best estimate for the true distance modulus to be 20.07 @xmath5 0.03 for the metallicity and reddening adopted above . this result agrees well with the distance found by smecker - hane et al . ( 1994b ) . more recently , mighell ( 1997 ) has proposed a true modulus of 19.87 @xmath5 0.11 for carina , with a visual extinction of 0.18 ( e@xmath3 = 0.051 ) , and [ fe / h ] = @xmath41.9 . a comparison of the resulting pl relations for these parameters is shown in figure 6 . there is a small systematic offset from the data for the carina variables ( 0.09 @xmath5 0.03 mag ) ; our best estimate of the true modulus is 19.96 @xmath5 0.03 for this reddening and metallicity . none of these results include realistic estimates of the uncertainties in the dc distance scale adopted by nnl which is tied directly to the distance of the metal - poor globular cluster m 15 . nnl adopted a true modulus of 15.03 ; for comparison , durrell and harris ( 1993 ) derive a true modulus of @xmath24 for the cluster . this change results in a carina modulus of 20.13 and 20.02 for the smecker - hane et al . ( 1994b ) and mighell ( 1997 ) reddening / metallicity parameters , respectively . the uncertainties of the distance scale adopted by mcnamara ( 1995 ) are difficult to quantify because the calibrating stars are taken from a number of different sources and clusters ; it is likely that this scale is also uncertain at the 10 - 15% level . we conclude that the best estimate of the distance to carina is approximately @xmath25 based on the dcs alone and assuming [ fe / h ] = @xmath22 and e(b@xmath4v ) = 0.025 . figure 7 shows a region enclosing the upper portion of the main sequence in carina and all 20 newly - discovered dwarf cepheids . we know from paper i that the completeness in this region is quite high and uniform , and that the total field contamination is quite small . thus , we have simply counted the total number of stars in this region ( 844 ) to estimate the frequency of dcs in carina . we find that approximately one of 42 stars in this region of the color - magnitude diagram is a dwarf cepheid , or about 2.5% of all upper main - sequence stars in carina . this is a lower limit because we have probably missed some small - amplitude variables . one consequence of this result has to do with the suitability of dcs as tracers of the sort of stellar population present in carina . there are numerous hints suggesting that the halo has considerable substructure , possibly from the accretion of small dwarfs ( rodgers et al . 1981 ; majewski 1992 ; ct et al . 1993 ; arnold and gilmore 1992 ; preston et al . the discovery of the sgr dwarf galaxy ( ibata et al . 1994 ) is a spectacular example that such accretion does occur even now to some extent . systematically identifying and tracing the substructure has proven to be quite difficult . can dcs help ? in the case of carina , the total area we surveyed for dcs is 0.013 deg@xmath26 , implying a mean surface density of dcs of 1540 @xmath5 340 deg@xmath27 ( poisson error ) . the mean @xmath10-band surface brightness of carina in these fields is about 25.5 mag arcsec@xmath27 . if instead we had viewed the same stellar population found in carina in a structure spread out to an average @xmath10-band surface brightness of 30.5 mag arcsec@xmath28 , we would have found approximately 15 dcs per square degree . such a faint surface brightness is unobservable in any conventional sense , but the detection of the associated dcs _ is _ possible . a deep , wide - field search for dcs might be useful to uncover extended structure beyond the tidal radius of carina , or perhaps other extended structures with similar populations elsewhere in the halo . closer halo structures would be more spread out , consequently requiring less depth but greater areal coverage to identify the dwarf cepheids . in figure 7 we plot the boundaries of the dc instability strip ( is ) from breger ( 1979 ) and from mateo ( 1993 ) superimposed on the cmd of carina . the locations of the carina dcs are consistent with the breger boundaries . most of the dcs lie within the more confined boundaries from mateo which was based on the locations of these variables in other intermediate - age and old stellar systems ( in the latter case , globular clusters with blue stragglers ) . the breger is boundaries are based on the locations of dcs associated with younger populations , and in particular the so - called @xmath7-sct stars which generally have very low amplitudes . these stars are generally drawn from field samples ; random errors in their reddenings and distances contribute to uncertainties in the location of the instability strip edges . because of the limited @xmath12-band phase coverage for the carina variables , the mean colors are sometimes in error by up to 0.1 - 0.2 magnitudes . this introduces some of the scatter visible in figure 7 . the close conformance of the carina dcs to the is boundaries is further evidence that these variables are normal for their class . an obvious feature of figure 7 basically a corollary of the point raised in section 4.2 above is that not all stars within the dwarf - cepheid instability strip are actually dwarf cepheids . we have looked at the light curves of a number of these nonvariables , and they are certainly constant to the precision of the data , or about 0.03 - 0.05 mag at the mean magnitude range of the detected dcs . among rr lyr stars and classical and type ii cepheids , it is generally agreed that _ any _ star in the is will begin to pulsate and be seen as the appropriate sort of variable . this is clearly not true of large - amplitude dcs where we find that only 2 - 3% of the stars in the is are variable with v - band amplitudes larger than 0.03 - 0.05 mag . saha et al . ( 1986 ; hereafter ssm ) first searched for variable stars in carina using photographic plates with the ctio 4 m telescope . their search targeted rr lyr stars and anomalous cepheids in carina and was not deep enough to detect the dcs described here . our data do include some measurements of the rr lyr stars and candidates identified by ssm . a summary of the results of our observations of these stars is provided in table 4 . we confirm the two rr lyr stars for which ssm derived periods and light curves : ssm 74 and ssm 116 . a third variable ssm 65 was not included in our dataset due to its proximity to the central bright star in field 3 . of the 10 suspected variables from ssm that have adequate data in our observations we find that nine are certainly non - variable to @xmath29 mag . for the tenth star ssm 87 our data can not confirm variability because the star is too close to the edge of field 2 . we also find one candidate rr lyr variable that was not identified by ssm which we call ssm 176 to continue ssm s numbering sequence . the locations in the color - magnitude diagram of the two well - observed rr lyr stars from ssm and our two additional candidate variables are shown in figure 4 . their locations in each carina field are noted in figure 3 along with the locations of the other apparently non - variables from ssm . our data are inadequate to determine periods much longer than 0.25 days without severe aliasing , so we can not comment on the periods for ssm 74 and 116 , and we are unable to determine a reliable period for ssm 176 . the location of ssm 176 in the cmd ( figure 4 ) is consistent with its identification as a possible rr lyr star . we have identified an additional variable , ssm 177 , that appears to exhibit portions of two eclipses . we can not determine its period with any precision and its image appears to be composite in many of the individual ccd frames . we regard this star as only a weak candidate eclipsing contact binary that clearly requires confirmation . the photometric data for both ssm 176 and 177 are listed in table 4 and their locations are marked in figure 3 . if the numbers of rr lyr stars and dcs in our fields are typical of carina as a whole , we find the ratio of these variables to be @xmath30 , or 0.13 @xmath5 0.08 ( the fraction in the numerator represents ssm 176 which is the uncertain rr lyr star found in our study ) . of course , the frequency of rr lyr stars is highly variable and the numbers of dcs requires a significant intermediate - age population or blue stragglers . nonetheless , this result does suggest that dcs can greatly outnumber rr lyr stars in populations of satellites in the outer galactic halo . we have discovered 20 dwarf cepheids in the carina dsph galaxy . these are by far the most distant ( @xmath0100 kpc ) and faintest ( @xmath31 ) dcs known . the carina dcs obey a well - defined period - luminosity relation , allowing us to distinguish between overtone and fundamental pulsators in nearly every case . unlike rr lyr stars , the pulsation mode turns out to be uncorrelated with light - curve shape , nor do the overtone pulsators tend towards shorter periods compared to the fundamental pulsators . using the pl relations from nnl , we find good agreement with the distance / metallicity parameters advocated by smecker - hane et al . ( 1994b ) : @xmath32 , e@xmath3 = 0.025 , and [ fe / h ] = @xmath22 and for the results of mighell ( 1997 ) using slightly different parameters . our best estimate of the carina true distance modulus is @xmath33 , where the uncertainty includes the errors of the dc distance scale ( nnl ; mcnamara 1995 ) . this corresponds to a distance of @xmath34 kpc . the frequency of dcs among upper main sequence stars in carina is approximately 3% in the region enclosed in figure 4 . we find that the ratio of dwarf cepheids to rr lyr stars in carina is 0.13 @xmath5 0.10 , though of course the numbers of such stars are highly sensitive to the details of the star - formation history of carina and the evolution of the horizontal branch , respectively . the fact that we could discover these faint , short - period variables using a relatively poor ccd in mediocre seeing conditions ( by modern standards ) on a 4 m telescope suggests that these stars could conceivably be used as tracers of extended structure throughout the galactic halo . about one out of 40 upper main sequence stars in carina is a dc . if carina has a tidal tail or an extended halo ( see lynden - bell and lynden - bell 1995 for some motivation why this may occur ; kuhn et al . 1996 ) , dcs may prove to be an excellent way to efficiently trace such extended structures . as noted in section 4.2 , if carina were spread out to 1% of its current surface brightness we would expect to see about 15 of these stars per square degree . more generally , any survey of the galactic halo should identify one dcs per 40 so - called blue - metal - poor ( bmp ) star . these stars were first described by preston et al . ( 1994 ; though their lineage extends back at least to rodgers et al . 1981 ) and may represent an intermediate - age population that comprises up to 10% of the galactic halo . based on one detected dc , preston and landolt ( 1998 ) find a frequency of about half this in their sample of spectroscopically studied bmp stars ; this result is not statistically different from the frequency we find in carina . in an on - going survey of the halo field , harding , morrison , olszewski and mm find approximately 15 bmp stars per square degree to a limiting magnitude of @xmath35 . thus , one halo dc should occur in every three - square degree region covered by this survey , more frequently for deeper surveys . eventually , the numbers of bmp stars and dcs will fall off towards fainter magnitudes as the halo density drops and the frequency of all luminous halo stars becomes too low . precisely at what magnitude this occurs depends on the structural parameters of the halo . for a ` reasonable ' halo one expects the number of bmp and therefore dcs to become negligible by @xmath36 in a halo devoid of substructure . other dsph stars may also contain dcs ; the prerequisite seems to be the existence of a significant intermediate - age population . like carina , both leo i ( gallart et al . 1998 ) and fornax ( e.g. beauchamp et al . 1995 ; demers et al . 1998 ) contain large numbers of stars that appear to be considerably younger than globular clusters . all of the remaining dsph systems apart from umi contain at least a modest intermediate - age population ( da costa 1998 ) . it is therefore very likely that many of these systems contain dcs . leo i is probably too distant to survey for these variables from the ground , but such observations should be feasible for fornax . located about 35 kpc further than carina , the dcs in fornax are likely to have @xmath10-band magnitude of about 23 - 24 . though challenging , such observations are certainly possible using current ccds on 4m - class telescopes . much of what we have written about the carina dcs is necessarily a generalization based on the variables of only one galaxy , so it would be of great interest to broaden our understanding of these stars by searching in other systems . mar 11 1989 & 596.54913 & v & 1 & 420 & 1.09 & 1.6 mar 11 1989 & 596.55781 & b & 1 & 600 & 1.10 & 1.9 mar 12 1989 & 597.52663 & v & 1 & 420 & 1.08 & 1.3 mar 12 1989 & 597.53255 & v & 1 & 420 & 1.08 & 1.3 mar 12 1989 & 597.53847 & v & 1 & 420 & 1.08 & 1.3 mar 12 1989 & 597.54738 & v & 1 & 420 & 1.09 & 1.5 mar 12 1989 & 597.55332 & v & 1 & 420 & 1.10 & 1.4 mar 12 1989 & 597.55928 & v & 1 & 420 & 1.11 & 1.4 mar 12 1989 & 597.56528 & v & 1 & 420 & 1.12 & 1.5 mar 12 1989 & 597.57125 & v & 1 & 420 & 1.13 & 1.5 mar 12 1989 & 597.57784 & v & 1 & 500 & 1.14 & 1.4 mar 12 1989 & 597.58480 & b & 1 & 500 & 1.15 & 1.5 mar 12 1989 & 597.59195 & v & 1 & 500 & 1.17 & 1.5 mar 12 1989 & 597.59925 & b & 1 & 500 & 1.19 & 1.5 mar 12 1989 & 597.60628 & v & 1 & 500 & 1.21 & 1.4 mar 12 1989 & 597.61348 & b & 1 & 500 & 1.23 & 1.4 mar 12 1989 & 597.62072 & v & 1 & 500 & 1.26 & 1.6 mar 12 1989 & 597.62784 & b & 1 & 500 & 1.28 & 1.5 mar 12 1989 & 597.63496 & v & 1 & 500 & 1.31 & 1.5 mar 12 1989 & 597.64204 & b & 1 & 500 & 1.34 & 1.5 mar 12 1989 & 597.64928 & v & 1 & 500 & 1.38 & 1.5 mar 12 1989 & 597.65659 & b & 1 & 500 & 1.42 & 1.7 mar 12 1989 & 597.66439 & v & 1 & 500 & 1.46 & 1.8 mar 13 1989 & 598.50585 & b & 1 & 500 & 1.07 & 1.5 mar 13 1989 & 598.51289 & v & 1 & 500 & 1.07 & 1.3 mar 13 1989 & 598.51988 & b & 1 & 500 & 1.07 & 1.4 mar 13 1989 & 598.52684 & v & 1 & 500 & 1.08 & 1.3 mar 13 1989 & 598.53380 & b & 1 & 500 & 1.08 & 1.3 mar 13 1989 & 598.54074 & v & 1 & 500 & 1.09 & 1.3 mar 13 1989 & 598.54775 & b & 1 & 500 & 1.10 & 1.5 mar 13 1989 & 598.55469 & v & 1 & 500 & 1.10 & 1.4 mar 13 1989 & 598.56190 & b & 1 & 500 & 1.12 & 1.4 mar 13 1989 & 598.56885 & v & 1 & 500 & 1.13 & 1.5 mar 13 1989 & 598.57597 & b & 1 & 500 & 1.14 & 1.5 mar 13 1989 & 598.58990 & b & 1 & 500 & 1.17 & 1.4 mar 13 1989 & 598.59690 & v & 1 & 500 & 1.19 & 1.4 mar 13 1989 & 598.60393 & v & 1 & 500 & 1.21 & 1.6 mar 13 1989 & 598.61083 & v & 1 & 500 & 1.23 & 1.5 mar 13 1989 & 598.61762 & v & 1 & 500 & 1.25 & 1.5 mar 13 1989 & 598.62483 & v & 1 & 500 & 1.28 & 1.5 mar 13 1989 & 598.63167 & v & 1 & 500 & 1.31 & 1.4 mar 13 1989 & 598.63844 & v & 1 & 500 & 1.34 & 1.6 mar 13 1989 & 598.64522 & v & 1 & 500 & 1.37 & 1.5 mar 13 1989 & 598.65202 & v & 1 & 500 & 1.41 & 1.6 mar 13 1989 & 598.65881 & v & 1 & 500 & 1.45 & 1.5 apr 1 1990 & 982.53487 & b & 2 & 500 & 1.17 & 1.2 apr 1 1990 & 982.54197 & v & 2 & 500 & 1.19 & 1.2 apr 1 1990 & 982.54906 & b & 2 & 500 & 1.21 & 1.2 apr 1 1990 & 982.55601 & v & 2 & 500 & 1.23 & 1.2 apr 1 1990 & 982.56295 & b & 2 & 500 & 1.26 & 1.2 apr 1 1990 & 982.56981 & v & 2 & 500 & 1.28 & 1.2 apr 1 1990 & 982.57685 & b & 2 & 500 & 1.31 & 1.2 apr 1 1990 & 982.58374 & v & 2 & 500 & 1.34 & 1.2 apr 1 1990 & 982.59077 & b & 2 & 500 & 1.38 & 1.2 apr 1 1990 & 982.59763 & v & 2 & 500 & 1.41 & 1.3 apr 1 1990 & 982.60461 & b & 2 & 500 & 1.45 & 1.3 apr 1 1990 & 982.61153 & v & 2 & 500 & 1.50 & 1.2 apr 1 1990 & 982.62557 & v & 3 & 500 & 1.60 & 1.3 apr 1 1990 & 982.63237 & v & 3 & 500 & 1.66 & 1.3 apr 1 1990 & 982.63919 & v & 3 & 500 & 1.72 & 1.5 apr 1 1990 & 982.64587 & v & 3 & 500 & 1.78 & 1.4 apr 1 1990 & 982.65249 & v & 3 & 500 & 1.86 & 1.4 apr 1 1990 & 982.65912 & v & 3 & 500 & 1.93 & 1.4 apr 1 1990 & 982.66654 & v & 3 & 600 & 2.03 & 1.4 apr 1 1990 & 982.67445 & v & 3 & 600 & 2.15 & 1.3 apr 1 1990 & 982.68247 & v & 3 & 600 & 2.28 & 1.4 apr 2 1990 & 983.50585 & v & 3 & 500 & 1.12 & 1.2 apr 2 1990 & 983.52660 & b & 3 & 500 & 1.16 & 1.1 apr 2 1990 & 983.53704 & v & 3 & 500 & 1.19 & 1.1 apr 2 1990 & 983.54415 & v & 3 & 500 & 1.21 & 1.1 apr 2 1990 & 983.55078 & v & 3 & 500 & 1.23 & 1.1 apr 2 1990 & 983.55740 & v & 3 & 500 & 1.25 & 1.2 apr 2 1990 & 983.56403 & v & 3 & 500 & 1.27 & 1.2 apr 2 1990 & 983.57305 & b & 3 & 500 & 1.31 & 1.2 apr 2 1990 & 983.57987 & b & 3 & 500 & 1.34 & 1.3 apr 2 1990 & 983.58656 & b & 3 & 500 & 1.37 & 1.3 apr 2 1990 & 983.59320 & b & 3 & 500 & 1.40 & 1.3 apr 2 1990 & 983.59982 & b & 3 & 500 & 1.44 & 1.4 apr 2 1990 & 983.61250 & v & 2 & 500 & 1.52 & 1.4 apr 2 1990 & 983.61920 & v & 2 & 500 & 1.57 & 1.3 apr 2 1990 & 983.62591 & v & 2 & 500 & 1.62 & 1.5 apr 2 1990 & 983.63258 & v & 2 & 500 & 1.68 & 1.6 apr 2 1990 & 983.64425 & v & 2 & 500 & 1.80 & 1.2 vapr 2 1990 & 983.65091 & v & 2 & 500 & 1.87 & 1.2 apr 2 1990 & 983.65755 & v & 2 & 500 & 1.95 & 1.3 apr 2 1990 & 983.66424 & v & 2 & 500 & 2.04 & 1.4 ccccccccccc 1 & 1 & 6 39 50.0 & @xmath450 58 55 & 22.70 & 22.97 & 0.27 & 0.05960 & 597.4661 & 0.30 & o 2 & 1 & 6 39 57.1 & @xmath450 58 21 & 23.10 & 23.43 & 0.33 & 0.04792 & 597.4666 & 0.45 & ? 3 & 1 & 6 39 51.6 & @xmath450 58 16 & 22.92 & 23.21 & 0.29 & 0.07697 & 597.4565 & 0.30 & f 4 & 1 & 6 39 43.1 & @xmath450 57 46 & 23.02 & 23.48 & 0.21 & 0.05991 & 597.4537 & 0.45 & f 5 & 1 & 6 39 49.9 & @xmath450 57 28 & 23.06 & 23.34 & 0.28 & 0.05910 & 597.4596 & 0.50 & f 6 & 1 & 6 39 45.1 & @xmath450 57 15 & 22.79 & 23.10 & 0.31 & 0.07004 & 597.4868 & 0.35 & f 7 & 1 & 6 39 50.2 & @xmath450 56 44 & 23.20 & 23.53 & 0.33 & 0.04872 & 597.4875 & 0.45 & f 8 & 1 & 6 39 43.7 & @xmath450 56 44 & 23.23 & 23.44 & 0.21 & 0.05275 & 597.4465 & 0.50 & f 9 & 1 & 6 40 01.6 & @xmath450 56 35 & 22.79 & 23.10 & 0.31 & 0.07542 & 597.4659 & 0.45 & f 10 & 1 & 6 39 54.2 & @xmath450 56 22 & 23.03 & 23.37 & 0.34 & 0.05899 & 597.4725 & 0.45 & f 11 & 1 & 6 39 49.4 & @xmath450 55 59 & 22.75 & 23.00 & 0.25 & 0.07674 & 597.4478 & 0.50 & f 12 & 1 & 6 39 50.2 & @xmath450 55 45 & 22.83 & 22.96 & 0.12 & 0.05851 & 597.4562 & 0.35 & o 13 & 2 & 6 40 16.8 & @xmath450 54 40 & 22.93 & 23.12 & 0.19 & 0.06716 & 982.4454 & 0.60 & f 14 & 2 & 6 40 05.4 & @xmath450 53 42 & 22.45 & 22.66 & 0.21 & 0.0626 & 982.4820 & 0.85 & o 15 & 2 & 6 40 03.2 & @xmath450 52 45 & 22.97 & 23.29 & 0.32 & 0.0578 & 982.4359 & 0.80 & f 16 & 2 & 6 39 56.0 & @xmath450 51 29 & 22.81 & 23.05 & 0.24 & 0.065 & 982.5192 & 0.90 & f 17 & 3 & 6 40 47.3 & @xmath450 52 51 & 22.95 & 23.20 & 0.25 & 0.06033 & 982.5790 & 0.70 & f 18 & 3 & 6 40 44.0 & @xmath450 53 13 & 22.74 & 22.97 & 0.23 & 0.05514 & 982.5512 & 0.60 & o 19 & 3 & 6 40 42.5 & @xmath450 53 15 & 22.63 & 22.90 & 0.27 & 0.06489 & 982.5591 & 0.70 & o 20 & 3 & 6 40 33.7 & @xmath450 52 11 & 22.84 & 23.12 & 0.28 & 0.06738 & 982.5756 & 0.85 & f * figure 1 * a color - magnitude diagram of carina from the data presented in paper i. the solid curved line is an isochrone from bertelli et al . ( 1994 ) representing a population with an age of 15 gyr , a metallicity of [ fe / h ] = @xmath42.23 , a true distance modulus of 20.1 and e@xmath3 = 0.025 . the parallelogram encloses the region where pop ii dwarf cepheids ( sx phe stars ) are found in other systems from the data given in nemec et al . 1994 . * figure 2 * light curves of the 20 dwarf cepheids found in carina . for each star the upper plot shows the @xmath10-band light curve , and the lower plot shows the @xmath12-band light curve phased with the ephemeris determined from the @xmath10-band data . the period ( from table 2 ) and pulsation mode of each individual star is also noted . ` f ' represents stars pulsating in the fundamental mode , ` o ' represents overtone pulsators , and ` ? ' represents cases where the mode is uncertain . * figure 3 * charts of the three carina fields showing the locations of the newly - discovered dwarf cepheids ( denoted v1-v20 ) and the known rr lyr stars ( ssm 74 and ssm 116 ) and new rr lyr candidates ( rr 176 ) . the rr lyr candidates from saha et al . ( 1986 ) that we determine are _ not _ variable are denoted by their identification number from that paper , but with no alphabetic prefix . ssm 177 is a candidate eclipsing binary as discussed in the text . figures a , b , and c , show fields 1 , 2 , and 3 , respectively . note that v14 is the faint star located between the two brighter stars at its location . the 1950.0 coordinates of these three fields are @xmath38 , @xmath39 for field 1 , @xmath40 , @xmath41 for field 2 , and @xmath42 , @xmath43 for field 3 . for field 1 , north is up , and east is to the left ; for the charts for fields 2 and 3 , north is to the left , and east is down . each field is 3.9 arcmin on a side . further details can be found in paper i. * figure 4 * the color - magnitude diagram of carina from figure 1 . the square symbols show the locations of the fundamental - mode dcs using the mean magnitudes and colors from table 2 ; the circles denote overtone pulsators ; the triangle is the one star whose mode is uncertain from the pl relation ( figures 5 and 6 ) . the three rr lyr stars are shown as bold crosses using the data from table 4 and ssm 177 , a binary candidate , is shown as a large open square . the enclosed region is described in the text ; the horizontal dashed line represents our approximation of the faint magnitude limit to which we could identify dwarf cepheids with amplitudes @xmath44 0.25 mag . * figure 5 * the period - apparent magnitude relation for the dwarf cepheids in carina . the solid lines correspond to the plm relations proposed by nnl for fundamental - mode and overtone dwarf cepheids ( the lower and upper lines , respectively ) using the true distance modulus ( 20.09 ) , e@xmath3 ( 0.025 ) and metallicity for the younger population ( [ fe / h ] = @xmath42.0 ) proposed by smecker - hane et al . ( 1994b ) . the dashed lines represent the uncertainty of 0.06 mag in this distance modulus . the slopes and relative offsets of the plm relations agree well with the carina observations for these assumed distance / reddening / metallicity values . * figure 6 * the same as figure 6 except now for the parameters proposed by mighell ( 1997 ) : @xmath45 , e@xmath46 , and [ fe / h ] = @xmath41.9 . the mean offset of these pl relations from the data is 0.09 @xmath5 0.02 mag such that the best - fit distance modulus is 19.98 @xmath5 0.02 based on the dcs and the nnl plm relations alone .
certain strong relations between birational minimal models have long been suspected . kollr @xcite proved that 3-dimensional birational @xmath1-factorial terminal minimal models have isomorphic hodge structures on intersection cohomology groups by employing intersection cohomology theory . for higher dimensions , it was proven by batryev @xcite that birational calabi - yau manifolds have the same betti numbers by using p - adic integration . this result was extended by wang @xcite to the classes of k - equivalent smooth projective manifolds which in particular applies to birational smooth minimal models . with the aid of motivic integration developped by denef and loeser these results were further refined @xcite to the equivalence of hodge numbers . most of the results nowadays are either established by certain integration formalism to obtain `` numerical results '' @xcite , or are restricted to particular cases of flops to obtain rather strong `` geometric conclusions '' @xcite,@xcite . a series of conjectures concerning k - equivalent proper smooth varieties are proposed in ( * ? ? ? * section 6 ) and explained in ( * ? ? ? * section 4.3 ) . the first one of these conjecture is to find an algebraic correspondence between k - equivalent projective smooth varieties that identifies their chow motives . this article makes a progress in this direction . the main result ( [ main ] ) of this paper is of a style intermediating between these two ends , stating that two smooth projective varieties related by a _ semismall flop _ ( ) have isomorphic chow motives with coefficients in a local ring . following several ideas of @xcite , @xcite , @xcite and @xcite , we adopt perverse sheaves to compare the motives of smooth varietes under semismall flops . for motives with coefficients @xmath1 , the decomposition theorem is available and it suffices to compare local systems over respective strata ( [ intcmp ] ) . this can be achieved by employing arc spaces to compare irreducible components of the fibres . the decomposition theorem can fail , even for fields of positive characteristics . nevertheless , this difficulty can be remedied for coefficients in local rings by the extensibility of morphisms between sheaves from an zariski open subset ( [ cat ] ) and then to the ability to lift invertibles in the category of finite associative algebras over a noetherian commutative local ring ( [ lift ] ) . these properties permit an abstract gluing of isomorphisms of perverse sheaves across strata . an immediate consequence ( [ tor ] ) of the main result is that the singular cohomology groups @xmath2 of a smooth variety @xmath3 are unchanged under semismall flops . i am indebted to professor chin - lung wang for his supervision and innumerable inspiring discussions , and i am thankful to chen - yu chi , luc illusie , ming - lun hsieh , hui - wen lin , jeng - daw yu and many others for their advices . we will always work in the category of separated schemes of finite type over the complex numbers @xmath4 , an object whereof is simply called a variety henceforth . we will mostly compare complexes of sheaves in the bounded derived category @xmath5 of cohomologically constructible complexes sheaves of @xmath0 of finite tor - dimension on @xmath3 with coefficients in a ( commutative unital ) noetherian local ring @xmath0 . see @xcite and ( * ? ? ? * expos 2 ) for a reference . henceforth we will simply call sheaves in place of sheaves of @xmath0-modules we summerise briefly here important properties of this category . recall that an ( algebraic ) _ whitney stratification _ of an algebraic variety @xmath3 is a partition ( called stratification ) of @xmath3 into finite disjoint collection of zariski locally closed subsets @xmath6 , whereof each element ( called stratum ) @xmath7 is a smooth subvarieties such that the closure @xmath8 is a union of strata . every stratification admits a whitney stratification as refinement . according to @xcite , by a stractification of a algebraic variety @xmath3 we mean a whitney stratification of equidimensional strata of @xmath3 . we call a sheaf @xmath9 of @xmath0-modules on @xmath3 a _ locally constant constructible sheaf _ , if there is an open covering @xmath10 ( in the classical topology ) of @xmath3 such that the restriction of @xmath9 to each @xmath11 is a finitely generated free @xmath12-module . a sheaf @xmath9 is called _ constructible _ if there is a whitney stratification of @xmath3 along each stratum of which @xmath9 is locally constant constructible . a complex of sheaves is called _ cohomologically constructible _ if every cohomology sheaf of it is constructible . a complex of sheaves @xmath13 is called _ of finite torsion dimension _ if each of its stalk @xmath14 is isomorphic in the bounded derived category of @xmath0-modules @xmath15 to a complex of projective @xmath0-modules . let @xmath16 denote the full subcategory of @xmath17 of bounded cohomologically constructible complexes of finite torsion dimension ( ctf for short ) . amongst the most important topological properties of an algebraic map is the preservation of constructibility . given an morphism of varieties @xmath18 , the six operations are induced on the subcategories : f _ * , f_!d^b_(x , ) d^b_(y , ) , f^ * , f^!d^b_(y , ) d^b_(x , ) , as well as the derived tensor product ^ld^b_(x , ) d^b_(x , ) d^b_(x , ) and the internal homomorphism _ x ( , ) d^b_(y , ) ^d^b_(x , ) d^b_(x , ) . proofs of the constructibility can be found in @xcite . as an immediate consequence , @xmath19 is a ctf @xmath0-complex for any pair @xmath20 . there is a natural biduality functor @xmath21 defined by @xmath22 , where @xmath23 is the dualising complex , and @xmath24 is the constant map . perverse t - structures of middle perversity are defined on these categories of the following manner : ^0 = \ { kd^b_(x , ) : ( isx ) ( k > s)^k i^ * k = 0 } and ^0 = \ { kd^b_(x , ) : ( isx ) ( k < s)^k i^ ! k = 0 } . the heart @xmath25 of this t - structure is simply called the category of perverse sheaves on @xmath3 . we remark that @xmath25 is a noetherian abelian category and that it is not in general artinian . see @xcite for detailed properties of arc spaces . to each variety @xmath3 we can associate a projective system of schemes @xmath26 , called the _ @xmath27^th^ truncated arc space _ of @xmath3 , which represents the functor @xmath28/t^{m+1 } , x \right)$ ] that sends a @xmath4-scheme @xmath7 to @xmath29/t^{m+1 } , x \right)$ ] . let @xmath30 be the limit of this system , which represents the functor @xmath31 . when @xmath3 is smooth of equidimension @xmath32 , the natural projection @xmath33 is a zariski locally trivial @xmath34-fibration over @xmath3 . when there is a birational morphism @xmath18 between smooth varieties , one can read the structure of the system of induced morphisms @xmath35 from the relative canonical divisor @xmath36 and vice versa . we have an important lemma about the structure of a birational morphism ( * ? ? ? * lemma 3.4 ) . [ str ] let @xmath3 and @xmath37 be smooth algebraic varieties over @xmath38 , of pure dimension @xmath39 and let @xmath18 be a birational morphism . for @xmath38 in @xmath40 , let ( x)^k = \ { ( x ) : ( _ t ) ( ) = k ( f ) ( y)}. for @xmath41 , let @xmath42 be the morphism induced by @xmath43 , and let @xmath44 be the image of @xmath45 in @xmath26 . then , for all @xmath38 in @xmath40 with @xmath46 , the following holds . 1 . the set @xmath44 is a union of fibres of @xmath47 . the restriction of @xmath47 to @xmath44 is a piecewise trivial fibration with fibre @xmath48 onto its image . besides @xmath44 , we introduce the notations ( x)^k = \ { ( x ) : ( _ t ) ( ) k ( f ) ( y ) } , + ( x)^ > k = \ { ( x ) : ( _ t ) ( ) > k ( f ) ( y ) } , etc , to indicate the subvarieties of arcs of corresponding orders , and denote their images by @xmath49 , @xmath50 , etc . moreover , for a subset @xmath51 , we add the restriction sign @xmath52 to indicate those arcs @xmath53 on @xmath3 that are originated from @xmath54 , that is to say @xmath55 . for example , @xmath56 . notations concerning order of arcs in the previous paragraph also applies to this situation . in this section , we will consider complexes of sheaves of abelian groups and the bounded derived categories @xmath57 thereof . let @xmath3 and @xmath58 be smooth varieties of dimension @xmath32 , @xmath37 a varieties , and @xmath59 and @xmath60 be proper birational morphisms . suppose that @xmath3 and @xmath58 are k - equivalent through @xmath61 . that is to say , given any common resolution @xmath62 and @xmath63 , the relative canonical divisors @xmath64 and @xmath65 are equal . we suppose further that @xmath43 and @xmath66 are _ semismall _ in the sense that @xmath67 and @xmath68 . in this case @xmath69 is refered to as a _ semismall k - equivalence_. one important example we shall bear in mind is that when @xmath18 is a semismall crepant resolution and @xmath60 is a flop , @xmath3 and @xmath58 are k - equivalent through @xmath61 . in this case @xmath69 is refered to as a _ semismall flop_. a stratum @xmath70 is called _ @xmath43-relevant _ if @xmath71 . the goal of this section is to prove the following [ intcmp ] . [ intcmp ] for a sufficiently fine stratification @xmath72 of @xmath37 , on each relevant stratum @xmath73 of dimension @xmath39 , there is an isomorphism @xmath74 . taking the dual local system , [ dual ] under the assumptions of [ intcmp ] , let @xmath75 be the embedding . there is an isomorphism @xmath76 here is a plan of the proof . given an initial stratification @xmath72 , we shall recursively study the isomorphism and at the same time refine the stratification and the common resolution @xmath77 if necessary . + to be more precise , assume the comparison is done on a union of strata , which is zariski open in @xmath37 . we shall pick a stratum dense , say @xmath7 , in the complement , blow up @xmath77 accordingly , and establish the comparison on a zariski open subset of @xmath7 in case @xmath7 is relevant , by means of arc spaces . then we shall refine the stratification by splitting up @xmath7 into the zariski open subset and some other smooth zariski locally closed subvarieties . by noetherian induction , this will prove [ intcmp ] . let @xmath78 be a zariski locally closed smooth conncected subvariety , of dimension @xmath39 . suppose that @xmath79 and @xmath80 are ( classically ) locally constant constructible , and that @xmath43 and @xmath66 are flat over @xmath7 . we begin with some observations on the irreducible components . consider the projection _ m , f^-1(t)_m(x)\|_f^-1(t ) f^-1(t ) from the space of @xmath27^th^ truncated arcs with origin in @xmath81 , to @xmath81 . [ toarc ] under the above assumptions , the irreducible components of @xmath82 correspond to those of @xmath81 under @xmath83 , and that r_m , f^-1(t ) , ! . let @xmath84 be an irreducible component . then @xmath85 is a zariski locally trivial @xmath34-fibration over @xmath86 , and hence is irreducible . consequently , the irreducible components of @xmath82 correspond to those of @xmath81 under the projection @xmath87 . since it is a @xmath34-fibration , by the base change property for exceptional push - forwards , the projection @xmath83 induces r_m , f^-1(t ) , ! . in regards to [ toarc ] , we have r_m , f^-1(t ) , ! as well as r_m , f^-1(t ) , ! and r_m , h^-1(t ) , ! . the situation is indicated below & _ m(z)\|_h^-1(t ) & + _ m(x)\|_f^-1(t ) & h^-1(t ) & _ m(x)\|_f^-1(t ) + f^-1(t ) & & f^-1(t ) + the next step is to establish relations between the irreducible components of @xmath88 and those of @xmath82 . we shall see that each component of @xmath82 is dominated by one certain irreducible component of @xmath88 . for each irreducible component @xmath89 , let \(f ) = \{k : _ m(z)^kf } . [ cmpcomp ] under the above assumptions , assume in addition that @xmath90 for all irreducible components @xmath91 . then @xmath92 , and that the equality holds exactly when @xmath93 dominates an irreducible component of maximal dimension @xmath94 . by [ str ] , the additional assumption @xmath90 make arcs @xmath95 of contact order @xmath96 all stabilised(i.e . there is a piecewise affine fibration structure ) , in particular , @xmath97 is a zariski locally trivial @xmath98-fibration when restricted to @xmath99 with image @xmath100 . therefore ( _ m(z)^(f ) f)- ( _ m(x)^(f ) _ m(g)(f ) ) = ( f ) . on the other hand , @xmath101 is zariski open and dense in @xmath93 , and that @xmath102 is zariski open and dense in @xmath103 , so \(f ) & = ( _ m(z)^(f ) f ) - ( _ m(x)^(f ) _ m(g)(f ) ) + & = f - _ m(g)(f ) . recall that the assumption on semismallness forces that @xmath104 , and hence @xmath105 . \(f ) = f - _ m(g)(f ) f - 2 - mn . the equality holds exactly when @xmath106 . in that case , @xmath103 is of maximal dimension in @xmath82 . [ cmpcomp1 ] under the assumptions of [ cmpcomp ] , @xmath97 induces a bijection & \ { : e = 2 + mn } + & \ { : ( f ) = f - 2 - mn } in regards to [ cmpcomp ] , we are left to prove that each irreducible component @xmath107 is dominated by exactly one irreducible component of @xmath88 . this is clear by the surjectivity of @xmath97 and the fact that @xmath97 is an @xmath98-fibration over an zariski open set of @xmath86 . for convenience , we may and we shall refine the resolution as @xmath108 , which is an isomorphism outside the closure @xmath109 , such that @xmath110 is a sum of divisors , and that @xmath111 for all @xmath93 . the condition then becomes @xmath112 . we rename @xmath113 . [ cmpcomp1 ] describes explicitly the relation between components of importance , the difficulty to the comparison being that @xmath97 is only an @xmath114-fibration on an zariski open subset . this difficulty is evitable by restricting everything to an zariski open subset of @xmath7 . set u = \ { tt : ( _ m(x)^ > \|_f^-1(t ) ) < 2 + mn,(_m(x)^ > \| _ f^-1(t ) ) < 2 + mn}. [ res ] under the assumptions of [ cmpcomp ] , @xmath115 is zariski open and dense in @xmath7 . put v & = \ { tt : ( _ m(x)^ > \|_f^-1(t ) ) < 2 + mn } + v & = \ { tt : ( _ m(x)^ > \| _ f^-1(t ) ) < 2 + mn } , so that @xmath116 . then it suffices to verify that @xmath117 and @xmath118 are zariski open and dense in @xmath7 . we shall do that for @xmath117 only . we may assume that @xmath7 is irreducible . firstly , @xmath117 is nonempty . if @xmath119 , then by the piecewise fibration structure [ str ] @xmath120 for all @xmath121 , so ( _ m(z)^ > \|_h^-1(t ) ) + 1 + 2 + mn + d = ( m+1)n , which is absurd . thus @xmath122 . we study the set @xmath117 by applying @xmath123 to the sequence 0 _ _ m(x)^\|_f^-1(t ) _ _ m(x)\|_f^-1(t)__m(x)^ > \|_f^-1(t ) 0 . then r^(n - d)+2mn ( f_m , f^-1(t ) ) _ ! _ _ m(x)\|_f^-1(t)r^(n - d)+2mn ( f_m , f^-1(t ) ) _ ! _ _ \|_f^-1(t ) 0 . by [ toarc ] , the former sheave is r^(n - d)+2mn ( f_m , f^-1(t ) ) _ ! _ _ m(x)\|_f^-1(t)r^n - df__f^-1(t ) , which is locally constant constructible by assumption . according to the `` semicontinuity '' of quotients of locally constant sheaves , the set of @xmath124 at which the stalk is zero 0 = ( r^(n - d)+2mn ( f_m , f^-1(t ) ) _ ! _ _ m(x)^ > \|_f^-1(t))_t = h^(n - d ) + 2mn_c ( _ \|_f^-1(t ) , ) is zariski open in @xmath7 . since @xmath125 , @xmath126 exactly when @xmath127 . the set is exactly @xmath117 , whence @xmath117 is zariski open . by restricting to @xmath128 and its image @xmath129 , we arrive at the new situation indicated below : & _ m(z)^\|_h^-1(u ) & + _ m(x)^\|_f^-1(u ) & h^-1(u ) & _ m(x)^\|_f^-1(u ) + f^-1(u ) & & f^-1(u ) + & u & + according to the structure theorem ( [ str ] ) , the morphisms @xmath130 and @xmath131 are piecewise trivial @xmath114-fibrations . therefore there are zariski open dense @xmath132 and @xmath133 such that @xmath134 and @xmath135 are trivial @xmath114-fibrations . let @xmath136 and @xmath137 be the corresponding complements . finally , pick @xmath138 such that the fibre of @xmath139 and of @xmath140 over each point of @xmath141 is of dimension @xmath142 . [ cohomcmp ] under the assumptions of [ cmpcomp ] , in the above diagramme , there is an isomorphism on @xmath141 , @xmath143 . in this context , leray spectral sequence for @xmath144 then gives @xmath145 since @xmath97 is trivial @xmath114-fibration over @xmath117 . @xmath146 applying @xmath147 to the relative sequence on @xmath148 0_v _ _ m(x)^\|_f^-1(u ) _ v^c 0 , we have @xmath149 applying @xmath147 to the relative sequence on @xmath148 0__m(x)^\|_f^-1(u ) _ _ m(x)\|_f^-1(u ) _ _ m(x)^ > \|_f^-1(u ) 0 , we have then @xmath150 and the latter is isomorphic to @xmath151 combining equations [ eq1 ] , [ eq2 ] , [ eq3 ] , [ eq4 ] and [ eq5 ] , we get & ( r^(n - d ) + 2mn+ 2 ( h_m , h^-1(u ) ) _ ! _ _ m(z)^\|_h^-1(u))\|_w + & ( r^(n - d ) + 2mn+ 2 ( f_m , f^-1(u)_m(g ) ) _ ! + & ( r^(n - d ) + 2mn ( f_m , f^-1(u))_!_v)\| _ w + & ( r^(n - d ) + 2mn ( f_m , f^-1(u))_!__m(x)^\|_f^-1(u))\| _ w + & ( r^(n - d ) + 2mn ( f_m , f^-1(u))_!__m(x)\|_f^-1(u))\| _ w + & ( r^n - df _ _ f^-1(u))\| _ w + similarly , ( r^(n - d ) + 2mn+ 2 ( h_m , h^-1(u ) ) _ ! _ _ m(z)^\|_h^-1(u))\| _ w ( r^n - df _ * _ f^-1(u))\| _ w. therefore r^n - df _ * _ f^-1(w ) ( r^n - df _ * _ f^-1(u))\|_w ( r^n - df _ * _ f^-1(u))\|_w r^n - df _ * _ f^-1(w ) . we may redefine @xmath7 as @xmath141 and refine the stratification @xmath72 accordingly , so that r^n - df _ * r^n - df _ * . from the arguments above , one sees that [ dim ] under the assumptions of [ cmpcomp ] , the semismallness of @xmath43 implies that of @xmath66 , and that @xmath152 . moreover , if a stratum @xmath7 is @xmath43-relevant , then it is also @xmath66-relevant . finally , we can prove [ intcmp ] . as is sketched right after the statement of [ intcmp ] , the proof proceeds by recurrence on strata . we assign a partial order on the set of strata . there is a relation between two strata @xmath153 if @xmath154 . suppose that @xmath70 is a stratum , and that on every @xmath155 with @xmath156 but @xmath157 , the comparison @xmath158 has been established . apply [ cohomcmp ] to @xmath7 , so that there is a resolution @xmath159 and an zariski open @xmath160 with @xmath161 . refine the stratification on @xmath162 so that @xmath163 is one of the strata . this finishes the recursive step . clearly by the noetherian assumption on @xmath37 , the recurrence eventually stops . that finishes the proof . [ qmot ] in coefficient @xmath1 , we have rf _ * rf _ * consequently , if @xmath3 and @xmath58 are projective , then the chow motives in @xmath1-coefficients are isomorphic m(x ) _ m(x ) _ tensoring with @xmath1 the isomorphisms provided by [ intcmp ] , we have r^n - df_&r^n - df_(^l f^)r^n - df_^l + & r^n - df_^l r^n - df_(^l f^ * ) r^n - df_. then ( t , r^n - df _ ) ( t , r^n - df _ * ) . by the explicit decomposition theorem of @xcite , rf__t(t , r^n - df _ ) _ t(t , r^n - df _ * ) rf_. an argument in @xcite , which we will reproduce in the next section , shows that @xmath164 . as the decomposition theorem is available only with coefficients in fields of characteristic @xmath165 , in order to extend the result to more general coefficients , we shall study the extensions of perverse sheaves across strata in the next section and forth . provided with [ intcmp ] we are left with extending the isomorphisms across strata , in other words , to glue the isomorphisms of perverse sheaves over strata . the gluing is not automatic . for nice general expositions of gluing of perverse sheaves , confer @xcite or @xcite . now fix a noetherian coefficient ring @xmath0 . throughout this section , we work on the derived categories @xmath166 of ctf @xmath0-complexes with respect to a stratification @xmath72 of @xmath37 obtained in [ intcmp ] . according to the notations of @xcite , we will write @xmath167 for the derived functor , whereas standard cohomology sheaf functors will be denoted by @xmath168 , and the perverse cohomology sheafs by @xmath169 . the situation of semismall k - equivalence @xmath170 is continued in this section . according to @xcite , the sheaves @xmath171 $ ] and @xmath172 $ ] are perverse . that can be shown by estimation of supports using leray spectral sequences together with the self - duality of @xmath171 $ ] and of @xmath172 $ ] . we begin with a observation on the convolution algebra and morphisms between @xmath171 $ ] and @xmath172 $ ] . for @xmath173 open subset , _ d^b_c(u , ) ( f__f^-1(u ) , f__f^-1(u ) ) h^_2n(f^-1(u ) _ u f^-1(u ) , ) . and this isomorphism commutes with the obvious restriction morphisms of open subsets @xmath174 in the sense that the following diagramme commutes _ d^b_c(u , ) ( f__f^-1(u ) , f__f^-1(u ) ) & h^_2n(f^-1(u ) _ u f^-1(u ) , ) + _ d^b_c(v , ) ( f__f^-1(v ) , f__f^-1(v ) ) & h^_2n(f^-1(v ) _ v f^-1(v ) , ) when there is a third @xmath175 , the composition pairing _ d^b_c(u , ) ( f__f^-1(u ) , f _ * _ f^-1(u ) ) _ _ d^b_c(u , ) ( f _ * _ f^-1(u ) , f _ * _ f^-1(u ) ) + _ d^b_c(u , ) ( f__f^-1(u ) , f _ _ f^-1(u ) ) is compatible with the convolution product of cohomological correspondences h^_2n(f^-1(u ) _ u f^-1(u ) , ) _ h^_2n ( f^-1(u ) _ u f^-1(u ) , ) + h^_2n ( f^-1(u ) _ u f^-1(u ) , ) emphasis is put on the compatibility with restriction maps , whereof we will be using in a crucial way . observe also that in the situation of semismall k - equivalence , since the borel - moore homology groups are freely generated by the respective sets of irreducible components of dimension @xmath32 ( [ dim ] ) , the restriction map is surjective , and so is the other restriction map . besides , the compatibility of product structure provides a link between isomorphism of sheaves and of chow motives . [ corr ] in the situation of semismall k - equivalence @xmath176 , any isomorphism @xmath171\cong f'_\lambda[n]$ ] gives rise to a cohomological correspondence @xmath177 , which is the class of an algebraic cycle with coefficients in @xmath0 . when @xmath3 and @xmath58 are projective , @xmath178 provides an isomorphism of chow motives @xmath179 with coefficients in @xmath0 . we shall prove by recurrence on strata in @xmath72 that @xmath171 \cong f'_\lambda[n]$ ] . we are then reduced to the situation under which there are zariski open subsets @xmath180 with smooth complement @xmath181 of equidimension @xmath39 , on which @xmath182 $ ] and @xmath183 $ ] are locally constant constructible sheaves , isomorphisms @xmath184 \cong f'_\lambda _ { { f'}^{-1}(v)}[n]$ ] and @xmath182 \cong\operatorname { { \prescript{o } { } { \calh}}}^{-d}f ' _ \lambda[n]$ ] . lift the isomorphism @xmath185 to a morphism @xmath186 \to f'_\lambda _ { { f'}^{-1}(u)}[n]$ ] , and its inverse @xmath187 to @xmath188 \to f_\lambda_{f^{-1}(u)}[n]$ ] . we remark that @xmath189 and @xmath190 may not be isomorphisms anymore . then there are morphisms between distinguished triangles i_i^ ! f _ & f _ * & j_j^ f _ * & i_i^ ! f _ + i_i^ ! f _ & f _ * & j_j^ f _ * & i_i^ ! f _ + i_i^ ! f _ & f _ * & j_j^ f _ * & i_i^ ! f _ + recall that by the semismallness of @xmath43 and of @xmath66 , @xmath171 $ ] and @xmath172 $ ] are perverse . applying the perverse cohomolology functor @xmath191 to this diagramme , we have exact sequences 0 & ^0 i_i^ ! f _ & f _ * & ^0 j_j^ f _ * & ^1 i_i^ ! f _ & 0 + 0 & ^0 i_i^ ! f _ & f _ * & ^0 j_j^ f _ * & ^1 i_i^ ! f _ & 0 + 0 & ^0 i_i^ ! f _ & f _ * & ^0 j_j^ f _ * & ^1 i_i^ ! f _ & 0 + since @xmath192 , @xmath193 as well . similarly @xmath194 , so @xmath195 and @xmath196 are inverse to each other . the diagramme is reduced to 0 & ^0 i_i^ ! f _ & f _ * & d & 0 + 0 & ^0 i_i^ ! f _ & f _ * & d & 0 + 0 & ^0 i_i^ ! f _ & f _ * & d & 0 + where @xmath185 and @xmath197 are isomorphisms . since @xmath198 \cong \left(\operatorname { { \prescript{o } { } { \calh}}}^{n - d } i_i^ ! f _ \lambda\right)[d ] \cong \left(\operatorname { { \prescript{o } { } { \calh}}}^{n - d } i_i^ ! f ' _ \lambda\right)[d ] \cong \operatorname { { \prescript{\frakp } { } { \calh}}}^0 i_i^ ! f ' _ \lambda[n]$ ] by [ dual ] tensored with @xmath0 , it amounts now to a statement concerning extensions of perverse sheaves , which will be demonstrated in the next section . [ cat ] let @xmath12 and @xmath199 be objects in an abelian category @xmath200 locally finite over a noetherian local ring @xmath0 , such that @xmath201 . suppose we have a diagramme of short exact sequences 0 & a & b & c & 0 + 0 & a & b & c & 0 + 0 & a & b & c & 0 + where @xmath185 and @xmath197 are isomorphisms . then there exists an isomorphism @xmath202 which lifts @xmath185 . with this lemma in combination with [ corr ] , we have proven [ main ] let @xmath0 be a noetherian local ring . given projective birational morphisms @xmath18 and @xmath60 of complex algebraic varieties . suppose that @xmath43 and @xmath66 are semismall and that @xmath3 and @xmath58 are smooth varieties k - equivalent through @xmath69 ( e.g. when @xmath203 is a semismall flop ) . then rf _ * rf _ * . suppose further that @xmath3 and @xmath58 are projective varieties , then there is an isomorphism of motives in coefficient @xmath0 m(x)_m(x)_. [ tor ] under the assumptions of [ main ] , there are isomorphisms of singular cohomology groups h^(x , ) h^(x , ) and h^(x , _ p ) h^(x , _ p ) for all prime @xmath204 . since @xmath2 and @xmath205 are finitely generated abelian groups , it suffices to compare the ranks and the @xmath204-primary components @xmath206 $ ] and @xmath207 $ ] for all primes @xmath204 . [ main ] provides in particular @xmath208 for all local rings @xmath0 . taking @xmath209 , we see that @xmath210 . taking @xmath211 the ring of @xmath204-adic integers , we have on the other hand @xmath212\cong h^k(x , \bfz_p)\left [ p^{\infty } \right]\cong h^k(x ' , \bfz_p)\left [ p^{\infty } \right ] \cong h^k(x , \bfz)\left [ p^{\infty } \right]$ ] . this proves @xmath213 . the latter statement is similarly proven by taking @xmath214 . in this section we work abstractly on a noetherian abelian category @xmath200 _ locally finite _ over a noetherian local ring @xmath0 . this means that the noetherian abelian category @xmath200 has on every hom set @xmath215 a finitely generated @xmath0-module structure for every @xmath216 , such that each composition map given objects @xmath218 such that @xmath201 , we shall study the @xmath0-module @xmath219 of extensions of @xmath199 by @xmath12 . recall that elements of @xmath219 are short exact sequences in @xmath200 let @xmath222 and @xmath223 . then @xmath219 is endowed with a natural @xmath224-bimodule structure in the following manner : for any @xmath225 , the element @xmath226 is defined as the second row of the following diagramme 1 . suppose we have a diagramme of short exact sequences + 0 & a & b & c & 0 + 0 & a & b & c & 0 + then the left square is cocartesian . dually , suppose we have a diagramme of short exact sequences + 0 & a & b & c & 0 + 0 & a & b & c & 0 + then the right square is cartesian . 3 . given any diagramme of short exact sequence + 0 & a & b & c & 0 + 0 & a & b & c & 0 + there is a 3-step factorisation + 0 & a & b & c & 0 + 0 & a & a ^ ( , ) b & c & 0 + 0 & a & b _ ( , ) c & c & 0 + 0 & a & b & c & 0 + where @xmath229 is defined by + ( a , b ) = ( ( a ) + ( b ) , ( b ) ) . with @xmath187 . in other words , this amounts to @xmath236 = [ ( b ' , \gamma , \delta)]\cdot \varphi$ ] and @xmath237 = [ ( b , \alpha , \beta)]\cdot \varphi^{-1}$ ] in @xmath219 . we will invoke the following easy lemma . [ lift ] let @xmath0 be a noetherian commutative local ring , and @xmath12 be an ( associative unital ) @xmath0-algebra which is a finitely generate @xmath0-module . given any surjection of @xmath0-algebras @xmath238 , the induced group homomorphism @xmath239 is surjective as well . we prove firstly that @xmath240 , where @xmath241 is the jacobson radical of @xmath12 , and similarly that @xmath242 . the jacobson radical is the intersection of all maximal left ideals of @xmath12 . thus it suffices to show that for every maximal left ideal @xmath243 , the pullback @xmath244 in @xmath0 is equal to @xmath245 , or equivalently that @xmath246 is a field . to this effect , consider the irreducible left @xmath12-module @xmath247 , whose endomorphism algebra @xmath248 is a finitely generated division @xmath249-algebra . we identify @xmath246 with a subring of the centre of @xmath250 . let @xmath251 be an element . we shall find its inverse element in @xmath246 . firstly , @xmath252 has a two - sided inverse @xmath253 in @xmath250 . since @xmath250 is a finite @xmath249-algebra , every element is integral over @xmath246 . let @xmath254 be a monic polynomial relation for @xmath253 , with coefficients @xmath255 . multiplying it with @xmath256 , we obtain @xmath257 . hence @xmath253 is a inverse of @xmath252 in @xmath246 , so @xmath246 is a field . the inclusion @xmath240 shows in particular that @xmath258 is a quotient of @xmath259 , thus a semisimple artinian @xmath260-algebra . now given @xmath261 , the reduction @xmath262 is invertible as well . since @xmath263 is an epimorphism of semisimple artinian algebras , there clearly exists @xmath264 such that @xmath265 and @xmath266 . since @xmath267 , there exists @xmath268 such that @xmath269 . it remains to verify that @xmath270 . let @xmath271 such that @xmath272 . then @xmath273 , and that @xmath274 is invertible according to the characterisation . similarly , @xmath275 is invertible . therefore @xmath270 . let @xmath276 $ ] be the sub-@xmath277-module generated by @xmath278 $ ] . then @xmath279\cdot \varphi$ ] . thus @xmath280 induces an automorphism on @xmath281 . put @xmath282 and @xmath283 . then the lemma gives an invertible @xmath284 such that @xmath285 = [ ( b ' , \gamma , \delta)]$ ] . therefore there is a diagramme of exact sequences 99 _ lments de gomtrie algbrique ( rdigs avec la collaboration de jean dieudonn ) : iv . tude locale des schmas et des morphismes de schemas.,troisime partie . _ publications mathmatiques de li.h..s . , tome 28 ( 1966 ) , p. 5 http://www.numdam.org/item?id=pmihes_1966__28__5_0 _ sminaire de gomtrie algbrique : 4 1/2 . cohomologie tale . _ lecture notes in mathematics , vol . 569 , berlin ; new york : springer - verlag , 1977 . _ analyse et topologie sur les espaces singuliers ( i ) . _ socit mathmatique de france , astrisque 100 , 1982 . _ birational calabi yau n - folds have equal betti numbers . _ _ new trends in algebraic geometry _ ( warwick , 1996 ) , pp 1 - 11 , london math . lecture note ser . 264 , cambridge univ . press , cambridge 1999 . _ partial resolutions of nilpotent varieties . _ socit mathmatique de france , astrisque 101(102):23 - 74 , 1983 . _ the hard lefschetz theorem and the topology of semismall maps . _ annales scientifiques de lcole normale suprieure , 4^e^ srie , t. 35 , 2002 , p. 759 - 772 . _ the chow motive of semismall resolutions . _ mathematical research letters , 2004 , volume 11 , number 2 , pp 151 - 170 . _ representation theory and complex geometry . _ birkuser , boston , 1997 . _ germs of arcs on singular algebraic varieties and motivic integration . _ inventiones mathematicae , january 1999 , volume 135 , issue 1 , pp 201 - 232 . _ motivic and quantum invariance under stratified mukai flops . _ journal of differential geometry , 2008 , volume 80 , pp 261 - 80 . _ decomposition numbers for perverse sheaves . _ annales de linstitut fourier 59.3 , 2009 . _ flops . _ nagoya mathematical journal , 1989 , volume 113 , pp 15 - 36 . _ flops , motives , and invariance of quantum rings . _ annals of mathematics , july 2010 , second series , volume 172 - 1 , pp 243 - 290 . _ topology of singular spaces and constructible sheaves . _ monografie matematyczne , 2003 , volume 63 , birkhuser basel . _ on the topology of birational minimal models . _ j. differential geometry , 1998 , volume 50 , number 1 , pp 129 - 146 . _ k - equivalence in birational geometry and characterizations of complex elliptic genera . _ j. algebraic geometry , 2003 , volume 12 , number 2 , pp 285 - 306 . _ k - equivalence in birational geometry . _ in _ proceeding of the second international congress of chinese mathematicians _ ( grand hotel , taipei 2001 ) .
in the nearby universe , the population of galaxies with stellar masses greater than 10@xmath2 m@xmath5 is dominated by large early type galaxies ( baldry et al . 2004 ) with correspondingly large sizes ( shen et al . these nearby systems contain old and metal - rich stellar populations that formed quickly in the early universe ( heavens et al . 2004 ; feulner et al . 2005 ; thomas et al . however , it has remained unknown whether the stars in these galaxies were all assembled in the same system , or formed in lower mass galaxies that later merged . two major formation models have been proposed in order to explain the properties of these galaxies : the so - called monolithic collapse model ( eggen , lynden - bell & sandage 1962 ; larson 1975 ; arimoto & yoshii 1987 ; bressan , chiosi & fagotto 1994 ) and the hierarchical merging scenario ( toomre 1977 ; white & frenk 1991 ) . these two mechanisms have both observational and theoretical successes and drawbacks . for instance , evidence in favor of a fast and dissipative monolithic collapse is the fact that the bulk of stars in massive ellipticals are old ( mannucci et al . 2001 ) and have high [ @xmath11/fe ] ratios ( i.e. short star formation time - scales ; worthey et al . 1992 ) . in addition , the structural and dynamical properties of observed spheroid galaxies are well reproduced by cold dissipationless collapse ( van albada 1982 ; may & van albada 1984 ; mcglynn 1984 ; aguilar & merritt 1990 ; londrillo et al . 1991 ; udry 1993 ; hozumi et al . 2000 ; trenti et al . 2005 ) , a process that is expected to dominate the last stages of a highly dissipative collapse . on the other hand , supporting a hierarchical merger scenario , observations find a decline in the number of massive galaxies seen at higher redshifts . this decline is moderate since z@xmath61 and much stronger at even higher redshifts ( daddi , cimatti & renzini 2000 ; pozzetti et al . 2003 ; bell et al . 2004 ; drory et al . 2004 , 2005 ; daddi et al . 2005 ; saracco et al . 2005 ; faber et al . 2005 ; pozzetti et al . 2007 ) . moreover , new renditions of semi - analytical models where merging is the cornerstone of galaxy formation ( e.g. de lucia et al . 2006 ) are much better able to match the stellar population properties of elliptical galaxies . finally , it has been found theoretically that the scaling laws followed by elliptical galaxies are robust against merging ( ciotti et al . 2007 ) . exploring the assembly of massive galaxies with cosmic time through number density analysis or merger rate estimations are however difficult to conduct since , first , very massive galaxies are scarce and , second , their clustering properties make them strongly affected by field - to - field variance associated with limited observed volumes . these two difficulties make claims of the assembly of massive galaxies in the early universe based on number densities and merger rate estimations highly uncertain ( e.g. , cimatti , daddi & renzini 2006 ; renzini 2007 ) . a much more straightforward approach to the assembly problem of massive galaxies is to explore the size evolution of these systems at a given stellar mass . in a `` monolithic - like '' scenario where a galaxy is fully assembled after the formation of its stars , the stellar mass - size relation should remain unchanged as cosmic time evolves . in the hierarchical merging scenario , however , the stellar mass - size relation will evolve as a result of the increase in size after each galaxy merger . for instance , state - of - the - art hierarchical semi - analytical models predict a very strong ( a factor of 1.5 - 3 ) evolution in the size of very massive galaxies ( m@xmath0@xmath110@xmath2h@xmath3@xmath4m@xmath5 ) nine billion years ago ( khochfar & silk 2006b ) . this predicted evolution in size is a strong function of galaxy mass , with the more massive objects expected to have the largest increase in size . due to the lack of a large sample of very massive galaxies at high redshift , analysis of the evolution of the stellar mass - size relation of galaxies have mainly explored objects in the 10@xmath12@xmath8m@xmath13@xmath810@xmath2 m@xmath5 range . observations within this mass interval do not find a significant evolution of the stellar mass - size relation since z@xmath61 ( barden et al . 2005 ; mcintosh et al . 2005 ) . explorations of these objects at even earlier look - back times ( 1@xmath8z@xmath83 ) have been also attempted , and show a moderate decline in size for galaxies at a given stellar mass ( trujillo et al . 2004 ; 2006a ) . however , the size of very massive galaxies , m@xmath0@xmath110@xmath2m@xmath5 , is still largely unexplored , with only around a dozen objects studied in detail so far ( daddi et al . 2005 ; trujillo et al . 2006b ; longhetti et al . 2007 ; mcgrath et al . 2007 ) . definitive conclusions regarding the evolution of sizes for massive galaxies thus remain largely unknown . the goal of this paper is to shed some light on this issue by exploring the sizes of a large sample of very massive galaxies since z@xmath82 . this will help to clarify whether the evolutionary scenario for these objects is hierarchical or monolithic - like . the paper is structured as follows . in section 2 we give a brief summary of the palomar / deep2 data , and in section 3 we describe the determination of the stellar masses . size measurement technique and robustness estimations for our data are provided in section 4 . in section 5 we study the selection effects and in section 6 present the observed stellar mass size relations . we compare our results with other samples in section 7 , and finally , we discuss our results in section 8 . in what follows , we adopt a cosmology of @xmath14=0.3 , @xmath15=0.7 and h@xmath16=70 km s@xmath17 mpc@xmath17 . we use the palomar observatory wide - field infrared ( powir)/deep2 survey ( bundy et al . 2006 ; conselice et al . 2007a , b ; davis et al . 2003 ) to define a sample of 831 galaxies with masses larger than 10@xmath2h@xmath3@xmath4m@xmath5 located over @xmath6710 arcmin@xmath18 in the extended groth strip ( egs ) . this field ( 63 hubble space telescope tiles ) was imaged with the advanced camera for surveys ( acs ) in the v(f606w , 2660s ) and i - band ( f814w , 2100s ) . each tile was observed in 4 exposures that were combined to produce a pixel scale of 0.03 arcsec , with a point spread function ( psf ) of 0.12 arcsec full width half maximum ( fwhm ) . in addition to the hst data , optical imaging from the cfht 3.6-m telescope in the b , r and i bands taken with the cfh12k camera was used . integration times for these observations were 1 hour in b and r , and 2 hours in i. limiting magnitudes reached are b=24.5 ( ab , 8@xmath19 ) , r=24.2 ( ab , 8@xmath19 ) and i=23.5 ( ab , 8@xmath19 ) . the details of the data reduction for this data is provided in coil et al . ( 2004 ) . in the egs region , the palomar near - infrared k@xmath20-band imaging has a typical depth of greater than k@xmath21=22.5 mag ( 5 @xmath19 ) , and a j - band depth of j@xmath21=23.4 mag ( 5 @xmath19 ) . about 70% of the galaxies brighter than r@xmath21=24.1 mag in this field were targeted by the deep2 galaxy redshift survey , using the keck 10 m telescope , for a total spectroscopic redshift completeness of @xmath656% . in our case , 410 objects have a spectroscopic redshift ( @xmath650% of the full dataset ) , with a spectroscopic completeness of 61% for galaxies with z@xmath81 , and a completeness of 30% for galaxies with z@xmath91 . we supplemented our spectroscopic redshift catalogue with photometric redshifts with an accuracy for all galaxies in the total k - band survey of @xmath22z/(1+z)@xmath230.07 in the range 0.2@xmath8z@xmath81.4 and @xmath22z/(1+z)@xmath80.22 for galaxies at 1.4@xmath8z@xmath82 ( conselice et al . however , it is worth noting that , for the massive galaxies we use in this paper , the accuracy is @xmath24 at @xmath25 , and likely similarly lower for @xmath26 galaxies ( i.e. @xmath27 ) . the determination of stellar masses for each of our galaxies follows a standard multi - color ( brijk ) stellar population fitting technique , producing uncertainties of @xmath230.2 dex . the largest systematic source of error comes from the assumed initial mass function ( imf ) , in this paper we have used the chabrier imf for all stellar mass measurements . the details of these estimations are presented in bundy et al . ( 2006 ) and conselice et al . ( 2007b ) . however , because the measurement of stellar masses is an integral part of this paper , we give a brief description of how our masses are measured , and what systematics might be present . the basic mass determination method we use consists of fitting a grid of model seds constructed from bruzual & charlot ( 2003 ) ( bc03 ) stellar population synthesis models , with different star formation histories . we use an exponentially declining model to characterise the star formation , with various ages , metallicities and dust contents included . these models are parameterised by an age , and an e - folding time . these parameterisations are fairly simple , and it remains possible that stellar mass from older stars is missed under brighter younger populations . while the majority of our systems are passively evolving older stellar populations , it is possible that up to a factor of two in stellar mass is missed in any star bursting blue systems . however , stellar masses measured through our technique are roughly the expected factor of 5 - 10 smaller than dynamical masses at @xmath28 using a sample of disk galaxies ( conselice et al . 2005 ) , demonstrating their inherent reliability . we calculate the likely stellar mass , age , and absolute magnitudes for each galaxy at all star formation histories , and determine stellar masses based on this distribution . distributions with larger ranges of stellar masses have larger resulting uncertainties . it turns out that while parameters such as the age , e - folding time , metallicity , etc . are not likely accurately fit through these calculations due to various degeneracies , the stellar mass is robust . typical errors for our stellar masses are 0.2 dex from the width of the probability distributions . there are also uncertainties from the choice of the imf . our stellar masses utilise the chabrier imf , which can be converted to salpeter imf stellar masses by adding 0.25 dex . there are additional random uncertainties due to photometric errors . the resulting stellar masses thus have a total random error of 0.2 - 0.3 dex , roughly a factor of two . there is furthermore the issue of whether or not our stellar masses are overestimated based on using the bruzual & charlot ( 2003 ) models . it has recently been argued by maraston ( 2005 ) and bruzual ( 2007 ) that the exclusion of an updated treatment of thermal - pulsating agb stars in the bc03 models results in calculated stellar masses too high by a factor of a few . while we consider an uncertainty of a factor of two in our stellar masses , we must consider whether our sample is in the regime where the effects of tp - agb stars will influence our mass measurements . this has been investigated recently in maraston ( 2005 ) and bruzual ( 2007 ) who have both concluded that galaxies stellar masses computed using newer tp - agb star prescriptions are up to roughly 50 - 60% lower than without . this is particularly true for masses determined in the rest - frame infrared . this problem has also been recently investigated independently by kannappan & gawiser ( 2007 ) who come to similar conclusions , but do not advocate one model over another . furthermore , the effect of tp - agb stars is most pronounced in the rest - frame ir , and for young stellar population ages . our survey is k - selected , and the observed k - band is used as the flux in which the masses are computed . the rest - frame wavelength probed with the observed k - band ranges from 0.7@xmath29 m to 1.5@xmath29 m where the effects of tp - agb stars are minimised . the ages of our galaxies are also older than the ages where tp - agb stars have their most effect ( maraston 2005 ; bruzual 2007 ) . it is also worth noting that the effects of tp - agb stars are more important when normalising stellar masses further into the red . in this paper we find a strongest evolution at higher redshifts where we are probing the rest - frame optical at observed - k , and where the effects of tp - agb stars are minimised ( e.g. , bruzual 2007 ) . we investigate the effects of tp - agb stars in our estimations by determining how our stellar masses change between using the bc03 models and the updated charlot & bruzual ( 2007 , in prep ) models using the new tp - agb methods . we find for 1330 massive galaxies in the palomar sample a difference of @xmath60.07 in log m@xmath30 , which is neglectable . a similar conclusion has been recently achieved by mcgrath , stockton & canalizo ( 2007 ) using a sample of galaxies at z@xmath61.5 with stellar masses similar to those explored here . on comparing the new rendition of models from charlot & bruzual ( 2007 ) vs bc03 , they found that the inferred galaxy masses are slightly ( @xmath610% ) smaller using the 2007 models . although we do have spitzer data for our sources , we do not use this imaging for two reasons . the first is that by normalising the stellar masses with spitzer irac magnitudes , we are in a regime where the tp - agn stars are more pronounced , and thus would affect the stellar mass measurements to a degree even greater than using the observed k - band ( kannappan & gawiser 2007 ; bruzual 2007 ) . secondly , the large psf of the irac images makes it difficult to obtain accurate photometry for many of our sources due to contamination from other galaxies with overlapping psfs . while it is possible to correct for this , the large resulting random photometry uncertainties make the stellar masses less certain . another possible source of uncertainty is the photometric redshifts we use for our sample . while at @xmath25 about half of our sample have spectroscopic redshifts , at @xmath26 all of our systems have photometric redshifts . we can however determine the accuracy for those systems at @xmath25 . the agreement is very good for our massive systems with @xmath24 . this results in another @xmath820% uncertainty in the stellar mass measurements . for @xmath26 , if we assume @xmath31 , the uncertainty in the stellar masses would be @xmath632% . overall , however , these uncertainties can not account for the trends seen later in this paper . the structural parameters used in this paper were measured using the acs i - band filter . sizes ( as parameterised by the half - light or effective radius along the semi - major axis a@xmath32 ) were estimated using the galfit code ( peng et al . sizes were circularised , r@xmath32=a@xmath32@xmath33 , with @xmath34 being the ellipticity of the object . galfit convolves srsic ( 1968 ) r@xmath35 galaxy models with the psf of the images , and determines the best fit by comparing the convolved model with the galaxy surface brightness distribution using a levenberg - marquardt algorithm to minimise the @xmath36 of the fit . the srsic model is a flexible parametric description of the surface brightness distribution of the galaxies and contains the exponential ( n=1 ) and de vaucouleurs ( n=4 ) models as particular cases . in addition , this model is used in the structural analysis of the sdss galaxy sample ( our local comparison sample ; blanton et al . 2003 ; shen et al . 2003 ) . the srsic index n measures the shape of surface brightness profiles . in the nearby universe , galaxies with n@xmath82.5 are mostly disk - like objects , whereas galaxies with n@xmath92.5 are mainly spheroids ( ravindranath et al . we use this srsic index criterion to split our sample at higher redshifts and facilitate a comparison with the local galaxy population . during the fit , neighbouring galaxies were excluded using a mask , but in the case of closely neighbouring objects with overlapping isophotes , the objects were fitted simultaneously . the results of our fitting are shown in the appendix : table [ rawdata ] . the results presented in this paper rely on our ability to measure accurate structural parameters . to gauge the accuracy of our parameter determination realistic simulations were conducted . we have created 1000 artificial galaxies uniformly generated at random in the following ranges , matching the observed distribution of our galaxies : 18@xmath37i@xmath21@xmath3726 ( see fig . [ maghist ] ) , 0.03@xmath37r@xmath32@xmath372.55 arcsec ( which at z@xmath61 equals 0.25 - 20.5 h@xmath38 kpc ) , 0.5@xmath37n@xmath378 , and 0@xmath37@xmath34@xmath370.8 . to simulate the real conditions of our observations , we add a background sky image taken ( randomly at each time ) from a piece of the acs i - band image . finally , the galaxy models were convolved with the observed psf . the same procedure was used to retrieve the structural parameters in both the simulated and actual images . the results of these simulations are shown in figures [ simul1 ] and [ simul2 ] . fifty percent of our galaxies are brighter than i(ab)=21.9 mag , and for these galaxies we find dr@xmath32/r@xmath32@xmath85% and dn / n@xmath87% . for 95% of our galaxies which are at i(ab)@xmath824.2 mag , the uncertainties are dr@xmath32/r@xmath32@xmath830% and dn / n@xmath838% . as expected , at fainter apparent magnitudes the structural parameters are recovered with larger uncertainties . only for magnitudes fainter than i(ab)=24 mag is there a small bias ( @xmath620% ) of the index n towards smaller indices for galaxies with n@xmath39@xmath92.5 . recovery of the structural parameters with larger n are more affected than those with lower values . as shown in figure [ simul2 ] , we do not find any bias of our sizes or shape parameter as a function of the size of the objects . finally , we have also explored whether the variation of the psf along the image can affect the recovery of the size of the galaxies . using different stars in the images as a psf we find that the estimation of the sizes is robust to changes in the selected psf to analyze the data : the scatter between the sizes is @xmath4010% ( 1@xmath19 ) . the sizes of the galaxies presented in this paper were measured using the acs observed i - band , which implies that for galaxies at z@xmath91.3 sizes are retrieved in the rest - frame uv . to check whether this k - correction effect can affect our size estimates we have compared our sizes measured in the i - band with sizes obtained in the near infrared camera and multi - object spectrometer ( nicmos ) h - band ( f160w ) for a subset of 27 galaxies ( with 0.8@xmath8z@xmath81.8 and median z@xmath61.2 ) that were observed at both wavelengths . these nicmos data consist of 63 pointings of camera 3 ( 52@xmath4152 arcsec , 0.203 arcsec / pixel ) in the egs field . each pointing is the combination of four sub - pixel dithered exposures , with a total exposure time of 2600 sec . the final mosaic was assembled using a drizzle task and has a pixel scale of 0.10 arcsec . we found that sizes measured in the nicmos images , and those measured with the acs i - band data , are well correlated ( see figure [ nicmos ] ) . the scatter between both measurements is 32% ( 1@xmath19 ) . there seems to be a systematic ( although statistically non - significant ) bias of 19@xmath107% between both measurements : sizes measured in the redder band ( h - band ) are slightly smaller than those measured in the bluer band ( i - band ) . consequently , if any k - correction effect is affecting our results , the size evolution found here using the i - band filter would be an upper limit . interestingly , this trend of smaller sizes at redder bands dr@xmath42)/dlog@xmath43=-0.6(@xmath100.2 ) is in qualitative agreement to what is found in nearby galaxies -0.18@xmath9dr@xmath44(@xmath43)/dlog@xmath43@xmath9 - 0.25 ( barden et al . 2005 ; mcintosh et al . 2005 ) . however , due to the large uncertainty on the bias ( and consequently in the size corrections ) , and because we do not know how this correction could evolve with redshift we avoid making any k correction in our results . in any case , it is important to note that the trend towards smaller sizes observed using the h - band does not strongly affect the main result of this paper . in fact , the evolution found in this paper would be even stronger using the near - infrared band sizes at high redshift . another source of concern in the size determination is the presence of an active galactic nuclei ( agn ) in the center of the galaxy which can bias our measurement towards smaller sizes ( daddi et al . deep ( 200 ks per pointing ) x - ray observations ( l@xmath45@xmath4610@xmath47 erg s@xmath17 at z@xmath61 ) from the chandra telescope in this field ( nandra et al . 2007 ; conselice et al . 2007b ) only detect emission in 35 objects ( i.e. @xmath64.2% our sample ) . we remove these potential agns from our sample in what follows , but our essential results are unchanged . in practice , any galaxy survey has a surface brightness limit beyond which the sample is incomplete . characterising this limit is particularly important for high - z samples , where the effects of the cosmological surface brightness dimming are severe . at a given total flux limit , the surface brightness limit translates into an upper limit on the size in which a galaxy can be detected . we have explored whether our k@xmath20-band selected sample could be incomplete at large sizes by examining the detectability of our galaxies as a function of their apparent magnitudes and sizes . to determine the detection map of the palomar k@xmath20 band image , we have created two sets of 10@xmath48 mock sources each with intrinsic exponential or de vaucouleurs profiles uniformly distributed as follows : k@xmath20 band total magnitudes between 15.5 and 20.5 ( vega ) mag ( see fig . [ maghist ] ) , effective radius r@xmath32 between 0.0625 and 3.75 arcsec ( this will be equivalently to 0.5 - 30 h@xmath38 kpc at z@xmath61 ) , and ellipticities between 0 and 0.8 . the simulated sources are placed randomly on the real image and extracted as for the real source detection . we construct from these simulations detection maps giving the number of recovered sources over the number of input artificial sources per input magnitude and input log r@xmath32 bin ( see figure [ maps ] ) . at it is expected , galaxies with a de vaucouleurs profile , and consequently more centrally concentrated , are easier to detect at a given magnitude . over - plotted on the detection maps are the distribution of our sample galaxies . k@xmath20 band magnitudes are measured using the mag@xmath49 output from sextractor . the sizes are those estimated in the i - band . we are making the implicit assumption that sizes are similar in both bands ( i.e. that k - correction effects are not relevant , see previous section ) . at a given magnitude , the observed size distribution declines more rapidly to larger sizes than the detection limit . this indicates that our sample is not affected by incompleteness for the largest galaxies at a given magnitude . the accuracy of the structural parameters and the completeness of our sample has been demonstrated in the previous section . in this section we discuss the stellar mass size relations for our massive galaxy sample . these relations are presented in figures [ disks ] ( n@xmath82.5 ; disk - like objects ) and [ spheroids ] ( n@xmath92.5 ; spheroid - like objects ) . in each of these figures , we have divided our sample ( over 0.2@xmath8z@xmath82 ) into six redshift slices with a width of dz=0.3 . over - plotted on our observed distributions are the mean and dispersion of the distribution of the srsic half - light radii from the sdss ( york et al . 2000 ) galaxies . we use the sdss sample as the local reference . as it has been done for our high - z galaxies , the sdss galaxy sizes were determined from a srsic model fit ( blanton et al . the characteristics of the sdss sample used here are detailed in shen et al . the mean of the sdss galaxies redshift distribution used for comparison is 0.1 . we use the sizes and the shapes estimated in the observed sdss r - band as this closely matches the v - band rest - frame filter at z@xmath60.1 . sdss stellar masses were derived using a kroupa ( 2001 ) imf , which gives the same masses as the chabrier imf used to measure the higher redshift sample stellar masses . figures [ disks ] and [ spheroids ] show that at a given stellar mass galaxies are progressively smaller at higher redshift . this evolution is particularly strong for more concentrated ( n@xmath92.5 ) galaxies . we find the remarkable result that there are no spheroid - like objects at z@xmath91.5 on the local relation . to illustrate this size decrease at progressively higher z , figure [ mosaico ] shows the appearance of six of our n@xmath92.5 ( i.e. spheroid - like ) galaxies with the same stellar mass , but at different look - back times . the objects shown in figure [ mosaico ] have the mean structural properties of the galaxy population at each of the different redshift slices ( see the stellar mass - size relations in the above redshift intervals in figures [ disks ] and [ spheroids ] ) . to allow a fair comparison between the sizes of objects at very different redshifts , the limiting surface brightness in each panel is changed according to the cosmological surface bright dimming @xmath6(1+z)@xmath48 . for this reason the surface brightness of the object at z=0.35 is shown down to @xmath63.25 mag brighter than the object at z=1.85 . this figure visually illustrates that the most massive galaxies are progressively smaller at increasing redshift . to quantify the size evolution , we measure the ratio between the observed size , and the expected size at a given stellar mass by comparing with the sdss ( shen et al . 2003 ) distribution at different redshifts . to estimate the expected size from sdss at a given stellar mass , we interpolate linearly between the sdss points when necessary . the evolution of the median and the dispersion of the above ratio are shown in figure [ evolution ] and listed in table [ data ] . the observed size evolution of m@xmath13@xmath910@xmath2 m@xmath5 galaxies is stronger than the one found in previous work using less massive galaxies ( see the detailed comparison of this issue in the next section ) . following recent claims ( i.e. maraston et al . 2006 ) that stellar masses could be systematically overstimated by a factor of few at high z , we have repeated our analysis under the assumption that our masses could be overestimated by a factor of 2 . in this case , at z@xmath61.5 , our galaxies will still be more compact than present - day galaxies of the same stellar mass by a factor of 1.6 for n@xmath82.5 and a factor of 3 for n@xmath92.5 . in other words , even a systematic effect of factor of 2 in the stellar mass determination can not avoid a significant evolution in the size of these galaxies . in fact , to prevent a significant evolution of the sizes of galaxies at a given stellar mass we found that , at z@xmath61.5 , our stellar masses would need to be overestimated by a factor of @xmath610 for galaxies with n@xmath82.5 and overestimated by a factor of @xmath650 for galaxies with n@xmath92.5 . [ cols="^,^,^,^,^ " , ] an interesting point to explore is how the size evolution found here for the most massive galaxies compares with the size evolution observed for galaxies with lower masses . this comparison can be done in detail in the redshift range 0@xmath8z@xmath81 using the data from one of the largest sample currently available : the gems survey ( rix et al . the stellar mass - size relation of this survey has been already derived for late - type ( barden et al . 2005 ) and early - type ( mcintosh et al . 2005 ) galaxies . the gems late- and early - type separation criteria are based on the srsic index n. late - types are defined through n@xmath82.5 , and early types through n@xmath92.5 , and a colour within the `` red sequence '' ( bell et al . the comparison of the distribution of their data points with ours are shown in figure [ gemspalomar ] . it is encouraging to see that these two independent analyses and data sets match well where there is overlap in their stellar mass ( i.e. for the small subset of gems galaxies with m@xmath13@xmath910@xmath2 m@xmath5 ) . these two data sets allow us to compare the difference in size evolution in the 0@xmath8z@xmath81 redshift interval at two different mass ranges 10@xmath12@xmath8m@xmath13@xmath810@xmath2 m@xmath5 ( gems ) , and m@xmath13@xmath910@xmath2 m@xmath5 ( palomar ) . this is shown in figure [ gemspalomarevol ] . from this comparison we can see that more massive galaxies evolve in size much faster than lower mass objects ( particularly for disk - like galaxies ) . this mass dependent evolution was hinted in previous works ( trujillo et al . 2006a ) but our current large data set show this more clearly and robustly . at higher redshift , the amount of data is more limited . at 1@xmath8z@xmath82 our results are in good agreement with recent findings ( based on a few objects ) of massive compact galaxies at high - z ( waddington et al . 2002 ; daddi et al . 2005 ; di serego alighieri et al . 2005 ; trujillo et al . 2006b ; longhetti et al . for example , using the 10 of the most massive galaxies ( 1.2@xmath8z@xmath81.7 ) in the munics survey , trujillo et al . ( 2006b ) found that these galaxies were a factor of 4@xmath50 smaller that local counterparts . at even higher z ( i.e. z@xmath62.5 ) there has been also recent claims of very compacts ( r@xmath32@xmath401 kpc ) massive galaxies ( trujillo et al . 2006a ; zirm et al . 2007 ; toft et al . as shown in section 4 the compact nature of our most massive galaxies at high - z can not be interpreted as a k - correction or agn effect . in addition , there is no observational evidence in the local universe for galaxies as massive and compact as the ones in our sample . these two observational facts raise the following two questions : first , how can these objects be so dense in the past ? second , how do these objects evolve in stellar mass and/or size in order to reach the current local relation ? addressing the first question , recent theoretical results suggest that major galaxy mergers in the early universe had a much larger component of cold gas available than in the present ( khochfar & silk 2006a ) . these wet ( dissipative ) mergers generate very efficient and massive starbursts creating a very compact massive remnant . consequently , the very dense nature of our objects at high - z could reflect the much denser condition of the universe at the time of their formation . focusing on the second question , at lower redshifts the available amount of gas is less , and new ` dry ' ( dissipationless ) mergers ( van dokkum 2005 ; bell et al . 2006 ) would be the dominant mechanisms of size and stellar mass growing ( ciotti & van albada 2001 ; nipoti , londrillo & ciotti 2003 ; khochfar & burkert 2003 ; domnguez - tenreiro et al . 2006 ; naab , khochfar & burkert 2006 ; ciotti , lanzoni & volonteri 2007 ) . dry mergers are not efficient at forming new stars , but are efficient in increasing the size of the objects . a particular effective size evolutionary mechanism ( r@xmath32@xmath6m@xmath51 ) has been recently found ( boylan - kolchin , ma & quataert 2006 ) in mergers of galaxies with radial orbits along large - scale structure filaments . this kind of mechanism would be able to evolve our compact galaxies ( a factor of 4 smaller at z@xmath61.5 ) to the local relation with just two major ( equal - mass ) mergers . since @xmath52 very few gas rich mergers occur in massive galaxies ( e.g. , conselice et al . 2003 ; conselice 2006 ) , but a few dry mergers are possible based on pair counts ( e.g. , lin et al . in addition , dry mergers of similar mass objects ( and consequently , potentially similar ages and metallicities ) will also help to understand the age - uniformity found in the local massive ( spheroid ) galaxies . on the other hand , in agreement with our results here , new semi - analytical models ( somerville et al . 2007 ) find that disc - like galaxies ( i.e. those presumably with a relative quiet live ) evolve only mildly in size since high - z . summarising our results , we find that the size evolution of the most massive galaxies is not consistent with a scenario whereby massive galaxies were fully assembled in the early universe , and have subsequently evolved passively until today ( i.e. a pure `` monolithic '' scenario ) . in fact , our findings agree with a scenario where a fraction of the most massive galaxies possibly formed in a short `` monolithic - like '' collapse and then evolved through major gas - rich , or gas - poor merging . in this sense , the two scenarios would be just different phases of galaxy formation and evolution . our results , consequently , point to a scenario whereby the stellar populations of the most massive galaxies we observe today were located in different `` primordial '' massive galaxy pieces in the early universe . authors are grateful to m. barden and d. mcintosh for providing us with the gems data points to allow us a comparison of our results with their less massive galaxies . we acknowledge useful discussions with o. almaini and a. aragn - salamanca . we also thank the useful suggestions from an anonymous referee . the palomar and deep2 surveys would not have been completed without the active help of the staff at the palomar and keck observatories . the palomar survey was supported by nsf grant ast-0307859 and nasa stsci grant hst- ar-09920.01.a . support for the acs imaging of the egs in go program 10134 was provided by nasa through nasa grant hst - go-10134.13-a from the space telescope science institute . 99 aguilar l. a. , merritt d. , 1990 , apj , 354 , 33 arimoto n , & yoshii y. , 1987 , a&a , 173 , 23 baldry i. k. , glazebrook k. , brinkmann j. , ivezi , lupton r. h , nichol r. c. szalay a. s. , 2004 , apj , 600 , 681 barden et al . , 2005 , apj , 635 , 959 bell e. f. et al . , 2004 , apj , 608 , 752 bell e. f. et al . , 2006 , apj , 640 , 241 blanton m. r. , et al . , 2003 , apj , 592 , 819 boylan - kolchin m. , ma c - p . , quataert e. , 2006 , mnras , 369 , 1081 bressan a. , chiosi c. , fagotto f. , 1994 , apjs , 94 , 63 bruzual g. astro - ph/0703052 bruzual g. , charlot s. , 2003 , mnras , 344 , 1000 bundy k. et al . , 2006 , apj , 651 , 120 cimatti a. , daddi e. , renzini a. , 2006 , a&a , 453 , l29 ciotti l. & van albada t. s. , 2001 , apj , 552 , l13 ciotti l. , lanzoni b. , & volonteri m. , 2007 , apj , 658 , 65 coil a. l. , newman j. a. , kaiser n. , davis m. , ma c .- p . , kocevski d.d . , koo d. c. , 2004 , apj , 617 , 765 conselice c. j. 2006 , apj , 639 , 120 conselice , c.j . , bundy , k. , ellis , r. , brichmann , j. , vogt , n. , phillips , a. 2005 , apj , 628 , 160 conselice c. j. , bershady , m.a . , dickinson , m. , papovich , c. 2003 , aj , 126 , 1183 conselice c. j. et al . , 2007a , apj , 660 , l55 conselice c. j. et al . , 2007b , mnras , submitted daddi e. , cimatti a. , renzini a , 2000 , a&a , 362 , l45 daddi e. et al . , 2005 , apj , 626 , 680 davis m. et al . , 2003 , spie , 4834 , 161 de lucia , g , springel v. , white s. d. m. , croton d. , kauffmann g. , 2006 , mnras , 366 , 499 di serego alighieri s. , et al . , 2005 , a&a , 442 , 125 domnguez tenreiro , r. , oorbe , j. , siz , a. , artal , h. , serna , a. , 2006 , apj , 636 , l77 n. , bender r. , feulner g. , hopp u. , maraston c. , snigula j. , hill g. j. , 2004 , apj , 608 , 742 n. , salvato m. , gabasch a. , bender r. , hopp u. , feulner g. , pannella m. , 2005 , apj , 619 , l131 eggen o. j. , lynden bell d. , sandage a. r. , 1962 , apj , 136 , 748 faber s. m. , et al . 2005 , astro - ph/0506044 g. , gabasch a. , salvato m. , drory , n. , hopp u. , bender r. , 2005 , apj , 633 , l9 a.,panter b.,jimenez r. , dunlop j. , 2004 , nature , 428 , 625 hozumi s. , burkert a. , fujiwara t. , 2000 , mnras , 311 , 377 khochfar s. , burkert a. , 2003 , apj , 597 , l117 khochfar s. , silk j. , 2006a , mnras , 370 , 902 khochfar s. , silk j. , 2006b , apj , 648 , l21 kannappan , s.j . , & gawiser , e. 2007 , apj , 657 , l5 larson r. , 1975 , mnras , 173 , 671 lin , l. , et al . 2004 , apj , 617 , l9 londrillo p. , messina a. , stiavelli m. , 1991 , mnras , 250 , 54 longhetti m. et al . , 2007 , mnras , 374 , 614 mannucci f. , basile f. , poggianti b.m . , cimatti a. , daddi e. , pozzetti l. , vanzi l. , 2001 , mnras , 326 , 745 maraston , c. 2005 , mnras , 362 , 799 may a. , van albada t.s . , 1984 , mnras , 209 , 15 mcglynn t. a. , 1984 , apj , 281 , 13 mcgrath e. j. , stockton a. , canalizo g. , 2007a , apj , in press , astro - ph/0707.1046 mcgrath e. j. , stockton a. , canalizo g. , masanori i. , maihara t. , 2007b , apj , submitted , astro - ph/0707.1050 mcintosh d. et al . , 2005 , apj , 632 , 191 nandra k. et al . , 2007 , apjl , in press , astro - ph/0607270 nipoti c. , londrillo p. & ciotti l. , 2003 , mnras , 342 , 501 peng c.y . , ho l.c . , impey c.d . , rix h.w . , 2002 , aj , 124 , 266 pozzetti l. et al . 2003 , a&a , 402 , 387 pozzetti l. et al . 2007 , a&a , astro - ph/0704.1600 ravindranath s. et al . 2002 , apj , 604 , l9 renzini a. , 2007 , astro - ph/0702148 rix h .- w . et al . , 2004 , apj , 152 , 163 saracco p. et al . , 2005 , mnras , 357 , l40 srsic j .- l . 1968 , atlas de galaxias australes ( cordoba : observatorio astronomico ) shen s.,mo h. j. , white s. d. m.,blanton m. r. , kauffmann g.,voges w.,brinkmann j. , csabai i. , 2003 , mnras , 343 , 978 somerville r. , et al . , 2007 , apj , submitted , astro - ph/0612428 thomas d.,maraston c. , bender r. , de oliveira c. m. , 2005 , apj , 621 , 673 toft s. , et al . , 2007 , apj , in press toomre a. , 1977 , in evolution of galaxies and stellar populations , ed . b. m. tinsley & r. b. larson ( new haven : yale univ . obs . ) , 401 trenti m. , bertin g. , van albada t.s . , 2005 , a&a , 433 , 57 trujillo i. et al . , 2004 , apj , 604 , 521 trujillo i. et al . , 2006a , apj , 650 , 18 trujillo i. et al . , 2006b , mnras , 373 , l36 udry s. , 1993 , a&a , 268 , 35 van albada t. s. , 1982 , mnras , 201 , 939 van dokkum p. g. , 2005 , aj , 130 , 2647 white s.d.m , frenk c. s. , 1991 , 379 , 52 worthey g. , faber s. m. , gonzalez j. j. , 1992 , apj , 398 , 69 york d. , et al . 2000 , aj , 120 , 1579 zirm a. w. , et al . , 2007 , apj , 656 , 66 note : full data table is available on the online version of mnras or by request to the authors . the following table contains the information about the structural parameters of the sample of galaxies analysed in this paper . column 1 indicates the galaxy identification , column 2 lists the apparent k - band magnitude in the vega system , column 3 is the efffective radius along the semimajor axis , column 4 indicates the value of the srsic index of the fit , column 5 is the ellipticity of the source , column 6 is the stellar mass of the galaxy in units of 10@xmath12h@xmath3@xmath4m@xmath5 , column 7 is the measured redshift of the object , column 8 and 9 are the r.a . and dec . of the source and , finally , column 10 specifies whether the redshift was determined spectroscopically ( 1 ) or photometrically ( 0 ) . 11051619 & 16.75 & 0.90 & 3.60 & 0.38 & 11.94 & 0.46 & 214.258865 & 52.399944 & 0 + 12004106 & 17.13 & 0.80 & 6.72 & 0.11 & 26.72 & 0.74 & 214.379730 & 52.443295 & 1 + 12004308 & 16.95 & 0.89 & 1.23 & 0.58 & 18.03 & 0.64 & 214.266815 & 52.412926 & 0 + 12004443 & 17.06 & 0.48 & 5.04 & 0.21 & 26.96 & 0.77 & 214.318298 & 52.429092 & 1 +
an interesting framework for studying the non - perturbative quantum field theories is through the use of anti - de sitter / conformal field theory ( ads / cft ) correspondence @xcite . according to this duality , in principle , one can perform gravity calculations to find information about the field theory side or vice versa . in this context , the central charges of the dual theory ( cft ) relate to coupling constants of its dual gravity . therefore , einstein gravity with one coupling constant restricts the dual theory to a limited class of cft with equal central charges @xcite . for extension of the duality beyond this limit , one needs to involve higher curvature terms in the gravity action . it is clear that each correction term introduces a new coupling constant and therefore one may have cft theory with different central charges . indeed , this procedure leads to the richness of the cft theory @xcite . the most natural extension of general relativity with higher curvature terms and with the assumption of einstein that the left hand side of the field equations is the most general symmetric conserved tensor containing no more than two - derivatives of the metric is lovelock theory . lovelock @xcite found the most general symmetric conserved tensor satisfying this property . the resultant tensor is nonlinear in the riemann tensor and differs from the einstein tensor only if the spacetime has more than 4 dimensions . although lovelock gravity leads to second - order field equations and it has ghost free ads solution @xcite , it has been recently shown that quadratic and cubic gravities entail causality violation and there are stringent conditions on the coupling constants @xcite . the problem with the total action of einstein gravity is that it is divergent when evaluated on the solutions @xcite . due to this fact , all the other conserved quantities which is calculated through the use of this action is also divergent . one way of eliminating these divergences is through the use of background subtraction method of brown and york @xcite . in this method , the boundary surface is embedded in another ( background ) spacetime , and one subtracts the action evaluated on the embedded surface of the background spacetime from the total action . such a procedure causes the resulting physical quantities to depend on the choice of reference background . furthermore , it is not possible in general to embed the boundary surface into a background spacetime . for asymptotically ads solutions of einstein gravity , one may remove the non - logarithmic divergences in the action by adding a counterterm action which is a functional of the boundary curvature invariants @xcite . indeed , this counterterm method furnishes a means for calculating the action and conserved quantities intrinsically without reliance on any reference spacetime @xcite . although there may exist a very large number of possible invariants , only a finite number of them are non - vanishing in a given dimension on a boundary at infinity . this method has been applied to many cases such as black holes with rotation , nut charge , various topologies , rotating black strings with zero curvature horizons and rotating higher genus black branes @xcite . although the counterterm method applies for the case of a specially infinite boundary , it was also employed for the computation of the conserved and thermodynamic quantities in the case of a finite boundary @xcite . all of the works mentioned in the previous paragraph were limited to einstein gravity . although the counterterm of lovelock gravity with flat horizon has been introduced @xcite , only a few works related to the counterterm method have been done for lovelock gravity with curved horizon . this is due to the fact that even for einstein gravity , the systematic construction that provides the form of the counterterms becomes cumbersome for high enough dimensions @xcite . indeed in this method , one should reconstruct the spacetime metric by solving iteratively the field equations in the fefferman - graham frame @xcite : @xmath0 dx^{i}dx^{j } , \label{fefmet}\]]where @xmath1 is the boundary data of an initial - value problem governed by the equations of motion . however , even for einstein - hilbert theory , solving the coefficients @xmath2 in eq . ( [ fefmet ] ) as covariant functionals of @xmath3 is only possible for low enough dimensions . thus , it is expected that the holographic renormalization procedure would be even more complicated in lovelock gravity because of the nonlinearity of field equations . indeed , because of the nonlinearity of the field equations solving @xmath4 in eq . ( [ fefmet ] ) as covariant functionals of @xmath5 would be even more cumbersome . so , the authors in ref . @xcite presented an alternative construction of kounterterms . instead of adding counterterms to cancel the divergence at the boundary explained above , they circumvented the difficulties of the standard method by using kounterterms which depend on the intrinsic and the extrinsic curvatures of the boundary . they selected the kounterterms as the boundary terms which are regular on the asymptotic region . indeed , the regularization process is encoded in the boundary terms already presented and there is no need to add further counterterms . the exact rotating solutions of lovelock gravity with curved horizon are not introduced till now . indeed , only static solutions of lovelock gravity with different matter fields are known @xcite . so , because of the difficulties of the holographic renormalization procedure in lovelock gravity and the nonexistance of an exact rotating solution of this theory , we limit ourselves to the case of counterterms of lovelock gravity for static solutions . the counterterms of asymptotically ads static solutions of gauss - bonnet gravity have been introduced in ref . @xcite . also , the finite action and global charges of asymptotically de sitter static solutions has been obtained in ref . @xcite . here we like to apply the counterterm method to the case of the static solutions of the field equations of third order lovelock gravity with curved horizon . we define the cosmological constant in such a way that the maximally symmetric ads spacetime @xmath6be the vacuum solution of lovelock gravity . ( [ adsmet ] ) @xmath7 is the metric of an ( @xmath8)-dimensional maximally symmetric space with curvature constant @xmath9 and volume @xmath10 . indeed , this choice of cosmological constant makes the asymptotic form of the solutions of lovelock gravity to be exactly the same as that of einstein gravity . thus , we expect that the counterterm introduced for einstein gravity in @xcite may remove the power law divergences in the action of lovelock gravity . although the counterterms which should be added to gauss - bonnet gravity in order to remove the power law divergences of the action for static solutions are introduced in ref . @xcite , they depend on the gauss - bonnet coefficient . however , because of our choice of the cosmological constant , our counterterms are the same as those of einstein gravity and are independent of lovelock coefficients . in order to check our counterterms , we calculate the finite action and the mass of the black hole through the use of counterterm method . then , we use these finite quantities and the gibbs - duhem relation to obtain the entropy . we find that the calculated entropy of the black holes is consistent with the wald s formula @xcite . as another test of our counterterm method , we show that the mass obtained through the use of counterterm method satisfies the first law of thermodynamics . we , also , perform a stability analysis of the black hole solutions in canonical ensemble and investigate the effects of third order lovelock term on the stability . this paper is organized as follows . in section [ sol ] , we review the well - defined action of lovelock gravity . in section [ count ] , we introduce the counterterms for third order lovelock gravity for static spacetimes . we also , introduce the finite stress energy tensor of this theory . section [ therm ] is devoted to the thermodynamics of the black hole solutions of the theory . we calculate the finite action , the total mass , the temperature , the charge and the electric potential . we calculate the entropy through the use of gibbs - duhem relation and wald formula and find that they are consistent . we , also , investigate the first law of thermodynamics . in sec . [ stab ] , we investigate the thermal stability of the solutions in canonical ensemble . we finish our paper with some concluding remarks . the bulk action of lovelock gravity in @xmath11 dimensions may be written as @xcite @xmath12 } \alpha _ { p}\left ( \mathcal{l}_{p}-2\lambda _ { p}\right ) + i_{mat } \label{action1}\]]with @xmath13 $ ] denoting the integer part of @xmath14 , @xmath15 s ( @xmath16 ) are lovelock coefficients , @xmath17is the euler density of a @xmath18-dimensional manifold , @xmath19 is the general asymmetric kronecker delta and@xmath20the action ( [ action1 ] ) is written in such a way that the maximally symmetric ads spacetime ( [ adsmet ] ) is the vacuum solution of action ( action1 ) . in this notation , the independent coupling constants are @xmath21 and all the lovelock coefficients . from a geometric point of view the lagrangian of the action ( [ action1 ] ) in @xmath22 + 1 $ ] and @xmath22 + 2 $ ] dimensions is the most general lagrangian that yields second order field equations , as in the case of einstein - hilbert action which is the most general lagrangian producing second order field equations in three and four dimensions . in the rest of the paper , we work in a unit system with @xmath23 and the dimensionless lovelock coefficients @xmath24 defined as@xmath25with the definition ( [ alphahat ] ) , the cosmological constant for ads spacetime is@xmath26 } ( -1)^{p}\tilde{% \alpha}_{p}.\ ] ] in this paper , we consider the third order lovelock gravity in the presence of electromagnetic field . thus , the action of matter field is @xmath27}\partial ^{\lbrack \mu } a^{\nu ] } , \]]where @xmath28 is the electromagnetic potential . the first term in lovelock lagrangian is the einstein - hilbert term @xmath29 , the second term is the gauss - bonnet lagrangian @xmath30 , the third term is @xmath31and the cosmological constant is@xmath32 as in the case of einstein - hilbert action , the action ( [ action1 ] ) does not have a well - defined variational principle , since one encounters a total derivative that produces a surface integral involving the derivatives of @xmath33 normal to the boundary @xmath34 . these normal derivatives of @xmath35 can be canceled by the variation of the surface action @xcite @xmath36 } \alpha _ { p}q_{p},\]]where @xmath37}^{[jj_{1}\cdots j_{2p-1}]}\,k_{j_{1}}^{i_{1}}\,\times \notag \\ & & \times \left ( \frac{1}{2}\,\hat{r}_{j_{2}j_{3}}^{i_{2}i_{3}}(\gamma ) -t^{2}k_{j_{2}}^{i_{2}}k_{j_{3}}^{i_{3}}\right ) \cdots \left ( \frac{1}{2}\,% \hat{r}_{j_{2p-2}j_{2p-1}}^{i_{2p-2}i_{2p-1}}(\gamma ) -t^{2}\,k_{j_{2p-2}}^{i_{2p-2}}k_{j_{2p-1}}^{i_{2p-1}}\right ) . \label{surlag}\end{aligned}\]]in eq . ( [ surlag ] ) @xmath38 and @xmath39 are the induced metric and extrinsic curvature of the boundary @xmath34 , respectively . the explicit form of the first three terms of eq . ( [ surlag ] ) are @xcite : @xmath40where @xmath41 and @xmath42 are the traces of @xmath43and @xmath44k_{ab } \notag \\ & & -(4k^{3}-12kk_{ed}k^{ed}+8k_{de}k_{f}^{e}k^{fd})k_{ac}k_{b}^{c}-24kk_{ac}k^{cd}k_{de}k_{b}^{e } \notag \\ & & + ( 12k^{2}-12k_{ef}k^{ef})k_{ac}k^{cd}k_{db}+24k_{ac}k^{cd}k_{de}k^{ef}k_{bf}\ } , \label{pab}\end{aligned}\]]respectively . in eq . ( [ isurf ] ) @xmath45 is the einstein tensor , @xmath46 is the intrinsic curvature and @xmath47 is the gauss - bonnet tensor of the metric @xmath38 given as @xmath48 it is well known that the action @xmath49 is not finite for asymptotically ads solutions . inspired by ads / cft correspondence , one needs to add counterterms to the gravity action in order to get a finite action . these counterterms are made from the curvature invariants of the boundary metric with the coefficients of the higher curvature terms chosen so that power law divergences in the bulk are canceled for all possible boundary topologies permitted by the equations of motion . at any given dimension there are only a finite number of counterterms that do not vanish at infinity . this does not depend upon what the gravity theory is i.e. whether or not it is einstein , gauss - bonnet , 3rd order lovelock , etc . indeed , for asymptotically ads solutions , the boundary counterterms that cancel the divergences in einstein gravity may also cancel the divergences in lovelock gravity if one chooses the cosmological constant as in eq . this is due to the fact that the @xmath50th order lovelock lagrangian @xmath51 calculated for the metric ( [ adsmet ] ) is independent of lovelock coefficients . that is , the different orders of lovelock action do not mix with each other and one may find the counterterms for different orders of lovelock terms separately . this point makes the calculation easier . of course , the coefficients of the various counterterms for different lovelock terms will be different , depend only on @xmath21 and will be independent of lovelock coefficients . thus , using the counterterms of einstein gravity @xcite , we may write the counterterms of third order lovelock gravity as @xmath52where @xmath53the coefficients @xmath54 , @xmath55 , @xmath56 and @xmath57 in eq . ( [ ict - p ] ) should depend on @xmath21 . these coefficients for einstein gravity ( @xmath58 , @xmath59 ) are @xcite @xmath60one may note that the coefficients @xmath54 , @xmath55 , @xmath56 and @xmath57 are independent of lovelock coefficients . we apply the counterterms ( [ ict - p ] ) to various static solutions of gauss - bonnet and third order lovelock gravity with different topolgy and find that these coefficients are @xmath61and @xmath62for gauss - bonnet and third order lovelock gravity , respectively . the reason that we use exactly the counterterms of einstein gravity is as follows . first , the boundary at @xmath63 for static solutions is a constant curvature hypersurface and therefore no six - derivative term will be appeared in the counterterms . in other words , all the terms of a specific order of counterterms [ for example @xmath64 and @xmath65 in @xmath66 ) $ ] ] for static solutions are proportional to @xmath67 . also , @xmath68 is proportional to @xmath67 . therefore , in order to remove the divergences of the action which are proportional to @xmath69 , any combination of @xmath64 , @xmath65 and @xmath68 can be used . so , we just use exactly the counterterms of einstein gravity . second , @xmath54 * * , * * @xmath55 * * , * * @xmath56 * * , * * @xmath57 , for @xmath70 and @xmath71 are proportional to those of einstein counterterm independent of the dimensions . that is , the dimensional - dependence of these coefficients are the same as those in einstein gravity . third , as we will see in the next section , the entropy calculated through the use of gibbs - duhem relation and the mass and action calculated by our counterterms is consistent with the entropy obtained by use of wald s formula . fourth , the mass calculated by our counterterms satisfies the first law of thermodynamics . while the total action @xmath72 is appropriate in grand - canonical ensemble where @xmath73 is zero at the boundary , the appropriate action in the canonical ensemble where the electric charge is fixed is @xcite @xmath74thus both in canonical and grand - canonical ensemble , the variation of total action about the solutions of the field equations is @xmath75so , the energy - momentum tensor can be written as : @xmath76the explicit expressions of @xmath77 and @xmath78 are somewhat cumbersome so we give them in the appendix . to compute the conserved charges of the spacetime , we choose a spacelike hypersurface @xmath79 in @xmath34 with metric @xmath80 , and write the boundary metric in adm form : @xmath81where the coordinates @xmath82 are the angular variables parameterizing the hypersurface of constant @xmath83 around the origin , and @xmath84 and @xmath85 are the lapse and shift functions , respectively . when there is a killing vector field @xmath86 on the boundary , then the quasilocal conserved quantities associated with the stress tensors of eq . ( [ stres ] ) can be written as @xmath87where @xmath88 is the determinant of the metric @xmath89 , and @xmath90 is the timelike unit normal vector to the boundary @xmath79 . in the context of counterterm method , the limit in which the boundary @xmath79 becomes infinite ( @xmath91 ) is taken , and the counterterm prescription ensures that the action and conserved charges are finite . no embedding of the surface @xmath79 into a reference of spacetime is required and the quantities which are computed are intrinsic to the spacetimes . the field equation of third order lovelock gravity for the static metric @xmath92 in the presence of electromagnetic field may be written as @xmath93-\tilde{\alpha}_{2}[\psi ^{2}(r)-1]+[\psi ( r)-1]+\frac{ml^{2}}{r^{n}}-\frac{2}{(n-1)(n-2)}\frac{q^{2}l^{2}}{r^{2(n-1)}}% = 0 , \label{cubic}\]]where @xmath94l^{2}/r^{2}$ ] , @xmath95 is the charge of the black hole and @xmath96 is the integration constant which is related to the mass of the solution . the electromagnetic potential for the above metric is@xmath97the mass parameter @xmath96 may be written in terms of horizon radius and @xmath95 by using the fact that @xmath98 , and therefore @xmath99where@xmath100as one expects , @xmath101 ( @xmath102 ) is the root of eq . ( [ cubic ] ) for the vacuum spacetime ( @xmath103 and @xmath104 ) . the three solutions of the cubic equation ( [ cubic ] ) are @xmath105where @xmath106all of the above three roots could be real in the appropriate range of @xmath107 and @xmath108 . the second and third solutions are real provided @xmath109 . here , we will consider only the first solution @xmath110 which is real provided @xmath111 at any @xmath83 . now , we investigate the thermodynamics of the black hole solutions . the temperature of the event horizon may be calculated through the use of analytic continuation of the metric . one obtains @xmath112n% \frac{r_{+}^{2}}{l^{2}}+k(n\mu -2\eta ) -\frac{2q^{2}}{(n-1)r_{+}^{2(n-2)}}% \right ) , \label{temp}\]]where@xmath113the charge of the black holes per unit volume can be calculated by integrating the flux of the electric field as @xmath114the electric potential @xmath115 , measured at infinity with respect to the horizon is defined as @xmath116where @xmath117 is the null generator of the horizon . one obtains @xmath118 the finite total action in grand - canonical and canonical ensembles can be found through the use of the counterterm method introduced in the last section . it is a matter of straightforward calculations to show that the total action is finite . since we are interested in the stability of the solutions in canonical ensemble , we calculate the finite euclidian action per unit volume @xmath10 in this ensemble . one obtains @xmath119 \frac{r_{+}^{n}}{l^{2}}-k(n\mu -2\eta ) \frac{\xi } { % \eta } r_{+}^{n-2}\right\}+i_{0}\end{aligned}\]]where @xmath120 is the euclidean time period ( the inverse of temperature ) , @xmath121 is given in eq . ( [ eta ] ) , @xmath122 is@xmath123and @xmath124 is @xmath125one should note that @xmath124 appears only in odd dimensions ( even @xmath126 ) . for our static solution , there is a killing vector field @xmath127 on the boundary , and therefore the quasilocal conserved quantity associated with the stress tensors of eq . ( [ stres ] ) is the mass of the black hole . using the counterterm method introduced in the last section , it is a matter of calculation to obtain the mass of black hole as @xmath128where @xmath129 is the casimir energy for the vacuum ads metric per volume @xmath10 , which is nonzero only in odd dimensions . in order to have positive energy @xmath130 , one should restrict the range of lovelock coefficients in terms of @xmath131 , @xmath21 and @xmath132 as@xmath133that is , the third order lovelock coefficient has a lower limit in terms of @xmath107 , @xmath95 , @xmath21 and @xmath132 . using the mass and action per unit volume @xmath10 calculated through the use of counterterm method , one may calculate the entropy per unit volume through the use of gibbs - duhem relation @xmath134 as @xmath135it is worth to note that the entropy of the black hole solution per unit volume @xmath10 calculated by the use of gibbs - duhem relation is consistent with the calculation through the use of wald formula wald , myers2 @xmath136where the integration is done on the @xmath137-dimensional spacelike hypersurface of killing horizon , @xmath138 is the induced metric on it , @xmath139 is the determinant of @xmath138 and @xmath140 is the @xmath141nd order lovelock lagrangian of @xmath142 . also , one may note that the thermodynamic quantities calculated in this section satisfy the first law of thermodynamics@xmath143 now , we study the thermal stability of the black hole solutions of third order lovelock gravity in canonical ensemble . first , we investigate the conditions of having black hole solution in lovelock gravity . the solution given by eq . ( [ sol1 ] ) presents a black hole solution provided @xmath144 has at least one real positive root . this occurs if @xmath145 , or @xmath146 where @xmath147 and @xmath148 satisfy the following equation : @xmath149n\frac{r_{ext}^{2}}{l^{2}}% + k[(n-2)+k(n-4)\frac{l^{2}}{r_{ext}^{2}}+k^{2}(n-6)\frac{l^{4}}{r_{ext}^{4}}-% \frac{2q_{ext}^{2}}{(n-1)r_{ext}^{2(n-2)}}=0 . \label{ext}\ ] ] also , the temperature of a physical black hole should be positive . one may note that the temperature changes its sign at the root of @xmath150 ( @xmath151 ) . the radius @xmath152 depends on @xmath21 and lovelock coefficients , while @xmath147 depends on @xmath95 too . in fig . [ fig1 ] , the vertical line is @xmath153 line . this figure shows that one may have only small ( @xmath154 ) and large black holes , and there is no medium black hole solution ( @xmath155 ) . this feature does not occur for black holes of einstein gravity or lovelock gravity with positive lovelock coefficients , since @xmath156 . thus , for the case of lovelock gravity with negative lovelock coefficient(s ) one may divide the black holes into two classes with negative and positive @xmath157 . in the canonical ensemble @xmath131 is fixed and therefore the black hole solutions are stable provided @xmath158 in the range that @xmath159 is positive . using the expressions for mass and entropy given in eq . ( [ mass ] ) and ( [ ent ] ) , one may calculate @xmath160 : @xmath161where @xmath162 \\ & & -k^{2}r_{+}^2\left[n(\tilde{\alpha}_{2}(15\tilde{\alpha}_{3}+1)-15\tilde{% \alpha}_{3}(1 + \tilde{\alpha}_{3}))-8\tilde{\alpha}_{2}\right]l^2 - 2k\left[n({% \tilde{\alpha}_{2}}^2 - 2\tilde{\alpha}_{3})- 4{\tilde{\alpha}_{2}}^2 - 6\tilde{% \alpha}_{3}\right]l^4 \\ & & -3k^2\tilde{\alpha}_{3}\tilde{\alpha}_{2}r_{+}^{-2}(n-8)l^{6 } -3k{\tilde{% \alpha}_{3}}^2r_{+}^{-4}(n-6)l^8.\end{aligned}\ ] ] . versus @xmath132 for @xmath163 , @xmath164 , @xmath165 and @xmath166 in einstein gravity ( solid ) and lovelock gravity ( dotted ) with @xmath167 and @xmath168 . the vertical line is @xmath169 line.,title="fig : " ] ( dotted ) and @xmath170 ( solid ) versus @xmath132 for @xmath171 and @xmath166 . the vertical line is @xmath172 line.,title="fig : " ] ( dotted ) and @xmath173 ( solid ) versus @xmath174 for @xmath175 , @xmath176 , @xmath164 , @xmath165 , @xmath166 and @xmath177 . the vertical line is @xmath169 line.,title="fig : " ] ( dotted ) and @xmath178 ( solid ) versus @xmath174 for @xmath175 , @xmath176 , @xmath164 , @xmath165 , @xmath166 and @xmath179.,title="fig : " ] ( dotted ) and @xmath180 ( solid ) versus @xmath181 for @xmath182 , @xmath183 , @xmath104 , @xmath164 , @xmath165 and @xmath166 . the vertical line is @xmath169 line.,title="fig : " ] ( dotted ) and @xmath184 ( solid ) versus @xmath174 for @xmath185 , @xmath176 , @xmath164 , @xmath165 , @xmath166 and @xmath186.,title="fig : " ] ( dotted ) and @xmath187 ( solid ) versus @xmath174 for @xmath175 , @xmath176 , @xmath164 , @xmath165 , @xmath166 and @xmath177 . the vertical line is @xmath169 line.,title="fig : " ] to investigate the stability of black holes of third order lovelock gravity in canonical ensemble , we should find the sign of @xmath188 , when @xmath159 is positive . in order to investigate the stability , we plot @xmath159 and @xmath189 in one figure . the allowed region for investigation of @xmath190 is when @xmath159 is positive . in fig . [ fig2 ] , the vertical dotted - line is @xmath150 line . this figure shows that the small black holes ( @xmath154 ) divided into stable and unstable black holes . that is a hawking - page phase transition exists for small black holes between very small and small black holes . there is no phase transition for large black holes ( @xmath191 ) as one may see in fig . [ fig2 ] . for investigating the effect of third order term of lovelock gravity on the stability of the black hole solutions , we plot @xmath159 and @xmath188 versus @xmath108 for small and large black holes . figure [ fig3 ] shows that there is no large black holes for large negative @xmath108 , while the large black holes for small negative @xmath108 are not stable . as @xmath192 becomes larger , large stable black holes may exist . figure [ fig4 ] shows that there is no small black hole solution for positive @xmath192 , while the small black hole with negative @xmath108 are stable . figure [ fig5 ] shows that there is no negative-@xmath157 ( small ) uncharged black hole with event horizon , while the large black holes ( @xmath193 ) show a phase transition between large and very large ones . one may see in fig . [ fig5 ] that the large black holes with positive @xmath157 show a phase transition , while these black holes with negative @xmath157 are not stable . finally , figs . [ fig6 ] and [ fig7 ] show the temperature and @xmath194 versus @xmath108 for small and large uncharged black holes . figure [ fig6 ] shows that there is no uncharged black hole with negative @xmath157 , while the black holes with positive @xmath157 are stable . on the other hand , the large uncharged black holes with positive @xmath157 show a phase transition as @xmath192 becomes larger and these solutions are not stable for very large negative @xmath108 . the concepts of action plays a central role in gravitation theories , but the sum of the bulk and the surface terms diverges . in this paper , we introduced the counterterms that remove the non - logarithmic divergences of static solutions of third order lovelock gravity . we did this by defining the cosmological constant in such a way that the ads metric ( [ adsmet ] ) is the vacuum solution of lovelock gravity . indeed , the cosmological constant ( [ cos ] ) makes the asymptotic form of the solutions of lovelock gravity to be exactly the same as that of einstein gravity . thus , we employed the counterterms introduced for einstein gravity in @xcite and found that the power law divergences of static solutions in the action of lovelock gravity can be removed by suitable choice of coefficients . we found that the counterterms are independent of lovelock coefficients and the dimensionally dependent of them is the same as those of einstein gravity . the main difference of our work with that of ref . [ 16 ] is that the counterterms are exactly the same as those of einstein gravity and do not depend on lovelock coefficients . the only job which remains is that one needs to calculate the coefficients @xmath54 , @xmath55 , @xmath56 and @xmath57 . for example if one wants to remove the divergences of the action due to the gauss - bonnet term , one should calculate @xmath195 , @xmath196 , @xmath197 and @xmath198 that remove the divergences of @xmath199 , without regarding the other lovelock terms . this enables one to generalize this method to other higher curvature theories of gravity easily , including fourth order lovelock gravity or @xmath200 gravity . we also introduced the finite energy - momentum tensor in third order lovelock gravity . in addition , we employed these counterterms to calculate the finite action and mass of the static black hole solutions of third order lovelock gravity . calculating the temperature , the electric charge and electric potential and using the calculated finite action and mass , we showed that the entropy calculated through the use of gibbs - duhem relation is consistent with the calculated entropy by wald s formula . we , also , showed that the conserved and thermodynamic quantities satisfy the first law of thermodynamics . finally , we investigated the stability of charged black holes of lovelock gravity in canonical ensemble . we found that the black holes with respect to the sign of @xmath201 given in eq . ( [ eta ] ) may be divided into two classes . the negative @xmath201 ( small ) black holes show a phase transition between very small and small black holes , while the large black holes are stable . there is no black hole solution with medium size . of course by small black holes we mean the size of them with respect to the cosmological parameter @xmath21 . we , also , investigated the effects of third order lovelock term on the stability of the solutions . we found that there is no large black holes for large negative @xmath108 , while the black holes for small negative @xmath108 are not stable . we also found that as @xmath108 becomes larger , stable black holes may exist . this shows that there is a phase transition as the third order term of lovelock gravity becomes larger . finally , we found that there is no small black hole solution for positive @xmath108 , while the black hole with negative @xmath108 are stable . finally , we considered the uncharged black holes of lovelock gravity and found that negative @xmath201 solutions are not black holes with event horizon , while the positive @xmath201 black holes show a phase transition . the first term of eq . ( [ stres ] ) which is the energy - momentum tensor of the surface term is@xmath202 \\ & & + 3\alpha _ { 3}[5p_{ab}-p\gamma _ { ab}+2k\hat{g}_{ab}^{(2)}+\mathcal{l}% _ { 2}(k_{ab}+k\gamma _ { ab})+4j\hat{r}_{ab}-24j_{(a}{}^{c}\hat{r}_{b)c}+8k^{cd}% \hat{r}_{ac}\hat{r}_{bd}-8kk_{a}{}^{c}k_{b}{}^{d}\hat{r}_{cd } \\ & & + 8kk_{ab}k^{cd}\hat{r}_{cd}-8k^{cd}\hat{r}_{ab}\hat{r}_{cd}+16k_{(a}{}^{c}% \hat{r}_{b)}{}^{d}\hat{r}_{cd}+16k_{(a}{}^{c}k_{b)}{}^{d}k_{d}{}^{e}\hat{r}% _ { ce}-8k_{ab}k_{c}{}^{e}k^{cd}\hat{r}_{de}+6j_{ab}\hat{r } \\ & & -8k_{(a}{}^{c}\hat{r}_{b)c}\hat{r}-12j^{cd}\hat{r}_{acbd}-4k^{cd}\hat{r}% \hat{r}_{acbd}-8k_{a}{}^{c}k_{bc}k^{de}\hat{r}% _ { de}+16k_{c}{}^{d}k^{ef}k_{(a}{}^{c}\hat{r}% _ { b)edf}+16k_{d}{}^{f}k^{de}k_{(a}{}^{c}\hat{r}_{b)ecf } \\ & & + 16k_{(a}{}^{c}\hat{r}_{b)dce}\hat{r}^{de}-16kk^{de}k_{(a}{}^{c}\hat{r}% _ { b)dce}-16k^{cd}\hat{r}_{(a}{}^{e}\hat{r}_{b)cde}-4k^{cd}\hat{r}_{ac}{}^{ef}% \hat{r}_{bdef}+8k^{cd}\hat{r}_{c}{}^{e}\hat{r}_{aebd } \\ & & + 8k^{cd}\hat{r}_{c}{}^{e}\hat{r}_{adbe}-8k^{cd}\hat{r}% _ { a}{}^{e}{}_{c}{}^{f}\hat{r}_{bedf}-8k_{(a}{}^{c}\hat{r}_{b)}{}^{def}\hat{r}% _ { cdef}+8k_{a}{}^{c}k_{b}{}^{d}k^{ef}\hat{r}_{cedf}-4k_{ab}k^{cd}k^{ef}\hat{r% } _ { cedf } \\ & & + 8k^{cd}\hat{r}_{a}{}^{e}{}_{b}{}^{f}\hat{r}_{cedf}+2\gamma _ { ab}(\hat{g}% _ { cd}^{(2)}k^{cd}+6j^{cd}\hat{r}_{cd}-j\hat{r}+2kk^{cd}k^{ef}\hat{r}% _ { cedf}-4k_{c}{}^{e}k^{cd}k^{fh}\hat{r}_{dfeh})]\}\end{aligned}\]]and the second term corresponding to the counterterm is @xmath203 \right . \\ & & \left . -d_{p}\left [ \frac{2(3n+2)}{4(n-1)}\left [ \left ( \hat{g}_{ab}^{(1)}% \hat{r}^{cd}\hat{r}_{cd}\right ) -d_{a}d_{b}\left ( \hat{r}^{ef}\hat{r}% _ { ef}\right ) + \gamma _ { ab}d^{2}\left ( \hat{r}^{ef}\hat{r}_{ef}\right ) + 2\hat{% r}\hat{r}_{a}^{\;c}\hat{r}_{bc}\right . \\ & & \left . + \gamma _ { ab}d_{c}d_{d}\left ( \hat{r}\hat{r}^{cd}\right ) + d^{2}\left ( \hat{r}\hat{r}_{ab}\right ) -d^{c}d_{b}\left ( \hat{r}\hat{r}% _ { ac}\right ) -d^{c}d_{a}\left ( \hat{r}\hat{r}_{bc}\right ) \right ] -\frac{% n(n+2)}{16(n-1)^{2}}\left [ -\gamma _ { ab}\hat{r}^{3}\right . \right . \\ & & \left . + 6\left ( \hat{r}^{2}\hat{r}_{ab}-d_{a}d_{b}\hat{r}% ^{2}+\gamma _ { ab}d^{2}\hat{r}^{2}\right ) \right ] -2\left [ -\gamma _ { ab}\hat{r% } ^{ef}\hat{r}^{cd}\hat{r}_{ecfd}+3\left ( \hat{r}_{a}^{\;e}\hat{r}^{cd}\hat{r}% _ { ecbd}+\hat{r}_{b}^{\;e}\hat{r}^{cd}\hat{r}_{ecad}\right ) \right . \\ & & \left . -2d_{c}d_{d}\left ( \hat{r}_{ab}\hat{r}^{cd}\right ) + 2d_{c}d_{d}\left ( \hat{r}_{a}^{\;c}\hat{r}_{b}^{\;d}\right ) + 2\gamma _ { ab}d_{e}d^{f}\left ( \hat{r}^{cd}\hat{r}_{\;cfd}^{e}\right ) + 2d^{2}\left ( \hat{r}^{cd}\hat{r}_{acbd}\right ) \right . \\ & & \left . \left . -2d_{e}d_{a}\left ( \hat{r}^{cd}\hat{r}_{\;cbd}^{e}\right ) -2d_{e}d_{b}\left ( \hat{r}^{cd}\hat{r}_{\;cad}^{e}\right ) \right ] + \frac{n-2% } { 2(n-1)}\left [ -d_{a}\hat{r}d_{b}\hat{r}+\hat{r}_{b}{}^{c}d_{c}d_{a}\hat{r}% \right . \right . \\ & & \left . + \hat{r}_{a}{}^{c}d_{c}d_{b}(\hat{r})-2d_{c}d^{c}d_{b}d_{a}(\hat{r}% ) -\frac{7}{2}\hat{r}_{a}{}^{c}{}_{b}{}^{d}d_{c}d{d}(\hat{r})+2d_{a}(\hat{r}% _ { bc})d^{c}(\hat{r})+d_{b}(\hat{r}_{ac})d^{c}(\hat{r})\right . \\ & & \left . + d_{b}\hat{r}_{ac}d^{c}\hat{r}-2d_{c}\hat{r}_{ab}d^{c}\hat{r}-\hat{r% } _ { b}{}^{c}{}_{c}{}^{d}d_{a}d_{d}\hat{r}-\hat{r}% _ { a}{}^{c}{}_{c}{}^{d}d_{d}d_{b}\hat{r}\right . \\ & & \left . -\hat{r}_{a}{}^{c}{}_{b}{}^{d}d_{d}d_{c}(\hat{r})+2\hat{r}% _ { ab}d_{e}d^{e}(\hat{r})+\frac{1}{2}\gamma _ { ab}\left ( d_{e}\hat{r}d^{e}\hat{% r}+d^{4}\hat{r}\right ) \right ] \right . \\ & & -\left . \left [ 2d_{a}d_{b}(d^{2}\hat{r})-2d^{4}\hat{r}_{ab}+2d_{p}\hat{r}% _ { ( a}\hat{r}_{b)}^{p}-4\,\hat{r}^{pq}d_{a}d_{b}\hat{r}_{pq}+12\,\hat{r}% ^{pq}d_{(b}\hat{r}_{pa)q}-4(d^{2}\hat{r}^{pq})\hat{r}_{paqb}\right \\ & & \left . + 6d_{p}(\hat{r})d_{(b}\hat{r}_{a)}^{p}-2d_{a}(\hat{r}^{pq})d_{b}(% \hat{r}_{pq})+4d_{(a}\,d_{q}\left ( \hat{r}^{pq}\hat{r}_{b)p}\right ) + 16d^{r}(\,\hat{r}^{pq})d_{(b}\hat{r}_{\|rqp\|a)}\right . \\ & & \left . -4\,\hat{r}^{pq}\hat{r}_{pa}\hat{r}_{qb}+8\,\hat{r}^{pr}\hat{% r}_{r}^{q}\hat{r}_{paqb}+4\,\hat{r}^{pq}\hat{r}_{(a}^{r}\hat{r}_{\|rqp\|b)}% \right ] \right . \\ & & + \left . \gamma _ { ab}\left [ -\,d^{4}\hat{r}-4\,d_{p}d_{q}(\hat{r})% \hat{r}^{pq}+2\hat{r}_{pq}d^{2}(\hat{r}^{pq})+4\,d_{r}d_{s}(\hat{r}_{pq})% \hat{r}^{prqs}-\,d_{e}(\hat{r})d^{e}(\hat{r})\right . \\ & & \left . + 5\,d_{r}(\hat{r}_{pq})d^{r}(\hat{r}^{pq})-8\,d_{r}(\hat{r}% _ { pq})d^{q}(\hat{r}^{pr})-4\,\hat{r}_{pq}\hat{r}_{r}^{p}\hat{r}^{qr}+4\,\hat{% r}_{pq}\hat{r}_{rs}\hat{r}^{prqs}\right ] \right . \\ & & \left + \frac{1}{n-1}\left [ 2\,d_{a}d_{b}(d^{2}\hat{r})-2\,(d^{2}% \hat{r})\hat{r}_{ab}+d_{a}(\hat{r})d_{b}(\hat{r})+\gamma _ { ab}\left ( -2\,d^{4}\hat{r}-\frac{1}{2}\,d_{c}\hat{r}d^{c}\hat{r}\right ) \right ] \right ] \right\ } .\end{aligned}\]]where @xmath54 s , @xmath55 s , @xmath56 s and @xmath57 s are given in eqs . ( a1- [ a3 ] ) . the above counterterms remove the divergences of the energy - momentum tensor for @xmath204 . as in the case of einstein gravity , one should add more counterterms for @xmath205 . r. g. cai , z. y. nie and y. w. sun , phys . d * 78 * , 126007 ( 2008 ) ; r. g. cai , z. y. nie , n. ohta and y. w. sun , phys . d * 79 * , 066004 ( 2009 ) ; m. blau , k. narain and e. gava , j. high energy phys . * 09 * , 018 ( 1999 ) ; a. buchel , j. t. liu and a. o. starinets , nucl . phys . b * 707 * , 56 ( 2005 ) ; y. kats and p. petrov , j. high energy phys . * 01 * , 044 ( 2009 ) ; a. buchel , nucl . b * 803 * , 166 ( 2008 ) ; r. c. myers , m. f. paulos and a. sinha , phys . d * 79 * , 041901 ( 2009 ) ; a. buchel , r. c. myers , m. f. paulos and a. sinha , phys . b * 669 * , 364 ( 2008 ) ; d. m. hofman and j. maldacena , j. high energy phys . * 05 * , 012 ( 2008 ) ; d. m. hofman , nucl . b * 823 * , 174 ( 2009 ) ; j. de boer , m. kulaxizi and a. parnachev , arxiv:0912.1877 . m. h. dehghani and m. shamirzaie d * 74 * , 064023 ( 2006 ) ; m. h. dehghani , r. pourhasan , phys . rev . d * 79 * , 064015 ( 2009 ) ; n. dadhich , s. g. ghosh and s. jhingan , phys . b * 711 * , 196 ( 2012 ) . ; n. farhangkhah and m. h. dehghani , phys . d * 90 * , 044014 ( 2014 ) .
in this article we study a type of resampling procedure for markov chains . the procedure , called weighted ensemble ( we ) sampling @xcite , simulates some replicas of a markov chain @xmath0 , resampling from the replicas at certain time intervals . the distinguishing feature of we sampling , compared to other resampling techniques , is that we is designed so that the replicas of @xmath0 are evenly distributed throughout state space . this is achieved by dividing state space into bins and resampling in each bin so that the number of replicas in a bin remains roughly constant . the replicas carry probabilistic weights so that the resulting statistical distribution is unbiased . this distribution can be used , in principle , to estimate any function of of @xmath0 at a fixed time @xcite . our interest in we arises from longstanding problems in computational chemistry . in this setting , @xmath0 is a time discretization of some stochastic molecular dynamics ( md ) . md simulations have proven useful for understanding many chemical and biological processes ; see @xcite for an overview . however , such simulations are limited by time scale separation . many phenomena of interest occur at the laboratory time scale of microseconds , while md simulations have time steps that correspond to femtoseconds . in this case , straightforward md simulations are not practical . many methods exist for extending the time scale of md simulations ; we do not attempt to give a review of them here . however , we mention some related methods , including exact milestoning @xcite , neus @xcite , trajectory tilting @xcite , transition interface sampling @xcite , forward flux sampling @xcite , and boxed molecular dynamics @xcite . see for instance @xcite for review and comparison of these methods . we will compare we to exact milestoning in section [ sec : dynamics ] below . in this article we give a mathematical framework for we based on martingale theory . the framework allows us to prove that we is exact even when the bins and the number of replicas per bin are chosen adaptively , as noted earlier in @xcite . while we can be used with a broad range of stochastic processes , when the process is time homogenous and markovian as in many models of md , such as langevin dynamics we can be used to efficiently compute dynamical quantities like reaction rates without actually simulating the reactions @xcite . these computations rely on hill s relation @xcite , which we generalize in section [ sec : dynamics ] below . from hill s relation , obtaining reaction rates and probabilities requires a calculation on the stationary distribution of a nonreversible process . to speed up this stationary calculation , we is combined with a preconditioning step @xcite in which a markov state model ( msm ) @xcite is used to approximate the stationary distribution . this is sometimes called accelerated we @xcite . accelerated we begins with replicas evenly distributed in space , with weights chosen to match the stationary distribution of the msm . the replicas are then allowed to relax according to their exact dynamics , with we sampling ensuring that the replicas remain evenly distributed in state space . we show below that information from the msm can be used to optimize the we sampling in this relaxation step , in the sense that the variance in the appropriate stationary calculation is minimized . this optimization requires an adaptive number of replicas per bin , in contrast with traditional we sampling . we show in a simple model that this adaptive sampling can be significantly better than traditional we sampling . this article is organized as follows . in section [ sec : assumptions ] , we introduce notation for the replica processes , and we make assumptions on we sampling that we assume hold throughout . ( we refer to the replica processes as `` point and weight '' processes , following familiar notation in the mathematical resampling literature . ) we prove that we is unbiased in a general setting that includes adaptive binning procedures . in section [ sec : sampling ] , we introduce an explicit algorithm for we sampling ( algorithm [ alg1 ] ) and prove that it is unbiased ( proposition [ prop1 ] below ) . we give an exact formula for the mean squared error of this sampling ( proposition [ prop2 ] ) . then , using this error formula , we produce a heuristic near - optimal adaptive strategy . in section [ sec : stationary ] , we propose a version of accelerated we which uses this strategy ( algorithm [ alg2 ] ) . in section [ sec : example ] , we implement the adaptive strategy on a simple model , and compare with traditional we sampling as well as naive sampling . in section [ sec : dynamics ] , we show how to obtain certain dynamical averages , including mean first passage times and hitting probabilities , from the stationary distribution of a nonreversible markov chain . all proofs are in section [ sec : proofs ] . in this section we introduce a general mathematical framework for unbiased we sampling . throughout @xmath1 is a markov chain taking values in a measurable space @xmath2 , with transition kernels @xmath3 . we write @xmath4 for the expectation operator of @xmath1 . below we consider sampling schemes for @xmath1 of the following type . let @xmath5 and @xmath6 be a points and weights processes evolving by selection and mutation : @xmath7 here , each @xmath8 and @xmath9 . we refer to @xmath5 and @xmath6 as the points and weights at time @xmath10 . the expectation operator associated to the points and weights process will be denoted @xmath11 . below we describe rules which guarantee the points and weights process is unbiased . readers interested only in explicit algorithms can skip to the next section . before describing this process in more detail , we need some notation . throughout , @xmath12 indicates denote equality in law of random variables or processes . consider the information generated by the process up to time @xmath10 , that is , the @xmath13-algebra generated by the points and weights up to time @xmath10 , prior to the selection step : @xmath14 consider also the larger @xmath13-algebra obtained when the information from the points and weights in the selection step at time @xmath10 is included : @xmath15 we assume the points and weights process obeys rules ( a1)-(a5 ) below . the rules do not uniquely define the selection and mutation procedure , but they ensure the points and weights process has certain properties , like being unbiased for the law of @xmath1 ; see theorem [ thm1 ] below . we first consider the initialization of the process . the initial points and weights have the distribution of @xmath16 in the following sense : * the initial collection of points and weights satisfies @xmath17 w.p . @xmath18 and @xmath19 = { \mathbb e}[h(x_0)].\ ] ] for every bounded measurable @xmath20 . for example , the initial condition could be an average of a fixed number of samples of @xmath16 , that is , @xmath21 is deterministic and @xmath22 , and @xmath23 for @xmath24 . notice we do not require that the @xmath25 s are independent , so they can be generated by , for example , markov chain monte carlo or other sequential samplers . next we consider the selection step . in we sampling , points are divided according to `` bins '' that partition state space . we do not require bins here , but we assume that the collection of points @xmath26 at time @xmath10 can be divided into @xmath27 different collections , and that these collections are identifiable at time @xmath10 . we make the same assumption for the selected points . * there are partitions @xmath28 of @xmath29 and @xmath30 of @xmath31 that can be determined at time @xmath10 . that is , @xmath27 and @xmath28 are @xmath32-measurable , and @xmath30 are @xmath33-measurable . note that the partitions above need not be associated with a partition of state space . all that is required is a classification of the point indices into disjoint collections . we assume that the selection is unbiased with respect to this classification : * the selection in unbiased with respect to the partition in ( a2 ) , that is , @xmath34 = \sum_{i \in i_p^r } { \omega}_p^i h_p({\xi_p^i})\ ] ] for every bounded measurable @xmath35 . we now turn to the mutation step . in this step , each selected point evolves to form a point of the next generation , with its weight remaining the same . * the number of mutated points is the same as the number of selected points : @xmath36 and moreover the weights do not change upon mutation : @xmath37 finally , we assume the mutations are independent and follow the law of @xmath1 . * the mutations are obtained by independent draws from @xmath3 : @xmath38 at time @xmath10 , the points and weights process defines an empirical distribution @xmath39 consider the first time all the points have been killed , @xmath40 we adopt the convention that @xmath41 we will now show that @xmath42 is an unbiased approximation for the law of @xmath1 . this allows us to compute unbiased estimates of functions of @xmath1 at a fixed time @xmath10 . see the discussion after algorithm [ alg1 ] below . [ thm1 ] @xmath43 is an unbiased estimator of the law of @xmath44 . that is , @xmath45 \equiv { \mathbb e}^{emp}[\eta_n(f_n ) ] = { \mathbb e}[f_n(x_n)]\ ] ] for every bounded measurable @xmath46 . we now present a we algorithm that is consistent with the assumptions in section [ sec : assumptions ] . compared to the general setting above , our algorithm below has two specific features . first , we use bins to define the partition in ( a2 ) above . and second , the resampling in each bin is designed to produce points in the next generation that all have the same weight ( see also @xcite ) . this ensures that the weights do not usually vary by large amounts in small regions of state space . [ alg1 ] pick initial points @xmath47 and weights @xmath48 as in ( a1 ) . then iterate : * _ selection step : _ * choose bins @xmath49 forming a partition of @xmath50 , and a target number of points @xmath51 for each bin @xmath52 . the @xmath52 s and @xmath51 s must be deterministic functions of the points and weights up to and including the current time @xmath10 . * for @xmath53 , do the following . set a target weight in bin @xmath52 as @xmath54 for @xmath55 such that @xmath56 , let @xmath57 be a random variable with law @xmath58 let the number of selected points equal to @xmath56 be @xmath57 : @xmath59 assign the target weight @xmath60 to all the selected points in bin @xmath52 : @xmath61 let @xmath62 be the total number of selected points : @xmath63 * _ mutation step : _ * independently evolve the @xmath64 s to obtain the points at time @xmath65 : @xmath66 set the new weights equal to the selected weights : @xmath67 set the total number of new particles as @xmath68 . then update @xmath69 and return to 1 . some remarks are in order . * given the points and weights at time @xmath10 , the procedure in step 2 uniquely defines the law of the selected points and weights , up to index relabeling . from , each point @xmath26 in bin @xmath52 is selected on average @xmath70 times , that is , @xmath71 = \frac{\omega_p^i}{{\bar \omega}_p^r},\ ] ] so the expected number of selected points in bin @xmath52 is at time @xmath10 is @xmath72= \sum_{i:\xi_p^i \in b_p^r } \frac{\omega_p^i}{{\bar \omega}_p^r } = n_p^r.\ ] ] thus , the @xmath51 can be seen as parameters controlling the average number of points per bin . as we have seen in theorem [ thm1 ] , these numbers may be chosen adaptively . indeed , a key feature of algorithm [ alg1 ] is that the bins @xmath52 and the numbers @xmath51 may be chosen adaptively . * the algorithm generates empirical approximations @xmath73 of the distribution of @xmath1 . in proposition [ prop1 ] below , we show these empirical approximations are unbiased in the sense of theorem [ thm1 ] . * recall that @xmath74 is the first time that all the points have been killed . if @xmath75 is such that @xmath76 $ ] is too large , then the variance of the estimator @xmath77 of @xmath78 $ ] can be large as well . this is because when @xmath76 $ ] is large , in many simulations the value of @xmath77 will be @xmath79 , which is uninformative . in practice , to avoid this , the values of @xmath51 can be given some lower threshold . see the remarks after algorithm [ alg2 ] below . * in practice , the points may not be resampled at every time step . we have in mind the case where @xmath1 is obtained from an underlying markov chain at certain resampling times . that is , @xmath80 where @xmath0 is an underlying markov chain and @xmath81 are some resampling times . the @xmath81 can be deterministic times for instance @xmath82 , where @xmath83 is a constant lag time or they can be certain stopping times for @xmath0 . for stationary calculations based on hill s relation , the resampling times should be chosen so that @xmath1 is a time homogeneous markov chain . see the discussion in section [ sec : dynamics ] below . as noted above , algorithm [ alg1 ] produces unbiased estimates of the law of @xmath44 : [ prop1 ] algorithm [ alg1 ] satisfies ( a1)-(a5 ) above when @xmath84 in particular , @xmath43 is an unbiased estimator of the law of @xmath44 in the sense of theorem [ thm1 ] . in proposition [ prop2 ] below we present an explicit formula for the mean squared error in the estimate @xmath85 $ ] when @xmath43 is obtained from algorithm [ alg1 ] . below proposition [ prop2 ] , we use the error formula to present an adaptive sampling strategy . this strategy uses some information from a coarse model ( for instance , a msm ) to guide the sampling . [ prop2 ] let @xmath75 be a fixed time and @xmath46 a bounded measurable function . in algorithm [ alg1 ] , the mean squared error in the approximation @xmath77 of @xmath78 $ ] satisfies @xmath86\right)^2\right ] = { \mathbb e}^{emp}[(\eta_0(f_0)-{\mathbb e}[f_n(x_n)])^2]\\ & \qquad \qquad \qquad\qquad \qquad\qquad \qquad + \sum_{p=1}^n { \mathbb e}^{emp}\left[\left ( \eta_p(f_p)-\eta_{p-1}(f_{p-1})\right)^2 \right ] \end{split}\end{aligned}\ ] ] where , writing @xmath87 $ ] and @xmath88 , we have @xmath89\\ & \qquad = \sum_{r=1}^{r_p } \sum_{i:\xi_p^i \in b_p^r}\left [ { \bar \omega}_p^r\omega_p^i k_p g_{p+1}^2(\xi_p^i)- \left({\bar \omega}_p^r\omega_p^i + \left(\omega_p^i\right)^2 -\left({\bar \omega}_p^r\right)^2\sigma_p^{r , i}\right ) g_p(\xi_p^i)^2\right ] . \end{split}\end{aligned}\ ] ] given the points and weights at time @xmath10 , we may try to minimize the contributions to the mean squared error in over choices of @xmath52 and @xmath51 . we pursue this idea , for fixed bins but an adaptive target number of points per bin . suppose for given bins @xmath52 we want to optimize over the allocation @xmath51 of points , subject to the constraint that the average total number of points is fixed at @xmath90 . for the remainder of this section , we assume that @xmath1 is time homogeneous with transition kernel @xmath91 , that the bins @xmath92 do not change in time , with @xmath93 total bins , and that @xmath94 . using the mean squared error formula in proposition [ prop2 ] , we heuristically derive an adaptive sampling strategy below . notice that @xmath57 are defined so that they have minimal variance . we estimate @xmath95 \approx { \mathbb e}^{emp}\left[\left . c_p^{r , i}\right\| { \mathcal f}_p\right]^2 = \left(\frac{\omega_p^i}{{\bar \omega}_p^r}\right)^2.\ ] ] suppose we have approximations @xmath96 of @xmath97 inside @xmath98 , that is , @xmath99 in this case , rewrites as @xmath100\\ & = \sum_{r=1}^{r_p } \sum_{i:\xi_p^i \in b_p^r}\left [ { \bar \omega}_p^r\omega_p^i k g_{p+1}^2(\xi_p^i)- \left({\bar \omega}_p^r\omega_p^i + \left(\omega_p^i\right)^2 -\left({\bar \omega}_p^r\right)^2\sigma_p^{r , i}\right ) g_p(\xi_p^i)^2\right ] \\ & \approx \sum_{r=1}^{r_p } \sum_{i:\xi_p^i \in b_p^r } { \bar \omega}_p^r\left[\omega_p^i k g_{p+1}^2(\xi_p^i)- \omega_p^i g_p(\xi_p^i)^2\right ] \\ & \approx \sum_{r=1}^{r_p } { \bar \omega}_p^r v_p^r \sum_{i:\xi_p^i \in b_p^r } \omega_p^i \\ & = \sum_{r=1}^r \frac{v_p^r \left(\sum_{i:\xi_p^i \in b_p^r } \omega_p^i\right)^2}{n_p^r}. \end{split}\end{aligned}\ ] ] minimizing the last expression with the constraint @xmath101 leads to the formula @xmath102 the approximations @xmath96 can be obtained , for instance , by estimating a coarse probability @xmath103 of @xmath1 going from bin @xmath98 to bin @xmath104 in one time step . this essentially amounts to constructing a msm @xcite for @xmath1 , with each state corresponding to a bin , and with time step @xmath18 . of course , as with any msm , there is error arising from the markov assumption , since the probability @xmath103 for @xmath1 to transition between bins @xmath98 and @xmath104 depends on the initial distribution of @xmath1 in @xmath98 . however , the msm can be seen as a coarse model for @xmath1 , and in particular , when @xmath1 has stationary distribution @xmath105 , we will use the stationary distribution of the msm as an initial guess of @xmath105 . some comments on the term @xmath106 are in order . it is not hard to see that @xmath107.\ ] ] the heuristic strategy suggests the larger the value of @xmath108 , the more beneficial it is to select @xmath109 . it is also easy to check that for naive sampling that is , sampling with no selection step the contributions to the mean squared error are @xmath110 = \sum_{i=1}^n \left[k g_{p+1}^2(\xi_p^i ) - g_p^2(\xi_p^i)\right],\ ] ] with @xmath90 the ( constant ) total number of points . we explore the heuristic strategy above in section [ sec : stationary ] below , showing how it can be combined with existing we technologies to efficiently sample the stationary distribution of @xmath1 in the time homogeneous case . in this section we present a procedure , algorithm [ alg2 ] below , for computing stationary averages in the time homogeneous case . in this section , @xmath1 is a time homogeneous markov chain with values in a measurable space @xmath111 , and transition kernel @xmath112 . we assume @xmath1 has a unique stationary distribution @xmath105 . the algorithm below is designed to efficiently compute @xmath113 for a fixed bounded measurable function @xmath114 . [ alg2 ] choose a relaxation time @xmath75 , a sampling measure @xmath109 on @xmath111 , and bins @xmath115 forming a partition of @xmath116 . choose a target number of total points @xmath117 . then iterate : _ preconditioning step : _ * let @xmath118 estimate the probability for @xmath1 to go from @xmath98 to @xmath104 in one step : @xmath119\,\xi(dx)\ ] ] let @xmath120 estimate the value of @xmath121 inside bin @xmath98 : @xmath122 * let @xmath123 be the stationary distribution for the transition matrix @xmath124 . that is , @xmath125 is the left eigenvector of @xmath126 corresponding to the eigenvalue @xmath18 , normalized so that @xmath127 . * let @xmath96 be the @xmath128th entry of the vector @xmath129 , where @xmath130 is the vector estimate of @xmath121 and the squaring operations are entrywise . _ relaxation step : _ * choose initial points so that the number of points in @xmath98 is @xmath131 : @xmath132 choose initial weights so that the total weight inside each bin @xmath98 is @xmath133 : @xmath134 * proceed through algorithm [ alg1 ] from time @xmath135 to @xmath136 , with either @xmath137 or @xmath138 defining the target number of points in bin @xmath98 at time @xmath10 . * the output of the algorithm is @xmath139 . some remarks are in order . * the purpose of the preconditioning step is to obtain a coarse approximation @xmath123 of @xmath105 , the stationary distribution of @xmath1 . this approximation is then used as the initial condition of the relaxation step . note that @xmath103 and @xmath120 can be estimated by running many short trajectories in parallel . the sampling measure @xmath109 is the initial distribution for these short trajectories . * @xmath103 is essentially the transition matrix of a msm for @xmath1 defined so that its states are the bins @xmath140 , and its time step is exactly @xmath18 . if @xmath1 is obtained from an underlying markov chain @xmath0 via @xmath141 , then this can be viewed as an msm for @xmath0 with time step @xmath83 . under suitable conditions , the vector @xmath123 is the unique stationary distribution of the msm . to understand the expression for @xmath96 , note that since @xmath142 , @xmath143 thus , @xmath96 is defined using , where the approximation is obtained by replacing @xmath112 with @xmath126 and @xmath121 with @xmath144 on the right hand side of that equation . in particular , the definition of @xmath96 is consistent with the heuristic calculations in section [ sec : sampling ] . * the purpose of the relaxation step is to allow the coarse approximation @xmath123 to relax to the stationary distribution @xmath105 . when the relaxation time @xmath75 is appropriately chosen , @xmath145 is a good approximation of @xmath113 . since the relaxation step starts at the coarse approximation @xmath123 of @xmath105 , the relaxation time @xmath75 can be much smaller than the typical time to converge from an arbitrary initial distribution . * when @xmath146 as in , we recover the accelerated weighted ensemble technique of @xcite . on the other hand , when @xmath147 satisfies the sampling is adaptive . for comparisons of sampling with and , see the simulations in section [ sec : example ] below . in the adaptive case , the choice of @xmath51 depends on the parameter @xmath75 . thus , a simulation at a given value of @xmath75 can not be easily extended to a larger value of @xmath75 with the same formula defining @xmath51 . also , in the adaptive case the choice of @xmath51 depends on @xmath121 , and so separate simulations are likely needed for computing @xmath148 $ ] for different functions @xmath121 . * the @xmath149 is introduced so that @xmath150 with high probability . when @xmath149 is too small , it can happen that all the points are killed in a large fraction of simulations . in this situation , the variance can be large . this killing can occur if , for example , @xmath51 is nonzero only in regions where there are no points . on the other hand , if @xmath149 is too large , we can lose the benefits of the adaptive sampling . here we consider an underlying markov chain @xmath0 on state space @xmath151 with transition matrix @xmath152 where @xmath153 and @xmath154 is chosen so that @xmath155 is stochastic . we take resampling intervals @xmath156 , so @xmath157 and the transition matrix of @xmath1 is @xmath158 . the bins will be @xmath159 thus , there are @xmath160 fixed bins . let @xmath105 be the stationary distribution of @xmath1 , and @xmath161 we also let @xmath162 be its normalized version . we use algorithm [ alg2 ] to obtain empirical approximations @xmath145 of @xmath113 for a fixed relaxation time @xmath75 . throughout , we use a sampling measure @xmath109 which is uniform : @xmath163 for all @xmath55 . in the relaxation step of algorithm [ alg2 ] , the initial empirical distribution of points and weights can be considered a sample from @xmath164 we will use three types of sampling . the first type of sampling uses the adaptive allocation of points from equation above . we call this adaptive we sampling . in , we take @xmath165 and @xmath166 . we found this @xmath149 to be sufficiently large that @xmath150 in all our simulations . in the second type of sampling we used a fixed target number of points per bin . we call this traditional we sampling . it is the sampling method described in @xcite . here we use equation , taking @xmath167 . again , we had @xmath150 in all our simulations . notice for both adaptive and traditional we sampling , we have chosen @xmath90 so that the target number of points is @xmath168 . [ fig1]-190pt vs. @xmath75 from the example in section [ sec : example ] from adaptive , traditional and naive sampling . the crosses are exact values corresponding to @xmath169 , and the dotted line is the stationary value @xmath113 . bottom right : empirical standard deviations @xmath170 of @xmath145 , obtained from @xmath171 independent simulations . ( these are the error bar radii from the other plots . ) ] -170pt [ fig2]-195pt at time @xmath172 , compared to @xmath173 , the normalized version of @xmath121 . here , @xmath174 is the total number of points at time @xmath75 . bottom right : stationary distribution @xmath105 . ] -170pt the last type of sampling does not use resampling at all . we call this naive sampling . here , we skip steps ( ii ) and ( iii ) of the relaxation step of algorithm [ alg2 ] . instead , after the points @xmath175 and weights @xmath176 are chosen , we simply evolve these points independently until time @xmath75 , without changing the weights . for this sampling we take @xmath165 total points , to facilitate comparison with the traditional and adaptive sampling described above . results comparing the different relaxation techniques are in figures [ fig1]- [ fig2 ] . in figure [ fig1 ] , we plot @xmath145 vs. @xmath75 for various values of @xmath75 , showing convergence to the stationary value @xmath113 . we compute error bars as empirical standard deviations from @xmath171 independent simulations . we found that the error bars for adaptive we sampling were significantly smaller than that of traditional we and naive sampling . in figure [ fig2 ] , we plot histograms representing the average distribution of the points @xmath177 at time @xmath75 . note that traditional we sampling distributes the points roughly uniformly in space , as expected , while adaptive we sampling guides the points towards the region in state space relevant for computing @xmath121 . meanwhile , naive sampling distributes the points according to the stationary distribution @xmath105 . in figure [ fig3 ] , we plot the estimates @xmath96 from the adaptive sampling strategy for @xmath178 and @xmath179 where @xmath180 is the relaxation time . note that by time @xmath181 , the sampling is focused near the support of @xmath121 . when @xmath121 is a function with large values in regions of low @xmath105 probability , as in this example , naive sampling performs poorly compared to both traditional and adaptive we sampling . when state space is very large compared to the region @xmath182 where @xmath121 has large values ( or is non - negligible ) , we expect adaptive we sampling to perform much better than traditional we sampling , due to the fact that traditional we sampling will distribute the points very thinly throughout space , including in @xmath182 , while adaptive we sampling will push most of the points towards @xmath182 . a possible drawback of adaptive we sampling is that it requires more computations at the resampling times , compared to traditional we sampling . however , in practice the resampling times may be large enough so that this extra effort contributes little to the overall computational cost . finally , we note that the adaptive sampling above can also be used more generally to estimate time marginals of @xmath44 , that is , expectations of the form @xmath148 $ ] at fixed finite times @xmath75 , from an arbitrary initial distribution of @xmath16 . in this case , a msm is still required to guide the sampling . one of the advantages of the adaptive sampling in the stationary case is that an msm has already been computed as part of a preconditioning step . [ fig3]-210pt from vs. @xmath10 for the example in section [ sec : example ] . here , we take @xmath135 and @xmath183 , when @xmath180 is the final time . ] in this section we show how to compute certain dynamical averages of time homogeneous markov chains from stationary calculations on a related markov chain . the hill relation @xcite shows that a mean first passage time can be reformulated as a certain stationary average . we will generalize this relation below . assume that @xmath1 is obtained from an underlying markov chain @xmath0 at some resampling times @xmath81 , that is , @xmath80 . we assume @xmath1 and @xmath0 have values in a measurable space @xmath111 . @xmath0 can be continuous or discrete in time . for simplicity , and since in simulations time discretization is required anyway , we assume here it is discrete in time . to obtain @xmath1 from @xmath0 , we need to define the resampling times @xmath81 . we will assume they satisfy the following : * @xmath184 are stopping times for @xmath0 such that for all @xmath185 , @xmath186 informally , this means @xmath0 and the stopping time increments @xmath187 regenerate at each stopping time @xmath81 . for example , we can take @xmath81 to be @xmath10 multiplied by some deterministic lag time . alternatively , the @xmath81 can be a sequence of hitting times of coarse sets in state space , as in exact milestoning ; see the comments below . to be able to compute dynamical averages from stationary ones , we must introduce a source and sink . let @xmath188 and @xmath189 be disjoint initial and final sets , such that when @xmath0 arrives in @xmath189 at one of the resampling times , it immediately returns to @xmath188 . when it returns to @xmath188 , it starts at a fixed source distribution @xmath190 on @xmath188 : * there are disjoint measurable sets @xmath191 and a distribution @xmath190 on @xmath188 so that @xmath192 = \int { \mathbb p}^x[y_1 \in a]\rho(dx).\ ] ] we are interested in computing , for example , the mean first passage time of @xmath0 from @xmath188 to @xmath189 . we want to compute this from a stationary average , using the source - sink . the following technical assumption guarantees this computation is well defined : * with @xmath193 we have @xmath194 < \infty , \qquad \text{for all } x \in e.\ ] ] we assume ( b1)-(b3 ) hold throughout the remainder of this section . from ( b1 ) , it is easy to see @xmath195 is a stopping time for @xmath0 . since @xmath80 , by the strong markov property , ( b1 ) guarantees @xmath1 is a time homogeneous markov chain . from ( b2)-(b3 ) , @xmath1 has a unique stationary distribution @xmath105 . the latter can be proved by standard coupling arguments , based on @xmath1 reaching @xmath189 in finite expected time and then regenerating at @xmath190 . we omit proof . for the resampling times @xmath81 , we could take @xmath82 where @xmath83 is a fixed resampling time interval , as in standard we sampling . more generally , we could take @xmath196 and @xmath197 where @xmath198 is a stopping time for @xmath0 starting at @xmath199 , as in exact milestoning . in that setting , @xmath198 is the first time for @xmath0 to reach a milestone starting at @xmath199 ; see @xcite . we are now ready to generalize the hill relation @xcite . [ thm_dynamic ] with @xmath105 the unique stationary distribution of @xmath1 , @xmath200 = \frac{\int_f g(y)\pi(dy)}{\pi(f)}\ ] ] and @xmath201 = \frac{{\mathbb e}^\pi\left[\sum_{t=1}^{\tau_1 } g(y_t)\right ] } { \pi(f)}.\ ] ] in , if @xmath202 is the disjoint union of @xmath203 and @xmath204 and @xmath205 , we have @xmath206 = \frac{\pi(b)}{\pi(a\cup b)}.\ ] ] this is the probability to hit @xmath204 before @xmath203 , starting at @xmath190 . in , if @xmath207 , @xmath208 = \frac{{\mathbb e}^\pi[\tau_1]}{\pi(f)}.\ ] ] this is the mfpt from @xmath190 to @xmath189 . note that when @xmath209 is constant , this equation is simply the hill relation @xcite . when @xmath210 is random , this is the expression recently proved for exact milestoning @xcite . it is important to note that assumption ( b2 ) does not affect the averages on the left hand side of and . thus , @xmath0 can be any time homogeneous markov chain , and in particular it need not have a source and sink . the source and sink in ( b2 ) are only needed to define the stationary distribution @xmath105 used in the sampling . since here @xmath105 is the stationary distribution of a time homogeneous markov chain , algorithm [ alg2 ] can be used to compute the right hand side of . see the example in section [ sec : stationary ] above . although it is not obvious how to extend that algorithm to compute the right hand side of , in some cases a simple modification may suffice . for instance , in exact milestoning , one can apply algorithm [ alg2 ] to the function @xmath211 , where @xmath212 is the union of the milestones . the empirical distributions @xmath43 from the algorithm can be used as the initial distribution for computing the numerator of . we begin with the following preliminary result . [ lem1 ] suppose @xmath213 are iid random variables and @xmath90 is a nonnegative integer valued random variable independent of the @xmath214 s . then @xmath215 = { \mathbb e}[n]{\mathbb e}[\xi_1]\ ] ] and @xmath216 = \left({\mathbb e}[n^2]-{\mathbb e}[n]\right){\mathbb e}[\xi_1]^2 + { \mathbb e}[n]{\mathbb e}[\xi_1 ^ 2].\ ] ] we omit the proof , which is straightforward . we will also use the following result . [ lem2 ] let @xmath217 where @xmath218 . then we have @xmath219 = g_p({\hat \xi}_p^i)\ ] ] and @xmath220 = \begin{cases } g_p({\hat \xi}_p^i ) g_p({\hat \xi}_p^j ) , & i\ne j\\ k_p g_{p+1}^2({\hat \xi}_p^i ) , & i = j\end{cases}.\ ] ] first observe that , by ( a5 ) , @xmath221 \\ & = { \mathbb e}^{emp}\left[\left . \int k_{p+1}(\xi_{p+1}^i , dy_{p+2})\ldots k_{n-1}(y_{n-1},dy_{n})f(y_{n})\right\|\hat{{\mathcal f}_p}\right ] \\ & = \int k_p({\hat \xi}_p^i , dy ) k_{p+1}(y , dy_{p+2})\ldots k_{n-1}(y_{n-1},dy_{n})f(y_{n})= g_p({\hat \xi}_p^i).\end{split}\end{aligned}\ ] ] when @xmath222 , in the second equation , using ( a5 ) we have @xmath223 = \int k_p({\hat \xi}_p^i , dy ) g_{p+1}(y)^2 = k_p g_{p+1}^2({\hat \xi}_p^i).\ ] ] on the other hand , when @xmath224 , conditionally on @xmath225 , @xmath226 and @xmath227 are independent . thus , appealing to completes the proof . we now turn to the proof of theorem [ thm1 ] . let @xmath228 be as in lemma [ lem2 ] and recall that @xmath229 to emphasize that @xmath230 when @xmath231 , we write @xmath232 let @xmath233 note that @xmath234 is @xmath235-measurable and @xmath236 using also @xmath237 we find that @xmath238 & = { \mathbb e}^{emp}\left[\left.\mathbbm{1}_{\tau_{kill } > p+1}\sum_{i=1}^{n_{p+1 } } \omega_{p+1}^i g_{p+1}(\xi_{p+1}^i)\right\|{\mathcal f}_p\right ] \\ & = \mathbbm{1}_{\tau_{kill } > p}{\mathbb e}^{emp}\left[\left.{\mathbb e}^{emp}\left[\left.\sum_{i=1}^{n_{p+1 } } \omega_{p+1}^i g_{p+1}(\xi_{p+1}^i)\right\|{\hat{\mathcal f}}_p\right]\right\|{\mathcal f}_p\right ] \\ & = \mathbbm{1}_{\tau_{kill } > p } { \mathbb e}^{emp}\left[\left.\sum_{i=1}^{{\hat n}_{p } } { \hat \omega}_{p}^i { \mathbb e}^{emp}\left[\left . g_{p+1}(\xi_{p+1}^i)\right\|{\hat{\mathcal f}}_p\right]\right\|{\mathcal f}_p\right ] \qquad \textup { by ( a4 ) } \\ & = \mathbbm{1}_{\tau_{kill } > p}{\mathbb e}^{emp}\left[\left.\sum_{i=1}^{{\hat n}_p } { \hat \omega}_p^i g_p({\hat \xi}_p^i)\right\|{\mathcal f}_p\right ] \qquad \textup { by ( a5 ) and lemma~8}\\ & = \mathbbm{1}_{\tau_{kill } > p}\sum_{r=1}^{r_p}{\mathbb e}^{emp}\left[\left.\sum_{i\in { \hat i}_p^r } { \hat \omega}_p^i g_p({\hat \xi}_p^i)\right\|{\mathcal f}_p\right ] \qquad \text{by ( a2 ) } \\ & = \mathbbm{1}_{\tau_{kill } > p}\sum_{r=1}^{r_p}\sum_{i\in { i}_p^r } { \omega}_p^i g_p({\xi}_p^i ) \qquad \text{by ( a3)}\\ & = \mathbbm{1}_{\tau_{kill } > p}\sum_{i=1}^{n_p } { \omega}_p^i g_p({\xi}_p^i ) = m_p \qquad \text{by ( a2)}.\end{aligned}\ ] ] it follows that @xmath239 is a martingale , so @xmath240 = { \mathbb e}^{emp}[m_n]$ ] and @xmath241 & = { \mathbb e}\left[g_0(x_0)\right ] \qquad \\ & = { \mathbb e}^{emp}[m_0 ] \qquad \textup{by ( a1)}\\ & = { \mathbb e}^{emp}[m_n ] = { \mathbb e}^{emp}[\eta_n(f_n ) { \mathbbm{1}}_{\tau_{kill}>n}].\end{aligned}\ ] ] clearly , algorithm [ alg1 ] satisfies ( a1)-(a2 ) and ( a4)-(a5 ) . conditional on @xmath32 , for @xmath242 such that @xmath243 , the @xmath244 s are independent with expected value @xmath245 . conditional on @xmath32 , the expected number of such @xmath242 s is @xmath246 = { \mathbb e}^{emp}\left[\left . c_p^{r , i}\right\|{\mathcal f}_p\right ] = \frac{\omega_p^i}{{\bar \omega}_p^r}.\ ] ] thus , for any bounded measurable @xmath247 , @xmath248 & = { \bar \omega}_p^r{\mathbb e}^{emp}\left[\left.\sum_{i:{\hat \xi}_p^i \in b_p^r } h_p({{\hat \xi}_p^i } ) \right\| { \mathcal f}_p\right ] \\ & = { \bar \omega}_p^r \sum_{i : \xi_p^i \in b_p^r}{\mathbb e}^{emp}\left[\left . \sum_{j:{\hat \xi}_p^j = \xi_p^i } h_p({{\hat \xi}_p^j } ) \right\| { \mathcal f}_p\right ] \\ & = { \bar \omega}_p^r \sum_{i : \xi_p^i \in b_p^r } \frac{\omega_p^i}{{\bar \omega}_p^r}h_p ( { \xi}_p^i ) \qquad \text{by lemma } 7 \\ & = \sum_{i \in i_p^r } { \omega}_p^i h_p({\xi_p^i}).\end{aligned}\ ] ] define @xmath249 and @xmath250 as in the proof of theorem [ thm1 ] . there it was shown @xmath239 is a martingale , which leads to equation [ l2error ] . we turn to the expression in . using @xmath251 we find that @xmath252 \\ & = { \mathbb e}^{emp}\left[\left.\mathbbm{1}_{\tau_{kill}>p+1}\left(\sum_{i=1}^{n_{p+1 } } \omega_{p+1}^i g_{p+1}(\xi_{p+1}^i)\right)^2\right\|{\mathcal f}_{p}\right ] \\ & = \mathbbm{1}_{\tau_{kill}>p}{\mathbb e}^{emp}\left[\left.\sum_{i=1}^{n_{p+1 } } \sum_{j=1}^{n_{p+1 } } \omega_{p+1}^i \omega_{p+1}^j g_{p+1}(\xi_{p+1}^i ) g_{p+1}(\xi_{p+1}^j)\right\|{\mathcal f}_{p}\right ] \\ & = \mathbbm{1}_{\tau_{kill}>p}\sum_{r=1}^{r_{p } } \sum_{s=1}^{r_{p } } { \bar \omega}_{p}^r { \bar \omega}_{p}^s { \mathbb e}^{emp}\left[\left.\sum_{i:{\hat \xi}_{p}^i \in b_p^r } \ , \sum_{j:{\hat \xi}_{p}^j \in b_p^s } g_{p+1}(\xi_{p+1}^i ) g_{p+1}(\xi_{p+1}^j)\right\|{{\mathcal f}}_{p}\right]\\ & = \mathbbm{1}_{\tau_{kill}>p}\sum_{r=1}^{r_{p } } \sum_{s=1}^{r_{p}}{\bar \omega}_{p}^r { \bar \omega}_{p}^s \sum_{i:{\xi}_{p}^i \in b_p^r } \ , \sum_{j:{\xi}_{p}^j \in b_p^s } { \mathbb e}^{emp}\left[\left . \sum_{k : { \hat \xi}_{p}^k = \xi_{p}^i } \sum_{\ell:{\hat \xi}_{p}^\ell = \xi_{p}^j } g_{p+1}(\xi_{p+1}^k ) g_{p+1}(\xi_{p+1}^\ell)\right\|{{\mathcal f}}_{p}\right].\end{aligned}\ ] ] from lemma [ lem2 ] , @xmath253 = \begin{dcases } g_p({\hat \xi}_p^k)g_p({\hat \xi}_p^\ell ) , & k\ne \ell \\ k_p g_{p+1}^2({\hat \xi}_p^k ) , & k = \ell \end{dcases}\end{aligned}\ ] ] and recall that @xmath254= { \mathbb e}^{emp}\left[\left.\#\left\{k:{\hat \xi}_p^k = \xi_p^i \in b_p^r\right\}\right\|{\mathcal f}_p\right ] = \frac{\omega_p^i}{{\bar \omega}_p^r}.\ ] ] thus , for @xmath255 such that @xmath256 and @xmath257 , by lemma [ lem1 ] , @xmath258 = \frac{\omega_{p}^i}{{\bar \omega}_{p}^r } \frac{\omega_{p}^j}{{\bar \omega}_{p}^s } g_p(\xi_p^i)g_p(\xi_p^j).\end{aligned}\ ] ] on the other hand , when @xmath222 and @xmath259 , this becomes @xmath260 \\ & \qquad \qquad = \left(\sigma_p^{r , i}-\frac{\omega_p^i}{{\bar \omega}_p^r}\right ) g_p(\xi_p^i)^2 + \frac{\omega_p^i}{{\bar \omega}_p^r } k_p g_{p+1}^2(\xi_p^i).\end{aligned}\ ] ] where again we used lemma [ lem1 ] , and we recall that @xmath261.\ ] ] combining the above , @xmath252 = \mathbbm{1}_{\tau_{kill}>p}\sum_{r=1}^{r_p } \sum_{s=1}^{r_p}\sum_{\substack{i:\xi_p^i \in b_p^r\\ j:\xi_p^j \in b_p^s\\ i \ne j } } \omega_p^i \omega_p^j g_p(\xi_p^i)g_p(\xi_p^j ) \\ & \qquad \qquad+\mathbbm{1}_{\tau_{kill}>p}\sum_{r=1}^{r_p } \sum_{i:\xi_p^i \in b_p^r}\left [ \left(\left({\bar \omega}_p^r\right)^2\sigma_p^{r , i}-{\bar \omega}_p^r\omega_p^i\right ) g_p(\xi_p^i)^2 + { \bar \omega}_p^r\omega_p^i k_p g_{p+1}^2(\xi_p^i)\right],\end{aligned}\ ] ] and so , since @xmath262 we get @xmath263 \\ & = { \mathbb e}^{emp}\left[\left . m_{p+1}^2 \right\|{\mathcal f}_p\right]-m_p^2 \\ & = \mathbbm{1}_{\tau_{kill}>p}\sum_{r=1}^{r_p } \sum_{i:\xi_p^i \in b_p^r}\left [ { \bar \omega}_p^r\omega_p^i k_p g_{p+1}^2(\xi_p^i)- \left({\bar \omega}_p^r\omega_p^i + \left(\omega_p^i\right)^2 -\left({\bar \omega}_p^r\right)^2\sigma_p^{r , i}\right ) g_p(\xi_p^i)^2\right].\end{aligned}\ ] ] we begin with the proof of . recall @xmath264 is the transition kernel of @xmath265 . conditioning on @xmath266 and using assumption ( b1 ) , the lhs of rewrites @xmath267 & = \int { \mathbb e}^x\left[\left.g(y_{\tau_f})\right\|y_{\tau_1 } = y\right]k(x , dy ) \\ & = \int_{e \setminus f } { \mathbb e}^x\left[\left.g(y_{\tau_f})\right\|y_{\tau_1 } = y\right]k(x , dy ) + \int_f g(y)k(x , dy)\\ & = \int_{e \setminus f } { \mathbb e}^y[g(y_{\tau_f})]k(x , dy ) + \int_f g(y)k(x , dy).\end{aligned}\ ] ] integrating against @xmath105 and using stationarity , @xmath268 = \int_{e \setminus f } { \mathbb e}^y[g(y_{\tau_f})]\pi(dy ) + \int_f g(y)\pi(dy).\ ] ] cancelling terms and using the fact that @xmath269 , @xmath270\pi(dy ) = \pi(f ) { \mathbb e}^\rho[g(y_{\tau_f})].\ ] ] now we turn to the proof of . conditioning on @xmath266 and using assumption ( b1 ) , @xmath271 & = \int { \mathbb e}^x\left[\left.\sum_{t=1}^{\tau_f } g(y_t)\right\|y_{\tau_1 } = y\right]k(x , dy ) \\ & = \int { \mathbb e}^x\left[\left.\sum_{t=1}^{\tau_1 } g(y_t)\right\|y_{\tau_1 } = y\right]k(x , dy ) \\ & \qquad\qquad+\int_{e \setminus f}{\mathbb e}^x\left[\left.\sum_{t=\tau_1 + 1}^{\tau_f } g(y_t)\right\|y_{\tau_1 } = y\right]k(x , dy)\\ & = { \mathbb e}^x\left[\sum_{t=1}^{\tau_1 } g(y_t)\right ] + \int_{e \setminus f}{\mathbb e}^y\left[\sum_{t=1}^{\tau_f } g(y_t)\right]k(x , dy).\end{aligned}\ ] ] integrating against @xmath105 and using stationarity , @xmath272 = { \mathbb e}^\pi\left[\sum_{t=1}^{\tau_1 } g(y_t)\right ] + \int_{e \setminus f}{\mathbb e}^y\left[\sum_{t=1}^{\tau_f } g(y_t)\right ] \pi(dy).\ ] ] cancelling terms and using the fact that @xmath273 , @xmath274 = \int_f { \mathbb e}^y\left[\sum_{t=1}^{\tau_f}g(y_t)\right ] \pi(dy ) = \pi(f){\mathbb e}^\rho\left[\sum_{t=1}^{\tau_f}g(y_t)\right].\ ] ] d. aristoff would like to acknowledge enlightening conversations with tony lelivre , petr plech , mathias rousset , gideon simpson , ting wang , and dan zuckerman . d. aristoff also gratefully acknowledges support from the national science foundation via the award nsf - dms-1522398 .
the hot gas in clusters of galaxies seems like it should be easy to understand . because of the relatively low ratio of baryons to dark matter , the potential well of a cluster should be dark - matter dominated . the dynamical time within a cluster potential is shorter than a hubble time , so most clusters should be relaxed . also , the cooling time of the vast majority of intracluster gas is longer than a hubble time . it would appear that cluster structure ought to be scale - free , as long as the shape of a cluster s potential well does not depend systematically on its mass . if that were the case , then the global properties of clusters , such as halo mass , luminosity - weighted temperature , and x - ray luminosity , would scale self - similarly ( kaiser 1986 ) . in particular , the gas temperature would scale with cluster mass as @xmath0 and the bolometric x - ray luminosity would scale with temperature as @xmath1 in the bremsstrahlung - dominated regime above @xmath2 . indeed , numerical simulations that include gas dynamics but exclude non - gravitational processes such as radiative cooling and supernova heating produce clusters that obey these scaling laws ( e.g. , evrard , metzler , & navarro 1996 ; bryan & norman 1998 ; thomas 2001b ) . real clusters are not so simple . we have known for a decade that the observed luminosity - temperature relation is closer to @xmath3 ( e.g. , edge & stewart 1991 ) , indicating that non - gravitational processes must influence the density structure of a cluster s core , where most of the luminosity is generated ( kaiser 1991 ; evrard & henry 1991 ; navarro , frenk & white 1995 ; bryan & norman 1998 ) . the mass - temperature relation , on the other hand , seemed like it ought to be more fundamental and less sensitive to non - gravitational effects . yet , observations collected over the last few years indicate that this relation also disagrees with both the scale - free predictions and simulations that exclude non - gravitational processes ( horner , mushotzky , & scharf 1999 ; nevalainen , markevitch , & forman 2000 ; finoguenov , reiprich , & bhringer 2001 ; xu , jin , & wu 2001 ) . these results derive mostly from resolved x - ray and temperature profiles coupled with the assumption of hydrostatic equilibrium , but they do seem consistent with gravitational lensing measurements ( allen , schmidt , & fabian 2001 ) . understanding the scaling properties of clusters is of broad importance because these scaling laws are integral to determination of cosmological parameters . for example , the number density of clusters in the local universe reflects the amplitude of matter perturbations on @xmath4 mpc scales ( e.g. , henry & arnaud 1991 ; white , efstathiou , & frenk 1993 ; viana & liddle 1996 ; eke , cole & frenk 1996 ; kitayama & suto 1997 ; oukbir & blanchard 1997 ; henry 2000 ) . in order to measure this amplitude , one needs accurately measured cluster masses . gravitational lensing measurements of cluster mass have grown rapidly in sophistication during the last few years , but x - ray determined cluster temperatures , converted to cluster masses using the mass - temperature relation , remain the most commonly used mass measurement for cosmological studies . thus , any inaccuracies in the mass - temperature relation propagate into uncertainties in cosmological parameters derived from clusters ( e.g. , voit 2000 ) . a number of authors have recently re - evaluated the perturbation amplitude implied by cluster observations using a variety of techniques that circumvent the standard mass - temperature relation ( reiprich & bhringer 2001 ; van waerbeke 2001 ; seljak 2002 ; viana , nichol , & liddle 2002 ) . the results are disconcertingly different from those based on the standard relation , generally implying a much lower power - spectrum amplitude . clearly a better understanding of clusters is needed , and ideally , we would like to know how the present - day properties of clusters are linked to the physics of galaxy formation . the physical processes most likely to break the expected self - similarity of clusters are heating by supernovae or active galactic nuclei and radiative cooling , although other possibilities such as magnetic pressure or cosmic - ray pressure have not been ruled out . a large number of studies have investigated how the structure of the intracluster medium changes when these processes are included . because treating all of these processes in sufficient detail is extremely complicated , various approximations have been employed . these include numerical simulations that administer a particular amount of energy input at early times or adopt some other phenomenological prescription for injecting energy ( e.g. , metzler & evrard 1994 ; navarro , frenk , & white 1995 ; metzler & evrard 1997 ; bialek , evrard , & mohr 2001 ; borgani et al . 2001a ; da silva 2001 ; bryan & voit 2001 ) , analytical models for spherically symmetric accretion of preheated gas ( e.g. , cavaliere , menci & tozzi 1999 ; tozzi & norman 2001 ) , and semi - analytic models that consider the merger and star formation history of clusters ( e.g. , wu , fabian , & nulsen 1998 , 2000 ; balogh , babul , & patton 1999 ; bower 2001 ; babul , balogh , lewis , & poole 2002 ) . most of these papers have focused on heating of the gas , but recent work has also examined the possibility that radiative cooling is primarily responsible ( e.g. , knight & ponman 1997 ; suginohara & ostriker 1998 ; pearce 2000 ; bryan 2000 ; muanwong 2001 ; thomas 2001a ; lewis 2001 ) . two broad conclusions emerge from such studies : ( 1 ) radiative cooling without feedback probably locks too high a fraction of baryons into condensed objects , and ( 2 ) if heating is uniformly distributed , the amount required to produce realistic clusters is quite large , probably exceeding @xmath5 per baryon . introducing this much energy through supernova heating strains the bounds of plausibility , given how many metals are observed in the intracluster medium ( e.g. , balogh , babul , & patton 1999 ; valageas & silk 1999 ; kravtsov & yepes 2000 ; bower 2000 ; although see loewenstein 2000 ) . here , we revisit an approach pioneered by bower ( 1997 ) , focusing on entropy as the most fundamental characteristic of the intracluster medium . analyzing clusters in terms of entropy has two advantages . the first is that convection within clusters acts as an entropy - sorting device , shuttling low - entropy material to the cluster core and high - entropy material to the cluster s outskirts . convective stability is achieved when specific entropy becomes a monotonically increasing function of radius . thus , if we can determine the entropy distribution of the gas in a cluster and we know the structure of its dark - matter halo , we can predict the structure of that cluster in its relaxed state . in the idealized cases that we will consider , we will assume that clusters are spherically symmetric , dark - matter dominated , in hydrostatic equilibrium , and in convective equilibrium . the second advantageous property of entropy is that , in the absence of non - gravitational processes , the entropy distribution of a cluster can be readily computed using either hydrodynamical simulations or spherically symmetric analytical models . then , the effects of non - gravitational heating and radiative cooling can be implemented as modifications of that entropy distribution . in voit & bryan ( 2001 ) , we introduced a simple scheme for computing cluster models , taking advantage of these properties of the intracluster entropy distribution to show how heating and cooling conspire to produce a cluster @xmath6-@xmath7 relation quite similar to what is observed . this paper explores those modified - entropy models in more detail , showing that the observable properties of these model clusters are quite similar to those of real clusters on many counts . section 2 describes how we construct those models and analyzes how cluster properties depend on both halo concentration and entropy modification . section 3 uses relations between halo mass and halo concentration to define sets of cluster models determined entirely by the underlying cosmology . we then compare the model clusters with observations of the surface - brightness profiles of clusters , the temperature gradients of clusters , the mass - temperature relation , and the luminosity - temperature relation . in every case , our models reproduce the observations , with little need for adjustable parameters . in section 4 , we explore how the present - day intracluster entropy distribution emerges from hierarchical cosmological models . section 5 summarizes our conclusions . in this section we develop a family of modified - entropy models of clusters based on two parameters , one specifying the concentration of the cluster s dark - matter halo and the other specifying the entropy level at which cooling and feedback modify the intracluster entropy distribution . we begin by casting the equilibrium equations in dimensionless form and selecting appropriate boundary conditions . then we assume that the unmodified entropy distribution determined by gravitational processes alone would produce a gas density distribution similar to the dark - matter density distribution . because the innermost gas in such a cluster can cool in much less than a hubble time , we proceed to investigate several potential modifications of the entropy distribution by radiative cooling and subsequent feedback . in each case , we show how both entropy modification and the shape of the underlying dark - matter halo affect the surface - brightness profile of a cluster , its luminosity - weighted temperature , its temperature gradient , and its x - ray luminosity . two important results emerge from this analysis : ( 1 ) the temperature of a cluster of a given mass is determined primarily by halo concentration and is affected only modestly by entropy modification , and ( 2 ) altering the intracluster entropy distribution at the scale set by radiative cooling leads to an @xmath8-@xmath7 relation that has the observed slope at @xmath9 and that evolves very little with time . the equilibrium structure of the intracluster medium is determined by three things , the gravitational potential of the cluster s dark matter , the entropy distribution of the cluster s gas , and the confining pressure at the outer boundary of the cluster . in order to solve for that structure , one must integrate the equations of hydrostatic equilibrium and gas mass conservation , @xmath10 using the equation of state @xmath11 , where @xmath12 is the adiabatic index and @xmath13 specifies the adiabat of the gas . in this system of equations , @xmath14 is the total mass within radius @xmath15 , @xmath16 is the gas mass within that radius , and the other symbols have their usual meanings . throughout this paper , we will assume that the gas mass is gravitationally negligible and that @xmath17 , as appropriate for an ideal monatomic gas . in a sufficiently relaxed cluster , convection ensures that @xmath18 monotonically increases with radius . thus , if one knows @xmath19 , the entropy distribution of the intracluster medium expressed in terms of the mass of gas with @xmath20 , then one can use the inverse relation @xmath21 to solve the equilibrium equations that determine the cluster s structure . because this crucial quantity @xmath13 is so closely related to specific entropy , we will often refer to it as the `` entropy '' of the gas , even though the standard thermodynamic entropy per particle for an ideal monatomic gas is @xmath22 we can gain insight into the key parameters that control the structure of the intracluster medium by investigating a family of dimensionless models for clusters in hydrostatic and convective equilibrium . because the virial radius of a cluster lies close to the radius @xmath23 within which the mean matter density of the cluster is 200 times the critical density @xmath24 , we elect to represent radii in terms of @xmath25 and gas density in terms of @xmath26 , where @xmath27 and @xmath28 is the fractional contribution of baryons to the total mass of the universe . one can then rewrite the equilibrium equations in dimensionless form as follows : @xmath29 where @xmath30 is the virial mass , @xmath31 is the temperature in energy units is absorbed into @xmath7 throughout the paper . ] of the corresponding singular isothermal sphere , @xmath32 $ ] is the dimensionless pressure , @xmath33 is the dimensionless mass within @xmath34 , and @xmath35 is the fraction of a cluster s baryons in the intracluster medium within radius @xmath34 . the corresponding dimensionless entropy distribution is @xmath36 $ ] , and we will also make use of the quantity @xmath37 . general solutions to these equations can be found for a few idealized cases . for example , if the potential is a singular isothermal sphere and the gas is also isothermal , then we have @xmath38 , @xmath39 , and @xmath40 for a monatomic ideal gas . if instead the mass profile follows the form of navarro , frenk & white ( 1997 ; nfw hereafter ) , with @xmath41 $ ] where @xmath42 is the concentration parameter , then the equation of hydrostatic equilibrium becomes ( see wu , fabian , & nulsen 2000 ) @xmath43^{-1 } \frac { d } { d\hat{r } } \left [ \frac { \ln ( 1+c\hat{r } ) } { \hat{r } } \right ] \ ; \ ; .\ ] ] given an isentropic gas in which @xmath44 we thus obtain @xmath45^{3/2}$ ] , where @xmath46 is a constant of integration . a particular solution to the equilibrium equations can be found by choosing @xmath47 and @xmath48 at the origin and integrating outwards . the gas mass fraction at the origin is always zero by definition , but there is some freedom in the choice of @xmath49 . any given choice of @xmath50 corresponds to a unique icm structure with a unique luminosity and temperature profile , but which solutions are the physical ones ? here is where the confining pressure comes into play . it is generally assumed that accreting matter confines the icm in the neighborhood of @xmath23 . the pressure at @xmath23 would then be determined by the ram pressure of the infalling matter . now suppose that @xmath51 has been modified by non - gravitational processes . for example , heating might drive gas out of the cluster potential , pushing the accretion shock to a radius beyond @xmath23 . alternatively , cooling and condensation might reduce the total gas mass of the intracluster medium so that the accretion shock moves inward . in either case the radius at which the integration of hydrostatic equilibrium should terminate differs from @xmath23 . so what is the termination radius @xmath52 at which @xmath53 equals the total uncondensed gas mass , and what is the pressure at that radius ? in a freely falling accretion flow with a constant mass flux , the density should vary as @xmath54 , where @xmath55 is the free - fall velocity from @xmath56 in an nfw potential . we therefore assume that the accretion pressure scales with radius as @xmath57 , and we normalize that pressure to equal the unmodified nfw value ( see [ unmod ] ) at @xmath58 . various choices of @xmath59 will lead to pressure profiles that intersect @xmath60 at different values of @xmath61 , each corresponding to a unique value of @xmath62 determined by a particular integration . the correct solution is the one that gives the proper value for the uncondensed gas mass at the radius where the gas pressure equals the accretion pressure . in practice , the physically reasonable solutions are not particularly sensitive to the outer boundary condition . in the models that follow , applying the boundary condition @xmath63 at @xmath58 would lead to similar results . however , the value of @xmath48 derived at @xmath64 using this boundary condition is not physically significant and can even be inconsistent with models in which cooling and condensation significantly reduce the maximum value of @xmath48 . one example of a physically _ unreasonable _ boundary condition is requiring @xmath65 at @xmath58 . such a requirement does not allow intracluster gas to expand beyond @xmath23 even when heating substantially raises the intracluster entropy . in the limit of extreme heating , this condition therefore artificially boosts the central pressure , density , and temperature , leading to an unphysically large cluster luminosity . once a correct solution has been identified , one can compute two quantities that correspond to the dimensionless luminosity and luminosity - weighted temperature : @xmath66 as long as the cooling function @xmath67 remains sufficiently constant within the gas contributing the bulk of the luminosity , the actual luminosity - weighted temperature is @xmath68 , and the bolometric x - ray luminosity is @xmath69 where @xmath70 . scaling to a 10 kev cluster at @xmath9 gives @xmath71 where @xmath72 . the emission - measure profile and emissivity - weighted temperature profile are also of interest . thus , we define the following dimensionless analogs as functions of the projected radius @xmath73 : @xmath74 where @xmath75 is the dimensionless distance along the line of sight . before exploring the consequences of modifying the entropy of intracluster gas , we need to know what the entropy distribution would be if it were not modified . to simplify matters , we will assume that the underlying dark matter profile of the cluster is of nfw form with some concentration parameter @xmath42 . this parameter typically ranges from @xmath76 for hot clusters to @xmath77 for groups of galaxies ( see [ conparam ] ) . if the intracluster medium were collisionless , then its density profile would be identical to that of the dark matter . thus , we define the unmodified entropy distribution @xmath78 of a cluster of concentration @xmath42 to be that of gas in hydrostatic equilibrium in the cluster potential with a density profile identical to that of the dark matter . numerical experiments reveal that the entropy distribution of intracluster gas in clusters simulated without radiative cooling or supernova heating is quite similar to the nfw form throughout most of the cluster but is elevated above nfw within the cluster s core . figure [ denprofs ] shows the density distribution of a cluster simulated with an adaptive - mesh refinement code ( norman & bryan 1998 ; bryan 1999 ) for the santa barbara cluster comparision project ( frenk 1999 ) . the dotted line labeled `` nfw '' shows an nfw density distribution with @xmath79 , which is very close to the underlying dark - matter distribution of the cluster . the dashed line labeled `` simulation '' shows the gas - density distribution from that same simulation . agreement is quite close outside 10% of the virial radius , but the gas density levels off within that radius , implying that the lowest - entropy gas has a somewhat higher entropy than assumed in our unmodified distribution @xmath80 . however , this discrepancy at the low - entropy end of the distribution is inconsequential to the modified - entropy models that follow because the entropy of the discrepant gas will always be subject to further modification . the unmodified distribution is also somewhat discrepant with simulations at the high - entropy end . assuming hydrostatic equilibrium and an nfw gas - density law near the virial radius leads to gas temperatures that are 20 - 30% higher than those in our fiducial simulation . thus , our unmodified entropy levels at large radii are slightly higher than what a numerical simulation would produce . however , in order to properly reproduce the density profile near the virial radius in a hydrostatic model , we must retain these elevated entropy levels . this overestimate at the cluster s outskirts has virtually no effect on emissivity - weighted global quantities , such as @xmath6 and @xmath81 , but if one is interested in the temperature of cluster gas near @xmath23 , then departures from hydrostatic equilibrium , which are outside the scope of our models , must be taken into account . modification of this baseline entropy distribution is inevitable because clusters simulated without radiative cooling are physically inconsistent . in general , the gas at the center of such a simulated cluster can radiate many times its thermal energy within a hubble time , resulting in cluster luminosities that vastly exceed those observed ( e.g. , muanwong 2001 ; bryan & voit 2001 ) . condensation and removal of the lowest - entropy gas from the intracluster medium must happen at some level , because that is how the cluster s galaxies form stars . furthermore , the most massive of these stars must explode , resulting in supernova feedback . condensed gas accreting onto an active galactic nucleus can provide additional feedback ( heinz , reynolds , & begelman 1998 ; kaiser & alexander 1999 ; quilis , bower , & balogh 2001 ; reynolds , heinz , & begelman 2002 ; bhringer 2002 ) . all of these processes will modify the entropy distribution of the intracluster medium , but modeling them in detail is a daunting task . because the physics of feedback is so complex , we shall adopt a highly simplistic phenomenological approach to entropy modification . instead of trying to model all the consequences of heating and cooling we will restrict our investigation to three qualitatively different modifications of the entropy distribution @xmath80 : ( 1 ) truncation of the distribution , ( 2 ) shifting of the distribution , and ( 3 ) radiative losses from the distribution . each of these modifications depends on a single entropy threshold parameter @xmath82 , which we take to be the entropy at which the cooling time of gas of temperature @xmath83 equals the age of the universe . thus , for each type of entropy modification , we obtain a two - parameter family of modified - entropy models specified by a concentration @xmath42 and a dimensionless entropy @xmath84 . as in voit & bryan ( 2001 ) , we use an approximate cooling function for gas with a metallicity of one - third solar to determine the appropriate entropy threshold . according to this approximation , the threshold for cooling within 15 gyr can be expressed as @xmath85 with @xmath86 for @xmath87 and @xmath88 for @xmath89 . converting this threshold to dimensionless units yields @xmath90 as long as @xmath91 ( see [ emodtemp ] ) , then @xmath92 will be appropriate for the hottest clusters ( @xmath93 ) and @xmath94 will be appropriate for the coolest ( @xmath95 ) groups . one simple yet physically motivated way to modify the entropy distribution of intracluster gas is to truncate it at @xmath84 ( bryan 2000 ; voit & bryan 2001 ) . mathematically , we define the truncated distribution to be @xmath96 where @xmath97 is defined by @xmath98 . such a modification corresponds to removing all the gas with a cooling time less than a hubble time and ignoring the effects of cooling on the rest of the intracluster medium . this type of modification can viewed either as an extreme form of cooling , in which all the gas below the critical threshold condenses , or an extreme form of heating , in which all the gas below the threshold is heated to very high entropy ( @xmath99 ) and convects beyond the virial radius of the cluster . figure [ denprofs ] shows some gas density profiles resulting from different values of @xmath84 . the thin solid line labeled @xmath100 in figure [ kdist ] shows the truncated entropy distribution for @xmath101 and @xmath102 . notice that the maximum value of @xmath48 in these truncated models is less than unity because some of the gas has been removed from the intracluster medium . the thick solid line in figure [ rmax ] illustrates how the fraction @xmath103 of a cluster s baryons that remain in the intracluster medium depends on @xmath84 . for hot clusters this fraction is @xmath10490% , but for cool groups it drops below 50% . nevertheless , the outer radius @xmath61 of the intracluster gas remains close to @xmath23 , even when @xmath105 drops below 0.5 . figure [ denprofs ] shows the reason . as the entropy threshold for truncation rises , the central entropy of the cluster also rises the resulting density profile is therefore much flatter for large values of @xmath106 , enabling a smaller amount of gas + to fill a similar volume . note also that the density profile of a modified - entropy model with @xmath101 and @xmath107 is quite similar to that of a @xmath79 cluster simulated without cooling and feedback , suggesting that the cooling threshold may have only a modest effect on the structure of the hottest clusters . another way to modify intracluster entropy is to add a constant term @xmath84 to the unmodified entropy distribution ( voit & bryan 2001 ) . this type of modification is a convenient way to mimic `` preheated '' models for cluster formation that inject a fixed amount of entropy per particle into the intergalactic medium at some early time . in this case , the modified entropy distribution is defined by @xmath108 the dotted line in figure [ kdist ] shows this distribution for @xmath79 and @xmath102 . because entropy has been added but no gas has been removed , the outer radius for models with a shifted entropy distribution extends beyond the usual virial radius ( see figure [ rmax ] ) . qualitatively , this corresponds to heating that inflates the intracluster medium , driving the accretion shock to larger radii . comparing the structure of shifted models to truncated models reveals interesting similarities . figure [ denprofs_all ] depicts the dimensionless density profiles for @xmath109 , 0.3 , and 1.0 in clusters with @xmath79 . the run of density with radius is virtually identical for the shifted and truncated models when @xmath109 and is still quite similar when @xmath102 . thus , for most of the interesting range of entropy thresholds , there is little observable difference between shifted models and truncated models . however , as @xmath84 rises to 1.0 , the density profile of the shifted model becomes noticeably flatter than that of the truncated model . similarity between the truncated and shifted models at the lower entropy thresholds arises because the unmodified entropy @xmath110 depends approximately linearly on @xmath48 for small @xmath106 . thus , shifting the distribution by @xmath84 and truncating it at @xmath84 both produce similar modified entropy distributions and therefore similar density , temperature , and pressure distributions in the inner regions of the cluster . that is why entropy modification by heating and cooling can have similar effects on the luminosity and temperature of clusters as long as they establish similar values of the minimum entropy . observations show that many clusters contain gas with a cooling time less than a hubble time ( fabian 1994 ) , motivating us to consider slightly more complicated modifications to the entropy distribution . radiative cooling reduces the the specific entropy of a gas parcel according to @xmath111 this entropy equation can be rewritten to eliminate the dependence on gas density : @xmath112 in general , the rate at which entropy declines will depend on the prevailing hydrodynamic conditions , standard examples being isobaric cooling or isochoric cooling . however , for the purposes of this toy model , let us assume that the right - hand side of this equation remains constant . we make this rather arbitrary assumption for two physical reasons . first , within the temperature range @xmath113 , the quantity @xmath114 is approximately constant . second , as gas settles within a nearly isothermal potential , work done by compression will attempt to maintain the temperature of that gas near the characteristic temperature of the halo . thus , we are motivated to use the following modification to represent radiative losses : @xmath115^{3/2 } - \hat{k}_c^{3/2 } \}^{2/3 } \ ; \ ; .\ ] ] this modification corresponds to reducing the value of @xmath116 by the constant amount @xmath117 throughout the entropy distribution and discarding all the gas with @xmath118 . the short - dashed line in figure [ kdist ] shows an example of the resulting distribution for @xmath79 and @xmath102 , and a similar line in figure [ rmax ] shows how the maximum radius depends on @xmath84 . it will also be interesting to investigate a generalization of this radiative - loss model : @xmath119^{\alpha } - \hat{k}_c^{\alpha } \}^{1/\alpha } \ ; \ ; , \ ] ] in which the tunable parameter @xmath120 enables us to adjust the amount of gas below the cooling threshold . in the limit @xmath121 , this generalized distribution becomes identical to the truncated distribution . in what follows , we will focus on the intermediate case of @xmath122 , represented by the symbol @xmath123 and long - dashed lines in the relevant figures . figure [ denprofs_all ] compares the density distributions that result from the radiative - loss models with the truncated and shifted models . for @xmath109 , all the density profiles are very similar at @xmath124 , because the entropy profiles are nearly identical there . however , within that radius , entropy in the radiative - loss models drops below @xmath84 , allowing for considerably greater compression and higher central density . this trend becomes more pronounced at higher @xmath84 because the proportion of the intracluster medium below the cooling threshold is even larger . modified - entropy models generated through shifts and truncations of the unmodified entropy distribution produce surface - brightness profiles that are strikingly similar to the @xmath125-model profiles of many real clusters , at least within about 30% of the virial radius . figure [ sbright ] shows the dimensionless emission measure @xmath126 as a function of projected radius @xmath34 for three different truncated models , each with @xmath79 . as the threshold entropy progresses from @xmath109 , characteristic of clusters , through @xmath102 to @xmath127 , characteristic of groups , the @xmath128 profile substantially flattens . as long as @xmath67 does not change dramatically with radius , these emission - measure profiles will be nearly identical to the surface - brightness profiles . for comparision , figure [ sbright ] also shows some @xmath125-model profiles selected to match the modified - entropy models . these @xmath125-model profiles are defined by @xmath129^{-3\beta + 1/2}$ ] , where @xmath130 is the core radius and @xmath125 determines the asymptotic slope ( cavaliere & fusco - femiano 1978 ) . in each case , the @xmath125-model closely tracks the modified - entropy model within @xmath131 . beyond that radius , the emission measures of the @xmath125-models exceed those of the modified - entropy models , but owing to the low surface brightness at large radii , the @xmath125-model parameters of clusters and groups have generally been measured within @xmath132 . recent measurements by vikhlinin , forman , & jones ( 1999 ) suggest that cluster surface brightness does indeed fall below the best fitting @xmath125-model near the virial radius . in order to determine how the best - fitting @xmath125 and @xmath130 depend on @xmath42 and @xmath84 , we performed surface - brightness weighted fits over a grid of modified entropy models . because the best - fit parameters depend somewhat on the portion of the profile being fitted , we elected to do these fits over two different radial intervals : @xmath133 and @xmath134 . results for truncated ( @xmath100 ) models are shown in figure [ betafits_t ] , and those for shifted ( @xmath135 ) models are shown in figure [ betafits_s ] . because the profiles in the shifted models tend to be shallower , the corresponding @xmath125 values are somewhat smaller , particularly for low values of @xmath84 . however , in both cases , the best - fitting @xmath125 values span the observed range for clusters and groups of galaxies ( see [ sec - sbprofs ] ) . one consequence of raising the core entropy of a cluster is a boost in the luminosity - weighted temperature of the intracluster gas . figure [ tprofs ] shows that increasing @xmath84 raises the temperature of gas at all radii in both the shifted and truncated models . amplification of the central temperature can be quite extreme , exceeding @xmath136 when @xmath84 rises to 1.0 . however , the resulting enhancement of the luminosity - weighted temperature is not nearly as dramatic . the value of @xmath137 turns out to depend more critically on the concentration parameter @xmath42 than on the entropy threshold , as demonstrated in figure [ tlums_ts ] . the reason that raising the core entropy of gas in an nfw potential has only a modest effect on its luminosity - weighted temperature involves both the gas density profile and the dark - matter density profile . extra entropy flattens the density profile of intracluster gas , diminishing the relative luminosity of gas in the core . the characteristic radius of gas contributing the bulk of the cluster s luminosity therefore moves outward . because the shape of an nfw halo is steeper than isothermal at @xmath138 , the characteristic temperature of gas within such a halo declines outside that radius ( see figure [ tprofs ] ) . thus , the rise in gas temperature resulting from an increase in @xmath84 is largely offset by an increased contribution to the total luminosity from lower - temperature gas at larger radii . this effect is most pronounced in the shifted models , which have substantial amounts of relatively cool gas beyond @xmath58 . in those models , @xmath137 actually _ declines _ as @xmath84 approaches unity . the large central temperatures in the shifted and truncated cases also seem somewhat unrealistic in light of observations showing either level or radially increasing temperature gradients within cluster cores ( e.g. , arnaud 2001a , b ; allen 2001 ) . in order to compare these temperatures more directly with observations , we must look at @xmath139 , the luminosity - weighted temperature along various lines of sight through a cluster . figure [ tperp_03 ] illustrates how @xmath139 depends on projected radius @xmath73 in clusters with @xmath79 , @xmath102 , and four different kinds of entropy modification . projected temperatures at @xmath140 in both the shifted and truncated models are still quite high ( @xmath141 ) . however , the core temperatures in the radiative - loss models look more like those in real clusters ( see [ sec - tinner ] ) , rising with radius in the innermost regions and peaking between @xmath142 and @xmath143 at @xmath144 in these particular models . cool gas in the cores of clusters with entropy modified through radiative losses can significantly affect the luminosity - weighted temperature if the fraction of gas below @xmath84 is relatively large . figure [ tlums_r ] shows that @xmath137 for a given concentration @xmath42 and entropy threshold @xmath84 is @xmath10430% lower in the standard radiative - loss model ( @xmath145 ) than in the corresponding truncated model . this effect is also present but not nearly so strong in the generalized radiative - loss model with @xmath122 , where @xmath137 is @xmath10410% lower than in the truncated models ( see figure [ tlums_ta ] ) . all of our schemes for entropy modification substantially lower the luminosities of clusters . figure [ lints_8 ] shows how dimensionless luminosity @xmath146 depends on the entropy threshold @xmath84 for clusters with @xmath79 . in all cases , luminosity declines by over an order of magnitude as @xmath84 rises from 0.1 to 1 this decline is shallowest for the shifted model because the additional gas at large radii contributes significantly to the total luminosity when @xmath147 . note also that @xmath146 is over twice as large in the standard radiative - loss case ( @xmath145 ) because a large proportion of the luminosity in such models comes from the dense core . luminosity enhancement in the @xmath148 radiative - loss model is more moderate only a few tens of percent greater than in the truncated case . analyzing how @xmath146 depends on @xmath84 leads to some important insights into how the cooling threshold determines the luminosity - temperature relation of clusters and groups . figure [ lints_tx ] shows that @xmath149 is a good approximation to the @xmath146-@xmath84 relation for truncated models in the range @xmath150 . this scaling reflects the asymptotic slope of the nfw density profile . notice that the outer portions of the density profiles in figure [ denprofs ] track the underlying nfw profile of the dark matter , and the inner portions are nearly isentropic with @xmath151 . one can therefore approximate the density profile of a truncated model by joining an isentropic density profile to an nfw profile at the radius @xmath152 at which the density of the unmodified nfw profile is @xmath153 . the analytical isentropic solution from [ dimensionless ] tells us that @xmath154^{3/2}$ ] within @xmath152 , so the luminosity from the isentropic gas scales @xmath155 . the luminosity from the nfw portion of the profile scales in the same way for the radii of interest . thus , for larger values of @xmath34 , where the density of the nfw profile approaches @xmath156 , we have @xmath157 . this particular scaling of @xmath146 with @xmath84 has some very interesting consequences . section [ integrated ] showed that the physical luminosity of a cluster scales as @xmath158 . in order to substitute physical quantities for @xmath146 , we need to recognize that @xmath159 and that @xmath160 , where @xmath161 is the age of the universe . we therefore obtain @xmath162 where @xmath163 is the hubble constant at time @xmath161 . the following three features of this expression are particularly noteworthy : * the power - law slope it predicts for the x - ray @xmath164 relation is very close to the observed slope of @xmath165 ( e.g. , markevitch 1998 ; arnaud & evrard 1999 ) . because the halos of low - temperature groups have higher concentrations than high - temperature clusters ( see [ conparam ] ) , we expect @xmath137 to decline by a factor @xmath166 as @xmath81 rises by an order of magnitude . this dependence on concentration steepens the power - law slope implied by equation ( [ ltscal ] ) to @xmath167 . * the @xmath6-@xmath7 relation does not depend on the form of the cooling function , which scales as @xmath168 in the free - free cooling regime at @xmath87 and as @xmath169 in the line - cooling regime at @xmath170 . if the entropy threshold in real clusters did not depend on @xmath171 , then we would expect a distinct steepening of the @xmath6-@xmath7 relation as @xmath7 rises through @xmath2 , which is not observed . equation ( [ ltscal ] ) is indepedent of @xmath171 because the dimensionless gas mass within @xmath152 scales @xmath172 , and thus remains approximately constant in an nfw halo for @xmath173 . the cooling time of that gas is @xmath174 , so its physical luminosity scales @xmath175 . this scaling law also applies to the entire cluster , because the luminosity of gas outside @xmath152 scales in lockstep with gas inside @xmath152 . applying the relation @xmath176 thus leads back to equation ( [ ltscal ] ) , except for the @xmath137 factor . * finally , the @xmath6-@xmath7 relation implied by equation ( [ ltscal ] ) changes very little with time , again in agreement with observations ( mushotzky & scharf 1997 ; donahue 1999 ; della ceca 2000 ; borgani 2001b ) . that happens because the mass of a cluster of a given temperature declines @xmath177 with increasing redshift . according to the @xmath178 scaling , the resulting decline in the intracluster gas mass compensates almost precisely for the drop in the cooling threshold owing to the shorter cooling time @xmath161 . this scaling breaks down at the low - temperature end , where @xmath179 , because the entire intracluster ( or intragroup ) medium in the truncated and shifted cases becomes effectively isentropic . then , the gas density determined by the cooling threshold is @xmath180 , and the luminosity from within @xmath23 scales as @xmath181 . thus , we should expect the @xmath6-@xmath7 relation for low - temperature halos to be steeper than for clusters , roughly @xmath182 for @xmath183 , and we should expect the luminosity of objects at a given temperature to increase with time . in order to make further progress , we will need to understand how halo concentration depends on halo mass . the previous section explored the properties of dimensionless modified - entropy models for clusters depending on two parameters , the halo concentration parameter @xmath42 and an entropy - threshold parameter @xmath84 . as in voit & bryan ( 2001 ) , we argued that the crucial entropy threshold should be the entropy level at which intracluster gas would cool in a hubble time . now we will use a set of relations between halo concentration and halo mass to construct modified - entropy models for real clusters . in these models , the halo mass @xmath184 determines the concentration parameter @xmath42 . the mass and halo concentration then set the temperature of the cluster , and this temperature sets the level of the entropy threshold @xmath82 . cluster properties are therefore determined primarily by the overall cosmological model , because it fixes the global baryon fraction , the relation between halo mass and concentration , and the hubble time that governs @xmath185 . some of a cluster s properties , such as the temperature of the innermost gas , also depend on the chosen scheme for entropy modification . the properties of the resulting modified - entropy models turn out to be remarkably similar to those of observed clusters and groups . the section begins by defining the @xmath184-@xmath42 relations we will use and comparing them to some of the scarce data . then , we explore the structure of the intracluster medium in our modified - entropy cluster models , showing that the observed relationship between @xmath125 and @xmath81 may be more than just a selection effect and demonstrating how entropy modification can lead to temperature profiles similar to those observed in both the inner and outer regions of clusters . we go on to compare the mass - temperature relation from our modified - entropy models to observations , showing that the models reproduce the observational results of nevalainen ( 2000 ) and finoguenov ( 2001 ) . we also compare the @xmath8-@xmath81 relation from our models to cluster data , again showing excellent agreement . much of the dispersion of this relation , in both the models and the data , stems from differences in the amount of gas below the cooling threshold , and we conclude with some speculations about what governs these differences . the concentration of a dark matter halo is determined by its formation history . early formation generally leads to a denser core and thus a higher concentration . low - mass halos , which tend to collapse earlier in time , are therefore expected to be somewhat more concentrated than high - mass halos . because a halo s collapse and merger history determines its concentration , the relation between concentration and halo mass depends on the underlying cosmological model , specified by the matter density @xmath186 , the dark - energy density @xmath187 , the normalization @xmath188 of the perturbation spectrum , and the shape of that spectrum , often parameterized by @xmath189 ( bardeen 1986 ; sugiyama 1995 ) . in this paper , we will restrict our attention to currently - favored @xmath171cdm models , in which the resulting clusters turn out to be thoroughly consistent with observations . several versions of the concentration parameter @xmath190 can be found in the literature , each defined with respect to a radius @xmath191 within which the mean halo density is @xmath192 . here we will define @xmath42 with respect to @xmath23 , in order to be consistent with the models developed in the previous section ; thus , @xmath193 . figure [ c200 ] shows various relations between halo mass and @xmath194 that can be found in the lit- erature . the _ chandra _ observations of allen ( 2001 ) suggest that concentrations corresponding to @xmath195 are appropriate for massive ( @xmath196 ) clusters . concentrations of groups are less certain because they are more model - dependent , and the dispersion in @xmath194 is expected to increase at lower halo masses ( afshordi & cen 2002 ) . in this section , we will adopt the halo mass - concentration relations predicted by models @xmath197 and @xmath198 from eke , navarro , & steinmetz ( 2001 ) , which are consistent with the _ chandra _ data on massive clusters . both models are derived from @xmath199-body simulations of @xmath171cdm cosmologies with @xmath200 , @xmath201 , and @xmath202 . the power spectrum in each model has a shape parameter @xmath203 , but the normalization of the power spectrum differs : @xmath204 in model @xmath197 and @xmath205 in model @xmath198 . the higher normalization in model @xmath198 leads to halos that are about 30% more concentrated . figure [ kc_m200 ] shows the dependence of @xmath84 on halo mass that follows from model @xmath198 ; the entropy thresholds for model @xmath197 are very similar . the cluster models we analyze in this section are completely determined by the assumed relationship between halo mass and concentration ( either @xmath197 or @xmath198 ) and the chosen scheme for entropy modification ( @xmath206 , @xmath207 , @xmath208 , or @xmath209 ) . for each halo mass , we compute the bolometric x - ray luminosity @xmath8 , luminosity - weighted temperature @xmath81 , x - ray surface - brightness profile @xmath210 , and projected temperature profile @xmath211 . unlike in [ integrated ] , each integral includes a cooling function @xmath67 drawn from the models of raymond & smith ( 1977 ) for gas with 30% solar metallicity . we will first focus on the spatially resolved properties of these model clusters , showing how their surface - brightness profiles and temperature gradients compare with those of observed clusters . modified - entropy models of groups tend to have flatter surface - brightness profiles than hot clusters , in accordance with observations ( e.g. , ponman 1999 ; horner 1999 ) . because the surface - brightness profiles of both real clusters and our modified - entropy models are well described by @xmath125-models with @xmath212^{-3\beta + 1/2}$ ] inside of @xmath213 ( see [ emodsb ] ) , we compare our models with the data by comparing the best - fitting @xmath125-model parameters . figure [ bmods ] shows how @xmath125 and @xmath130 depend on @xmath81 in truncated modified - entropy models with relation @xmath198 and in the cluster data of finoguenov ( 2001 ) . the models reproduce the observed range of both parameters and the observed rise in each parameter with increasing temperature . however , the observed dispersion in each parameter at a given temperature is quite large . wu & xue ( 2002 ) find the same kind of temperature dependence in their cluster models , based on the work of bryan ( 2000 ) , which are very similar to the truncated models developed in this paper . some of the dispersion in @xmath125 and @xmath130 at a given temperature may stem from selection effects . because the surface - brightness profiles of modified - entropy clusters continually steepen with radius , unlike those of @xmath125-models , the best - fitting @xmath125 and @xmath130 depend somewhat on the range of radii being fit . figure [ bmods ] shows that the best - fitting values of @xmath125 and @xmath130 both decline as the surface - brightness threshold of the fit rises from @xmath214 through @xmath215 to @xmath216 @xmath217 . this effect is more severe for groups than for clusters because their overall surface brightness is smaller . the figure also shows that fits over the radial range @xmath218 yield slightly larger @xmath125 and @xmath130 than fits over @xmath219 . komatsu & seljak ( 2001 ) noted a similar surface - brightness selection effect in their polytropic models for clusters . however , the rise of @xmath125 with @xmath81 in their models is entirely due to this surface - brightness bias . in our models , it is largely a consequence of the greater impact of entropy modification on groups . puzzlingly , this selection effect becomes stronger at low surface brightness levels in the models of wu & xue ( 2002 ) , perhaps because their boundary condition leads to a less precipitous decline in @xmath220 near the virial radius . the observed dispersion in @xmath125 and @xmath130 may also depend in part on the amount of gas below the cooling threshold . figure [ bmods_all ] illustrates how @xmath125 and @xmath130 depend on the scheme for entropy modification . shifted models have core radii identical to those of the corresponding truncated models and slightly smaller values of @xmath125 . however , allowing for radiative losses reduces the best - fitting core radius by a factor of @xmath221 and flattens the @xmath125-@xmath81 relation , yielding @xmath222 for fits over the interval @xmath223 . other possible sources of dispersion include ellipticity and cluster substructure , which can not be investigated in the context of our spherically symmetric models . the temperature gradient within @xmath224 also appears to depend on the amount of gas below the cooling threshold . figure [ tinner ] , in which the model parameters are appropriate for hot clusters , shows that @xmath211 gradually declines with radius when the entropy distribution is simply truncated at the entropy threshold . our radiative - loss models , in contrast , have temperatures that rise outward from @xmath225 , then flatten , becoming virtually isothermal out to 30% of the virial radius . as the parameter @xmath120 drops from the very large values characteristic of truncated models to the canonical radiative - loss value of 3/2 , the region with a positive temperature gradient becomes progressively larger . the run of temperature with radius in our model with @xmath226 is very similar to the `` universal '' temperature profile observed by allen ( 2001 ) in clusters that appear to have strong cooling flows . in order to facilitate comparison of our models with that data , figure [ tinner ] shows radii in units of @xmath227 , the radius within which the mean matter density is @xmath228 , and temperature in units of @xmath229 , the mean mass - weighted temperature of intracluster gas within @xmath227 . for clusters of this concentration , @xmath230 . note that setting @xmath231 produces a cluster model with a temperature gradient remaining within 10% of isothermal from small radii to beyond @xmath131 . projected cluster temperatures at large radii gradually decline with radius in all of our modified - entropy models . _ asca _ observations of many clusters show evidence for such a decline , which is often expressed as an effective polytropic index @xmath232 ( e.g. , markevitch 1998 ; markevitch 1999 ; ettori & fabian 1999 ) . these observations indicate that @xmath233 outside the cores of clusters . however , recent _ xmm - newton _ observations have shown that at least some clusters are nearly isothermal within @xmath234 , except for the low - temperature core ( arnaud 2001a , b ) . the results of [ sec - tinner ] suggest that radiative losses may be responsible for the isothermal profiles found in these clusters . figure [ gamma_profs ] shows how @xmath235 varies with radius in truncated modified - entropy models and in radiative - loss model @xmath209 . the cores of the truncated models are nearly isentropic , with @xmath236 at @xmath237 . this isentropic region extends to larger radii in cooler halos because their entropy thresholds amount to a larger fraction of the halo s mean entropy . outside the isentropic region , we find @xmath238 , in agreement with observations . the origin of this value of @xmath235 can be traced back to our assumption that the unmodified gas and dark - matter density profiles are described by identical nfw laws . komatsu & seljak ( 2001 ) have exploited that same assumption to construct polytropic models for the intracluster medium , finding that @xmath239 produces the best agreement between these density profiles at large radii . the behavior of @xmath235 in the radiative - loss models is similar to the truncated models at large radii but quite different at small radii . figure [ gamma_profs ] shows that the effective polytropic index steadily rises from @xmath240 inside the core through @xmath241 near the core radius to @xmath239 at larger radii . the point at which a cluster makes this transition from an isothermal gradient to a mildly declining gradient depends somewhat on cluster temperature . hot clusters with massive halos tend to have smaller concentrations which allow for near - isothermality at larger radii . in order to compare our models with observations , we have evaluated @xmath235 at @xmath242 , essentially twice the core radius . this radius is large enough to escape the severe model - dependence of @xmath235 within the cluster core but small enough that the signal - to - noise in cluster observations still allows for accurate measurements . figure [ gammas ] compares those evaluations of @xmath235 with the data of finoguenov ( 2001 ) . our models agree with the data , within the large observational uncertainties , and also share the observed ( 2.5@xmath243 ) tendency for @xmath235 to decline with increasing temperature . two different effects combine to produce higher values of @xmath235 at low temperature . because the halos of groups tend to be more concentrated than those of clusters , the scaled temperature @xmath244 tends to be larger in the cores of groups , leading to higher @xmath235 . an additional enhancement of @xmath235 appears in truncated models , in which the entropy threshold has a more significant effect at @xmath245 on group scales ( see figure [ gamma_profs ] ) . many efforts to measure cosmological parameters such as @xmath186 and @xmath188 have relied heavily on the cluster mass - temperature relation to link cluster masses to observables like cluster temperature and luminosity ( e.g. , henry 1997 ; bahcall & fan 1998 ; eke 1998 ; donahue & voit 1999 ; borgani 1999 ; henry 2000 ) . until very recently , observers had to rely on @xmath246-@xmath7 relations that were calibrated with numerical simulations ( e.g. , evrard 1996 ; bryan & norman 1998 ) . these simulations provide the constant of proportionality @xmath137 in the scaling relation @xmath247 values of @xmath137 from simulations in the literature can differ by as much as 30% ; examples include @xmath248 from evrard ( 1996 ) , @xmath249 from bryan & norman ( 1998 ) , @xmath248 from thomas ( 2001 ) , and @xmath250 from the non - radiative simulations of muanwong ( 2001 ) . the vast majority of these simulations ignore radiative cooling and supernova heating , both of which can elevate @xmath137 . mass - temperature relations derived from spatially - resolved x - ray observations now suggest that the normalization constant @xmath137 derived from simulations is indeed too low ( horner 1999 ; nevalainen 2000 ; finoguenov 2001 ) . furthermore , cluster - formation simulations that include radiative cooling also yield a higher normalization constant ; for example , @xmath251 in the radiative simulations of muanwong ( 2001 ) . our modified - entropy models support these findings . figure [ mtrel ] compares two @xmath252-@xmath81 relations derived from our modified - entropy models with the data from finogeunov ( 2001 ) and nevalainen ( 2000 ) . in these relations , we scale to @xmath252 , the halo mass inside the radius @xmath253 within which the mean matter density is @xmath254 , because measurements of @xmath252 are thought to be more reliable than measurements of @xmath184 . the data have been corrected for any cooling flow that might be present in the cluster core , so we compare them to our truncated models with concentration laws @xmath197 and @xmath198 . the corresponding @xmath252-@xmath81 relations generated from shifted models are virtually identical . both the truncated and shifted models agree well with the data , but those based on concentration relation @xmath198 appear to match the data slightly better . this agreement between the observed @xmath252-@xmath81 relation and the modified - entropy models is quite compelling , considering that the models contain no adjustable parameters . for comparision , figure [ mtrel ] also shows the relation derived from simulations by evrard ( 1996 ) , which has a lower temperature normalization ( equivalent to a higher mass normalization ) and a shallower slope . the steeper slope of the @xmath252-@xmath7 relation from modified - entropy models arises primarily from the dependence of @xmath137 on halo concentration . the higher concentration of low - mass halos leads to a higher luminosity - weighted temperature for a given mass . komatsu & seljak ( 2001 ) find a similar effect in their polytropic cluster models . entropy modification produces some additional steeping in the relation and shifts its normalization to higher temperatures ( see figure [ tlums_ts ] ) . the greater dispersion and offset toward lower masses at @xmath255 seen in figure [ mtrel ] could be due to surface - brightness bias ( lloyd - davies , bower , & ponman 2002 ) . because the gas temperature of intracluster gas in our models tends to decline with radius at large radii , the luminosity - weighted temperature depends on the maximum radius over which that temperature is measured . this effect is negligible in clusters because the bulk of the emission is concentrated around the cluster s core , but it can be significant in groups . figure [ mtrel_t_sb ] shows how the observed @xmath252-@xmath81 relation would depend on the limiting surface - brightness for truncated models with @xmath198 concentrations . increasing the surface brightness threshold changes the derived relation in the sense that is observed . figure [ mtrel_sb5e15 ] shows the relations for different entropy modification schemes at a fixed surface - brightness threshold of @xmath256 . truncated and shifted models are affected similarly , but surface - brightness bias has less effect on the radiative - loss model ( @xmath123 ) because its central temperature gradient is closer to isothermal . some of the scatter to higher temperatures may also arise from the scatter in concentrations expected in low - mass halos ( e.g. , afshordi & cen 2002 ) , but the analysis of lloyd - davies ( 2002 ) suggests that halo concentration alone can not account for the large dispersion of the mass - temperature relation for groups . despite the excellent agreement between our models and the observations , a word of caution is in order . both the modified - entropy models and the x - ray mass derivations assume that clusters are in hydrostatic equilibrium . simulations , on the other hand , suggest the intracluster velocity field is not completely relaxed , with turbulent velocities @xmath257% of the sound speed ( ricker & sarazin 2001 ) . furthermore , the luminosity - weighted temperature derived from theoretical modeling is not necessarily identical to the spectral - fit temperature derived from observations ( mathiesen & evrard 2001 ) . nevertheless , our results strongly imply that entropy modification and the cosmology - dependent relationship between halo mass and concentration must be accounted for in simulations designed to normalize the cluster mass - temperature relation . our previous work on modified - entropy models and the cooling threshold demonstrated how well these models account for the observed slope and normalization of the cluster luminosity - temperature relation ( voit & bryan 2001 ) . that paper compared truncated and shifted modified - entropy models to cluster data that were corrected for the effects of cooling flows . here , we revisit those models and show how the @xmath8-@xmath81 relation changes when gas below the cooling threshold is included . figure [ ltrel ] shows the @xmath8-@xmath81 relation for clusters without cooling flows . of the three models shown , the truncated model with concentration relation @xmath198 best fits the data , but all three models faithfully describe the @xmath8-@xmath81 relation equally well on cluster scales . substituting concentration relation @xmath197 shifts the models to slightly lower temperatures , but the agreement with groups is almost as good . using shifted instead of truncated entropy distributions elevates the x - ray luminosity on group scales by a factor @xmath258 . while there are essentially no free parameters in these models , the cluster luminosity does depend on our assumed value of @xmath186 . raising @xmath186 would lower the global ratio of baryons to dark matter , thereby lowering the luminosities of clusters in our models @xmath259 ( see eq . [ eq - lx ] ) . conversely , lowering @xmath186 would raise the cluster luminosities . the excellent agreement found here for @xmath260 should therefore reinforce confidence in similar measurements of the matter density from the baryon - to - dark - matter ratios of clusters ( e.g. , evrard 1997 ) . our models imply that groups are less suitable for such measurements because their baryon - to - dark - matter ratio within @xmath23 may be @xmath261% of the global value ( see figure [ rmax ] ) . the tendency of the data at @xmath262 to scatter toward the low-@xmath8 , high-@xmath81 side of the model relations may be due to the same effects that cause similar scatter at those temperatures in the observed @xmath252-@xmath263 relation . figure [ ltrel_t_sb ] shows how surface - brightness bias affects the luminosity - temperature relations derived from our models . raising the surface - brightness threshold within which luminosity and temperature are measured lowers @xmath8 while raising @xmath81 , producing the same trend seen in the data . measuring both of these quantities only within regions where @xmath264 produces relations that track the lower envelope of the group data . unlike the cluster measurements , the data on groups have not been corrected for the effects of gas below the cooling threshold , which could counterbalance the surface - brightness bias by raising @xmath8 and lowering @xmath81 ( see figure [ ltrel_sb5e15 ] ) . we therefore suspect that the dispersion in the @xmath6-@xmath7 data of helsdon & ponman ( 2000 ) arises from a combination of differing surface - brightness biases and differing amounts of gas below the cooling threshold . reproducing the behavior of clusters and groups with significant amounts of gas below the cooling threshold requires an additional parameter . in our models , that parameter is @xmath120 , which characterizes the slope of the entropy distribution function below the cooling threshold . figure [ lt_mark ] shows the relationship between @xmath81 and @xmath265 , the rosat - band ( 0.1 - 2.4 kev ) luminosity , in the data from markevitch ( 1998 ) and in models with differing values of @xmath120 . triangles show data corrected to remove the effects of gas below the cooling threshold , and the squares connected to those triangles show the uncorrected data . while the corrected data cluster around the line representing truncated models , many of the uncorrected data points lie closer to the line representing radiative - loss models with @xmath122 . a few uncorrected data points stray even farther from the truncated models , toward models with @xmath266 . notice that the lines connecting data points with the largest corrections are generally parallel to the lines connecting modified - entropy models with identical halo mass . this agreement suggests that the parameter @xmath120 adequately characterizes the offset in @xmath8-@xmath81 space owing to intracluster gas below the cooling threshold . an interesting test of these radiative - loss models would be to compare the value of @xmath120 inferred from the luminosity - temperature offset to that implied by the inner temperature gradient . the modified - entropy models we have constructed suggest that most of the observable properties of clusters depend on only two parameters . the halo mass @xmath184 determines a sort of cluster `` main sequence '' with well - defined @xmath252-@xmath81 and @xmath8-@xmath81 relations . deviations from those relations and the severity of the inner temperature gradient both depend on a second parameter , @xmath120 , related to the amount of intracluster gas below the cooling threshold , but what physical processes determine the value of @xmath120 ? one is tempted to interpret @xmath120 in terms of a mass cooling rate @xmath267 ; however , _ xmm - newton _ observations show little evidence for the low - temperature line emission expected in the standard cooling - flow picture ( peterson 2001 ) . thus , instead of viewing @xmath120 as simply a measure of cooling - flow strength , we would like to suggest a wider range of possibilities : * entropy history . our simplistic schemes for entropy modification treat clusters as if they were assembled very early in time , with some standard initial entropy distribution . in fact , the entropy history of the intracluster medium is likely to be far more complex , influenced in differing ways at various times by merger shocks , radiative cooling , and feedback . the dispersion seen in @xmath120 could stem from differences in the entropy history of clusters , with some evolutionary paths leading to abundant gas below the cooling threshold and other paths leading to very little . if this is the case , then a cluster s value of @xmath120 would be an important window into its past . * incomplete relaxation . _ chandra _ observations of several clusters have revealed `` cold fronts '' thought to be evidence of low - entropy gas sloshing around in the cores of those clusters ( markevitch 2000 ; vikhlinin , markevitch , & murray 2001 ) . our models assume that the lowest - entropy gas in a cluster has already settled into hydrostatic equilibrium at the cluster s center . some of the apparent dispersion in @xmath120 could be due to differences in the degree to which low - entropy gas has settled within cluster cores following the last major merger . * recent feedback . cooling and condensation of low - entropy gas at the center of a cluster is likely to lead to star formation and perhaps enhanced activity in the nucleus of the central galaxy ( fabian 1994 ) . both star formation and radio jets can provide feedback , imparting entropy to the cluster core that inhibits further condensation ( bhringer 2002 ) . perhaps these feedback processes control the value of @xmath120 . * thermal conduction . the presence of significant temperature gradients within the cores of some clusters has often been cited as evidence that magnetic fields strongly suppress electron thermal conduction in clusters ( fabian 1994 ) . recent _ chandra _ observations of some particularly sharp cold fronts have reinforced that supposition ( ettori & fabian 2000 ) . however , these observations constrain conduction in only one dimension and do not necessarily rule out conduction in the other two dimensions . furthermore , recent theoretical work suggests that magnetic fields might not suppress conduction as effectively as previously believed ( malyshkin & kulsrud 2001 ; malyshkin 2001 ; narayan & medvedev 2001 ) . it is possible that the magnetic field geometry in a cluster s core places a lower limit on the value of @xmath120 . investigating all of these possibilities is beyond the scope of this paper , but in the next section we explore how one might self - consistently calculate the intracluster entropy distribution within a hierarchical merging scenario . we have shown that some very simple prescriptions for the entropy distribution in present - day clusters lead to a remarkably realistic set of cluster models . however , the astrophysics that leads to these entropy distributions is certainly more complicated than we have assumed . understanding how the the physics of structure formation , cooling , and feedback determine the intracluster entropy distribution will require a more complete theory based on hierarchical merging . this section outlines a few of the basic issues involved in constructing such a theory . considering the entropy history of a single gas parcel affords some insight into how radiative cooling establishes an entropy threshold ( voit & bryan 2001 ) . suppose that the parcel begins with entropy @xmath268 in a halo of temperature @xmath269 at some early time @xmath270 . in a hierarchical merging scenario , the parcel s halo will eventually merge with another halo at time @xmath271 . merger shocks will then raise the parcel s entropy to some new value @xmath272 , and as that parcel settles within the new halo , it will approach a new temperature @xmath273 , similar to the characteristic temperature of that new halo . each subsequent merger will also raise the parcel s temperature and entropy . thus , the trajectory of a parcel through the entropy - temperature plane can be schematically represented by a sequence of points like that in figure [ entprop ] . the dashed lines in figure [ entprop ] show how the entropy threshold associated with cooling rises with time until it achieves its present value , indicated by the solid line . as long as a parcel s trajectory through the entropy - temperature plane remains above the cooling threshold , then radiative cooling and subsequent feedback will not significantly affect that parcel s entropy . however , any parcel that spends a large fraction of time below that threshold will be subject to substantial entropy loss , condensation , and whatever feedback ensues . feedback can continue as long as there are gas parcels below the cooling threshold , but once cooling and feedback have eliminated all gas below the cooling threshold , feedback must cease . thus , the threshold for entropy modification established by both cooling and feedback corresponds to the cooling threshold defined by the hubble time @xmath274 . although cooling sets the threshold for entropy modification , the model we are describing is not a cooling - flow model . hierarchical structure formation will ensure that mergers disperse the products of condensation throughout the final cluster . stars and galaxies form out of the lowest - entropy gas long before the cluster itself forms . hence , much of the low - entropy gas that would eventually find its way into the cluster core in a simulation with no radiative cooling decouples from the intracluster medium before it can participate in a centrally - focused cooling flow . finally , for reasons we will describe next , a certain amount of feedback is needed to prevent the incipient intracluster medium from overcooling at high redshift . overcooling is one of the classic problems that models of hierarchical structure formation must overcome ( cole 1991 ; white & frenk 1991 ; blanchard , valls - gabaud , & mamon 1992 ; balogh 2001 ) . if radiative cooling were allowed to proceed unchecked as structure formed , then a substantial percentage ( @xmath275% ) of the universe s baryons would have been locked into condensed objects before the era of cluster formation . because the fraction of uncondensed baryons in present - day clusters is @xmath276% , feedback processes like supernova heating are presumed to inhibit cooling and condensation of high - redshift baryons ( white & rees 1978 ; cole 1991 ; white & frenk 1991 ) . in our models , overcooling manifests itself as an entropy threshold @xmath82 that is much higher than the characteristic entropy of an unmodified high - redshift halo . to illustrate this effect , let us assume that the progenitor halo of a present - day cluster has a temperature @xmath277 that depends on redshift . the entropy threshold for that halo is then @xmath278 \ , t_h(z ) \}^{2/3}$ ] , and the characteristic entropy within @xmath23 in the unmodified halo is @xmath279 . the ratio of these two quantities in the high - redshift limit is thus @xmath280 } { t(z ) } \right\}^{2/3 } [ h(z ) t_h(z)]^{2/3 } h^{2/3}(z ) \ ; \ ; .\ ] ] because @xmath281 rises and @xmath277 declines with increasing redshift , this ratio must exceed unity at early times . we can use extended press - schechter theory ( bond 1991 ; bower 1991 ) to approximate @xmath277 and estimate the temperature , entropy , and redshift scales at which overcooling is problematic . first , to demonstrate the principle with simple scaling relations , we consider a scdm ( @xmath282 ) universe with a power spectrum given by an @xmath283 power law . in that case , the characteristic mass scale evolves like @xmath284 , and assuming that @xmath285 yields @xmath286 . at low redshifts , where @xmath7 is high , we have @xmath287 and @xmath288 . however , at high redshifts , when the progenitor temperature is smaller , we have @xmath169 and @xmath289 . thus , the progenitor of a halo with @xmath290 at @xmath291 would have @xmath292 at @xmath293 . figure [ ent_evol ] illustrates this effect more precisely for a @xmath171cdm universe seeded by a power spectrum with shape parameter @xmath294 . the dashed line shows @xmath295 for a halo that reaches @xmath296 at @xmath9 . the solid line shows the entropy threshold @xmath82 in that evolving halo for a metallicity of one - third solar . feedback must impart an entropy @xmath297 to prevent overcooling at @xmath298 . feedback can solve the overcooling problem , but it introduces another , more subtle problem into our framework . our modified - entropy models assume that gas above the cooling threshold follows the unmodified distribution @xmath78 . however , the need for feedback at early times implies that a large proportion of intracluster gas has experienced entropy modification at least once in its history . somehow , the process of cluster formation needs to restore the unmodified distribution to gas above the threshold at @xmath9 . exactly how that happens is unclear . consider what happens in a merger between two halos containing isothermal gas at temperatures @xmath269 and @xmath273 that collide at velocity @xmath299 . merger shocks with mach number @xmath300 will propagate through each halo s gas , raising the entropy of that gas by some more - or - less uniform factor depending on @xmath301 . this mechanism could potentially amplify the effects of early feedback , which must raise the intrahalo entropy to at least @xmath82 to avoid overcooling . subsequent shocks would then boost the entropy even further , along a path parallel to the dashed line in figure [ ent_evol ] . if this mechanism dominates , then very little gas should remain near @xmath82 at @xmath302 . however , early feedback is likely to diminish the effects of mergers , particularly in gas with low entropy . if feedback raises all the gas in the progenitor halo of a cluster to @xmath303 , then the intrahalo medium will be nearly isentropic , with high - temperature gas at the center of the halo and lower - temperature gas at the outskirts ( see figure [ tprofs ] ) . when mergers occur , mach numbers associated with lower - entropy gas in the halo s core will therefore be lower than those associated with higher - entropy gas near the virial radius . hence , merger shocks will act to steepen any entropy gradient that exists . dynamical friction may also add to this effect , because the cores of merging halos tend to lose orbital energy before they collide and merge . another effect to consider is the quasi - continuous accretion of gas associated with small halos , for which the mach number is likely to be quite high . this type of accretion will enhance the amount of high - entropy gas in the final cluster . thus , it seems possible that the modified - entropy models we have developed can successfully be linked with hierarchical structure formation , although many details remain to be worked out . several recent analyses of similarity breaking in groups and clusters have cast the problem in terms of energy rather than entropy ( e.g. , wu 1998 , 2000 ; lloyd - davies , ponman , & cannon 2000 ; loewenstein 2000 ; bower 2001 ; lloyd - davies 2002 ) . for example , if one knows what the configuration of the intracluster gas would be in the absence of non - gravitational heating and cooling processes , then one can opt to define the `` excess energy '' of a cluster s gas to be the difference in total energy between that unmodified configuration and the actual configuration . depending on what is assumed about the unmodified configuration , these excess energies can range from @xmath304 for groups ( e.g. , lloyd - davies 2000 , 2002 ) to over 2 kev for clusters ( e.g. , wu 2000 ) . the results of this paper indicate that some combination of heating and cooling is responsible for similarity breaking because cooling is needed to explain the core entropies of clusters and groups and heating is needed to prevent overcooling at early times . therefore , we prefer to compute excess energy in a way that explicitly accounts for the separate contributions of cooling and heating . first , we assume that the baseline state of a cluster of mass @xmath184 is a truncated model from which cooling and condensation has removed a fraction @xmath305 of the lowest - entropy gas . in other words , the initial configuration of the intracluster medium has an entropy floor equal to @xmath306 and contains a fraction @xmath307 of the cluster s baryons . then , we assume that the final state of the cluster is a shifted model with the same proportion of baryons but an entropy floor equal to the cooling threshold , so that it obeys the observed @xmath246-@xmath7 and @xmath6-@xmath7 relations . formally , the entropy distribution of the final state is @xmath308 with an outer boundary where @xmath309 . because the amount of hot baryonic gas in both models is the same , the amount of heat input required to explain the cluster scaling relations after a fraction @xmath305 of the intracluster gas has condensed equals the energy input needed to convert the initial configuration to the final configuration , including work done at the outer boundary of the intracluster medium . figure [ xs_energy ] shows the energy difference between these configurations as a function of halo mass for different values of the condensed baryon fraction @xmath305 . to facilitate comparisons with other treatments of excess energy , we divide this energy difference by the total number of particles in an unmodified cluster ( @xmath310 ) , giving the quantity @xmath311 in units of kev per particle . when the condensed baryon fraction is small ( @xmath312 ) , the excess energy needed to account for the scaling relations of massive clusters ( @xmath196 ) approaches the @xmath313 level . however , this large amount of excess energy depends heavily on the assumed form of the unmodified entropy distribution , which is likely to be a poor approximation for these purposes . the large difference in the excess energy of massive clusters between models with @xmath312 and @xmath314 indicates that a large amount of energy is needed to lift the lowest - entropy gas out of the very center of the gravitational potential . yet , figure [ denprofs ] shows that the unmodified intracluster density distribution is unlikely to be as centrally condensed as we have assumed . even in the absence of cooling , the true unmodified profile probably corresponds more closely to a truncated model with @xmath315 , which substantially reduces the excess energy burden . thus , when calculating excess energies for clusters , one must make sure that the baseline model accurately represents the cluster configuration in the absence of heating . the curve with @xmath316 is likely to be a more realistic representation of the heat input needed to explain the scaling properties of massive clusters for two reasons : ( 1 ) this value of @xmath305 roughly corresponds to the fraction of cluster baryons that have condensed into stars , and ( 2 ) when @xmath317 , truncation eliminates the innermost part of the entropy distribution that is discrepant with simulations . at @xmath318 , this excess energy curve suggests that only a fraction of a kev per particle is needed to explain cluster scaling relations . that is because a shifted model with an entropy shift determined by the cooling threshold is virtually identical to a truncated model with the innermost @xmath305 of gas removed . likewise , the mass scales at which curves with larger values of @xmath305 intercept the zero - energy level also indicate cases in which the truncated and shifted models are virtually identical , requiring no excess heating to produce the observed scaling relations . in the limiting case of zero supernova heating ( e.g. , bryan 2000 , muanwong 2001 ) , these intercept points show the fraction of condensed baryons implied by the @xmath6-@xmath7 relation as a function of halo mass . if the condensed baryon fractions of groups are similar to those of clusters , then @xmath319 of excess energy is needed to explain the @xmath246-@xmath7 and @xmath6-@xmath7 relations at @xmath262 . the main difference between the initial configuration and the final configuration of groups with @xmath320 is that the gas in the final configuration is considerably more extended , with substantially more gravitational potential energy . because models with higher @xmath305 contain less intragroup gas , they are less extended , and less energy is needed to change their configuration . we note that the excess energy needs of groups are comparable to estimates of the supernova energy released over the course of the group s history ( e.g. , ponman 1999 , loewenstein 2000 ; pipino 2002 ) , which may mean that no other source of energy is needed to explain their scaling properties , if that energy efficiently heats the intragroup medium . the bottom line here is that calculations of the heat input needed to account for the scaling relations of clusters and groups depend critically on what is assumed to happen in the absence of heat input . most calculations of excess energy have assumed that the intracluster gas density obeys either a simple polytropic equation of state ( e.g. , wu 2000 ; loewenstein 2000 ) or a @xmath125-model density profile ( e.g. , bower 2001 ; lloyd - davies 2002 ) whose parameters are adjusted according to the amount of heat input . the modified - entropy models discussed in this paper allow for cooling and condensation of the lowest - entropy gas to occur before heat is added . figure [ xs_energy ] demonstrates that the amount of heating needed to explain the scaling relations depends strongly on the amount of cooling that occurs prior to heat input and that realistic amounts of `` precooling '' substantially reduce the required amount of non - gravitational heating . a realistic family of models for clusters in hydrostatic and convective equilibrium can be constructed using the navarro , frenk , & white ( 1997 ) density profile and some rudimentary prescriptions for how radiative cooling and feedback induced by that cooling modify the entropy distribution of intracluster gas . the models presume that the intracluster entropy distribution in the absence of cooling and feedback would be identical to that of gas that takes on the same density distribution as the dark matter as it settles into the cluster s potential well an assumption supported by numerical simulations . because the lowest - entropy gas within that distribution can cool within a hubble time , its entropy must somehow be modified . the threshold entropy in our prescriptions for entropy modification corresponds to the entropy at which the cooling time of intracluster gas equals the age of the universe . the prescriptions themselves include truncation of the entropy distribution at the threshold entropy , shifting of the entropy distribution by adding the threshold entropy to the entire distribution , and a qualitative implementation of radiative cooling . exploring the properties of dimensionless models depending only on the halo concentration @xmath42 , the dimensionless entropy threshold @xmath84 , and the prescription for entropy modification reveals that : * removal of low - entropy gas acts to flatten the core density profile , regardless of whether cooling or heating eliminates that low - entropy gas ( see figures [ denprofs ] and [ denprofs_all ] ) . * the observable properties of clusters depend more critically on the threshold entropy than on the mode of entropy modification . for example , truncation of the intracluster entropy distribution at a given value of @xmath84 and shifting of the entropy distribution by that same amount lead to very similar pressure , density , and temperature profiles within the cluster s virial radius ( see figures [ denprofs_all ] and [ tprofs ] ) . * within about 30% of the virial radius , the surface - brightness profiles of truncated and shifted models are very similar to @xmath125-models ( see figure [ sbright ] ) . raising the entropy threshold tends to lower the best - fitting value of @xmath125 , while raising the halo concentration tends to decrease the core radius ( see figures [ betafits_t ] and [ betafits_s ] ) . * all of our prescriptions for entropy modification tend to augment the luminosity - weighted temperature of a cluster as the entropy threshold rises . however , that temperature increase is relatively small because the increased central entropy decreases the luminosity of high - temperature gas in the core . thus , modifiying the entropy of a cluster shifts the bulk of a cluster s luminosity to larger radii , where gas temperatures tend to be smaller , mitigating the effects of the entropy increase ( see [ emodtemp ] ) . * all of our prescriptions for entropy modification reduce the x - ray luminosity of a cluster as the entropy threshold rises . in the high - threshold limit , dimensionless luminosity scales as @xmath149 , owing to the asymptotic @xmath321 scaling of the unmodified nfw density distribution at large radius ( see [ emodlum ] ) . when luminosity scales with the entropy threshold in this way , the @xmath6-@xmath7 relation should not flatten below @xmath322 , where line cooling dominates free - free emission , and the relation should also evolve little with time . the observed @xmath6-@xmath7 relation shares both of these features . furthermore , this scaling of luminosity with the entropy threshold leads to an @xmath6-@xmath7 relation slightly steeper than @xmath323 , with the extra steepening coming from the tendency for low - temperature halos to be more concentrated , also in agreement with observations . adopting relations between halo concentration @xmath42 and halo mass @xmath184 that are drawn from simulations and supported by observations enables us to generate models for real clusters that reproduce many of their observable properties : * fitting @xmath125-models to our modified - entropy clusters yields @xmath125 and core - radius values that are similar to those of observed clusters . our models reproduce the observed tendency for @xmath125 to be lower in low - temperature halos , a behavior that arises because the entropy threshold determined by cooling has a much larger impact on low - temperature halos . our models also reproduce the observed relationship between core radius and temperature , implying that @xmath324 , regardless of temperature . despite the flattening of the density profile in low - temperature halos , the core radius remains near @xmath325 because of the increased halo concentration ( see figures [ betafits_t ] and [ betafits_s ] ) . the large dispersion observed in @xmath125 and @xmath130 could arise from a number of effects , some involving observational systematics such as surface - brightness bias and the range of radii in the fit , others involving physical differences such as cluster - to - cluster variations in the amount of gas below the cooling threshold ( see figures [ bmods ] and [ bmods_all ] ) . * the temperature gradient at small radii in our model clusters depends sensitively on the amount of gas below the cooling threshold , closely related to the parameter @xmath120 in our radiative - loss models . in our most extreme radiative - loss model , with @xmath226 , the temperature gradient at @xmath326 is very similar to the `` universal '' temperature gradient observed by allen ( 2001 ) . as the parameter @xmath120 rises to infinity , the limit in which no gas lies below the cooling threshold , the inner temperature gradient flattens and then reverses , monotonically decreasing with radius in the large @xmath120 limit ( see figure [ tinner ] ) . * all of our models have a negative temperature gradient at large radii because the underlying potential is steeper than isothermal at those radii . in order to compare our models with observations , we compute the effective polytropic index @xmath327 as a function of radius . at the virial radius , we find @xmath328 , but the behavior of @xmath235 at smaller radii depends strongly on the nature of entropy modification . in models with no gas below the cooling threshold , the cluster core is nearly isentropic , with @xmath329 . in models with significant amounts of gas below that threshold , we find @xmath330 within the core radius , a consequence of the positive temperature gradient ( see figure [ gamma_profs ] ) . values of @xmath235 evaluated at @xmath245 agree with observations , within the large observational uncertainties ( see figure [ gammas ] ) . * the mass - temperature relations derived from our truncated and shifted modified - entropy models , which have no gas below the cooling threshold , agree well with those derived from cluster observations that have been corrected for the presence of cooling flows ( see figure [ mtrel ] ) . the slope of this relation is steeper than the @xmath331 expectation from self - similar scaling because low - temperature halos are more concentrated , leading to a slightly higher temperature for a given mass ( see figure [ tlums_ts ] ) , a feature these models share with those of komatsu & seljak ( 2001 ) . clusters of a given temperature are less massive than those in simulations without cooling and feedback because entropy modification owing to these processes shifts all of our model clusters to higher temperature . however , the agreement between observations and our models does not necessarily imply that this @xmath246-@xmath7 relation is without problems , as both the observational determinations and the models assume that the intracluster medium is an ideal gas in hydrostatic equilibrium , which may not be the case . furthermore , this relation does not apply to clusters that have not been corrected for cooling flows the temperature normalization at a given mass in our radiative - loss models can be up to 30% lower , depending on the value of @xmath120 ( see figures [ tlums_r ] and [ tlums_ta ] ) . * the luminosity - temperature relations derived from truncated and shifted models also agree well with those derived from cooling - flow corrected observations of clusters ( see figure [ ltrel ] ) . while there are no adjustable parameters in these models , the baryon - to - dark - matter ratio does depend on our assumed value of @xmath260 . thus , this agreement can be taken as additional evidence for a low value of @xmath186 . * gas below the cooling threshold both raises the x - ray luminosity and lowers the luminosity - weighted temperature of our model clusters . hence , the @xmath6-@xmath7 relations derived from our radiative - loss models lie above and to the left of the truncated and shifted models in the @xmath6-@xmath7 plane . changing @xmath120 leads to a displacement almost perpendicular to the @xmath6-@xmath7 relations themselves ( see figure [ lt_mark ] ) . correcting for cooling - flow emission leads to a very similar displacement in @xmath6-@xmath7 space , suggesting that the @xmath6-@xmath7 relation for uncorrected clusters may depend almost entirely on the two parameters @xmath184 and @xmath120 . it will be interesting to see whether values of @xmath120 derived from displacement in @xmath6-@xmath7 space agree with those implied by the inner temperature gradient . the modified - entropy models we have derived can account for many of the observed properties of present - day clusters , but they are obviously too simplistic to tell us how those clusters got to be the way they are . somehow , hierarchical structure formation generated intracluster entropy distributions similar to our simple prescriptions for entropy modification . cooling , feedback , and merging all have important roles to play . the consistency of our models with the data strongly suggests that radiative cooling sets the level of the intracluster entropy threshold and substantially reduces the amount of heat input needed to explain the observed scaling relations . however , radiative cooling unchecked by feedback would lead to overcooling of intrahalo gas at @xmath332 ( see [ overcool ] ) . at least some feedback is needed to keep @xmath276% of intergalactic baryons in gaseous form , as observed . mergers following an episode of feedback can potentially amplify the entropy generated by that feedback , when low - density gas associated with low - temperature halos is shocked . however , the overall impact of mergers on intracluster entropy remains unclear . determining just how cooling , mergers , and feedback conspire to produce the present - day entropy distribution of clusters will require high - resolution numerical simulations and semi - analytical modeling focusing on the intergalactic entropy distribution and how it evolves . we acknowledge megan donahue and don horner for helpful conversations and trevor ponman and stefano borgani for comments on the original manuscript . mlb is supported by a pparc rolling grant for extragalactic astronomy and cosmology at the university of durham .
massive galaxy clusters provide several means for independently examining any viable cosmological model . cluster samples used for this purpose are usually complete in terms of x - ray and redshift measurements such as the rdcs ( rosati et al . 1998 , borgani et al . 1999 ) and the macs survey ( ebeling , edge , & henry 2001 , ebeling et al . 2007 ) that we examine here with lensing . ideally , controlled samples of clusters with reliable lensing masses would allow the most direct comparison with theory but these are presently still in their infancy with only a handful of clusters discovered in wide - field weak lensing searches ( taylor et al . 2004 , wittman et al . 2006 , massey et al . 2007 , miyazaki et al . 2007 ) , and may be often subject to projection biases ( e.g. , hennawi & spergel 2005 ) . weak lensing surveys should have the advantage of being approximately volume - limited over the redshift range 0.1<@xmath3<1.0 , where the lensing distance ratio has a broad maximum . ongoing measurements of the sunyaev - zeldovich ( sz ) effect ( which has already been detected in many tens of clusters , including some at relatively high redshifts @xmath11 , e.g. , laroque et al . 2006 ) in a large number of clusters over a wide redshift range should improve upon x - ray selection , due to the redshift - independence of the effect ( though clearly high redshift clusters are relatively less well resolved ) . also , since the insensitivity of the effect to internal gas density variations will not be biased as with x - ray flux selection towards systems with luminous shocked gas , measurements should better correlate with cluster masses . strong gravitational lensing ( sl ) is nearly always seen in sufficiently detailed observations of massive clusters . einstein radii derived from such observations provide a reliable projected mass within the observed einstein radius , which depends only on fundamental constants , g and c , and knowledge of the lens and source distances . constraining the inner mass profile from sl is much harder , requiring many sets of multiple images covering a wide range of source redshift , to overcome inherent lensing degeneracy between the gradient of the mass profile and the scaling of the bend angle with source distance , so that sl based mass profiles have been usefully constrained for only several clusters producing reliable mass maps ( e.g. , abell 370 , kneib et al . 1993 , richard et al . 2009a ; abell 901 , deb et al . 2010 ; abell 1689 , broadhurst et al . 2005 , coe et al . 2010 ; abell 1703 , limousin et al . 2008 ; cl0024 + 1654 , liesenborgs et al . 2008 , zitrin et al . 2009b ; ms 2137.3 - 2353 , gavazzi et al . 2003 , merten et al . 2009 ; rxj1347 , brada et al . 2008a , halkola et al . 2008 ; sdss j1004 + 4112 , sharon et al . 2005 , liesenborgs et al . 2009 ; `` the bullet cluster '' , brada et al . 2006 ) . in quite a few of these clusters there is a significant inconsistency between the einstein radius directly measured from sl analyses and the einstein radius predicted from model profiles fitted to weak - lensing measurements of the same clusters , often subject to background selection and dilution problems over a wide radial range ( e.g. , medezinski et al . 2010 ) , though more so towards the sl regime where the degree of contamination by cluster members is often poorly corrected . the importance of a sl analysis of a significant sample is clear , especially in the statistical aspect , where the modelling method is similar for all examined clusters and thus supplies a coherent view for objective comparison ( e.g. , richard et al . 2009b ) . the development of sl modelling - methods has increased in response to the improvement in data quality since the discoveries of the first giant arcs ( e.g. , lynds & petrosian 1986 , soucail et al . 1987 , kneib et al . most methods can be classified as `` parametric '' if based on physical parameterisation , and as `` non - parametric '' if they are `` grid - based '' ( see also 4.4 in coe et al . 2008 , and references therein ) . still , many of these methods include too many parameters to be well - constrained by the number of initially known multiply - lensed systems . here we use acs imaging to identify new multiply - lensed systems in order to constrain the mass distributions and define the critical curves of the sample , motivated by the successful minimalistic approach of broadhurst et al . ( 2005 ) to lens modelling . in following work ( zitrin et al . 2009b ) we presented an improved modelling method which involved only 6 free parameters enabling easier constraint , since the number of constraints has to be equal or larger to the number of parameters in order to get a reliable fit . two of these parameters are primarily set to reasonable values so only 4 of these parameters have to be constrained initially , which sets a very reliable starting - point using obvious systems . the mass distribution is therefore primarily well constrained , uncovering many multiple - images which can be then iteratively incorporated into the model , by using their redshift estimation and location in the image - plane . the massive cluster survey ( macs ) has been aimed to create a complete sample of the very x - ray luminous clusters in the universe , successfully increasing the number of known such clusters to hundreds , particularly at @xmath12 ( ebeling , edge , & henry 2001 ) . from this a complete sample of 12 high-@xmath3 macs clusters ( @xmath0 ) was defined by ebeling et al . ( 2007 ) which have proved very interesting in several follow - up studies including deep x - ray , sz and hst imaging . detection of a large - scale filament has been reported in the case of macs j0717.5 + 3745 by ebeling et al . ( 2004 ) , for which many multiply - lensed images have been recently identified by zitrin et al . ( 2009a ) , revealing this object to be the largest known lens , with an einstein radius equivalent to 55@xmath13 ( for a source at @xmath14 ) . in the case of macs j1149.5 + 2223 ( zitrin & broadhurst 2009 ) , a background spiral galaxy at @xmath15 ( smith et al . 2009 ) has been shown to be multiply - lensed into several very large images , requiring a very shallow , unrelaxed central mass distribution . another large multiply - lensed sub - mm source at a redshift of @xmath16 has been identified in macs j0454.1 - 0300 ( also referred to as ms 0451.6 - 0305 ; takata et al . 2003 , borys et al . 2004 , berciano alba et al . 2007 , 2009 ) , macs j0025.4 - 1222 was found to be a `` bullet cluster''-like ( brada et al . 2008b ) , and other macs clusters have been recently used for an extensive arc statistics study ( horesh et al . 2010 ) . the x - ray data available for this sample ( see ebeling et al . 2007 ) along with the high - resolution hst / acs imaging and sz imaging ( e.g. , laroque et al . 2003 ) make these macs targets particularly useful for understanding the nature of the most massive clusters . here we complete a full sl analysis of this 12 cluster sample with the available deep hst / acs imaging , principally to derive their einstein radii and also to help constrain the central mass distributions . the effective einstein radii derived here are the corresponding radii of circles of equivalent areas to those enclosed within the critical curves , or similarly , the radii within which the averaged @xmath17 is equal to 1 . in addition , this work can also supply detailed magnification maps to help motivate deeper searches for high-@xmath3 galaxies behind these clusters . another major motivation for pursuing accurate lensing - maps is the increased precision of model predictions for cluster- size massive halos in the standard @xmath9cdm model and close variance ( see umetsu & broadhurst 2008 , e.g. , bullock et al . 2001 , spergel et al . 2003 , 2007 , tegmark et al . 2004 , hennawi et al . 2007 , neto et al . 2007 , duffy et al . 2008 , keselman , nusser & peebles 2009 , meneghetti et al . 2010 , fedeli et al . 2010 ) . in the standard hierarchical model , large massive bodies are due to collapse recently and thus are not expected to be found in large numbers at high redshifts . on the other hand , the larger volume available with increasing distance means that in practice we can not expect to reside next to the most massive cluster ( see also zitrin et al . the increasing number of accurately analysed clusters have revealed larger lenses than predicted by the standard @xmath18 model ( broadhurst & barkana 2008 , puchwein & hilbert 2009 ) . this discrepancy is empirically supported by the surprisingly concentrated mass profiles measured for such clusters , when combining the inner strong lensing with the outer weak lensing signal ( gavazzi et al . 2003 , broadhurst et al . 2005 & 2008 , limousin et al . 2008 , donnarumma et al . 2009 , oguri et al . 2009 , umetsu et al . 2009 , zitrin et al . 2009b,2010 ) , or independently , when using the internal dynamics of cluster members ( lemze et al . 2009 ) and deep x - ray data ( lemze et al . 2008 ) . geometrically , lensing is considered to be optimised at intermediate redshifts , where for a given mass the critical density for lensing is minimal , but this is partly offset by the late hierarchical growth of high - mass systems . this trade - off results in estimates of the amplitude of strong lensing to favour the redshift range @xmath19 . our recent analysis of two x - ray selected macs clusters reveals very large lenses ( e.g. , macs j1149.5 + 2223 , zitrin & broadhurst 2009 ; macs 0717.5 + 3745 , zitrin et al . 2009a ) at high redshift , increasing the tension with @xmath9cdm predictions , since large einstein radius clusters at high-@xmath3 require earlier formation than implied by the standard @xmath9cdm model ( sadeh & rephaeli 2008 ) . the largest lensing clusters have proven to be excellent targets for accessing the faint early universe due to their large magnification consistently providing the highest redshift galaxies ( ebbels et al . 1996 , franx et al . 1997 , frye & broadhurst 1998 , bouwens et al . 2004 , kneib et al . 2004 , bradley et al . 2008 , zheng et al . 2009 ) . the statistics of large einstein radii provide an important opportunity to test the standard @xmath9cdm paradigm , as it probes both the high - mass end of the cluster mass function and central mass distributions of massive clusters ( oguri & blandford 2009 ) . increasing numbers of theoretical and observational works have been done recently , deriving einstein radius distributions of various samples . for example , cypriano et al . ( 2003 ) have derived the mass distributions of 24 x - ray selected abell clusters via weak - lensing . by fitting to sis profiles they obtain a mean einstein radius of @xmath20 . more recently , hoekstra ( 2007 ) finds by weak - lensing analysis and sis fitting a mean einstein radius of @xmath21 for a sample of 20 x - ray luminous clusters . okabe et al . ( 2009 ) find the mass distribution and einstein radii for a sample of @xmath22 locuss clusters by fitting the weak - lensing data to cis profiles , obtaining a mean value of @xmath23 ( see okabe et al . 2009 and references therein ) . note that these values quoted here are calculated from the corresponding tables in these papers , and are not necessarily scaled or normalised to certain lens and source redshifts . other recent work more relevant to our study summarises sl analysis for a sample of 20 , mostly relaxed _ undisturbed _ clusters ( richard et al . they find that the einstein radii are distributed log - normally with a peak at @xmath24 and a corresponding @xmath25 enclosed mass ( within r@xmath26250 kpc ) , and show that the predicted distribution of einstein radii from @xmath9cdm cosmology falls short of the observed einstein radii by a factor of 2 . it has been claimed that much larger einstein radii can be contemplated only with mass distributions which are highly prolate and aligned along the line of sight ( corless & king 2007 , oguri & blandford 2009 , see also hennawi et al . 2007 , meneghetti et al . 2007 , sereno , jetzer & lubini 2010 ) . analysing the 12 x - ray selected , high-@xmath3 macs clusters is important since they are predicted to be massive and should form efficient lenses , whose properties can be then compared to such studies , playing an important role in probing the @xmath9cdm scenario . the paper is organised as follows : in 2 we describe the sample and observations ; the lensing analysis is described in 3 detailing each cluster separately . our overall results are presented and discussed in 4 , followed by a conclusion ( 5 ) . throughout the paper we adopt the standard cosmology ( @xmath27 , @xmath28,@xmath29 ) . the @xmath0 macs clusters have been imaged with the wide field channel ( wfc ) of the acs installed on hst , mainly in the framework of proposal i d 9722 ( pi : ebeling ) . this mainly comprises two - filter observations in the f555w and f814w bands , taken during 2003 and 2004 with typical exposure times of @xmath30s in each filter . we make use of some other hst data available in the hubble legacy archive ( hla ) , which we list in table [ sample ] . in addition , available sextractor ( bertin & arnouts 1996 ) photometry catalogues were also downloaded from the hla , and were used to construct colour - magnitude diagrams for identifying the red cluster sequence galaxies belonging to each cluster , as a starting point for each lensing model , as detailed in [ model ] . [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] the 12 clusters have effective einstein radii in the range , 11@xmath13 to 55@xmath13 , with a median value of @xmath31 , ( and a mean value of @xmath32 ) assuming a fixed source redshift of @xmath33 , see figure [ hist ] . this corresponds to a range of physical radii , 68 kpc to 353 kpc , with a median value of 180 kpc , when transforming to the redshift of each lens . in the hierarchical model large - scale perturbations collapse recently and thus should be found relatively locally . this sample includes very large einstein radii , exceeding even the most impressive lenses previously studied at lower redshifts as analysed by broadhurst & barkana ( 2008 ) . more recently , richard et al . ( 2009b ) carefully studied the lensing properties of 20 mainly undisturbed clusters at lower-@xmath3 , selected to have sl features in hubble data and measured x - ray data , for which their mean einstein radius is @xmath24 , a factor of @xmath34two smaller than our sample . this difference reflects the larger masses of the macs sub - sample that we studied , which is purely x - ray flux selected and restricted to @xmath0 . this macs sample therefore does not suffer from a lensing - selection bias , but the effect of projection bias must be taken into account when evaluating 2d lensing observations with 3d mass profiles predicted by theory ( e.g. , hennawi et al . 2007 , oguri & blandford 2009 , sereno , jetzer & lubini 2010 ) , as discussed in [ cdm ] below . to @xmath35 . the solid curve is the expected distribution of einstein radii as calculate in [ cdm ] for the @xmath9cdm model , incorporating both the scatter on the @xmath36 relation and the projection bias from lensing . we multiply the resulting theoretical @xmath37 curve by the width of the bins to normalise it . the observed distribution is skewed to larger einstein radii than predicted . the median values of both the observed and the theoretical distributions are plotted on the histogram in a blue dashed line , and a black dash - dotted line , respectively.,width=302 ] cdm model ( black solid curve ) described in [ cdm].the median values of both the observed and the theoretical cumulative distributions are plotted on the histogram in a blue dashed line , and a black dash - dotted line , respectively.,width=302 ] , is tightly followed by the data , indicating that in general the central mass distributions of these clusters are well behaved.,width=302 ] with the system scale size is apparent . , width=302 ] most of these clusters have resisted analysis by strong lensing despite the long availability of the hubble data . our success in `` cracking '' these irregular lenses and defining their critical curves is encouraging for the application of our approach to complex unrelaxed clusters more generally . for such clusters , where the deflection fields are not symmetric , the identification of multiple images must be aided by models that allow for flexibility in describing the mass distribution , as described in section [ model ] , free of the strictures imposed by the use of idealised elliptical potentials in conventional models . the masses enclosed by the critical curves range from @xmath38 to @xmath39 with a median value of @xmath40 , and a mean value of @xmath41 as shown in table [ results_table ] . these are calculated for each cluster by integrating the surface mass density distribution within the 2d model critical curve , scaled to a source redshift of @xmath42 . we examine the relation between these einstein masses , @xmath43 enclosed within the critical curves , and the effective einstein radii derived above . theoretically this should of course simply scale as @xmath44 for symmetric lenses . here the lenses are not symmetric , but as can be seen in figure [ reme ] the quadratic relation is tightly followed , indicating that asymmetry is not very significant in terms of the mass distribution , and furthermore the measured or assumed redshifts of @xmath45 for the relevant systems is a reliable estimate . in fact , we see clearly in that the 2d mass distributions are in general noticeably rounder than the critical curves which are very sensitive to substructure , as can be seen by comparing the 2d mass distributions with the critical curves , shown in figures [ curves0018 ] - [ contours2214 ] . further encouragement for the accuracy of sl mass estimates is presented in recent work by meneghetti et al . ( 2010 ) , showing that sl methods based on parametric modelling are accurate at the level of few percent at predicting the projected inner mass . richard et al . ( 2009b ) found a mean mass of @xmath25 enclosed mass ( within r@xmath26250 kpc ) for their sample of 20 clusters , while smith et al . ( 2005 ) examined the mass distribution of 10 x - ray selected clusters , and found a similar mean mass of @xmath46 ( see also richard et al . these values are in good agreement with our sample median mass ( @xmath40 ) , and slightly lower than our sample mean mass which is boosted by few extremely massive clusters ( e.g. , macs j0717.5 + 3745 ) . we examine the behaviour of the central mass - to - light ratio , @xmath47 , with the einstein radius and shown in figure [ ml ] . the luminosity is summed over all cluster sequence galaxies identified as described in section [ model ] . a clear correlation is obtained such that @xmath47 increases with the einstein radius , and with values quite typical of other massive clusters ( sarazin 1988 , rines et al . 2004 , medezinski et al . this scaling towards higher @xmath48 provides confidence in our method . at the high mass end our result is similar to the peak m / l found in the detailed weak lensing profile studies of medezinski et al . measuring @xmath47 in high-@xmath3 clusters is of interest also for the added insight on the amount of dm initially associated with individual galaxies as compared with the overall cluster dm component ( see also sarazin 1988 ) . as can be seen in table [ results_table ] , most of the sample clusters provide large regions ( @xmath49 ) of highly magnified ( @xmath50 ) sky , useful for the search of the first stars and galaxies ( e.g. , zackrisson et al . one has to be cautious when making such a statement since unlike the critical curves and the enclosed mass , the magnification is sensitive to the mass profile which requires source redshift estimation for several sources , which we have here for only a few clusters ( section [ model ] ) . still , as explained in 2 , reasonable constraints are put on the mass profile slope by the image - plane minimisation , which generally has a broad minimum at the same point in the parameter space where the slope is also approximately correct . in addition , we are able to put a lower boundary on the high magnification area , by choosing the steeper profiles which naturally generate lower magnifications , or by assuming a boosted redshift for the main multiply - lensed system decreasing the lensing distance ratio for high-@xmath3 galaxies . thus we constrain the area of high magnification for a high-@xmath3 source at the current limit of @xmath8 , showing that most of this sample clusters are excellent targets for high-@xmath3 galaxy searches due to these large high - magnification areas and lens power . we estimate , based on previously analysed clusters ( e.g. , cl 0024 + 1654 , zitrin et al . 2009b ; macs j1149.5 + 2223 , zitrin & broadhurst 2009 ; macs j0717.5 + 3745 , zitrin et al . 2009a ) that in practice the areas of high magnification are about 10 - 20% higher than the lower limits presented in table [ results_table ] , with an upper limit of about 40 - 50% . we generally follow broadhurst & barkana ( 2008 ; see also sadeh & rephaeli 2008 ) in constructing the theoretically predicted distribution of einstein radii . in particular , we adopt the nfw parameters measured by neto et al . ( 2007 ) for simulated halos , although we reduce @xmath51 by @xmath52 according to the results of duffy et al . ( 2008 ) who used the more recently measured values of the cosmological parameters . studies based on large numerical simulations ( zhao et al . 2003,2009 , gao et al . 2008 ) have found for massive halos a rather weak decline of @xmath51 with increasing redshift ; we neglect this decline , which renders our results ( in terms of the disagreement with observations ) conservative . we then correct the concentration parameter distribution using hennawi et al . ( 2007 ) to obtain the effective parameters for the population of clusters observed in projection . note though that here we are considering a sample selected by x - ray flux ( not lensing cross - section ) , so there is no lensing bias , but there still is the projection bias ; this refers to the fact that when halos are seen in projection ( at randomly distributed angles ) , a different effective projected @xmath51 is measured than would be obtained from a spherical analysis of the halo density profile . in particular , projection increases the mean @xmath51 and adds substantially to the scatter compared to the distribution of the spherically - measured @xmath51 . neto et al . ( 2007 ) split their halo sample into two groups ( `` relaxed '' and `` unrelaxed '' ) and analysed the statistics of each group separately , so we do the same , and in the end combine the two groups according to their relative numbers in the simulation by neto et al . ( 2007 ) , obtaining our predicted results for the total , combined population of halos . this can then be compared with the observed sample , which does not select for relaxed clusters but includes both relaxed and unrelaxed ones . in order to approximately simulate a flux - selected sample , we need to convert @xmath53 to halo mass . we use the power - law relation from reiprich & bhringer ( 1999 ; with an updated hubble constant ) , along with the observed scatter of @xmath53 for a given mass , which we model as a lognormal distribution with a typical scatter that corresponds to a factor of 1.5 . note that this observed relation is based on @xmath54 , while the results from the above simulations refer to @xmath55 , so we must include the relation between @xmath54 and @xmath55 which is itself a function of @xmath51 . an important caveat is that the observed @xmath56 relation was measured for relaxed clusters ( whose mass could be estimated from the x - ray emission assuming hydrostatic equilibrium ) . we expect that unrelaxed clusters would tend to have an unusually high @xmath53 for a given mass ( e.g. , in a post - merger phase ) , which would insert more low - mass clusters within the sample than we assume , leading to smaller predicted einstein angles , increasing the discrepancy between the theory and the observations ( again making our results conservative ) . on the other hand , it should be noted that much of the scatter in the @xmath56 relation probably arises from cool cores in the center ( since it is significantly reduced if the inner core is excised ) , implying a possible bias towards relaxed and more concentrated systems . however , such a bias seems less probable in our case due to the high x - ray luminosity , the relatively high redshift , and the clear spread - out appearance of most of the clusters analysed here . the observed properties of each halo ( flux , einstein radius distribution , etc . ) change slowly with redshift , so to simplify the calculation we approximate the clusters as all lying at a typical redshift @xmath57 . however , the halo abundance changes rapidly with redshift , so we calculate the integrated halo mass function for all halos observed at @xmath0 , in the fraction of the sky corresponding to the macs survey ( around a quarter of the sky ) . we use the sheth & tormen ( 1999 ) formula , which accurately fits the halo mass function ( i.e. , the distribution of @xmath55 ) measured in simulations . this halo mass function was convolved with with the @xmath51 distribution to obtain the distribution of @xmath54 , and then add the @xmath58 scatter in order to obtain the halo mass function corresponding to selection by a given minimum flux . since the @xmath58 relation is not strictly valid for unrelaxed clusters , we only use the power - law and the scatter from this relation but allow some flexibility in the normalisation , choosing an effective minimum flux that corresponds to a predicted abundance of 12 clusters ( as in the observed sample ) . this is reasonable also since we are interested here in testing the predicted @xmath59 distribution , not in testing the predicted cluster abundance . note that the flux selection does not affect the distribution of the very highest @xmath59 values , since all clusters within the survey solid angle at @xmath0 that can produce the largest @xmath59 s will be well above the flux threshold . given our predicted , x - ray selected mass function in terms of @xmath54 , we must convolve it with the @xmath59 distribution of each halo of mass @xmath54 . from the simulations we have the @xmath51 distribution ( and thus also the @xmath59 and @xmath54 distribution ) for a given @xmath55 , but we can invert this conditional probability using bayes theorem . note that the @xmath55 to @xmath54 conversion should not really include the projection bias that is included in @xmath59 , but having two different @xmath51 distributions for each @xmath55 would makes things far more complicated . our simplification of using a single @xmath51 distribution is reasonable since the projection bias typically affects @xmath54 ( for a given @xmath55 ) only by @xmath60 . also , since @xmath51 affects both @xmath54 and @xmath59 in the same direction , the effect on our final result ( which derives @xmath59 for a given @xmath54 ) is much smaller . our predicted @xmath59 distribution is compared to the observations in figure [ histr ] . we compare the two cumulative distributions @xmath61 , i.e. , the total number of clusters expected above @xmath59 , where the theoretical distribution has been normalised ( as noted above ) to agree with the total of 12 clusters in the sample . the observed distribution clearly lies at higher einstein radii compared with the predicted distribution . one indication of this is the most discrepant cluster , j0717.5 + 3745 with @xmath62 . the theoretical calculation yields a probability of only @xmath63 of finding such a large @xmath59 in this @xmath0 cluster sample , giving a significant ( though not extreme ) 2-@xmath64 discrepancy . more generally , the observed distribution resembles the predicted one except offset towards higher angles by about a factor of 1.4 ( comparing , for example , the two medians ) . the standard k - s statistic for comparing two distributions gives a probability of @xmath65 that the observed distribution is drawn from the predicted one . this is also around a 2-@xmath64 discrepancy , largely independent of the above number since the k - s statistic focuses on the middle portion of the probability distributions and is insensitive to the edges . the relatively low significance of the k - s discrepancy is an inevitable result of the poisson fluctuations expected with a sample of only 12 clusters . the discrepancy that we find , of einstein radii that are too high by about a factor of 1.4 , appears smaller than the roughly factor of 2 difference found by broadhurst & barkana ( 2008 ) . this may be explained at least partially by the difference between a lensing selected and an x - ray selected sample . a more robust comparison would be possible with an x - ray selected cluster sample in which the clusters have reliable virial masses measured using weak lensing . another caveat is that we have compared the observed einstein radii with the predictions based on pure dark matter simulations . the possible effect of baryons on the halo density profile has been difficult to study with simulations , since hydrodynamic simulations that produce a large effect do so along with a central baryon concentration that disagrees with observations and is due to the `` overcooling '' problem in simulations . analytical models suggest that even without producing a central baryon concentration , the complex coupled history of baryon and dark matter accretion may allow the baryons to significantly affect the final central density profile , potentially alleviating the discrepancy at least partially ( barkana & loeb 2010 ) . on the other hand , recent simulations suggest that the discrepancy actually increases when agn feedback is taken into account to overcome the overcooling problem ( duffy et al . 2010 ) . in this work we determine the einstein radii and projected masses for the 12 clusters of the @xmath0 macs sample , and compare these to simulations . we make use of our well - established modelling method to identify many sets of multiply - lensed images which are in turn used to constrain the models , and determine the einstein radii and projected masses for @xmath66 . typical uncertainties in the source redshift estimates of @xmath67 impose only minor uncertainties on the comparison with @xmath9cdm . firstly , such an uncertainty does not enable a unique determination of the profile slope , which we do not attempt to fully constrain here since both the einstein radius and the projected mass are indifferent to the mass profile slope and are determined by the data . secondly , overestimating a source redshift would in practice increase the projected mass and the observed einstein radius for a source at @xmath68 , while maintaining the relation seen in figure [ reme ] and thus resulting only in growth of the discrepancy from @xmath9cdm simulations which are not dependent on the observed mass . on the other hand , underestimating a source redshift would in practice decrease the observed einstein radius and projected mass for a source at @xmath68 by less than 10% , thus insignificantly influencing the results . the macs @xmath0 sample has proven to be of great value for many very different studies spanning the x - ray through radio spectrum . we have extracted additional useful information from this sample by solving the strong lensing for these clusters . previous discoveries from this sample include the largest lensed images of a highly - magnified distant spiral galaxy ( macs j1149.5 + 2223 ; zitrin & broadhurst 2009 , smith et al . 2009 ) , the largest known lens ( macs j0717.5 + 3745 ; zitrin et al . 2009a ) , lensed sub - mm sources ( takata et al . 2003 , borys et al . 2004 , berciano alba et al . 2007 , 2009 ) , and another `` bullet cluster '' ( macs j0025.4 - 1222 ; brada et al . 2008b ) . we presented mass models and the corresponding critical curves for the 12 high-@xmath3 x - ray luminous clusters of the sample via strong - lensing analysis in hst / acs images . several of these clusters have only a few strongly - lensed images and corresponding low masses and einstein radii , while most are very massive and rich with multiply - lensed arcs due to large critical area and high lens power . the equivalent einstein radii of this sample range from 11@xmath13 to 55@xmath13 with a median value of @xmath1 ( similar to the mean value ) , or 180 kpc . the corresponding masses enclosed within these curves range from @xmath38 to @xmath39 with a median value of @xmath40 , and a mean value of @xmath41 . we find that the enclosed mass follows tightly the quadratic relation with the equivalent einstein radius , indicating that the deviation from symmetry is not prominent in these clusters and that our measurements are highly accurate given the measured or assumed redshifts . in addition , we have shown that @xmath47 increases as expected with the scale size of the system , or einstein radius . we compare these results to the predictions of @xmath9cdm taking into account projection biases and find that these predictions fall short of our measured einstein radii by a factor of @xmath10 . in addition , the standard k - s statistic gives a probability of only @xmath65 that the observed distribution is drawn from the predicted one , corresponding to a @xmath69 discrepancy . it is apparent that the observed einstein radii and implied high masses are not likely at high redshifts in the context of the @xmath9cdm model when no lensing - selection biases are involved , showing again possible tension with the prediction of the standard model . finally , as a result of their unrelaxed mass distributions , most of these clusters cover a very large area ( @xmath70 ) of high magnification ( @xmath50 ) making them primary targets for high-@xmath3 galaxy searches , for which substantial hst time will shortly be forth - coming . it should be acknowledged that theoretical predictions may clearly entail further uncertainties than those taken into account here , and a revised and more thorough analysis is needed in order to obtain higher significance results once source redshifts are available . such comparisons should also be independently applied to other large samples , so that a significant number of clusters are compared to the @xmath9cdm model in order to statistically support any claimed discrepancy . we thank the anonymous reviewer of this work for very useful comments . az acknowledges eran ofek and salman rogers for their publicly available matlab scripts used in part of this work . az thanks sharon sadeh for useful discussions , and elinor medezinski for sharing some subaru data . this research is being supported by the israel science foundation ( isf ) grant 1400/10 . rb is supported by the isf grant 823/09 . acs was developed under nasa contract nas 5 - 32865 . this research is based on observations provided in the hubble legacy archive which is a collaboration between the space telescope science institute ( stsci / nasa ) , the space telescope european coordinating facility ( st - ecf / esa ) and the canadian astronomy data centre ( cadc / nrc / csa ) . part of this work is based on data collected at the subaru telescope , which is operated by the national astronomical society of japan . barkana , r. & loeb , a. , 2010 , mnras , in press bentez , n. , et al . , 2009 , apj , 691 , 241 berciano alba , a. , garrett , m.a . , koopmans , l.v.e . , wucknitz , o. , 2007 , a&a , 462 , 903 berciano alba , a. , koopmans , l.v.e . , garrett , m.a . , wucknitz , o. , limousin , m. , 2009 , arxiv:0910.1213 bertin , e. & arnouts , s. , 1996 , a&as , 117 , 393 borgani , s. , rosati , p. , tozzi , p. , norman , c. , 1999 , apj , 517 , 40 borys , c. , et al . , 2004 , mnras , 352 , 759 bouwens , r.j . , et al . , 2004 , apj , 616 , 79 brada , m. , et al . , 2008a , apj , 681 , 187 brada , m. , allen , s.w . , treu , t. , ebeling , h. , massey , r. , morris , r.g . , von der linden , a. , applegate , d. , 2008b , apj , 687 , 959 brada , m. , et al . , 2006 , apj , 652 , 937 bradley , l.d . , et al . , 2008 , apj , 678 , 647 broadhurst , t. , et al . , 2005 , apj , 621 , 53 broadhurst , t.j . & barkana r. , 2008 , mnras , 390 , 1647 broadhurst , t , umetsu , k , medezinski , e. , oguri , m . , rephaeli , y. , 2008 , apj 685 , l9 bullock , j.s . , kolatt , t.s . , sigad , y. , somerville , r.s . , kravtsov , a.v . , klypin , a.a . , primack , j.r . , dekel , a. 2001 , mnras , 321 , 559 clowe , d. , luppino , g.a . , kaiser , n. , gioia , i.m . , 2000 , apj , 539 , 540 clowe , d. , luppino , g.a . , kaiser , n. , 2003 , a&a , 409 , 851 coe , d. , fuselier , e. , bentez , n. , broadhurst , t. , frye , b. , ford , h. , 2008 , apj , 681 , 814 coe , d. , bentez , n. , broadhurst , t. , moustakas , l. , ford , h. , 2010 , arxiv:1005.0398 corless , v.l . , king , l.j . , 2007 , mnras , 380 , 149 deb , s. , goldberg , d.m . , heymans , c. , morandi , a. , 2010 , arxiv:0912.4260 donnarumma , a. , ettori , s. , meneghetti , m. , moscardini , l. , 2009 , mnras , 398 , 438 dressler , a. , gunn , j.e . , 1992 , apjs , 78 , 1 dressler , a. , smail , i. , poggianti , b.m . , butcher , h. , couch , w.j . , ellis , r.s . , oemler , a.jr . , 1999 , apjs , 122 , 51 duffy , a.r . , schaye , j. , kay , s.t . , & dalla vecchia , c. 2008 , mnras , 390 , 64 duffy , a.r . , schaye , j. , kay , s.t . , dalla vecchia , c. , battye , r.a . , booth , c.m . , 2010 , arxiv:1001.3447 ebbels , t.m.d . , le borgne , j .- f . , pello , r. , ellis , r.s . , kneib , j .- p . , smail , i. , sanahuja , b. , 1996 , mnras , 281 , 75 ebeling , h. , edge , a.c . , henry , j.p . , 2001 , apj , 553 , 668 ebeling , h. , barrett , e. , donovan , d. , 2004 , apj , 609l , 49 ebeling , h. , barrett , e. , donovan , d. , ma , c.j . , edge , a.c . , van speybroeck , l. , 2007 , apj , 661l , 33 ellingson , e. , yee , h.k.c . , abraham , r.g . , morris , s.l . , carlberg , r.g . , 1998 , apjs , 116 , 247 ellis , s.c . , jones , l.r . , 2002 , mnras , 330 , 631 fedeli , c. , meneghetti , m. , gottloeber , s. , yepes , g. , 2010 , arxiv:1007.1551 franx , m. , illingworth , g.d . , kelson , d.d . , van dokkum , p.g . & tran , k. , 1997 , apj , 486 , 75 frye , b. & broadhurst , t. , 1998 , apj , 499 , 115 gao , l. , navarro , j.f . , cole , s. , frenk , c.s . , white , s.d.m . , springel , v. , jenkins , a. , neto , a.f . , 2008 , mnras , 387 , 536 gavazzi , r. , fort , b. , mellier , y. , pello , r. , dantel - fort , m. , 2003 , a&a , 403 , 11 geach , j.e . , et al . , 2006 , apj , 649 , 661 halkola , a. , hildebrandt , h. , schrabback , t. , lombardi , m. , brada , m. , erben , t. , schneider , p. , wuttke , d. , 2008 , a&a , 481 , 65 hennawi , j.f . & spergel , d.n . , 2005 , apj , 624 , 59 hennawi j.f . , dalal n. , bode p. , ostriker j.p . , 2007 , apj , 654 , 714 hoekstra , h. , 2007 , mnras , 379 , 317 horesh , a. , maoz , d. , ebeling , h. , seidel , g. , bartelmann , m. , 2010 , mnras in press , arxiv:1004.2048 jones , t. , swinbank , m. , ellis , r. , richard , j. , stark , d. , 2009 , arxiv:0910.4488 keselman , j.a . , nusser , a. , peebles , p.j.e . , 2009 , arxiv:0912.4177 kneib , j.p . , mellier , y. , fort , b. , mathez , g. , 1993 , a&a , 273 , 367 kneib , j .- , ellis , r.s . , smail , i. , couch , w.j . , sharples , r.m . , 1996 , apj , 471 , 643 kneib , j .- p . , ellis , r.s . , santos , m.r . , richard , j. , 2004 , apj , 607 , 697 laroque , s.j . , et al . , 2003 , apj , 583 , 559 laroque , s.j . , bonamente , m. , carlstrom , j.e . , joy , m.k . , nagai , d. reese , e.d . , dawson , k.s . , 2006 , apj , 652 , 917 lavery , r.j . , 1996 , aj , 112 , 1812 lemze , d. , barkana , r. , broadhurst , t. , rephaeli , y. , 2008 , mnras , 386 , 1092 lemze , d. , broadhurst , t. , rephaeli , y. , barkana , r. , umetsu , k. , 2009 , apj , 701 , 1336 liesenborgs , j. , de rijcke , s. , dejonghe , h. , 2006 , mnras , 367 , 1209 liesenborgs , j. , de rijcke , s. , dejonghe , h. , bekaert , p. , 2008 , mnras , 389 , 415 liesenborgs , j. , de rijcke , s. , dejonghe , h. , bekaert , p. , 2009 , mnras , 397 , 341 limousin , m. , et al . , 2008 , a&a , 489 , 23 limousin , m. , et al . , 2009 , arxiv:0911.4125 luppino , g.a . , gioia , i.m . , hammer , f. , le fevre , o. , annis , j.a . , 1999 , a&as , 136 , 117 lynds , r. & petrosian , v. , 1986 , baas , 18r , 1014 massey , r. , et al . , 2007 , apjs , 172 , 239 medezinski , e. , broadhurst , t. , umetsu , k. , oguri , m. , rephaeli , y. , bentez , n. , 2010 , mnras , 405 , 257 meneghetti , m. , argazzi , r. , pace , f. , moscardini , l. , dolag , k. , bartelmann , m. , li , g. , oguri , m. , 2007 , a&a , 461 , 25 meneghetti , m. , rasia , e. , merten , j. , bellagamba , f. , ettori , s. , mazzotta , p. , dolag , k. , 2010 , arxiv:0912.1343 meneghetti , m. , fedeli , c. , pace , f. , gottloeber , s. , yepes , g. , 2010 , arxiv:1003.4544 merten , j. , cacciato , m. , meneghetti , m. , mignone , c. , bartelmann , m. , 2009 , a&a , 500 , 681 miyazaki , s. , hamana , t. , ellis , r.s . , kashikawa , n. , massey , r.j . , taylor , j. , refregier , a. , 2007 , apj , 669 , 714 molnar , s.m . , hughes , j.p . , donahue , m. , joy , m. , 2002 , apj , 573l , 91 neto , a.f . , et al . , 2007 , mnras , 381 , 1450 oguri , m. , blandford , r.d . , 2009 , mnras , 392 , 930 oguri , m. , et al . , 2009 , apj , 699 , 1038 okabe , n. , takada , m. , umetsu , k. , futamase , t. , smith , g.p . , 2009 , arxiv:0903.1103o puchwein , e. & hilbert , s. , 2009 , 2009 , mnras , 398 , 1298 reiprich , t.h . & bhringer , h. , 1999 , an , 320 , 296 richard , j. , kneib j.p . , limousin , m. , edge . , a. , jullo , e. , 2009a , arxiv:0910.5553 richard , j. , et al . , 2009b , arxiv:0911.3302 ricker , p.m. & sarazin , c.l . , 2001 , apj , 561 , 621 rines , k. , geller , m.j , diaferio , a. , kurtz , m.j . , jarrett , t.h . , 2004 , aj , 128 , 1078 rosati , p. , della ceca , r. , norman , c. , giacconi , r. , 1998 , apj , 492l , 21 sadeh , s. & rephaeli , y. , 2008 , mnras , 388 , 1759 sarazin , c.l . , 1988 , x - ray emission from clusters of galaxies , ed . sarazin sereno , m. , jetzer , ph . , lubini , m. , 2010 , arxiv:1001.1696 sharon , k. , et al . , 2005 , apj , 629 , 73 sheth , r.k . & tormen , g. , 1999 , mnras , 308 , 119 soucail , g. , fort , b. , mellier , y. , picat , j. p. , 1987 , a&a , 172 , 14 smail , i. , ellis , r.s . , fitchett , m.j . , edge , a.c . , 1995 , mnras , 273 , 277 smail , i. , ellis , r.s . , dressler , a. , couch , w.j . , oemler , a.jr . , sharples , r.m . , butcher , h. , 1997 , apj , 479 , 70 smith , g.p . , kneib , j .- , smail , i. , mazzotta , p. , ebeling , h. , czoske , o. , 2005 , mnras , 359 , 417 smith , g.p . , et al . , 2009 , arxiv:0911.2003 spergel , d.n . , et al . , 2003 , apjs , 148 , 175 spergel d.n . , et al . , 2007 , apjs , 170 , 377 tegmark , m. , et al . , 2004 , apj , 606 , 702 takata , t. , et al . , 2003 , pasj , 55 , 789 tanaka , m. , hoshi , t. , kodama , t. , kashikawa , n. , 2007 , mnras , 379 , 1546 taylor , a. n. , 2004 , mnras , 353 , 1176 umetsu , k. & broadhurst , t. , 2008 , apj , 684 , 177 umetsu , k. , medezinski , e. , broadhurst , t. , zitrin , a. , okabe , n. , hsieh , b.c . , molnar , s.m . , 2009 , arxiv:0908.0069 williams , l.l.r . , navarro , j.f . , bartelmann , m. , 1999 , apj , 527 , 535 wittman , d. , dellantonio , i.p . , hughes , j.p . , margoniner , v.e . , tyson , j.a . , cohen , j.g . , norman , d. , 2006 , apj , 643 , 128 worrall , d.m . , birkinshaw , m. , 2003 , mnras , 340 , 1261 zackrisson , e. , et al . , 2010 , arxiv:1002.3368 zhao , d.h . , jing , y.p . , mo , h.j . , brner , g. , 2003 , apj , 597 , l9 zhao , d.h . , jing , y.p . , mo , h.j . , brner , g. , 2009 , apj , 707 , 354 zheng , w. , et al . , 2009 , apj , 697 , 1907 zitrin , a. & broadhurst , t. , 2009 , apj , 703l , 132 zitrin , a. , broadhurst , t. , rephaeli , y. , sadeh , s. , 2009a , apj , 707l , 102 zitrin , a. , et al . , 2009b , mnras , 396 , 1985 zitrin , a. , et al . , 2010 ,
high - energy heavy - ion collisions can produce a hot , dense , and deconfined nuclear medium , dubbed the quark - gluon plasma ( qgp ) . the thermodynamic states of a qgp can be specified by the axial chemical potential @xmath0 , besides the temperature @xmath1 and the vector chemical potential @xmath2 . @xmath0 characterizes the imbalance of right - handed and left - handed fermions in a system , and a _ chiral _ system bears a nonzero @xmath0 . chiral domains may be created locally in heavy - ion collisions through various mechanisms on an event - by - event basis ( e.g. topological fluctuations in the gluonic sector , glasma flux tubes , or fluctuations in the quark sector ) @xcite . in a noncentral collision , a strong magnetic field ( @xmath3 t ) can be produced ( mostly by energetic spectator protons ) @xcite , and will induce an electric current along @xmath4 in chiral domains , @xmath5 , which is called the chiral magnetic effect ( cme ) @xcite . on average , @xmath4 is perpendicular to the so - called reaction plane ( @xmath6 ) that contains the impact parameter and the beam momenta , as depicted in fig . [ fig : overlap ] . hence the cme will manifest a charge transport across the reaction plane . in the presence of the cme and other modes of collective motions , we can fourier decompose the azimuthal distribution of particles of given transverse momentum ( @xmath7 ) and pseudorapidity ( @xmath8 ) : 1 + 2v_1 , ( ) + 2v_2,(2 ) + ... + 2a_1 , ( ) + ... , [ equ : fourier_expansion ] where @xmath9 is the azimuthal angle of a particle , and @xmath10 . here the subscript @xmath11 ( @xmath12 or @xmath13 ) denotes the charge sign of the particle . conventionally @xmath14 is called directed flow " and @xmath15 elliptic flow " @xcite . the parameter @xmath16 ( with @xmath17 ) quantifies the electric charge separation due to the cme . an anomalous transport effect can also occur when a chiral system undergoes a global rotation . the fluid rotation can be quantified by a vorticity @xmath18 , where @xmath19 is the flow velocity field . for a given vorticity @xmath20 , the chiral vortical effect ( cve ) induces a vector current @xmath21 @xcite . while the cme is driven by @xmath4 , the cve is driven by @xmath22 in a chiral medium . here the subscript v " means vector " , and can be , for example , @xmath23 " ( baryon ) or @xmath24 " ( electron ) . in heavy - ion collisions , @xmath25 is typically larger than @xmath26 by an order of magnitude , making it easier to search for the cve via the baryonic charge separation than the electric charge separation . hence the subscript @xmath11 in eq . [ equ : fourier_expansion ] represents baryon or anti - baryon in the cve search . another complementary transport phenomenon to the cme has been found and named the chiral separation effect ( cse ) @xcite , in which chiral charges are separated along the axis of the magnetic field in the presence of finite density of vector charge : @xmath27 . in a chirally symmetric phase , the cme and cse form a collective excitation , the chiral magnetic wave ( cmw ) , a long wavelength hydrodynamic mode of chiral charge densities @xcite . the cmw , a signature of the chiral symmetry restoration , manifests itself in a finite electric quadrupole moment of the collision system , where the poles " ( equator " ) of the produced fireball acquire additional positive ( negative ) charge @xcite . this effect , if present , will be reflected in the measurements of charge - dependent elliptic flow . there are other chiral magnetic / vortical effects such as the chiral electric separation effect ( cese ) @xcite and the chiral vortical wave ( cvw ) @xcite ; see ref @xcite for a recent review on these effects . this article reviews the experimental results in the past decade to search for the chiral magnetic / vortical effects in high - energy heavy - ion collisions : evidence for the initial magnetic field and vorticity in sec . [ sec : drive ] , the observation of the electric ( baryonic ) charge separation in sec . [ sec : cme ] ( sec . [ sec : cve ] ) , and the manifestation of the electric quadrupole moment in sec . [ sec : cmw ] . an outlook for future development is discussed in sec . [ sec : outlook ] . we may intuitively regard the magnetic field ( vorticity ) as the driving force of the cme ( cve ) , while the chirality imbalance is the initial condition , and the electric ( baryonic ) charge separation is the manifestation . a rough estimate of the initial magnetic field gives @xmath28 , where @xmath29 , @xmath30 is the impact parameter , and @xmath31 is the lorentz factor . therefore a typical au+au collision at @xmath32 gev produces @xmath33 . many computations have attempted to quantify the electromagnetic field on the event - by - event basis ( see e.g. @xcite ) , in terms of the spatial distribution , the orientation fluctuation as well as the dependence on colliding nuclei , centrality and beam energy . a major uncertainty in theoretical calculations of the magnetic field @xmath4 is its duration in the qcd fluid created in the heavy - ion collision ( see e.g. @xcite ) . the time dependence of @xmath4 after the impact of the two nuclei crucially depends on whether / when / how a conducting medium may form and the lifetime of the magnetic field may be elongated . the electric conductivity @xcite and the time evolution of the quark densities @xcite can be studied via directed flow of charged hadrons in asymmetric collisions , such as cu+au . figure [ fig : cuau ] illustrates the transverse plane for a cu+au collision with @xmath34 fm @xcite . the difference in the number of protons creates a strong electric field in the initial stage of the collision , pointing along the arrow in the figure . the lifetime of the electric field might be very short ( e.g. @xmath35 fm/@xmath36 from ref . @xcite ) , but the electric charges from quarks and antiquarks that are present in the early stage of the collision would experience the coulomb force and lift the degeneracy in @xmath14 between positively and negatively charged particles @xcite : v_1^ = v_1 d_e , where @xmath37 denotes the azimuthal angle of the electric field , and the coefficient @xmath38 characterizes the strength of dipole deformation induced by the electric field and is proportional to the electric conductivity of the medium . here @xmath14 represents the rapidity - even component of directed flow that is dominant in asymmetric collisions , while in symmetric collisions @xmath14 conventionally denotes the rapidity - odd component . of positive and negative particles and the difference between the two as functions of @xmath7 in @xmath39 cu+au and au+au collisions @xcite . the phsd model calculations @xcite for charged pions with and without the initial electric field ( ef ) are presented for comparison . ] of positive and negative particles and the difference between the two as functions of @xmath7 in @xmath39 cu+au and au+au collisions @xcite . the phsd model calculations @xcite for charged pions with and without the initial electric field ( ef ) are presented for comparison . ] figure [ fig : v1 ] shows recent star measurements of charge - dependent @xmath40 and the difference @xmath41 as functions of @xmath7 in @xmath39 cu+au and au+au collisions @xcite . for @xmath42 gev/@xmath36 , @xmath41 seems to increase with @xmath7 . the @xmath40 results from au+au collisions have much smaller values , roughly by a factor of 10 , than those in cu+au . note that @xmath43 in au+au collisions is similarly small @xcite . @xmath41 in au+au is consistent with zero . calculations for charged pions from the parton - hadron - string - dynamics ( phsd ) model @xcite , which is a dynamical transport approach in the partonic and hadronic phases , are compared with the data . the phsd model calculates two cases : charge - dependent @xmath40 with and without the initial electric field ( ef ) . in the case with the ef switched on , the model assumes that all electric charges are affected by the ef and this results in a large separation of @xmath40 between positive and negative particles as shown in fig . [ fig : v1](a ) . in fig . [ fig : v1](b ) , the calculations of the @xmath40 with and without the ef are shown together , but note that the ef - on calculation points are scaled by @xmath44 . after scaling by @xmath44 , the model describes rather well the @xmath7 dependence of the measured data for @xmath42 gev/@xmath36 . this qualitative evidence for the strong initial electric field in asymmetric collisions provides an indirect evidence for the strong initial magnetic field in heavy - ion collisions that could leave an imprint on the final - stage particles . the vorticity is induced by the global rotation of the qgp in heavy - ion collisions . in a noncentral collision , the majority of the global angular momentum , @xmath45 , is carried away by spectator nucleons . however , a considerable fraction ( about @xmath46 of @xmath45 could remain in the qgp and be approximately conserved in time @xcite . this implies a relatively long duration of the vortical effects . on average , the angular momentum is pointing in the out - of - plane direction , so the cme and cve are very much alike in terms of their experimental observables . attempts to computate local vorticity @xmath20 and its space - time distribution have also been made @xcite . experimentally global polarization of hyperons such as @xmath47 provides a measure for both the plasma vorticity and the magnetic field . whereas the vortical effects will generate a positive polarization for both @xmath47 and @xmath48 , the coupling of the hadronic magnetic dipole moment to the magnetic field will generate a positive contribution for @xmath47 and a negative one for @xmath48 . therefore , a splitting between @xmath47 and @xmath48 polarization will be a direct evidence for the strong initial magnetic field . recently , preliminary star measurements @xcite have reported the first observation of global @xmath47 and @xmath48 polarization in heavy - ion collisions . at @xmath49 gev , the signal is on the order of a few percent , and displays a weak beam - energy dependence . the average polarization over @xmath47 and @xmath48 evidences the plasma vorticity , while the splitting observation requires much higher statistics to be delivered in the second beam energy scan ( bes - ii ) program at rhic @xcite to signify the magnetic field . the future search for evidence for the initial magnetic field ( vorticity ) is proposed via photon ( vector meson ) polarization measurements @xcite . the initial magnetic helicity ( @xmath50 ) of the collision system can be quite large and bears opposite signs in the upper and lower hemispheres . owing to the chiral anomaly , the helicity can be transferred back and forth between the magnetic flux and fermions as the collision system evolves , so that the magnetic helicity could last long enough to yield photons with opposite circular polarizations in the hemispheres above and below the rp @xcite . a similar asymmetry in photon polarization can also result from the initial global quark polarization @xcite , which could effectively lead to a polarization of photons @xcite . this local imbalance of photon circular polarization could be observed in experiments , e.g. , by studying the polarization preference with respect to the rp for photons that convert into @xmath51 pairs @xcite . similarly , vector mesons that decay into two daughters can also have their polarization preferences measured with the scheme outlined in ref . @xcite , and the helicity separation in this case originates from vorticity @xcite . the discovery of the cme in high - energy heavy - ion collisions would confirm the simultaneous existence of ultra - strong magnetic fields , chiral symmetry restoration and topological charge changing transitions . the experimental searches for the cme have been carried out extensively in the past decade at rhic and the lhc . this section will introduce the observables pertinent to the electric charge separation induced by the cme , present the experimental results and discuss the background contributions due to the coupling of elliptic flow and other physics mechanisms . from event to event , the signs of the @xmath0 values are equally likely , and the signs of finite @xmath52 and @xmath53 will flip accordingly , leading to @xmath54 . figure [ fig : a1 ] presents the star measurements of @xmath55 with the @xmath56 harmonic event plane reconstructed from spectator neutrons @xcite . these results indicate no significant charge dependence in all centrality intervals , with the typical difference between positive and negative charges less than @xmath57 . for positive and negative charges versus centrality for au+au collisions at @xmath58 gev @xcite . , scaledwidth=45.0% ] one therefore has to search for the cme with charge - separation _ fluctuations _ perpendicular to the reaction plane , e.g. , with a three - point correlator @xcite , @xmath59 , where the averaging is done over all particles in an event and over all events . in practice , the reaction plane is approximated with the event plane " ( @xmath60 ) reconstructed with measured particles , and then the measurement is corrected for the finite event plane resolution . the expansion of the @xmath31 correlator , @xmath61 reveals the difference between the _ in - plane _ and _ out - of - plane _ projections of the correlations . the first term ( @xmath62 ) in the expansion provides a baseline unrelated to the magnetic field . the background contribution ( @xmath63 ) is suppressed to a level close to the magnitude of @xmath15 @xcite . the star collaboration first measured the @xmath31 correlator with the @xmath64 harmonic event plane for au+au ( shown with crosses in fig . [ fig : gamma ] ) and cu+cu ( not shown here ) collisions at @xmath65 and @xmath66 gev with data from the 2004/2005 rhic runs @xcite . all the results have been found to be in qualitative expectation with the cme : the opposite - charge ( @xmath67 ) and the same - charge ( @xmath68 ) correlations display the right " ordering . the opposite - charge correlations in cu+cu collisions are stronger than those in au+au , possibly reflecting the suppression of the correlations among oppositely moving particles in a larger system . star also presented @xmath7 and @xmath69 dependences of the signal . the signal has a @xmath69 width of about one unit of rapidity , consistent with small chiral domains . the signal is found to increase with the pair average @xmath7 , and it was later shown @xcite that the radial flow expansion can explain this feature . similar @xmath31 results for 200 gev au+au and 2.76 tev pb+pb were observed by the phenix collaboration @xcite and the alice collaboration @xcite , respectively . phenix also employed a multiparticle charge - sensitive correlator , @xmath70 @xcite , and their preliminary results showed a concave @xmath70 distribution , also evidencing the charge separation effect . the background from conventional physics was studied with heavy - ion event generators mevsim @xcite , urqmd @xcite and hijing @xcite ( with and without an elliptic flow afterburner implemented ) . none of these generators could achieve reasonable agreement with the data . however , these generators do not provide particularly good descriptions of heavy - ion collisions , so the fact that they fail to describe such subtle effects as charge separation provides only somewhat limited support for a cme interpretation of the data . gev @xcite . the grey bars reflect the conditions of @xmath71 gev/@xmath36 and @xmath72 applied to @xmath31 . ] gev @xcite . the grey bars reflect the conditions of @xmath71 gev/@xmath36 and @xmath72 applied to @xmath31 . ] the charge - separation signal was cross - checked with data from the 2007 rhic run ( shown in fig . [ fig : gamma ] ) @xcite . the @xmath31 correlations from these data were measured with respect to both the @xmath56 harmonic event plane ( of spectators at large rapidity ) and the @xmath64 harmonic event planes at mid - rapidity . using the @xmath56 harmonic event plane determined by spectator neutrons ensures that the signal is not coming from three - particle background correlations , and is due to genuine correlations to the reaction plane . another test was carried out by replacing one of the two charged particles in @xmath31 with a neutral particle , e.g. @xmath73 , and the results show no separation between @xmath74 and @xmath75 @xcite . thus the charge separation observed in the @xmath31 correlation between two charged particles is indeed due to the electric charge . femtoscopic correlations are visible in the differential measurements of @xmath31 at low relative momenta , which are related to quantum interference ( hbt " ) and final - state - interactions ( coulomb dominated ) @xcite . the difference between in - plane and out - of - plane correlations in the femtoscopic region can be due to a difference in the emission volumes probed by in- and out - of - plane parts . such a difference may arrise from an azimuthally anisotropic freeze - out distribution coupled with flow . to suppress the contribution from femtoscopic correlations , the conditions of @xmath71 gev/@xmath36 and @xmath72 were applied to the three - point correlator , shown with the grey bars in fig . [ fig : msc ] . excluding pairs with low relative momenta significantly reduces the positive contributions to opposite - charge correlations in peripheral collisions , but the difference between same- and opposite - charge correlations remains largely unchanged and consistent with the expectations of the cme . the @xmath31 correlator weights different azimuthal regions of charge separation differently , i.e. oppositely charged pairs emitted azimuthally at @xmath76 from the event plane ( maximally out - of - plane ) are weighted more heavily than those emitted only a few degrees from the event plane ( minimally out - of - plane ) . it is a good test to modify the @xmath31 correlator such that all azimuthal regions of charge separation are weighted identically . this may be done by first rewriting eq . [ eq : threepoint ] as ( _ + _ -2_rp ) = ( m_m_s_s_)_in - ( m_m_s_s_)_out , [ eq : mmss ] where @xmath77 and @xmath78 stand for the absolute magnitude ( @xmath79 ) and the sign ( @xmath80 ) of the sine or cosine function , respectively . in represents the cosine part of eq . [ eq : threepoint ] ( in - plane ) and out represents the sine part ( out - of - plane ) . a modulated sign correlation ( msc ) is obtained by reducing the @xmath31 correlator @xcite : ( ) ^2(s_s _ _ in - s_s__out ) . [ eq : msc ] the modulated sign correlations are compared with the three - point correlator for au+au collisions at 200 gev in fig . [ fig : msc ] . it is evident that the msc is able to reproduce the same trend as the three - point correlator although their magnitudes differ slightly . star also carried out another approach called the charge multiplicity asymmetry correlation ( cmac ) , whose methodology is similar to the msc , and yielded very similar results @xcite . . charge independent results from the model calculations of mevsim @xcite are shown as grey curves . ] a further understanding of the origin of the observed charge separation could be achieved with a study of the beam - energy dependence of the correlation . the charge separation effect depends strongly on the formation of the quark gluon plasma and chiral symmetry restoration @xcite , and the signal can be greatly suppressed or completely absent at low collision energies where a qgp has significantly shortened lifetime or not even formed . taking into account that the lifetime of the strong magnetic field is larger at smaller collision energies , this could lead to an almost threshold effect : with decreasing collision energy , the signal might slowly increase with an abrupt drop thereafter . unfortunately , the exact energy dependence of the cme is not calculated yet . figure [ fig : lpv_bes ] presents @xmath67 and @xmath68 correlators as functions of centrality for au+au collisions at @xmath81 gev measured by star @xcite , and for pb+pb collisions at 2.76 tev by alice @xcite . in most cases , the difference between @xmath67 and @xmath68 is still present with the right " ordering , manifesting extra charge - separation fluctuations perpendicular to the reaction plane . with decreased beam energy , both @xmath67 and @xmath68 tend to rise up starting from peripheral collisions . this feature seems to be charge independent , and can be explained by momentum conservation and elliptic flow @xcite . momentum conservation forces all produced particles , regardless of charge , to separate from each other , while elliptic flow works in the opposite sense . for peripheral collisions , the multiplicity ( @xmath82 ) is small , and momentum conservation dominates . the lower beam energy , the smaller @xmath82 , and the higher @xmath67 and @xmath68 . for more central collisions where the multiplicity is large enough , this type of charge - independent background can be estimated with @xmath83 @xcite . mevsim is a monte carlo event generator developed for star simulations @xcite . in fig . [ fig : lpv_bes ] , we also show the model calculations of mevsim with the implementation of @xmath15 and momentum conservation , which qualitatively describe the beam - energy dependence of the charge - independent background . the difference between @xmath67 and @xmath68 seems to vanish at low collision energies , but the interpretation involves an ambiguity to be discussed in the sec [ sec : kappa ] . the @xmath31 correlator by construction contains the background terms @xmath84 and @xmath85 , and their difference was originally studied for the flowing cluster " case @xcite : v_2,cl , where @xmath86 is the cluster emission azimuthal angle , and @xmath87 and @xmath88 are the azimuthal angles of two decay products . the flowing cluster can be generalized to a larger portion of or even the whole event , through the mechanisms of transverse momentum conservation ( tmc ) @xcite and/or local charge conservation ( lcc ) @xcite . one useful tool to study the background is the two - particle correlator , @xmath89 , which ideally should be protortional to @xmath90 , but in reality is dominated by backgrounds . for example , the tmc effect leads to the following pertinent correlation terms in @xmath91 and @xmath31 @xcite : & & - , + & & - + & & v_2 , , where @xmath92 , and @xmath93 and @xmath94 represent the @xmath7- and @xmath95-weighted moments of @xmath15 , respectively . the subscript f " denotes an average of all produced particles in the full phase space ; the actual measurements will be only in a fraction of the full space , denoted by @xmath96 " . the background contribution due to the lcc effect has a similar characteristic structure as the above @xcite . it is convenient to express @xmath31 and @xmath91 with a two - component framework @xcite : & & = v_2 b - h , + & & = b + h , where @xmath97 and @xmath23 are the cme and background contributions , respectively . the background - subtracted correlator , @xmath97 , can be obtained from the ensemble averages of several observables : h^= ( v_2 - ) /(1+v_2 ) . the major uncertainty in the above expression , the coefficient @xmath98 , depends on particle charge combination and particle transverse momentum . it may also depend on centrality and collision energy , reflecting slightly different particle production mechanism in different conditions . as a function of beam energy for @xmath99 au+au ( pb+pb ) collisions @xcite . the systematic errors ( grey bars ) bear the same meaning as in fig . [ fig : msc ] . ] as a function of beam energy for @xmath99 au+au ( pb+pb ) collisions @xcite . the systematic errors ( grey bars ) bear the same meaning as in fig . [ fig : msc ] . ] figure [ fig : kappa ] shows the @xmath98 values estimated for au+au collisions at 200 gev @xcite , with the @xmath15 measurements by the phobos collaboration @xcite , and with the @xmath15 calculations from the ampt model @xcite . here only the tmc effect has been taken in account , and @xmath98 is typically within @xmath100 $ ] for @xmath101 . the @xmath98 values attained this way will vary slightly if a smaller @xmath102 acceptance is involved . in reality , @xmath98 should be averaged over various mechanisms such as tmc , lcc and resonance decays . the ampt model gives a more comprehensive estimate in fig . [ fig : kappa ] via @xmath103 , where the numerator is solely due to flow backgrounds . for the centrality range of @xmath104 , where the statistical uncertainties are small , the @xmath98 values thus obtained are close to those estimated with the @xmath15 information . figure [ fig : h_bes ] shows @xmath105 as a function of beam energy for @xmath99 au+au ( pb+pb ) collisions @xcite . @xmath106 is depicted with the solid line . in both cases of @xmath98 , @xmath107 demonstrates a weak energy dependence above 19.6 gev , and tends to diminish from 19.6 to 7.7 gev , though the statistical errors are large for 7.7 gev . this may be explained by the probable domination of hadronic interactions over partonic ones at low energies . a more definitive conclusion may be reached with a more accurate estimation of @xmath98 and with higher statistics at lower energies in the proposed phase ii of the rhic beam energy scan program , as illustrated by the shaded band in fig . [ fig : h_bes ] . as a function of @xmath15 for various centrality selections within the @xmath108 centrality range in u+u collisions ( left ) and au+au collisions ( right ) . @xcite ] as a function of @xmath15 for various centrality selections within the @xmath108 centrality range in u+u collisions ( left ) and au+au collisions ( right ) . @xcite ] uranium nuclei have been collided at rhic in order to study the dependence of multiplicity production , flow , and the cme on the initial overlap geometry @xcite . early ideas hold that the prolate shape of uranium nuclei would make it possible to select nearly fully overlapping events with large elliptic flow values , but with small magnetic fields . however , owing to fluctuations , the square of the magnetic field is not particularly small . measurements of very central collisions also demonstrated that the number of produced particles does not depend as strongly on the configuration of the collisions as anticipated in the two - component multiplicity model , leaving the experiments with a significantly reduced ability to independently manipulate the flow and the magnetic field @xcite . figure [ fig : gamma_v2 ] shows measurements of @xmath109 for the @xmath108 centrality range in 193 gev u+u ( left ) and 200 gev au+au collisions ( right ) @xcite . in both u+u and au+au collisions , the signal increases roughly with @xmath15 . this initial observation may suggest that the charge - separation observable is dominated by a @xmath15 dependence . however , the charge separation goes to zero while @xmath15 is still sizable in central au+au and central u+u collisions . model calculations show that the quantity @xmath110}$ ] as a function of eccentricity exhibits the same trend @xcite : although @xmath111 remains large owing to fluctuations , @xmath112}$ ] goes to zero as @xmath113 and @xmath6 become decorrelated in very central collisions . so while fluctuations in central collisions force the participant eccentricity ( positive - definite ) away from zero , the decorrelation of @xmath113 and @xmath6 drives @xmath110}$ ] to zero . the data therefore appear to be in better agreement with a cme interpretation than a flow background interpretation . a phenomenological study for extrapolating both signal and background from au+au to u+u collisions was done in ref . @xcite . another measurement that can clarify the origin of the charge - dependent correlations and the role of the background was suggested @xcite : the correlations measured with respect to the @xmath114 harmonic event plane , @xmath115 , should not contain any contribution from the cme , but it should include the effect of the flow - related background . the correlations due to the background in this case are expected to be somewhat smaller in magnitude as the @xmath114 harmonic flow is not that strong as the elliptic flow . the preliminary results of such measurements by alice are presented in fig . [ fig:4th - harmonic ] @xcite with the same - charge and opposite - charge pair correlations relative to the @xmath114 harmonic event plane as functions of centrality ( left ) , and the charge - dependent parts with respect to the @xmath64 and @xmath114 harmonic event planes ( right ) . the correlations relative to the @xmath114 harmonic event plane are very weak and suggestive of small background contributions . detailed simulations have to be performed to draw more definite conclusion from this measurement . harmonic event plane as functions of centrality . ( right panel ) comparison of the charge - dependent parts in correlations with respect to the @xmath64 and @xmath114 harmonic event planes . @xcite ] harmonic event plane as functions of centrality . ( right panel ) comparison of the charge - dependent parts in correlations with respect to the @xmath64 and @xmath114 harmonic event planes . the chiral vortical effect ( cve ) is related to the cme , and its experimental manifestation is the baryonic charge separation , instead of the electric charge separation , perpendicular to the reaction plane . as a result , the three - point @xmath31 correlator is still applicable , only now between two ( anti)baryons . however , if both particles are ( anti)protons that carry also electric charges , there will be an ambiguity due to the possible signal arising from the cme . the study of the @xmath31 correlator with an electrically neutral baryon , such as @xmath47 , will provide more conclusive evidence for the baryonic charge separation effect . correlation of @xmath47-@xmath116 ( @xmath48-@xmath117 ) and @xmath47-@xmath117 ( @xmath48-@xmath116 ) as functions of centrality in au+au collisions at 200 gev @xcite . ] correlation of @xmath47-@xmath116 ( @xmath48-@xmath117 ) and @xmath47-@xmath117 ( @xmath48-@xmath116 ) as functions of centrality in au+au collisions at 200 gev @xcite . ] although ( anti)@xmath47 s are electrically neutral , it is still a question whether the strange quarks behave the same way as the up / down quarks in the chiral dynamics during the collision . if the answer is no , then ( anti)@xmath47 s may still act like electrically charged particles in the @xmath31 correlation . figure [ fig : lambda_h ] shows the @xmath31 correlation of @xmath47-@xmath118 ( @xmath48-@xmath119 ) and @xmath47-@xmath119 ( @xmath48-@xmath118 ) as functions of centrality in au+au collisions at 200 gev @xcite . note that ( anti)protons have been excluded from the charged hadrons in the correlation to avoid any possible cve contribution . tentatively assuming @xmath47s ( @xmath48s ) are positively(negatively)-charged , we find that the same - charge " and opposite - charge " correlations are consistent with each other , which means no charge - dependent effect . the message is twofold : first , from the @xmath120 correlations @xcite we learn that the different behaviors of same - charge and opposite - charge particle correlations as shown in fig . [ fig : gamma ] are really due to the electric charge , and therefore the null charge - separation effect in @xmath47-@xmath121 indicates that ( anti)@xmath47s manifest no electric charges in the @xmath31 correlation . so the strange quarks inside seem to behave the same way as the up / down quarks in the chiral dynamics . second , the @xmath47-@xmath121 correlation provides a baseline check for the @xmath47-@xmath116 correlation , and any possible signal in the latter should not come from the cme contribution . figure [ fig : lambda_p ] shows @xmath31 correlation of @xmath47-@xmath116 ( @xmath48-@xmath117 ) and @xmath47-@xmath117 ( @xmath48-@xmath116 ) as functions of centrality in au+au collisions at 200 gev @xcite . the same - baryonic - charge correlation has a different behavior from the opposite - baryonic - charge correlation from mid - central to peripheral collisions . this baryonic charge separation with respect to the event plane is consistent with the expectation from the cve . more investigations into the background contribution are needed . for example , in analog with the local charge conservation , there could be the local baryonic charge conservation that plays a similar role as lcc when coupled to @xmath15 . the magnitudes of the @xmath47-@xmath116 correlations are much larger than those of the @xmath121-@xmath121 correlations . this is partially because the @xmath122 of baryons is higher than that of mesons , and the correlation strength increases with the average @xmath7 of the two particles in the correlation . a future differential measurement vs the pair average @xmath7 and further correlations between identified particles may provide a better comparison of the correlation strength between the cme- and cve - related correlations . the cmw is a signature of the chiral symmetry restoration in the qgp , and consists of actually two chiral gapless modes traveling at the same speed @xcite : the right - handed ( left - handed ) wave transports the right - handed ( left - handed ) density and current in the direction parallel ( antiparallel ) to the @xmath4 direction . a more general analysis @xcite studied various possible collective modes based on a non - neutral - background qgp ( i.e. with nonzero @xmath2 and/or @xmath0 ) in external electric and/or magnetic fields , and found a new type of collective motion , the chiral electric wave ( cew ) , arising from cese and propagating in parallel / antiparallel to the @xmath123 field . in symmetric collisions there should be no net electric field on average , but asymmetric collisions like cu+au could provide a test ground for the cew measurements . as a function of observed charge asymmetry and ( b ) @xmath15 difference between @xmath124 and @xmath125 as a function of charge asymmetry with the tracking efficiency correction , for @xmath126-@xmath127 au+au collisions at 200 gev @xcite . ] the cmw will induce a finite electric quadrupole moment of the collision system , with additional positive ( negative ) charge at the poles " ( equator " ) of the produced fireball @xcite . this electric quadrupole , if boosted by radial flow , will lead to charge - dependent elliptic flow . taking pions as an example , on top of the baseline @xmath128 , the cmw will lead to @xcite v_2(^ ) = v_2^base(^ ) ( ) a_ch , [ eq : v2_slope ] where @xmath129 , @xmath130 and @xmath131 are the quadrupole moment , the net charge density and the charge asymmetry of the collision event , respectively . as @xmath132 is always positive , the @xmath133-integrated @xmath15 of @xmath124 ( @xmath125 ) should be above ( below ) the baseline owing to the cmw . however , the baseline @xmath15 may be different for @xmath125 and @xmath124 because of several other physics mechanisms @xcite . therefore , it is less ambiguous to study the cmw via the @xmath133 dependence of pion @xmath15 than via the @xmath133-integrated @xmath15 . taking @xmath126-@xmath127 200 gev au+au for example , pion @xmath15 is shown as a function of @xmath133 in panel ( a ) of fig . [ fig : v2_a ] @xcite . @xmath124 @xmath15 increases with @xmath133 while @xmath125 @xmath15 decreases with a similar magnitude of the slope . note that @xmath15 was integrated over a narrow low @xmath7 range ( @xmath134 gev/@xmath36 ) to focus on the soft physics of the cmw . such a @xmath7 selection also makes sure that the @xmath122 is independent of @xmath133 and is the same for @xmath125 and @xmath124 , so that the @xmath15 splitting is not a trivial effect due to the @xmath122 variation . this @xmath15 splitting was also confirmed by alice results for pb+pb collisions at 2.76 tev @xcite . the @xmath15 difference between @xmath124 and @xmath125 is fitted with a straight line in panel ( b ) . the slope parameter @xmath135 , or presumably @xmath136 from eq . [ eq : v2_slope ] , is positive and qualitatively consistent with the expectation of the cmw picture . the fit function is nonzero at @xmath132 , indicating the @xmath133-integrated @xmath15 for @xmath124 and @xmath125 are different , which was also observed in ref . @xcite . the same procedure as above was followed by star to retrieve the slope parameter @xmath135 as a function of centrality for au+au collisions at 200 , 62.4 , 39 , 27 , 19.6 , 11.5 and 7.7 gev , as shown in fig . [ fig : cmw_bes ] @xcite . a similar rise - and - fall trend is observed in the centrality dependence of the slope parameter for all the beam energies except 11.5 and 7.7 gev , where the slopes are consistent with zero with large statistical uncertainties . it was argued @xcite that at lower beam energies the @xmath133-integrated @xmath15 difference between particles and anti - particles can be explained by the effect of quark transport from the projectile nucleons to mid - rapidity , assuming that the @xmath15 of transported quarks is larger than that of produced ones . the same model , however , when used to study @xmath137 as a function of @xmath133 , suggested a negative slope @xcite , which is contradicted by the data . as a function of centrality for au+au collisions at 7.7 - 200 gev @xcite . the grey bands represent the systematic errors . for comparison , we also show the urqmd calculations @xcite and the calculations of the cmw @xcite with different duration times . ] to check if the observed slope parameters come from conventional physics , the same analysis of the monte carlo events from urqmd @xcite was carried out . for au+au collisions at 200 gev , the slopes extracted from urqmd events are consistent with zero for the @xmath138-@xmath139 centrality range , where the signal from the real data is prominent . similarly , the ampt event generator @xcite also yields slopes consistent with zero ( not shown here ) . on the other hand , the simplified cmw calculations @xcite demonstrate a centrality dependence of the slope parameter similar to the data . recently a more realistic implementation of the cmw @xcite confirmed that the cmw contribution to @xmath135 is sizable , and the centrality dependence of @xmath135 is qualitatively similar to the data . a quantitative comparison between data and theory requires further work on both sides to match the kinematic regions used in the analyses . for example , the measured @xmath133 only represents the charge asymmetry of a pseudorapidity slice ( @xmath140 ) of an event , instead of that of the whole collision system . we expect @xmath133 for these two cases to be proportional to each other , but the determination of the ratio will be model dependent . figure [ fig : slope_alice ] shows a comparison in the slope parameter @xmath135 between star results for 200 gev au+au @xcite and alice results for 2.76 tev pb+pb @xcite . overall , the slopes are surprisingly similar when considering the different collision energies and multiplicities , as well as the different kinematic acceptance : the star data estimated @xmath15 for charged pions with @xmath141 gev/@xmath36 and @xmath142 , while the alice data are for unidentified hadrons with @xmath143 gev/@xmath36 and @xmath144 . one drawback with the measurement of @xmath145 is that the observed @xmath133 requires a correction factor due to the finite detector tracking efficiency . a novel correlator @xcite that is independent of efficiency was proposed , q_3 = q_3 - q_3 . [ daq ] here @xmath146 and @xmath147 are the azimuthal angles of particles 1 and 2 , and @xmath148 is the charge ( @xmath149 ) of particle 3 . the single brackets represent the average over particles and events , and the double bracket denotes the cumulant . in the absence of charge dependent correlations , the correlator should be equal to zero . note that when the charge of the third particle is averaged over all particles in the event ( in the specified kinematic acceptance ) , the mean is equal to the charge asymmetry , i.e. @xmath150 . the three - particle correlator ( as in eq . [ daq ] ) for the @xmath64 harmonic was measured by alice and is presented in fig . [ fig : threepart_alice ] as a function of centrality in pb+pb collisions at 2.76 tev gev @xcite . a substantial increase in the correlation strength is seen as the collisions become more peripheral , which can be caused by a combination of several factors . for example , the magnetic field strength increases as the impact parameter increases , and this would cause the stronger correlations due to the cmw . additionally , the lcc effect could play a role @xcite , and neither of these necessarily comes at the expense of the other ; in principle the observable could have contributions from both of these and/or additional contributions from as yet unknown sources of correlation . it was pointed out in ref . @xcite that local charge conservation at freeze - out , when convoluted with the characteristic shape of @xmath151 and @xmath152 , may provide a qualitative explanation for the finite @xmath15 slope observed from data . a realistic estimate of the contribution of this mechanism turns out to be smaller than the measurment by an order of magnitude @xcite . @xcite also proposes a test with the @xmath153 measurement , and the corresponding slope parameters for @xmath153 were reported by star to be consistent with zero @xcite , which further suggests the smallness of this effect . alice measured the three - particle correlator multiplied by @xmath154 for the @xmath155 harmonic and the @xmath114 harmonic @xcite . in both cases , the centrality dependence of the charge dependence is flat , in contrast to the @xmath64 harmonic that has a significant centrality dependence . this may suggest a different nature of the correlation , or reflect a weaker centrality dependence of @xmath153 compared with that of @xmath15 . future measurements of these higher harmonics with better precision will shed light on the true origin of this correlator . a recent hydrodynamic study @xcite suggested that simple viscous transport of charges , combined with certain specific initial conditions , might lead to a sizable contribution to the observed @xmath15 splitting of charged pions . in order for the results of pion splitting to resemble data , the authors had to assume a crucial relation between isospin chemical potential and the electric charge asymmetry , which needs to be verified . furthermore , certain predictions of this model ( e.g. splitting for kaons ) appear to be not in line with current experimental information @xcite . clearly whether such an idea works or not , would need to be thoroughly vetted by realistic viscous hydrodynamic simulations . but all that said , this study poses a very important question : to make a firm case for the observation of anomalous charge transport via the cmw , the normal ( viscous hydrodynamical ) transport of charges should be quantitatively understood . the confirmation of the experimental observation of several chiral anomalous effects will bring forth an exciting program to directly study the non - perturbative sector of qcd . future experimental measurements should aim at more detailed study of the observed signals as well as understanding the background effects . previous sections have covered a few such topics : initial magnetic field and vorticity , correlations with identified particles , higher - harmonic correlations , bes - ii and u+u collisions . in the following , we will focus on the event - shape engineering ( ese ) and isobaric collisions . flow - related backgrounds could be potentially removed via ese @xcite , with which _ spherical _ events or sub - events are selected , so that the particles of interest therein carry zero @xmath15 . a previous attempt was made with the charge - separation observable of cmac ( roughly equivalent to @xmath31 ) , as a function of event - by - event observed @xmath15 " @xcite . however , there are several issues in this approach that prevent a clear interpretation of the result . ref . @xcite studied the flow vector @xmath156 of the sub - event of interest , a : q_x^a & = & _ i^n ( 2_i^a ) [ qx ] + q_y^a & = & _ i^n ( 2_i^a ) , [ qy ] and found that @xmath157 is a good handle on event shape . ( upper ) and @xmath158 ( lower ) as functions of @xmath157 , from ampt simulations . @xcite the solid ( dashed ) line in the lower panel is a linear fit of the full ( open ) data points . ] ( upper ) and @xmath158 ( lower ) as functions of @xmath157 , from ampt simulations . @xcite the solid ( dashed ) line in the lower panel is a linear fit of the full ( open ) data points . ] figure [ fig : ampt_v2_q2 ] shows the event plane resolution ( @xmath159 ) for the sub - events b1 and b2 ( upper ) , and the true elliptic flow @xmath160 and the corrected @xmath161 as functions of @xmath157 ( lower ) , from ampt simulations of @xmath162 au+au collisions at @xmath163 gev @xcite . each ampt event has been divided into three sub - events according to pseudorapidity , @xmath8 : sub - event a contains particles of interest with @xmath101 , and sub - event b1(b2 ) serves as a sub - event plane using particles with @xmath164 ( @xmath165 ) . flow fluctuation causes a positive correlation in flow between sub - events in the same event , and as a result , @xmath159 for sub - event b1 ( b2 ) increases with @xmath157 for sub - event a. the lower panel displays a discrepancy between @xmath160 and the corrected @xmath161 , owing to the difference between the reaction plane and the participant plane @xcite , in terms of non - flow and flow fluctuation . what matters more is the fact that both @xmath15 values decrease with @xmath157 , and drop to @xmath166 , which demonstrates @xmath167 s capability of selecting spherical sub - events in the second harmonic . the upper panel of fig . [ fig : ampt_gamma_q2 ] presents the @xmath31 correlators multiplied by the number of participating nucleons , @xmath168 , as functions of @xmath157 , for @xmath162 ampt events of au+au collisions at 200 gev @xcite . for both the same - charge and the opposite - charge correlators , the true @xmath169 and the corrected @xmath170 are consistent with each other within the statistical uncertainties . this indicates that compared with @xmath15 , @xmath31 is less sensitive to non - flow or flow fluctuation . at larger @xmath157 , the opposite - charge correlators are above the same - charge correlators , suggesting a finite flow - related background . the opposite- and same - charge correlators converge at small @xmath157 . the lower panel shows @xmath171 vs @xmath157 , and again , the two observables seem to coincide . linear fits to both observables yield small intercepts that are consistent with zero . the finite @xmath172 values in ampt events are solely due to background contributions , so the disappearance of background is demonstrated when the correctable " observable ( @xmath172 ) is projected to zero @xmath157 . ref . @xcite has designed a promising recipe for future measurements to effectively remove flow backgrounds and restore the ensemble average of the cme signal . to disentangle the possible cme signal and the flow - related backgrounds , one can utilize experimental setups to either vary the backgrounds with the signal fixed , or vary the signal with the backgrounds fixed . the former approach was carried out by exploiting the prolate shape of the uranium nuclei @xcite . however , it was found that the total multiplicity of detected hadrons is far less dependent on the number of binary collisions than expected @xcite , so it is very hard to isolate tip - tip collisions ( that generate small @xmath15 ) from body - body collisions ( that generate large @xmath15 ) . this significantly reduces the lever arm available to manipulate @xmath15 in order to separate flow backgrounds from the cme . the latter approach ( with the @xmath15-driven backgrounds fixed ) can be realized , especially for mid - central / mid - peripheral events , with collisions of isobaric nuclei , such as @xmath173ru and @xmath174zr @xcite . ru+ru and zr+zr collisions at the same beam energy are almost identical in terms of particle production @xcite , while the charge difference between ru and zr nuclei provides a handle on the initial magnetic field . our current knowledge of the deformity ( @xmath175 ) of ru and zr is incomplete : e - a scattering experiments ( case 1 ) @xcite state that ru is more deformed ( @xmath176 ) than zr ( @xmath177 ) , while comprehensive model deductions ( case 2 ) @xcite tell the opposite , that @xmath178 is smaller than @xmath179 . this systematic uncertainty has different impacts on the signal ( via the initial magnetic field ) and the background ( via the initial eccentricity ) to be discussed later . as a by - product , @xmath15 measurements in central collisions will discern which information source is more reliable regarding the deformity of the zr and ru nuclei . figure [ fig_mag](a ) presents the theoretical calculation @xcite of the initial magnetic field squared with correction from azimuthal fluctuation of the magnetic field orientation , @xmath180}$ ] ( with @xmath181 the pion mass and @xmath182 the azimuthal angle of the magnetic field ) , for the two collision systems at 200 gev , using the hijing model @xcite . @xmath183 quantifies the magnetic field s capability of driving the cme signal in the @xmath31 correlator . for the same centrality bin , the ru+ru collision produces a significantly stronger magnetic field than zr+zr . panel ( b ) of fig . [ fig_mag ] shows that the relative difference in @xmath183 between ru+ru and zr+zr collisions is approaching @xmath184 ( case 1 ) or @xmath185 ( case 2 ) for peripheral events , and reduces to about @xmath186 ( both cases ) for central events . figure [ fig_mag](b ) shows the relative difference in the initial eccentricity , @xmath187 , obtained from the monte carlo glauber simulation . @xmath187 is highly consistent with 0 for peripheral events , and goes above ( below ) 0 for the parameter set of case 1 ( case 2 ) in central collisions , because the ru ( zr ) nucleus is more deformed . the relative difference in @xmath15 should closely follow that in eccentricity , so for the centrality range of interest , @xmath162 , the @xmath15-related backgrounds stay almost the same for ru+ru and zr+zr collisions . @xcite further carried out the projection for the @xmath31 measurements in ru+ru and zr+zr at 200 gev ( 400 million events for each collision type ) , and concluded that a @xmath188 significance can be achieved for the relative difference in the observable between the two collision systems , assuming the flow backgrounds take up to two - thirds of the observable . the results strongly suggest that the isobaric collisions can serve as an ideal tool to disentangle the signal of the chiral magnetic effect from the @xmath15-driven backgrounds . the isobaric collisions may also be used to disentangle the signal of the cmw from background effects . the physics of anomalous transport is at the heart of qcd as a non - abelian gauge theory . the interplay of quantum anomalies with magnetic field and vorticity induces a variety of novel transport phenomena in chiral systems . in heavy - ion collisions , these phenomena make a unique probe to the topological properties of the qgp by measuring the charge dependence of the azimuthal distributions of the produced hadrons . the experimental data from relativistic heavy ion collider at bnl and the large hadron collider at cern provide an evidence for the predicted effects , with magnitude consistent with theoretical estimates . there exist known conventional backgrounds to all of these experimental observables . however at present there is no compelling alternative explanation that can describe all of the data without invoking the anomalous chiral effects . nevertheless , much remains to be done both in experiment and theory to substantiate the existing evidence , and we outlined a few such programs that hopefully will be accomplished in the near future . we thank huan huang and other members of the ucla heavy ion physics group for discussions . this work is supported by a grant ( no . de - fg02 - 88er40424 ) from u.s . department of energy , office of nuclear physics . the authors declare that there is no conflict of interest regarding the publication of this paper . d. e. kharzeev , l. d. mclerran and h. j. warringa , nucl . a * 803 * , 227 ( 2008 ) . d. kharzeev , phys . b * 633 * , 260 ( 2006 ) . d. kharzeev and a. zhitnitsky , nucl . a * 797 * , 67 ( 2007 ) . d. kharzeev , a. krasnitz and r. venugopalan , phys . b * 545 * , 298 ( 2002 ) . i. iatrakis , s. lin and y. yin , phys . lett . * 114 * , 252301 ( 2015 ) . k. fukushima , d. e. kharzeev and h. j. warringa , phys . 104 , 212001 ( 2010 ) . a. m. poskanzer and s. voloshin , phys . c * 58 * , 1671 ( 1998 ) . d. e. kharzeev and d. t. son , phys * 106 * ( 2011 ) 062301 . d. t. son and a. r. zhitnitsky , phys d * 70 * 074018 ( 2004 ) . m. a. metlitski and a. r. zhitnitsky , phys . d * 72 * 045011 ( 2005 ) . y. burnier , d. e. kharzeev , j. liao and h .- u . yee , 052303 ( 2011 ) . g. m. newman , jhep * 0601 * 158 ( 2006 ) . x. g. huang and j. liao , phys . lett . * 110 * , 232302 ( 2013 ) . y. jiang , x. g. huang and j. liao , phys . d * 91 * , 045001 ( 2015 ) . y. jiang , x. g. huang and j. liao , phys . d * 92 * , 071501 ( 2015 ) . kharzeev , j. liao , s.a . voloshin and g. wang , prog . 88 * , 1 ( 2016 ) . a. bzdak , v. skokov , phys . b * 710 * , 171 ( 2012 ) . deng , x.g . huang , phys . c * 85 * , 044907 ( 2012 ) ; phys . b * 742 * , 296 ( 2015 ) . j. bloczynski , x.g . huang , x. zhang , j. liao , phys . b * 718 * , 1529 ( 2013 ) . l. mclerran , v. skokov , nuclear phys . a * 929 * , 184 ( 2014 ) . x. guo , s. shi , n. xu , z. xu , p. zhuang . u. gursoy , d. kharzeev , k. rajagopal , phys . c * 89(5 ) * , 054905 ( 2014 ) . k. tuchin , adv . high energy phys . * 2013 * , 490495 ( 2013 ) . y. hirono , m. hongo , and t. hirano , phys . c * 90 * , 021903 ( 2014 ) . v. voronyuk , v. d. toneev , s. a. voloshin , and w. cassing , phys . 064903 ( 2014 ) . l. adamczyk _ et al . _ [ star collaboration ] , arxiv:1608.04100 . b. i. abelev _ et al . _ [ star collaboration ] , phys . * 101 * , 252301 ( 2008 ) . y. jiang , z.w . lin , j. liao , arxiv:1602.06580 . baznat , k.k . gudima , a.s . sorin , o.v . teryaev , phys . c * 93 * , 031902 ( 2016 ) . f. becattini _ _ , arxiv:1501.04468 . f. becattini , l. csernai , d.j . wang , phys . c * 88 ( 3 ) * , 034905 ( 2013 ) . s. floerchinger , u.a . wiedemann , j. high energy phys . * 1111 * , 100 ( 2011 ) . gao , b. qi , s.y . wang , phys . d * 90 ( 8) * , 083001 ( 2014 ) . isaac upsal _ [ star collaboration ] , j. phys . * 736 * , 012016 ( 2016 ) . star collaboration , star note * 0598 * , https://drupal.star.bnl.gov/star/starnotes/public/sn0598 . a.h . tang and g. wang , phys . c * 94 * , 024920 ( 2016 ) . y. hirono , d. e. kharzeev , and y. yin , phys . d * 92 * , 125031 ( 2015 ) . a. ipp , a. di piazza , j. evers , and c. h. keitel , phys . lett . b * 666 * , 315 ( 2008 ) . k. a. mamo and h. u. yee , phys . d * 88 * , 114029 ( 2013 ) . liang and x .- wang , phys . * 94 * , 102301 ( 2005 ) . m. baznat , k. gudima , a. sorin , and o. teryaev , phys . c * 88 * , 061901 ( 2013 ) . deng and x .- g . huang , phys . c * 93 * , 064907 ( 2016 ) l. adamczyk _ et al . _ [ star collaboration ] , phys . c * 88 * 064911 ( 2013 ) . s. a. voloshin , phys . c * 70 * , 057901 ( 2004 ) [ hep - ph/0406311 ] . b. i. abelev _ et al . _ [ star collaboration ] , phys . 103 * 251601 ( 2009 ) . 7 i. abelev _ et al . _ [ star collaboration ] , phys . c * 81 * 54908 ( 2010 ) . a. bzdak , v. koch and j. liao , phys . c * 81 * 031901 ( 2010 ) . n. n. ajitanand , s. esumi , r. a. lacey ( phenix collaboration ) , in : proc . of the rbrc workshops , vol.96 , 230 ( 2010 ) : p- and cp - odd ects in hot and dense matter " . b. i. abelev , 021301 ( 2013 ) . n. n. ajitanand , r. a. lacey , a. taranenko and j. m. alexander , phys . c * 83 * 011901 ( 2011 ) . r. l. ray and r. s. longacre , arxiv : nucl - ex/0008009 . s. a. bass _ et al . * 41 * 255 ( 1998 ) . m. gyulassy and x .- wang , comput . * 83 * 307 ( 1994);x.n . wang and m.gyulassy , phys . d * 44 * 3501 ( 1991 ) . f. zhao ( star collaboration ) , nucl . a * 931 * ( 2014 ) 746 . l. adamczyk _ et al . _ [ star collaboration ] , phys . c * 89 * , 044908 ( 2014 ) . l. adamczyk _ et al . _ [ star collaboration ] , phys . * 113 * ( 2014 ) 052302 . a. bzdak , v. koch and j. liao , phys . c * 83 * 014905 ( 2011 ) . s. pratt , s. schlichting and s. gavin , phys . c * 84 * , 024909 ( 2011 ) . a. bzdak , v. koch and j. liao , lect . notes phys . * 871 * 503 ( 2013 ) [ arxiv:1207.7327 [ nucl - th ] ] . s. schlichting and s. pratt , phys . c * 83 * , 014913 ( 2011 ) . f. wen , l. wen and g. wang , arxiv:1608.03205 . [ phobos collaboration ] , phys . c * 83 * , 024913 ( 2011 ) . b.b back _ et al . _ [ phobos collaboration ] , phys . c * 72 * , 051901(r ) ( 2005 ) . b. zhang , c.m . ko , b .- a . li and z .- w . lin , phys . c * 61 * , 067901 ( 2000 ) . lin , c.m . ko , b .- a . li and b. zhang , phys . c * 72 * , 064901 ( 2005 ) . lin and c.m . ko , phys . c * 65 * , 034904 ( 2002 ) ; a. j. kuhlman and u. w. heinz , phys . rev . c * 72 * , 037901 ( 2005 ) . s. a. voloshin , phys . lett . * 105 * , 172301 ( 2010 ) . l. adamczyk _ et al . _ [ star collaboration ] , phys . * 115 * , 222301 ( 2015 ) . gang wang _ [ star collaboration ] , nucl . phys a * 904 - 905 * , 248c ( 2013 ) . p. tribedy [ star collaboration ] , charge sensitive cumulants and flow in u+u collisions " , workshop on chirality , vorticity and magnetic field in heavy ion collisions , 2016 ; http://starmeetings.physics.ucla.edu/sites/default/files/pritwish_tribedy.pdf . v. skokov , p. sorensen , v. koch , s. schlichting , j. thomas , s. voloshin , g. wang and h .- u . yee , arxiv:1608.00982 . j. bloczynski , x. g. huang , x. zhang and j. liao , nucl . a * 939 * , 85 ( 2015 ) . s. a. voloshin , prog . 67 * , 541 ( 2012 ) . s. a. voloshin [ alice collaboration ] , nucl . phys . a * 904 - 905 * , 90c ( 2013 ) . gorbar , v.a . miransky , i.a . shovkovy , x. wang , phys . d * 88 ( 2 ) * , 025025 ( 2013 ) . j. c. dunlop , m. a. lisa and p. sorensen , phys . c * 84 * , ( 2011 ) 044914 . j. xu , l. -w . chen , c. m. ko and z. -w . lin , phys . c * 85 * , ( 2012 ) 041901 ( 2012 ) . l. adamczyk _ ( star collaboration ) , phys . * 114 * , ( 2015 ) 252302 . ( alice collaboration ) , phys . c * 93 * , 044903 ( 2016 ) . l. adamczyk et al . [ star collaboration ] , phys . lett . * 110 * , 142301 ( 2013 ) . j. m. campbell and m. a. lisa , journal of physics : conference series * 446 * , 012014 ( 2013 ) . y. burnier , d.e . kharzeev , j. liao , h .- u . yee , arxiv : 1208.2537 ( 2012 ) . h. yee and y. yin , phys . c * 89 * ( 2014 ) 044909 . s. a. voloshin and r. belmont , nucl . phys . a * 931 * , 992 ( 2014 ) . a. bzdak and p. bozek , arxiv:1303.1138 . q .- y . shou [ star collaboration ] , nucl . a * 931 * , 758 ( 2014 ) . y. hatta , a. monnai and b. w. xiao , nucl . a * 947 * , 155 ( 2016 ) . shou [ star collaboration ] , charge asymmetry dependence of @xmath189 @xmath15 in au+au collisions at star " , workshop on chirality , vorticity and magnetic field in heavy ion collisions , 2016 ; http://starmeetings.physics.ucla.edu/sites/default/files/qiye_shou.pdf j. schukraft , a. timmins and s. a. voloshin , phys . b * 719 * , 394 ( 2013 ) . j. -y . ollitrault , a. m. poskanzer and s. a. voloshin , phys . c * 80 * , 014904 ( 2009 ) . deng , x .- huang , g .- ma , g. wang , arxiv:1607.04697 . s. raman , c. w. g. nestor , jr and p. tikkanen , atom . data nucl . data tabl . * 78 * , 1 ( 2001 ) . b. pritychenko , m. birch , b. singh and m. horoi , atom . data nucl . data tabl . * 107 * , 1 ( 2016 ) . p. moller , j. r. nix , w. d. myers and w. j. swiatecki , atom . data nucl . data tabl . * 59 * , 185 ( 1995 ) .
unveiling the nature of the unidentified gamma - ray sources ( ugss ) ( e.g. , * ? ? ? * ) is one of the biggest challenges in contemporary gamma - ray astronomy . since the era of the compton gamma - ray observatory many @xmath1-ray objects have not been conclusively associated with counterparts at other frequencies @xcite , although various classes have been investigated to understand whether they are likely to be detected at @xmath1-ray energies or not ( e.g. , * ? ? ? according to the second _ fermi _ large area telescope ( lat ) catalog ( 2fgl ; * ? ? ? * ) , @xmath01/3 of the @xmath1-ray detected sources are still unassociated with their low energy counterparts . moreover a large fraction of the ugss are likely to be of blazars , the rarest class of radio loud active galactic nuclei , because their emission dominates the @xmath1-ray sky ( e.g. , * ? ? ? * ; * ? ? ? * ) . however , due to the incompleteness of the current radio and x - ray surveys on the basis of the usual @xmath1-ray association method is not always possible to find the blazar - like counterpart of an ugs . additional attempts have also been recently developed to associate or to characterize the ugss using either pointed _ swift _ observations ( e.g. , * ? ? ? * ; * ? ? ? * ) or statistical approaches ( e.g. * ? ? ? * ; * ? ? ? blazar emission is characterized by high and variable polarization , apparent superluminal motions , and high luminosities , generally combined with a flat radio spectrum that steepens toward the infrared - optical bands and together with rapid flux variability from the radio to @xmath1-rays ( e.g. , * ? ? ? their spectral energy distributions show two main broad components : a low - energy one peaking in the range from the ir to the x - ray band , and a high - energy component peaking from mev to tev energies ( e.g. , * ? ? ? blazars are divided in two main classes : the low luminosity class constituted by the bl lac objects and characterized by featureless optical spectra , and the second class composed of flat - spectrum radio quasars that show optical emission lines , typical of quasar spectra @xcite . in the following we label the bl lac objects as bzbs and the flat - spectrum radio quasars as bzqs , following the nomenclature of the multifrequency catalogue of blazars ( roma - bzcat , * ? ? ? * ; * ? ? ? * ; * ? ? ? on the basis of the preliminary data release of the wide - field infrared survey explorer ( _ wise _ , see * ? ? ? * for more details ) , we discovered that in the 3-dimensional ir color space @xmath1-ray emitting blazars lie in a distinct region , well separated from other extragalactic sources whose ir emission is dominated by thermal radiation ( e.g. , * ? ? ? * ; * ? ? ? according to dabrusco et al . ( 2013 ) we refer to the 3-dimensional region occupied by @xmath1-ray emitting blazars as the @xmath2 , to its 2-dimensional projection in the [ 3.4]-[4.6]-[12 ] @xmath3 m color - color diagram as the _ wise _ gamma - ray strip . this _ wise _ analysis led to the development of a new association method to recognize @xmath1-ray blazar candidates for the unidentified @xmath1-ray sources listed in the 2fgl @xcite , as well as in the 4@xmath4 _ integral _ catalog @xcite . in the present paper we adopt several improvements recently made on the association procedure and we use a more conservative approach ( see * ? ? ? * for more details ) , mostly based on the _ wise _ full archive , available since march 2012 ( see also * ? ? ? we successfully tested the association procedure on all the blazars listed in the second _ fermi _ lat catalog of active galactic nuclei ( 2lac ; * ? ? ? * ) and in the 2fgl catalogs , to estimate its efficiency and its completeness . in this paper we apply this method to the ugss and to sample of the active galactic nuclei of uncertain type ( agus ) that have still unclear classification ( see 2fgl and also section [ sec : agu ] for specific definition of the class ) , both listed in the 2fgl . we also performed an extensive literature search looking for multifrequency information on the @xmath1-ray blazar candidates selected on the basis of their _ wise _ colors to confirm their nature . as we show below this research is crucial to determine whether or not there are classes of galactic and extragalactic sources that , having ir colors similar to those of blazars , could be a contaminants of the association method . the paper is organized as follows : in section [ sec : sample ] we describe the sample selected . in section [ sec : method ] we illustrate the basic details of the association procedure and highlight the improvements with respect to the previous version . in section [ sec : results ] we describe the results obtained . section [ sec : counterparts ] is dedicated to the correlating our results with several databases at radio , infrared , optical and x - ray frequencies to characterize the multifrequency behavior of the @xmath1-ray blazar candidates . we then compare our results on the associated sources with those based on statistical methods developed by ackermann et al . ( 2012 ) in section [ sec : comparison ] . finally , section [ sec : summary ] is devoted to our conclusions . the most frequent acronyms used in the paper are listed in table [ tab : acronym ] . multifrequency catalog of blazars & roma - bzcat + first _ fermi _ large area telescope catalog & 1fgl + second _ fermi _ large area telescope catalog & 2fgl + second _ fermi _ lat catalog of active galaxies & 2lac + bl lac object & bzb + flat spectrum radio quasar & bzq + blazar of uncertain type & bzu + unidentified gamma - ray source & ugs + active galactic nucleus of uncertain type & agus + + [ tab : acronym ] our primary sample of ugss consists of all the sources for which no counterpart was assigned at low energies in the 2fgl or in the 2lac ( * ? ? ? * ; * ? ? ? * respectively ) , for a total of 590 @xmath1-ray objects . we considered and analyzed independently two subsamples of ugss , distinguishing the 299 _ fermi _ sources without any @xmath1-ray analysis flags from the other 291 objects that have a warning in their @xmath1-ray detection . this distinction has been performed because future releases of the _ fermi _ catalogs based on improvements of the _ fermi _ response matrices and revised analyses , could make their detection more reliable , as occurred for a handful of sources flagged in the first _ fermi _ lat catalog ( 1fgl , * ? ? ? * ; * ? ? ? according to the definition of the 2lac and 2fgl catalogs , active galaxies of uncertain type ( agus ) are @xmath1-ray emitting sources with at least one of the following criteria : 1 . they do not have a good optical spectrum available or with an uncertain classification , as for example , sources classified as blazars of uncertain type ( bzu ) in the roma - bzcat ; 2 . they have been selected as candidate counterparts on the basis of the @xmath5 - @xmath6 and the likelihood ratio methods described in the 2lac and applied to several radio catalogs : including the at20 g @xcite , crates @xcite , or class @xcite ( see * ? ? ? * for details ) ; 3 . they are coincident with a radio and a x - ray source selected by the likelihood ratio method . the number of agus in the 2fgl , that have been analyzed is 210 ; excluding @xmath1-ray sources with analysis flags ( defined according to both the 2fgl or the 2lac descriptions ) . the complete description of our association procedure together with the estimates of its efficiency and its completeness can be found in dabrusco et al . ( 2013 ) where we discuss a new and improved version of the association method based on a 3-dimensional parametrization of the @xmath2 occupied by @xmath1-ray emitting blazars with _ wise _ counterparts . here we provide only an overview . we note that the results of the improved method are in agreement with those of the previous parametrization , thus superseding the previous procedure @xcite . the new association procedure was built to improve the efficiency of recognizing @xmath1-ray blazar candidates , to decrease the number of possible contaminants and , at the same time , to determine if a selected @xmath1-ray blazar counterpart is more likely to be a bzb or a bzq . the main differences between the two association methods reside in the parameter space where the @xmath2 has been defined ( ir color space for the old version and principal component space for the new one ) and in the assignment criteria of the classes for the @xmath1-ray blazar candidates ( see * ? ? ? the new method also takes into account of the correction for galactic extinction for all the _ wise _ magnitudes , [ 4.6 ] , [ 12 ] , [ 22 ] @xmath3 m nominal _ wise _ bands are in the vega system . ] according to the draine ( 2003 ) relation . as shown in dabrusco et al . ( 2013 ) , this correction affects only marginally the [ 3.4]-[4.6 ] color , in particular at low galactic latitudes ( i.e. , @xmath715 deg ) . the principal component analysis is designed to reduce the dimensionality of a dataset consisting of usually large number of correlated variables while retaining as much as possible of the variance present in the data in the smallest possible number of orthogonal parameters . this is achieved by transforming the observed parameter into a new set of variables , the principal components . they are ordered so that the first accounts the largest possible variance of the original dataset and the others in turn have the highest variance possible under the constraint of being orthogonal to the preceding ones ( e.g. , * ? ? ? * ; * ? ? ? thus our new parametrization of the @xmath2 in the pc space , where the maximum variance is contained along only one axis , is simpler than any other possible representation in the ir color space . for each @xmath1-ray source we defined a _ search region _ : a circular region of radius @xmath8 equal to the semi - major axis of the ellipse corresponding to the positional uncertainty region of the _ fermi _ source at 95% level of confidence and centered at the 2fgl position of the @xmath1-ray source ( e.g. , * ? ? ? we selected and calculated the ir colors for the _ wise _ sources within the _ search region _ detected in all four bands . to compare the infrared colors of generic infrared sources that lie in the _ search region _ with those of the @xmath1-ray emitting ones , we developed a 3-dimensional parametrization of the @xmath2 in the parameter space of its principal components . the @xmath2 was described as a cylinder in the space of the principal components . this choice simplifies and improves the previous description built using irregular quadrilaterals on all the color - color diagrams @xcite . moreover , the cylinder axis is aligned along the first pc axis , which accounts for the larger fraction possible of the variance of the dataset in the ir color space , is the simplest parametrization available . we then assign to each source _ score _ value @xmath9 that is a proxy of the distance between the @xmath2 surface and the source location in the 3-dimensional parameter space of the principal components . the values of @xmath9 allow to to evaluate if the ir colors of a generic source are consistent with those of the known @xmath1-ray emitting blazars . they were weighted taking into account of all the color errors and they are also normalized between 0 and 1 . we define three classes ( i.e. , a , b , c ) of reliability for the @xmath1-ray blazar candidates . a generic source is assigned to class a , class b or class c when its @xmath10 his higher than the threshold values defined by the 90% , 60% and 30% percentiles of the @xmath10 distributions of all the @xmath1-ray blazars that constitute the @xmath2 , respectively . we consider reliable @xmath1-ray blazar candidates only those having the score higher than 70% of their distributions . thus sources with high values of the @xmath10 ( e.g. , @xmath110.8 ) are very likely to be blazars and belong to class a , while sources with @xmath10 values @xmath00.5 belong to class c and are less probable @xmath1-ray blazars . ir sources that having @xmath10 values null or extremely low ( e.g. , @xmath00.1 ) were marked as _ outliers _ and were not considered as @xmath1-ray blazar candidates ( see * ? ? ? * for an extensive explanation on the class definitions ) . the @xmath2 was divided in subregions on the basis of the space density of bzbs and bzqs in the parameter space of its principal components , thereby permitting us to determine if a selected @xmath1-ray blazar candidate is more likely to be a bzb or a bzq . finally , we ranked all the _ wise _ sources within each _ search region _ and selected as best candidate counterpart for the ugs the one with the highest class ; when more than one candidate of the same class was present , we chose the one closest to the @xmath1-ray position as best one . for the ugss without @xmath1-ray analysis flags we found 75 @xmath1-ray blazar candidates out of the 299 objects analyzed : 8 sources have 2 candidates , 1 source has 3 , and 1 source has 4 candidates , while 52 associations are unique . we found 2 @xmath1-ray blazar candidates of class a , 12 of class b and 61 of class c , respectively , in the whole sample of 75 sources ; 32 of them are classified as bzb type , 29 as bzq type and the remaining 14 are still uncertain ( see * ? ? ? * for more details ) . all our @xmath1-ray blazar candidates have a _ signal - to - noise ratio _ systematically larger than 10.9 in the _ wise _ band centered at 12@xmath3 m and larger than @xmath020 for the 3.4@xmath3 m and 4.6@xmath3 m nominal bands . for all these 75 sources we performed a cross correlation with the major radio , infrared , optical , and x - ray surveys ( see section [ sec : counterparts ] ) . in the sample of ugss with @xmath1-ray analysis flags we found 71 @xmath1-ray blazar candidates out of the 291 objects investigated : 6 sources have 2 candidates , 4 sources have 3 candidates , 2 sources have 4 and 6 candidates , respectively , while 35 associations are unique . we found 8 @xmath1-ray blazar candidates of class a , 20 of class b and 43 of class c , respectively , in the whole sample of 71 sources ; 36 of them are classified as bzb type , 22 as bzq type and the remaining 13 are still uncertain ( see * ? ? ? * ) . we also performed the cross correlation with the major radio , infrared , optical , and x - ray databes for these 71 ugss listed in the section [ sec : counterparts ] . for the agu sample we found 125 @xmath1-ray blazar candidates out of the 210 sources analyzed : 10 sources have 2 candidates within their _ search region _ , while the remaining 105 candidates have unique associations . there are 10 @xmath1-ray blazar candidates of class a , 39 of class b and 76 of class c , respectively , in the whole sample of 125 sources ; 52 out of 125 are classified as bzb type on the basis of the ir colors of blazars of similar type , 39 as bzq type and the remaining 34 are still uncertain ( see * ? ? ? * for more details ) . eighty - seven sources out of 125 associations correspond to those reported in the 2lac or in the 2fgl . all our @xmath1-ray blazar candidates have a _ signal - to - noise ratio _ systematically larger than 10.9 in the _ wise _ band centered at 12@xmath3 m and larger than @xmath020 for the 3.4@xmath3 m and 4.6@xmath3 m nominal bands . in these case we did not provide any additional radio or x - ray information since it is already present in both the 2lac and the 2fgl , while a multifrequency investigation has been performed for the remaining 38 . additional ir information for all the agus associated will be discussed in section [ sec : infrared ] . the fraction of sources for which we have been able to find a @xmath1-ray blazar counterpart is about @xmath015 - 20% lower than presented in previous analyses of ugss @xcite and agus @xcite , respectively . this difference occurs because a more conservative approach has been adopted in the new parametrization of the @xmath2 . we not limit blazar candidates to those having the _ scores _ higher than 30% of the entire distribution of @xmath1-ray emitting blazars @xcite , rather than 10% as in the previous analysis . these choices made our association method more efficient , so decreasing the number of _ wise _ sources with ir colors similar to those of the @xmath1-ray blazar population . in addition , we now use a _ search region _ of radius @xmath8 instead of that at 99.9% level of confidence , to be consistent with the associations of the 2fgl and the 2lac catalogs . all the sources listed in this work as @xmath1-ray blazar candidates were also selected in our previous analysis based on _ wise _ preliminary data analysis @xcite . we note that only three ir _ wise _ sources have the contamination and confusion " flag that might indicate a _ wise _ spurious detection of an artifact in all bands ( e.g. , * ? ? ? it occurs for _ wise _ j085238.73 - 575529.4 within the agus , _ wise _ j084121.63 - 355505.9 in the ugs sample , and wise j125357.07 - 583322.3 among the ugs with @xmath1-ray analysis flags . the large majority ( i.e. , @xmath090% ) of the _ wise _ sources considered do not show any _ wise _ analysis flags , with 10% clean in at least two ir bands . finally , we remark that several @xmath1-ray pulsars have been identified since the release of the 2fgl where they were listed as ugss however , we tested these ugss and we did not find any _ wise _ blazar - like counterpart associable to them . thus , in agreement with other gamma - ray pulsars listed in the the public list of lat - detected gamma - ray pulsars.[multiblock footnote omitted ] we searched in the following major radio , infrared , optical and x - ray surveys as well as in the nasa extragalactic database ( ned)[multiblock footnote omitted ] for possible counterparts within 3.3 of our @xmath1-ray blazar candidates , selected with the _ wise_association method , to see if additional information could confirm their blazar - like nature . the angular separation of 3.3 from the _ wise _ position was chosen on the basis of the statistical analysis previously performed to assign a _ wise _ counterpart to each roma - bzcat source @xcite developed following the approach described in maselli et al . ( 2012a , 2012b ) . in particular , we found that for all radii larger than 3.3 the increase in the number of ir sources positionally associated with roma - bzcat blazars becomes systematically lower than the increase in number of random associations . this choice of radius results in zero multiple matches . for the radio counterparts we used the nrao vla sky survey ( nvss ; * ? ? ? * - n ) , the vla faint images of the radio sky at twenty - centimeters ( first ; * ? ? ? * ; * ? ? ? * - f ) , the sydney university molonglo sky survey ( sumss ; * ? ? ? * - s ) and the australia telescope 20 ghz survey ( at20 g ; * ? ? ? * - a ) ; for the infrared we used the two micron all sky survey ( 2mass ; * ? ? ? * - m ) since each _ wise _ source is already associated with the closest 2mass source by the default catalog ( see * ? ? ? * for more details ) . we also marked sources that are variable when having the _ variability flag _ higher than 5 in at least one band as in the _ wise _ all - sky catalog @xcite . then , we also searched for optical counterparts , with possible spectra available , in the sloan digital sky survey ( sdss ; e.g. * ? ? ? * ; * ? ? ? * - s ) , in the six - degree - field galaxy redshift survey ( 6dfgs ; * ? ? ? * ; * ? ? ? * - 6 ) ; while for the high energy we looked in the soft x - rays using the rosat all - sky survey ( rass ; * ? ? * - x ) . a deeper x - ray analysis based on the pointed observations present in the _ xmm - newton _ , _ chandra _ , _ swift _ and _ suzaku _ archives will be performed in a forthcoming paper @xcite . we also considered ned for additional information . we also searched in the usno - b catalog @xcite for the optical counterparts of our @xmath1-ray blazar candidates within 3.3 ; this cross correlation will be useful to prepare future follow up observations and the complete list of sources together with their optical magnitudes is reported in appendix . in table [ tab : sum ] we summarize all the multifrequency information for the ugs samples , without and with the @xmath1-ray analysis flags , respectively , while all the details are given in in table [ tab : ugs ] and table [ tab : ugf ] . in table [ tab : agu1 ] and table [ tab : agu2 ] we report our findings the agus . in each table we report the 2fgl source name , together with that of the _ wise _ associated counterpart and a generic one from the surveys cited above . we also report the ir _ wise _ colors , the type and the class of each candidate derived by our association procedure , the notes regarding the multifrequency archival analysis , as the optical classification , and , if known , the redshift . in table [ tab : agu1 ] and table [ tab : agu2 ] , we also indicate if the selected source is the same associated by the 2fgl and the 2lac . figure [ fig : wise ] shows the 3-dimensional color plot comparing the ir colors of the selected @xmath1-ray blazar candidates with the blazar population that constitutes the @xmath2 . ( known @xmath1-ray blazars are indicated in yellow ) in comparison with the selected @xmath1-ray blazar candidates : ugss ( red ) and agus ( black).,width=332,height=340 ] & & ugs ( no @xmath1-flags ) & ugss ( @xmath1-flags ) + nvss & radio & 19/75 & 4/71 + first & radio & 6/75 & 1/71 + sumss & radio & 7/75 & 0/71 + 2mass & infrared & 43/75 & 47/71 + 6dfgs & optical & 1/75 & 1/71 + sdss & optical & 13/75 & 1/71 + rosat & x - ray & 3/75 & 1/71 + + [ tab : sum ] in the ugs sample of sources without @xmath1-ray analysis flags , 19 have a counterpart in the nvss ; 7 in the sumss and 6 only in the first ( 5 in common with the previous 19 in the nvss ) . in the list of ugss with @xmath1-ray analysis flags , we found only 4 sources having a radio counterpart in the nvss , one also detected in the first , but none in the sumss or in the at20 g catalogs . in figure [ fig : radio ] we show the archival nvss radio image of _ wise _ j134706.89 - 295842.3 , the candidate low - energy counterpart of 2fglj1347.0 - 2956 . within the agu sample , 12 sources out the 38 new associations proposed have unique counterparts in one of the considered radio survey . two of them : bzuj1239 + 0730 and bzuj1351 - 2912 , were also classified as blazars of uncertain type in the roma - bzcat ( e.g. , * ? ? ? * ) , while the remaining one are divided as 6 in the nvss , 1 in the first , 2 in the sumss and 1 in the at20 g . in the ugs sample of sources without @xmath1-ray analysis flags , there are 43 _ wise _ candidates with counterparts in the 2mass catalog : 10 out of 75 are variable infrared sources according to the same criterion previously described . the large majority ( 47 out of 71 ) of the ugss , in the sample with @xmath1-ray analysis flags , have counterparts in the 2mass catalog and 15 out of 71 are variable according to the _ wise_all - sky catalog . of the 125 _ wise _ candidates counterparts of the agus , 59 are detected in 2mass , as generally expected for blazars ( e.g. , * ? ? ? in addition , 25 @xmath1-ray blazar candidates out of 125 have the variability flag in the _ wise _ catalog with a value higher than 5 in at least one band , suggesting that their ir emission is not likely arising from dust . in the sample of ugss without @xmath1-ray analysis flags , 13 sources have been found with a counterpart in the sdss , 4 with spectroscopic information ( table [ tab : ugs ] ) . among these 4 sources , two are broad line quasars , promising to be blazar - like sources of bzq type . one is a seyfert galaxy : sdss j015910.05 + 010514.5 is a contaminant of our association procedure ( although our method suggests a better candidate , for 2fglj0158.4 + 0107 ) . the remaining one , nvss j161543 + 471126 shows the optical spectrum similar to that of an x - ray bright , optically normal galaxy ( xbong * ? ? ? the source sdss j015836.23 + 010632.0 , another candidate counterpart of 2fglj0158.4 + 0107 is described as a quasar at redshift 0.723 in schneider et al . ( 2007 ) and hu et al . in addition to these 4 sources , spectroscopic information is also available for wise j230010.16 - 360159.9 a possible low - energy counterpart of 2fglj2300.0 - 3553 , classified as quasar according to jones et al . a quasar - like spectrum is then available for sdss j161434.67 + 470420.0 candidate counterpart of 2fglj1614.8 + 4703 . the search for the optical counterparts for ugss with @xmath1-ray analysis flags was less successful . only one source has an optical , counterpart : wise j131552.98 - 073301.9 , associated with 2fglj1315.6 - 0730 . this source has a counterpart in both the nvss and in the first radio survey . according to bauer et al . ( 2009 ) , this source is also variable in the optical and it was therefore selected as a blazar candidate . in figure [ fig : optical2 ] we show the archival sdss spectrum of the _ wise _ j161434.67 + 470420.1 candidate as the low energy counterpart of 2fglj1614.8 + 4703 . we found only 1 @xmath1-ray blazar candidate in the agu sample with a counterpart in the sdss , while 4 of them have a 6dfgs source lying 3.3 from their _ wise _ position . in the case of wise j033200.72 - 111456.1 associated with 2fglj0332.5 - 1118 , we also found that its 6dfgs optical spectrum appear to be featureless suggesting a bl lac classification @xcite . the same information has been found for wise j001920.58 - 815251.3 associated with 2fglj0018.8 - 8154 , for which the noisy , featureless 6dfgs optical spectrum points to a bl lac classification @xcite . 2fglj0823.0 + 4041 and 2fglj0858.1 - 1952 appear to be associated , both by the 2fgl catalog and our method to broad line quasars . wise j085805.36 - 195036.8 associated with 2fglj0858.1 - 1952 is also classified as a quasar at redshift 0.6597 by white et al . the archival 6dfgs spectrum of _ wise _ j001920.58 - 815251.3 the candidate low energy counterpart of 2fglj0018.8 - 8154 is available on ned ; the absence of features allows us to classify the source as a bzb . in the ugs sample without @xmath1-ray analysis flags , only 3 objects have x - ray counterparts in the rosat all - sky catalog : the seyfert 1 galaxy sdss j015910.05 + 010514.5 , the quasar sdss j161434.67 + 470420.0 ( both described in section [ sec : optical ] ) and wise j164619.95 + 435631.0 associated with 2fglj1647.0 + 4351 . in addition , sdss j161434.67 + 470420.0 is also detected in the _ chandra _ source catalog : cxo j161434.7 + 470419 as occurs nvss j161543 + 471126 , alias cxo j161541.2 + 471111 @xcite . in the ugs list of sources with @xmath1-ray analysis flags , there is only a single object detected in the rosat all - sky survey , namely wise j043947.48 + 260140.5 uniquely associated with the _ fermi _ source 2fglj0440.5 + 2554c and with a x - ray counterpart also in the _ chandra _ source catalog cxo j043947.5 + 260140 @xcite . in addition , wise j060659.94 - 061641.5 the unique counterpart of 2fglj0607.5 - 0618c , has the _ chandra _ counterpart cxo j060700.1 - 061641 @xcite . finally , in the agu sample of 38 new @xmath1-ray blazar candidates we found only 1 source in the rosat catalog , namely wise j181037.99 + 533501.5 , associated with the x - ray object 1rxs j181038.5 + 533458 and having a radio counterpart in the nvss . according to ned , wise j182352.33 + 431452.5 , associated with 2fglj1823.8 + 4312 , is also detected in the x - rays by _ chandra _ : cxo j182352.2 + 431452 @xcite . [ 12]-[22 ] & type & class & notes & z + name & name & name & mag & mag & mag & & & & + j0039.1 + 4331 & j003858.27 + 432947.0 & & 0.98(0.04 ) & 2.36(0.05 ) & 2.30(0.15 ) & bzb & c & m & ? + j0116.6 - 6153 & j011619.59 - 615343.5 & sumss j011619 - 615343 & 0.85(0.04 ) & 2.34(0.06 ) & 1.85(0.25 ) & bzb & c & s & ? + j0133.4 - 4408 & j013321.36 - 441319.4 & & 1.12(0.05 ) & 3.20(0.06 ) & 2.62(0.15 ) & bzq & c & & ? + & j013306.35 - 441421.3 & sumss j013306 - 441422 & 0.83(0.03 ) & 2.25(0.05 ) & 1.92(0.22 ) & bzb & c & s & ? + j0143.6 - 5844 & j014347.39 - 584551.3 & sumss j014347 - 584550 & 0.68(0.03 ) & 1.93(0.06 ) & 1.89(0.29 ) & bzb & c & s , m & ? + j0158.4 + 0107 & j015836.25 + 010632.1 & sdss j015836.23 + 010632.0 & 1.05(0.03 ) & 2.50(0.05 ) & 2.33(0.11 ) & und & c & m , s , qso & 0.723 ? + & j015757.45 + 011547.8 & & 1.06(0.04 ) & 2.52(0.06 ) & 2.18(0.18 ) & und & c & & ? + & j015910.05 + 010514.7 & sdss j015910.05 + 010514.5 & 0.91(0.03 ) & 2.47(0.03 ) & 2.48(0.05 ) & und & c & m , s , x , sy1 & 0.217 + j0158.6 + 8558 & j015550.16 + 854745.1 & & 1.12(0.04 ) & 2.50(0.06 ) & 2.23(0.18 ) & und & c & & ? + & j014935.30 + 860115.3 & & 0.68(0.03 ) & 2.32(0.05 ) & 2.00(0.17 ) & bzb & c & m & ? + j0227.7 + 2249 & j022744.35 + 224834.3 & nvss j022744 + 224834 & 0.93(0.03 ) & 2.58(0.03 ) & 2.09(0.08 ) & bzb & b & n , m , v & ? + j0316.1 - 6434 & j031614.31 - 643731.4 & sumss j031614 - 643732 & 0.74(0.03 ) & 2.09(0.06 ) & 1.84(0.26 ) & bzb & c & s , m & ? + j0332.1 + 6309 & j033153.90 + 630814.1 & nvss j033153 + 630814 & 0.77(0.03 ) & 2.38(0.04 ) & 1.96(0.11 ) & bzb & b & n , m & ? + j0409.8 - 0357 & j040946.57 - 040003.4 & nvss j040946 - 040003 & 0.88(0.03 ) & 2.36(0.04 ) & 1.94(0.12 ) & bzb & b & n , m & ? + j0414.9 - 0855 & j041457.01 - 085652.0 & apmuks 041232.66 - 090420.3 & 1.00(0.04 ) & 2.63(0.08 ) & 2.31(0.23 ) & und & c & & ? + j0416.0 - 4355 & j041605.81 - 435514.6 & sumss j041605 - 435516 & 1.11(0.03 ) & 2.89(0.04 ) & 2.44(0.08 ) & bzq & b & s , m & ? + j0431.5 + 3622 & j043103.34 + 362158.7 & & 1.14(0.06 ) & 2.72(0.11 ) & 2.44(0.26 ) & bzq & c & & ? + j0555.9 - 4348 & j055531.59 - 435030.7 & & 1.36(0.05 ) & 3.14(0.06 ) & 2.39(0.18 ) & bzq & c & & ? + & j055618.74 - 435146.1 & sumss j055618 - 435146 & 0.90(0.03 ) & 2.49(0.04 ) & 2.14(0.13 ) & bzb & b & s , m & ? + j0602.7 - 4011 & j060237.10 - 401453.2 & & 0.96(0.03 ) & 2.47(0.03 ) & 2.35(0.06 ) & und & b & m & ? + j0644.6 + 6034 & j064459.38 + 603131.7 & & 0.97(0.04 ) & 2.59(0.06 ) & 2.50(0.15 ) & und & c & m & ? + j0713.5 - 0952 & j071223.28 - 094536.3 & & 1.16(0.04 ) & 3.12(0.05 ) & 2.27(0.08 ) & bzq & c & m & ? + j0723.9 + 2901 & j072354.83 + 285929.9 & nvss j072354 + 285930 & 1.14(0.05 ) & 2.89(0.05 ) & 2.40(0.11 ) & bzq & c & n , f & ? + j0744.1 - 2523 & j074401.10 - 252205.9 & & 1.09(0.04 ) & 2.57(0.03 ) & 2.69(0.05 ) & bzq & c & & ? + & j074402.19 - 252146.0 & & 1.00(0.03 ) & 2.00(0.03 ) & 2.16(0.03 ) & bzb & b & m & ? + j0746.0 - 0222 & j074627.03 - 022549.3 & nvss j074627 - 022549 & 0.67(0.04 ) & 2.09(0.07 ) & 1.99(0.30 ) & bzb & c & n , m & ? + j0756.3 - 6433 & j075624.60 - 643030.6 & & 0.87(0.03 ) & 2.21(0.06 ) & 2.11(0.18 ) & bzb & c & m , v & ? + j0807.0 - 6511 & j080729.66 - 650910.3 & & 1.12(0.05 ) & 3.08(0.07 ) & 2.46(0.17 ) & bzq & c & & ? + j0838.8 - 2828 & j083842.77 - 282830.9 & & 1.03(0.05 ) & 2.73(0.09 ) & 2.24(0.27 ) & und & c & & ? + j0841.3 - 3556 & j084121.63 - 355505.9 & nvss j084121 - 355506 & 0.79(0.03 ) & 2.23(0.03 ) & 1.77(0.11 ) & bzb & b & n , m & ? + j0844.9 + 6214 & j084406.81 + 621458.6 & sdss j084406.83 + 621458.5 & 0.68(0.03 ) & 2.07(0.05 ) & 2.23(0.14 ) & bzb & c & m , s & ? + j0855.4 - 4625 & j085548.43 - 462244.3 & & 1.19(0.03 ) & 2.32(0.03 ) & 2.16(0.04 ) & und & c & & ? + j0858.3 - 4333 & j085839.22 - 432642.7 & & 0.74(0.03 ) & 1.82(0.03 ) & 2.29(0.04 ) & bzb & c & m & ? + j0900.9 + 6736 & j090121.65 + 673955.8 & & 0.94(0.05 ) & 2.84(0.08 ) & 2.16(0.26 ) & und & c & m & ? + j0955.0 - 3949 & j095458.30 - 394655.0 & & 0.77(0.04 ) & 2.28(0.05 ) & 2.03(0.18 ) & bzb & c & m & ? + j1013.6 + 3434 & j101256.54 + 343648.8 & sdss j101256.54 + 343648.7 & 0.89(0.04 ) & 2.62(0.07 ) & 2.09(0.26 ) & bzb & c & m , s & ? + j1016.1 + 5600 & j101544.44 + 555100.7 & nvss j101544 + 555100 & 1.05(0.06 ) & 3.08(0.09 ) & 2.57(0.25 ) & bzq & c & n , f , s & ? + j1029.5 - 2022 & j102946.66 - 201812.6 & & 1.26(0.06 ) & 2.88(0.10 ) & 2.59(0.26 ) & bzq & c & m & ? + j1032.9 - 8401 & j103015.35 - 840308.7 & sumss j103014 - 840307 & 0.96(0.04 ) & 2.59(0.05 ) & 2.07(0.15 ) & bzb & c & s , v & ? + j1038.2 - 2423 & j103754.92 - 242544.5 & & 1.22(0.04 ) & 3.26(0.04 ) & 2.58(0.10 ) & bzq & c & m & ? + j1207.3 - 5055 & j120750.50 - 510314.9 & & 1.05(0.05 ) & 2.97(0.08 ) & 2.55(0.19 ) & bzq & c & & ? + & j120746.43 - 505948.6 & & 1.14(0.05 ) & 2.87(0.07 ) & 2.53(0.15 ) & bzq & c & & ? + j1254.2 - 2203 & j125422.47 - 220413.6 & nvss j125422 - 220413 & 0.66(0.04 ) & 2.31(0.08 ) & 1.77(0.36 ) & bzb & c & n , m , v & ? + j1259.8 - 3749 & j125949.80 - 374858.1 & nvss j125949 - 374856 & 0.70(0.04 ) & 2.10(0.08 ) & 1.99(0.33 ) & bzb & c & n , s , m , v & ? + j1340.5 - 0412 & j134042.02 - 041006.8 & nvss j134042 - 041006 & 0.71(0.04 ) & 2.11(0.08 ) & 1.85(0.41 ) & bzb & c & n , v & ? + j1346.0 - 2605 & j134621.08 - 255642.3 & & 1.14(0.06 ) & 2.94(0.10 ) & 2.58(0.26 ) & bzq & c & & ? + j1347.0 - 2956 & j134706.89 - 295842.3 & nvss j134706 - 295840 & 0.78(0.03 ) & 2.10(0.06 ) & 1.91(0.23 ) & bzb & c & n , s , m , v & ? + j1404.0 - 5244 & j140313.11 - 524839.5 & & 1.22(0.05 ) & 2.97(0.07 ) & 2.63(0.15 ) & bzq & c & & ? + j1517.2 + 3645 & j151649.26 + 365022.9 & nvss j151649 + 365023 & 0.95(0.03 ) & 2.63(0.04 ) & 2.07(0.12 ) & bzb & c & n , f , s , v & ? + j1612.0 + 1403 & j161118.10 + 140328.9 & sdss j161118.10 + 140328.7 & 1.06(0.06 ) & 3.15(0.09 ) & 2.55(0.22 ) & bzq & c & s & ? + j1614.8 + 4703 & j161450.96 + 465953.7 & sdss j161450.91 + 465953.6 & 1.20(0.04 ) & 2.78(0.06 ) & 2.54(0.16 ) & bzq & c & s & ? + & j161513.04 + 471355.2 & sdss j161513.04 + 471355.4 & 1.17(0.05 ) & 2.86(0.10 ) & 2.61(0.24 ) & bzq & c & s & ? + & j161434.67 + 470420.1 & sdss j161434.67 + 470420.0 & 1.08(0.03 ) & 3.12(0.03 ) & 2.19(0.04 ) & bzq & c & f , m , s , x , x , qso&1.86 + & j161541.22 + 471111.8 & nvss j161543 + 471126 & 0.74(0.03 ) & 2.28(0.04 ) & 2.28(0.09 ) & bzb & c & n , f , m , s , x , xb?&0.199 + j1622.8 - 0314 & j162225.35 - 031439.6 & & 1.10(0.06 ) & 3.19(0.09 ) & 2.67(0.18 ) & bzq & c & & ? + j1627.8 + 3219 & j162800.40 + 322414.0 & sdss j162800.39 + 322413.9 & 1.13(0.05 ) & 2.83(0.10 ) & 2.62(0.24 ) & bzq & c & s & ? + j1647.0 + 4351 & j164619.95 + 435631.0 & nvss j164619 + 435631 & 0.77(0.04 ) & 2.09(0.09 ) & 2.11(0.39 ) & bzb & c & n , f , s , x & ? + j1730.6 - 0353 & j173052.86 - 035247.1 & & 1.16(0.04 ) & 2.92(0.04 ) & 2.28(0.07 ) & bzq & b & m & ? + j1742.5 - 3323 & j174201.11 - 332607.9 & & 0.64(0.05 ) & 1.73(0.03 ) & 1.60(0.03 ) & bzb & b & m & ? + j1745.6 + 0203 & j174507.82 + 015442.5 & nvss j174507 + 015445 & 1.26(0.03 ) & 3.40(0.03 ) & 2.45(0.03 ) & bzq & a & + & j174526.95 + 020532.7 & & 0.94(0.03 ) & 2.57(0.03 ) & 2.27(0.07 ) & und & b & m & ? + j1759.2 - 3853 & j175903.29 - 384739.5 & & 0.58(0.04 ) & 1.93(0.03 ) & 1.50(0.02 ) & bzb & a & m & ? + j1842.3 + 2740 & j184201.25 + 274239.2 & & 1.22(0.06 ) & 3.03(0.08 ) & 2.43(0.22 ) & bzq & c & & ? + j1904.8 - 0705 & j190444.57 - 070740.1 & & 0.91(0.05 ) & 2.79(0.09 ) & 2.45(0.19 ) & und & c & m & ? + j1924.9 - 1036 & j192501.63 - 104316.3 & & 1.24(0.05 ) & 3.25(0.05 ) & 2.66(0.09 ) & bzq & c & m & ? + j2004.6 + 7004 & j200503.41 + 700236.3 & & 1.17(0.05 ) & 2.95(0.07 ) & 2.21(0.24 ) & bzq & c & & ? + & j200506.02 + 700439.3 & nvss j200506 + 700440 & 0.70(0.03 ) & 2.11(0.05 ) & 2.11(0.18 ) & bzb & c & n , v & ? + j2021.5 + 0632 & j202154.66 + 062908.7 & & 0.96(0.04 ) & 2.67(0.06 ) & 2.47(0.10 ) & und & c & m & ? + & j202155.45 + 062913.7 & nvss j202155 + 062914 & 0.80(0.03 ) & 2.09(0.05 ) & 1.78(0.17 ) & bzb & c & n , m & ? + j2114.1 + 5440 & j211508.92 + 544815.7 & & 1.11(0.03 ) & 2.50(0.03 ) & 2.52(0.04 ) & bzq & b & m & ? + j2133.9 + 6645 & j213349.21 + 664704.3 & nvss j213349 + 664706 & 0.67(0.04 ) & 2.12(0.06 ) & 1.86(0.22 ) & bzb & c & n , v & ? + j2134.6 - 2130 & j213430.18 - 213032.6 & nvss j213430 - 213032 & 0.77(0.04 ) & 2.26(0.08 ) & 1.78(0.42 ) & bzb & c & n , m & ? + j2300.0 - 3553 & j230010.16 - 360159.9 & 6df j2300101 - 360200 & 1.17(0.06 ) & 3.36(0.08 ) & 2.43(0.22 ) & bzq & c & m,6,qso & 2.357 + j2319.3 - 3830 & j232000.11 - 383511.4 & mrss 347 - 103293 & 1.12(0.05 ) & 3.06(0.07 ) & 2.60(0.16 ) & bzq & c & & ? + j2358.4 - 1811 & j235828.61 - 181526.6 & & 1.09(0.04 ) & 2.60(0.06 ) & 2.46(0.15 ) & und & c & m & ? + + col . ( 1 ) 2fgl name ( 2 ) _ wise _ name . ( 3 ) other name if present in literature and in the following order : roma - bzcat , nvss , sdss , at20 g , ned . ( 4,5,6 ) infrared colors from the _ wise _ all sky catalog corrected for galactic extinciton . values in parentheses are 1@xmath12 uncertainties ( 7 ) type of candidate according to our method : bzb - bzq - und ( undetermined ) ( 8) class of candidate according to our method . ( 9 ) notes : n = nvss , f = first , s = sumss , a = at20 g , m = 2mass , s = sdss dr9 , 6 = 6dfgs , x = _ xmm - newton _ or _ chandra _ , x = rosat ; qso = quasar , sy = seyfert , lnr = liner , bl = bl lac , xb = x - ray bright optically inactive galaxies ; v = variability in _ wise _ ( var_flag @xmath11 5 in at least one band ) . + col . ( 10 ) redshift : ( ? ) = unknown , ( number ? ) = uncertain . [ tab : ugs ] [ 12]-[22 ] & type & class & notes & z + name & name & name & mag & mag & mag & & & & + j0233.9 + 6238c & j023418.09 + 624207.8 & & 0.80(0.03 ) & 2.02(0.04 ) & 2.31(0.06 ) & bzb & c & m & ? + & j023238.07 + 623651.9 & & 1.25(0.05 ) & 3.05(0.06 ) & 2.46(0.11 ) & bzq & c & & ? + j0341.8 + 3148c & j034204.35 + 314711.4 & & 0.48(0.03 ) & 1.70(0.03 ) & 1.78(0.08 ) & bzb & b & m & ? + & j034141.11 + 314804.5 & & 0.93(0.03 ) & 2.47(0.05 ) & 2.20(0.13 ) & und & c & m , v & ? + & j034158.52 + 314855.7 & & 0.64(0.04 ) & 1.37(0.03 ) & 1.50(0.02 ) & bzb & c & m & ? + j0423.4 + 5612 & j042430.44 + 561525.8 & & 1.19(0.06 ) & 2.98(0.13 ) & 2.47(0.26 ) & bzq & c & & ? + j0440.5 + 2554c & j043947.48 + 260140.5 & cxo j043947.5 + 260140 & 0.47(0.03 ) & 1.47(0.03 ) & 1.90(0.04 ) & bzb & a & m , x , x & ? + j0543.2 - 0120c & j054324.78 - 011545.6 & & 0.69(0.03 ) & 1.64(0.04 ) & 1.86(0.11 ) & bzb & b & m & ? + j0547.5 - 0141c & j054758.25 - 013616.6 & & 1.06(0.04 ) & 2.60(0.06 ) & 2.20(0.19 ) & und & c & & ? + j0607.5 - 0618c & j060659.94 - 061641.5 & cxo j060700.1 - 061641 & 1.17(0.04 ) & 2.45(0.07 ) & 2.33(0.18 ) & ? & c & x & ? + j0616.6 + 2425 & j061549.51 + 241654.0 & & 1.12(0.06 ) & 2.94(0.09 ) & 2.44(0.23 ) & bzq & c & & ? + & j061641.04 + 241138.4 & & 0.87(0.03 ) & 1.76(0.07 ) & 1.88(0.23 ) & bzb & c & m & ? + j0620.8 - 2556 & j062108.68 - 255757.9 & nvss j062108 - 255757 & 0.95(0.05 ) & 2.75(0.08 ) & 2.26(0.26 ) & ? & c & n & ? + j0631.7 + 0428 & j063104.12 + 042012.6 & & 1.08(0.03 ) & 2.44(0.03 ) & 2.56(0.03 ) & bzq & a & m & ? + j0634.3 + 0356c & j063459.32 + 040808.6 & & 1.13(0.03 ) & 2.11(0.03 ) & 2.73(0.04 ) & bzq & c & m , v & ? + & j063519.84 + 035047.1 & & 1.36(0.04 ) & 2.73(0.05 ) & 2.36(0.08 ) & bzq & c & m & ? + j0637.0 + 0416c & j063705.96 + 042537.2 & & 0.82(0.03 ) & 2.17(0.04 ) & 2.11(0.06 ) & bzb & b & m & ? + & j063701.93 + 042037.2 & & 0.86(0.03 ) & 2.17(0.03 ) & 2.17(0.03 ) & bzb & a & m , v & ? + & j063703.09 + 042146.1 & & 0.60(0.03 ) & 1.99(0.07 ) & 2.06(0.12 ) & bzb & c & m , v & ? + & j063647.19 + 042058.7 & & 0.52(0.03 ) & 1.79(0.05 ) & 1.93(0.18 ) & bzb & c & m & ? + j0708.5 - 1020c & j070807.98 - 102743.9 & & 0.55(0.03 ) & 1.55(0.05 ) & 1.97(0.11 ) & bzb & c & m & ? + & j070809.69 - 102805.8 & & 1.35(0.03 ) & 2.91(0.03 ) & 2.50(0.04 ) & bzq & b & & ? + & j070806.04 - 102736.2 & & 0.62(0.03 ) & 1.87(0.03 ) & 2.07(0.05 ) & bzb & b & m & ? + & j070813.96 - 102840.2 & & 0.89(0.03 ) & 1.73(0.05 ) & 2.21(0.09 ) & bzb & c & & ? + & j070826.05 - 103001.2 & & 1.18(0.04 ) & 2.88(0.04 ) & 2.48(0.05 ) & bzq & b & m , v & ? + & j070816.06 - 102832.0 & & 0.72(0.03 ) & 1.51(0.06 ) & 1.95(0.18 ) & bzb & c & m & ? + j0742.7 - 3113 & j074303.55 - 312057.3 & & 1.04(0.05 ) & 2.57(0.10 ) & 2.22(0.34 ) & und & c & & ? + j0748.5 - 2204 & j074835.46 - 215740.0 & & 1.32(0.05 ) & 2.80(0.07 ) & 2.68(0.14 ) & bzq & c & & ? + j0900.5 - 4441c & j090039.83 - 443510.0 & 2masx j09003978 - 4435094 & 1.08(0.03 ) & 1.48(0.03 ) & 2.32(0.02 ) & bzb & c & m , v & ? + j0914.1 - 4756 & j091319.84 - 475730.8 & & 1.08(0.03 ) & 1.71(0.03 ) & 2.20(0.03 ) & bzb & a & & ? + j0922.2 - 5214c & j092154.24 - 521236.1 & & 1.19(0.03 ) & 2.43(0.03 ) & 2.52(0.02 ) & bzq & a & m & ? + j1027.4 - 5730c & j102703.87 - 572830.7 & & 1.14(0.04 ) & 2.81(0.05 ) & 2.59(0.06 ) & bzq & b & m , v & ? + j1059.9 - 2051 & j110025.72 - 205333.4 & 2masx j11002568 - 2053333 & 0.89(0.03 ) & 2.46(0.05 ) & 1.99(0.14 ) & bzb & b & m & ? + j1208.6 - 2257 & j120816.33 - 224921.9 & & 0.98(0.04 ) & 2.25(0.05 ) & 2.00(0.18 ) & bzb & c & m & ? + j1248.6 - 5510 & j124946.07 - 550758.6 & & 0.95(0.04 ) & 2.71(0.05 ) & 2.52(0.12 ) & und & c & & ? + j1255.8 - 5828 & j125459.44 - 582009.5 & & 1.20(0.03 ) & 2.61(0.04 ) & 2.25(0.05 ) & und & c & m & ? + & j125357.07 - 583322.3 & & 0.62(0.04 ) & 2.04(0.03 ) & 1.56(0.02 ) & bzb & a & m , v & ? + j1315.6 - 0730 & j131543.62 - 073659.0 & & 0.93(0.04 ) & 2.39(0.06 ) & 2.15(0.21 ) & bzb & c & m & ? + & j131552.98 - 073301.9 & nvss j131552 - 073301 & 0.87(0.03 ) & 2.27(0.04 ) & 2.04(0.08 ) & bzb & b & n , f , m,6,v , bl ? & ? + j1324.4 - 5411 & j132415.49 - 541104.4 & & 1.33(0.05 ) & 2.88(0.07 ) & 2.46(0.18 ) & bzq & c & & ? + j1345.8 - 3356 & j134543.05 - 335643.3 & nvss j134543 - 335643 & 0.82(0.04 ) & 2.31(0.06 ) & 2.09(0.20 ) & bzb & c & n , m & ? + j1407.4 - 2948 & j140818.86 - 294203.2 & & 1.10(0.06 ) & 2.82(0.12 ) & 2.70(0.28 ) & bzq & c & & ? + j1512.5 - 6247c & j151156.72 - 625231.1 & & 1.20(0.10 ) & 2.68(0.09 ) & 2.68(0.18 ) & bzq & c & & ? + j1624.2 - 2124 & j162343.89 - 210707.0 & & 1.34(0.05 ) & 2.86(0.08 ) & 2.74(0.16 ) & bzq & c & & ? + j1632.6 - 2328c & j163306.53 - 233207.3 & & 1.07(0.03 ) & 2.71(0.04 ) & 2.43(0.09 ) & und & c & m & ? + j1639.8 - 4921c & j163907.38 - 492605.6 & & 0.64(0.05 ) & 1.66(0.03 ) & 1.64(0.03 ) & bzb & b & m , v & ? + j1716.6 - 0526c & j171717.66 - 052520.1 & & 1.25(0.04 ) & 2.95(0.04 ) & 2.37(0.06 ) & bzq & b & m & ? + j1747.2 - 3507 & j174741.23 - 350334.3 & & 0.58(0.05 ) & 1.80(0.03 ) & 1.41(0.02 ) & bzb & b & m & ? + j1749.7 - 3134c & j174943.85 - 314054.7 & & 0.92(0.03 ) & 1.15(0.03 ) & 1.74(0.02 ) & bzb & c & m , v & ? + & j174949.84 - 313045.6 & & 1.10(0.03 ) & 1.95(0.03 ) & 2.30(0.02 ) & ? & a & m & ? + j1754.1 - 2930 & j175414.40 - 293326.6 & & 0.68(0.06 ) & 1.81(0.04 ) & 1.48(0.05 ) & bzb & c & m & ? + j1829.8 - 0204c & j182927.72 - 020531.5 & & 0.55(0.03 ) & 1.52(0.03 ) & 1.94(0.04 ) & bzb & b & m & ? + & j182958.23 - 015805.1 & & 0.71(0.03 ) & 1.24(0.03 ) & 1.77(0.04 ) & bzb & a & m & ? + & j183009.32 - 020723.2 & & 0.75(0.03 ) & 1.48(0.04 ) & 1.84(0.06 ) & bzb & b & m , v & ? + j1835.4 + 1036 & j183551.92 + 103056.8 & & 0.94(0.03 ) & 2.82(0.04 ) & 2.56(0.06 ) & und & c & m & ? + j1835.4 + 1349 & j183522.00 + 135733.9 & & 1.12(0.04 ) & 2.61(0.05 ) & 2.20(0.14 ) & und & c & m & ? + & j183535.34 + 134848.8 & & 0.65(0.04 ) & 2.14(0.05 ) & 2.10(0.15 ) & bzb & c & m & ? + j1837.9 + 3821 & j183828.80 + 382704.3 & & 1.04(0.05 ) & 2.64(0.09 ) & 2.52(0.24 ) & und & c & & ? + & j183656.31 + 382232.8 & & 1.15(0.04 ) & 2.97(0.04 ) & 2.53(0.08 ) & bzq & b & & ? + & j183837.16 + 381900.5 & & 0.90(0.04 ) & 2.14(0.08 ) & 1.84(0.34 ) & bzb & c & & ? + j1844.3 + 1548 & j184425.36 + 154645.9 & nvss j184425 + 154646 & 0.80(0.03 ) & 2.30(0.04 ) & 2.06(0.08 ) & bzb & b & n , m & ? + j1844.9 - 1116 & j184456.29 - 111352.1 & & 0.67(0.06 ) & 1.63(0.03 ) & 1.50(0.02 ) & bzb & b & m & ? + j1857.6 + 0211 & j185727.36 + 021216.4 & & 0.77(0.07 ) & 2.26(0.07 ) & 1.81(0.02 ) & bzb & b & m , v & ? + j1944.3 + 7325 & j194343.83 + 731738.7 & & 1.30(0.04 ) & 2.84(0.06 ) & 2.36(0.16 ) & bzq & c & & ? + j1958.6 + 4020 & j195842.28 + 401125.8 & & 1.11(0.05 ) & 2.82(0.06 ) & 2.41(0.14 ) & bzq & c & & ? + j2042.0 + 4252c & j204241.58 + 424150.3 & & 1.28(0.05 ) & 2.38(0.07 ) & 2.63(0.07 ) & bzq & c & m , v & ? + j2124.0 - 1513 & j212423.63 - 152558.2 & & 1.14(0.06 ) & 3.15(0.11 ) & 2.56(0.26 ) & bzq & c & & ? + j2128.7 + 5824 & j212810.83 + 583336.6 & & 1.21(0.04 ) & 2.82(0.06 ) & 2.22(0.17 ) & bzq & c & & ? + & j212820.65 + 582053.4 & & 0.54(0.03 ) & 1.86(0.03 ) & 1.78(0.10 ) & bzb & b & m , v & ? + & j212900.37 + 583128.0 & & 1.04(0.04 ) & 2.37(0.13 ) & 2.43(0.29 ) & und & c & & ? + j2201.2 + 5926 & j215953.42 + 591227.1 & & 0.94(0.04 ) & 2.13(0.05 ) & 2.17(0.10 ) & bzb & b & m & ? + + col . ( 1 ) 2fgl name ( 2 ) _ wise _ ( 3 ) other name if present in literature and in the following order : roma - bzcat , nvss , sdss , at20 g , ned ( 4,5,6 ) infrared colors from the _ wise _ all sky catalog corrected for galactic extinciton . values in parentheses are 1@xmath12 uncertainties . + col . ( 7 ) type of candidate according to our method : bzb - bzq - und ( undetermined ) . ( 8) class of candidate according to our method . ( 9 ) notes : n = nvss , f = first , s = sumss , a = at20 g , m = 2mass , s = sdss dr9 , 6 = 6dfgs , x = _ xmm - newton _ or _ chandra _ , x = rosat ; qso = quasar , sy = seyfert , lnr = liner , bl = bl lac ; v = variability in _ wise _ ( var_flag @xmath11 5 in at least one band ) . + col . ( 10 ) redshift : ( ? ) = unknown , ( number ? ) = uncertain . [ tab : ugf ] [ 12]-[22 ] & type & class & notes & z & reassoc . + name & name & name & mag & mag & mag & & & & & flag + j0009.1 + 5030 & j000922.76 + 503028.8 & nvssj000922 + 503028 & 0.72(0.03 ) & 2.19(0.05 ) & 1.95(0.21 ) & bzb & c & m , v & ? & yes + j0018.8 - 8154 & j001920.58 - 815251.3 & sumssj001921 - 815251 & 0.87(0.03 ) & 2.30(0.03 ) & 1.98(0.07 ) & bzb & b & m , bl & ? & yes + j0022.2 - 1853 & j002209.25 - 185334.7 & nvssj002209 - 185332 & 0.87(0.03 ) & 2.27(0.04 ) & 1.83(0.12 ) & bzb & b & m & ? & yes + j0022.3 - 5141 & j002200.08 - 514024.2 & sumssj002159 - 514026 & 0.81(0.03 ) & 2.23(0.04 ) & 1.87(0.14 ) & bzb & b & m , v & ? & yes + j0045.5 + 1218 & j004543.33 + 121712.0 & nvssj004543 + 121710 & 0.80(0.03 ) & 2.23(0.05 ) & 2.03(0.14 ) & bzb & b & m & ? & yes + j0051.4 - 6241 & j005116.62 - 624204.3 & sumssj005116 - 624205 & 0.69(0.04 ) & 2.20(0.06 ) & 1.75(0.32 ) & bzb & c & m & ? & yes + j0059.7 - 5700 & j005846.56 - 565911.4 & sumssj005846 - 565912 & 1.03(0.03 ) & 2.92(0.04 ) & 2.64(0.07 ) & bzq & b & m & ? & yes + j0110.3 + 6805 & j011012.84 + 680541.1 & bzuj0110 + 6805 & 0.67(0.03 ) & 2.12(0.03 ) & 2.02(0.04 ) & bzb & a & m & ? & yes + j0134.4 + 2636 & j013428.19 + 263843.0 & nvssj013427 + 263842 & 0.80(0.04 ) & 2.00(0.08 ) & 1.93(0.33 ) & bzb & c & m , v & ? & yes + j0156.4 + 3909 & j015631.40 + 391430.9 & nvssj015631 + 391431 & 1.13(0.04 ) & 2.76(0.05 ) & 2.47(0.11 ) & bzq & c & m , v & ? & yes + j0156.5 - 2419 & j015606.46 - 241754.3 & & 1.21(0.06 ) & 2.93(0.11 ) & 2.63(0.26 ) & bzq & c & & ? & no + j0207.9 - 6832 & j020750.91 - 683755.1 & sumssj020750 - 683755 & 1.03(0.04 ) & 2.70(0.05 ) & 2.28(0.12 ) & und & c & & ? & yes + j0210.7 - 5102 & j021046.19 - 510101.8 & bzuj0210 - 5101 & 1.13(0.03 ) & 2.97(0.03 ) & 2.39(0.04 ) & bzq & a & m & 1.003 & yes + j0238.2 - 3905 & j023749.42 - 390050.3 & & 0.92(0.03 ) & 2.40(0.04 ) & 2.36(0.08 ) & bzb & c & m,6 & 0.323 ? & no + & j023800.62 - 390504.6 & nvssj023800 - 390504 & 0.68(0.04 ) & 2.18(0.07 ) & 1.91(0.27 ) & bzb & c & m & ? & yes + j0248.6 + 8440 & j024948.30 + 843556.9 & nvssj024948 + 843556 & 0.92(0.03 ) & 2.63(0.04 ) & 2.13(0.09 ) & bzb & c & m & ? & yes + j0253.4 + 3218 & j025333.64 + 321720.8 & nvssj025333 + 321721 & 1.12(0.04 ) & 2.87(0.06 ) & 2.46(0.16 ) & bzq & c & m & ? & yes + j0309.3 - 0743 & j030943.23 - 074427.4 & nvssj030943 - 074427 & 0.70(0.03 ) & 2.15(0.06 ) & 2.01(0.21 ) & bzb & c & m , v & ? & yes + j0332.5 - 1118 & j033200.72 - 111456.1 & 6dfj0332006 - 111456 & 1.30(0.03 ) & 3.05(0.03 ) & 2.63(0.04 ) & bzq & a & m,6,bl & ? & no + & j033223.25 - 111950.6 & nvssj033223 - 111951 & 0.96(0.03 ) & 2.58(0.03 ) & 2.23(0.05 ) & und & c & m , v & ? & yes + j0333.7 + 2918 & j033349.00 + 291631.6 & nvssj033349 + 291631 & 0.77(0.03 ) & 2.02(0.05 ) & 2.18(0.13 ) & bzb & c & m & ? & yes + j0334.3 + 6538 & j033356.74 + 653656.0 & nvssj033356 + 653656 & 0.75(0.03 ) & 2.07(0.05 ) & 1.73(0.17 ) & bzb & c & m & ? & yes + j0424.3 - 5332 & j042347.22 - 533026.6 & & 1.27(0.04 ) & 2.86(0.04 ) & 2.30(0.12 ) & bzq & c & m & ? & no + & j042504.26 - 533158.3 & sumssj042504 - 533158 & 0.99(0.03 ) & 2.58(0.03 ) & 2.12(0.04 ) & bzb & c & m , v & ? & yes + j0433.4 - 6029 & j043334.08 - 603013.7 & sumssj043333 - 603014 & 1.15(0.04 ) & 3.04(0.04 ) & 2.49(0.07 ) & bzq & b & m & ? & yes + j0433.9 - 5726 & j043344.12 - 572613.3 & & 0.90(0.04 ) & 2.23(0.07 ) & 2.20(0.27 ) & bzb & c & & ? & no + j0438.8 - 4521 & j043900.84 - 452222.6 & sumssj043900 - 452223 & 1.14(0.04 ) & 3.09(0.04 ) & 2.45(0.08 ) & bzq & b & & ? & yes + j0440.4 + 1433 & j044021.14 + 143757.0 & nvssj044021 + 143757 & 1.13(0.05 ) & 2.88(0.08 ) & 2.54(0.20 ) & bzq & c & & ? & yes + j0456.5 + 2658 & j045617.36 + 270221.1 & nvssj045617 + 270221 & 0.95(0.04 ) & 2.68(0.07 ) & 2.35(0.17 ) & und & c & & ? & yes + j0505.9 + 6116 & j050558.78 + 611335.9 & nvssj050558 + 611336 & 0.65(0.04 ) & 1.83(0.08 ) & 1.69(0.45 ) & bzb & c & m & ? & yes + j0506.7 - 5435 & j050657.80 - 543503.9 & sumssj050657 - 543459 & 0.74(0.03 ) & 1.89(0.05 ) & 2.02(0.19 ) & bzb & c & m & ? & yes + j0508.1 - 1936 & j050805.75 - 194721.6 & & 1.11(0.04 ) & 2.79(0.07 ) & 2.72(0.13 ) & bzq & c & m & ? & no + j0512.9 + 4040 & j051252.53 + 404143.7 & nvssj051252 + 404143 & 0.90(0.03 ) & 2.50(0.03 ) & 2.13(0.03 ) & bzb & a & m & ? & yes + j0525.5 - 6011 & j052537.74 - 601732.0 & & 1.11(0.04 ) & 3.25(0.05 ) & 2.68(0.12 ) & bzq & c & & ? & no + j0532.0 - 4826 & j053158.61 - 482736.0 & sumssj053158 - 482737 & 0.89(0.03 ) & 2.46(0.03 ) & 2.06(0.03 ) & bzb & a & v & ? & yes + j0537.7 - 5716 & j053748.95 - 571830.2 & sumssj053748 - 571828 & 0.83(0.03 ) & 2.34(0.04 ) & 2.16(0.09 ) & bzb & b & m & ? & yes + j0609.4 - 0248 & j060915.06 - 024754.6 & nvssj060915 - 024754 & 0.78(0.04 ) & 1.97(0.08 ) & 1.71(0.37 ) & bzb & c & m & ? & yes + j0621.9 + 3750 & j062157.63 + 375057.0 & nvssj062157 + 375056 & 1.14(0.05 ) & 2.94(0.07 ) & 2.52(0.17 ) & bzq & c & & ? & yes + j0644.2 - 6713 & j064428.06 - 671257.3 & sumssj064427 - 671257 & 1.06(0.03 ) & 2.78(0.02 ) & 2.26(0.03 ) & und & a & & ? & yes + j0647.8 - 6102 & j064740.85 - 605805.2 & sumssj064740 - 605804 & 1.10(0.04 ) & 2.96(0.06 ) & 2.34(0.16 ) & bzq & c & v & ? & yes + & j064806.55 - 610507.4 & & 1.38(0.04 ) & 3.03(0.05 ) & 2.48(0.12 ) & bzq & c & & ? & no + j0653.7 + 2818 & j065344.26 + 281547.5 & nvssj065343 + 281546 & 0.81(0.04 ) & 2.30(0.07 ) & 1.98(0.26 ) & bzb & c & n , m & ? & no + j0700.3 + 1710 & j070001.50 + 170921.8 & nvssj070001 + 170922 & 1.11(0.04 ) & 2.97(0.04 ) & 2.50(0.07 ) & bzq & b & & ? & yes + & j070046.29 + 171019.8 & & 1.08(0.04 ) & 2.62(0.08 ) & 2.18(0.24 ) & und & c & m & ? & no + j0702.7 - 1951 & j070242.90 - 195122.2 & nvssj070242 - 195123 & 1.02(0.04 ) & 2.70(0.04 ) & 2.15(0.10 ) & ? & b & m & ? & yes + j0703.1 - 3912 & j070312.64 - 391418.9 & nvssj070312 - 391418 & 0.94(0.03 ) & 2.51(0.04 ) & 2.15(0.09 ) & bzb & b & m & ? & yes + j0706.5 + 7741 & j070651.32 + 774137.0 & nvssj070651 + 774137 & 0.90(0.03 ) & 2.46(0.03 ) & 1.91(0.08 ) & bzb & b & m & ? & yes + j0706.7 - 4845 & j070549.12 - 483911.4 & & 0.95(0.04 ) & 2.86(0.06 ) & 2.38(0.19 ) & und & c & m & ? & no + j0709.3 - 0256 & j070945.05 - 025517.4 & nvssj070945 - 025517 & 1.09(0.03 ) & 2.91(0.03 ) & 2.30(0.06 ) & und & c & v & ? & yes + j0726.0 - 0053 & j072550.63 - 005456.4 & bzuj0725 - 0054 & 1.08(0.03 ) & 2.98(0.03 ) & 2.45(0.03 ) & bzq & a & m & 0.128 & yes + j0734.2 - 7706 & j073443.44 - 771113.4 & sumssj073441 - 771113 & 1.04(0.04 ) & 3.20(0.05 ) & 2.48(0.11 ) & bzq & c & & ? & yes + j0746.5 - 0713 & j074627.48 - 070949.7 & nvssj074627 - 070951 & 1.10(0.04 ) & 3.04(0.07 ) & 2.33(0.18 ) & und & c & m & ? & yes + j0746.5 - 4758 & j074642.30 - 475455.2 & sumssj074642 - 475455 & 1.15(0.03 ) & 2.91(0.06 ) & 1.89(0.16 ) & bzb & c & m & ? & yes + j0816.7 - 2420 & j081639.46 - 242635.4 & & 1.13(0.07 ) & 3.20(0.11 ) & 2.59(0.30 ) & bzq & c & & ? & no + & j081640.41 - 242106.6 & nvssj081640 - 242105 & 1.10(0.04 ) & 3.04(0.04 ) & 2.45(0.09 ) & bzq & b & & ? & yes + j0823.0 + 4041 & j082257.55 + 404149.8 & nvssj082257 + 404149 & 1.15(0.04 ) & 2.91(0.04 ) & 2.32(0.07 ) & bzq & c & qso & 0.8657 & yes + j0844.8 - 5459 & j084502.47 - 545808.5 & at20gj084502 - 545808 & 1.02(0.03 ) & 2.82(0.03 ) & 2.31(0.04 ) & und & b & m & ? & yes + j0849.9 - 3540 & j084945.61 - 354101.2 & nvssj084945 - 354102 & 1.01(0.03 ) & 2.53(0.04 ) & 2.35(0.08 ) & und & b & v & ? & yes + j0852.4 - 5756 & j085238.73 - 575529.4 & at20gj085238 - 575530 & 1.23(0.03 ) & 2.87(0.03 ) & 2.38(0.03 ) & bzq & a & & ? & yes + j0853.1 - 3659 & j085310.50 - 365823.1 & nvssj085310 - 365820 & 0.77(0.04 ) & 2.32(0.04 ) & 1.98(0.08 ) & bzb & b & n , m & ? & no + j0855.1 - 0712 & j085435.20 - 071837.5 & & 1.10(0.03 ) & 3.03(0.03 ) & 2.48(0.05 ) & bzq & b & m & ? & no + j0856.0 + 7136 & j085654.85 + 714623.8 & nvssj085654 + 714624 & 1.09(0.04 ) & 2.97(0.04 ) & 2.49(0.07 ) & bzq & b & m & ? & yes + j0858.1 - 1952 & j085805.36 - 195036.8 & nvssj085805 - 195036 & 1.16(0.04 ) & 2.80(0.05 ) & 2.25(0.11 ) & und & c & m , qso & 0.6597 & yes + j0904.8 - 3513 & j090423.42 - 351203.0 & & 1.24(0.06 ) & 2.82(0.10 ) & 2.37(0.32 ) & bzq & c & & ? & no + j0906.2 - 0906 & j090618.05 - 090544.9 & nvssj090618 - 090544 & 1.01(0.04 ) & 2.56(0.05 ) & 2.36(0.12 ) & und & c & m & ? & yes + j0919.3 - 2203 & j092002.74 - 215835.0 & & 1.01(0.04 ) & 2.76(0.06 ) & 2.21(0.19 ) & und & c & m & ? & no + j0940.8 - 6105 & j094047.33 - 610728.5 & at20gj094047 - 610726 & 0.87(0.03 ) & 2.57(0.03 ) & 2.17(0.06 ) & bzb & b & m , v & ? & yes + j0941.9 - 0755 & j094221.46 - 075953.1 & nvssj094221 - 075953 & 0.95(0.03 ) & 2.50(0.03 ) & 2.21(0.07 ) & bzb & b & v & ? & yes + j0946.9 - 2541 & j094709.52 - 254100.0 & nvssj094709 - 254100 & 0.72(0.04 ) & 2.16(0.06 ) & 2.06(0.23 ) & bzb & c & m & ? & yes + j1016.2 - 0638 & j101542.96 - 063055.1 & & 0.93(0.04 ) & 2.91(0.07 ) & 2.48(0.16 ) & und & c & & ? & no + & j101626.98 - 063625.2 & nvssj101626 - 06362 & 0.91(0.04 ) & 2.88(0.07 ) & 2.38(0.18 ) & und & c & n , f & ? & no + j1045.5 - 2931 & j104540.62 - 292726.4 & nvssj104540 - 292725 & 1.09(0.05 ) & 3.13(0.06 ) & 2.33(0.17 ) & bzq & c & & ? & yes + j1103.9 - 5356 & j110352.22 - 535700.7 & at20gj110352 - 535700 & 0.98(0.03 ) & 2.88(0.03 ) & 2.22(0.03 ) & ? & a & m , v & & yes + j1106.3 - 3643 & j110624.04 - 364659.0 & nvssj110624 - 364659 & 1.08(0.04 ) & 2.65(0.06 ) & 2.39(0.16 ) & ? & c & & ? & yes + j1154.1 - 3242 & j115406.16 - 324243.0 & nvssj115406 - 324242 & 1.02(0.04 ) & 2.87(0.05 ) & 2.41(0.12 ) & ? & c & m , v & ? & yes + j1230.2 - 5258 & j122939.88 - 530332.1 & at20gj122939 - 530332 & 0.66(0.04 ) & 2.22(0.07 ) & 1.90(0.27 ) & bzb & c & m & ? & no + j1238.1 - 1953 & j123824.40 - 195913.4 & nvssj123824 - 195913 & 0.88(0.03 ) & 2.46(0.04 ) & 2.13(0.11 ) & bzb & b & m & ? & yes + j1239.5 + 0728 & j123924.58 + 073017.2 & bzuj1239 + 0730 & 1.07(0.03 ) & 2.96(0.04 ) & 2.32(0.07 ) & ? & c & m & 0.4 & no + + col . ( 1 ) 2fgl name ( 2 ) _ wise _ name . ( 3 ) other name if present in literature and in the following order : roma - bzcat , nvss , sdss , at20 g , ned . ( 4,5,6 ) infrared colors from the _ wise _ all sky catalog corrected for galactic extinciton . values in parentheses are 1@xmath12 uncertainties ( 7 ) type of candidate according to our method : bzb - bzq - und ( undetermined ) . ( 8) class of candidate according to our method . ( 9 ) notes : n = nvss , f = first , s = sumss , a = at20 g , m = 2mass , s = sdss dr9 , 6 = 6dfgs , x = _ xmm - newton _ or _ chandra _ , x = rosat ; qso = quasar , sy = seyfert , lnr = liner , bl = bl lac ; v = variability in _ wise _ ( var_flag @xmath11 5 in at least one band ) ( 10 ) redshift : ( ? ) = unknown , ( number ? ) = uncertain . ( 11 ) re - association flag : `` yes '' if the association of our method corresponds to the one provided in the 2fgl , `` no '' otherwise . [ tab : agu1 ] [ 12]-[22 ] & type & class & notes & z & reassoc . + name & name & name & mag & mag & mag & & & & & flag + j1301.6 + 3331 & j130147.03 + 332236.3 & sdssj130147.01 + 332236.4 & 1.04(0.03 ) & 2.46(0.04 ) & 2.27(0.08 ) & und & b & m , s & ? & no + j1303.8 - 5537 & j130349.23 - 554031.6 & at20gj130349 - 554031 & 1.01(0.03 ) & 2.87(0.03 ) & 2.35(0.03 ) & und & a & m , v & ? & yes + j1304.1 - 2415 & j130416.70 - 241216.6 & nvssj130416 - 24121 & 0.89(0.03 ) & 2.37(0.05 ) & 2.01(0.15 ) & bzb & c & n , m & ? & no + j1304.3 - 4353 & j130421.01 - 435310.2 & sumssj130420 - 435308 & 0.84(0.03 ) & 2.24(0.03 ) & 1.93(0.05 ) & bzb & b & m , v & ? & yes + j1307.5 - 4300 & j130737.98 - 425938.9 & sumssj130737 - 42594 & 0.73(0.03 ) & 2.04(0.03 ) & 1.90(0.08 ) & bzb & b & s , m,6,v & ? & no + j1307.6 - 6704 & j130817.51 - 670705.8 & at20gj130817 - 670704 & 0.77(0.03 ) & 2.55(0.04 ) & 2.28(0.06 ) & bzb & b & m , v & ? & yes + j1329.2 - 5608 & j132901.16 - 560802.5 & at20gj132901 - 560802 & 1.05(0.03 ) & 2.78(0.03 ) & 2.22(0.07 ) & und & b & m , v & ? & yes + j1330.1 - 7002 & j133011.34 - 700312.7 & at20gj133010 - 700313 & 0.78(0.03 ) & 2.48(0.03 ) & 2.25(0.04 ) & bzb & b & & ? & yes + j1351.3 - 2909 & j135146.85 - 291217.4 & bzuj1351 - 2912 & 1.12(0.03 ) & 2.87(0.03 ) & 2.48(0.05 ) & bzq & b & v & 1.034 ? & no + j1406.2 - 2510 & j140609.60 - 250809.2 & nvssj140609 - 250808 & 0.83(0.04 ) & 2.37(0.07 ) & 2.02(0.25 ) & bzb & c & m , v & ? & yes + j1416.3 - 2415 & j141554.91 - 241925.8 & nvssj141554 - 241924 & 1.26(0.05 ) & 3.13(0.08 ) & 2.51(0.22 ) & bzq & c & n & ? & no + & j141642.23 - 241021.2 & & 0.92(0.04 ) & 2.71(0.08 ) & 2.39(0.23 ) & und & c & m & ? & no + j1419.4 - 0835 & j141922.56 - 083831.9 & nvssj141922 - 083830 & 1.01(0.03 ) & 2.78(0.04 ) & 2.15(0.09 ) & und & b & v & ? & yes + j1514.6 - 4751 & j151440.03 - 474829.7 & at20gj151440 - 474828 & 1.14(0.04 ) & 3.07(0.04 ) & 2.43(0.06 ) & bzq & b & m & ? & yes + j1518.2 - 2733 & j151803.60 - 273131.0 & nvssj151803 - 273131 & 0.62(0.03 ) & 2.14(0.04 ) & 2.18(0.10 ) & bzb & c & m & ? & yes + j1553.2 - 2424 & j155331.62 - 242206.0 & nvssj155331 - 242206 & 1.16(0.04 ) & 2.95(0.07 ) & 2.66(0.14 ) & bzq & c & & 0.332 ? & yes + j1558.3 + 8513 & j160031.76 + 850949.2 & nvssj160031 + 850948 & 0.82(0.03 ) & 2.18(0.04 ) & 2.02(0.14 ) & bzb & b & n , m & ? & no + j1604.5 - 4442 & j160431.03 - 444131.9 & at20gj160431 - 444131 & 0.98(0.03 ) & 2.80(0.03 ) & 2.46(0.05 ) & und & b & m & ? & yes + j1626.0 - 7636 & j162638.17 - 763855.4 & sumssj162639 - 763856 & 0.56(0.03 ) & 2.21(0.05 ) & 1.97(0.19 ) & bzb & c & m , v & ? & yes + j1650.1 - 5044 & j165016.63 - 504448.2 & at20gj165016 - 504446 & 0.95(0.04 ) & 2.65(0.04 ) & 2.22(0.09 ) & und & b & m , v & ? & yes + j1725.1 - 7714 & j172350.86 - 771350.3 & sumssj172350 - 771350 & 1.01(0.03 ) & 2.67(0.04 ) & 2.33(0.08 ) & und & b & m , v & ? & yes + j1759.2 - 4819 & j175858.45 - 482112.4 & sumssj175858 - 482112 & 0.91(0.03 ) & 2.62(0.03 ) & 2.21(0.05 ) & und & c & & ? & yes + j1811.0 + 5340 & j181037.99 + 533501.5 & nvssj181038 + 533501 & 0.83(0.03 ) & 2.22(0.05 ) & 2.15(0.15 ) & bzb & c & n , m , x & ? & no + j1815.6 - 6407 & j181425.96 - 641008.8 & & 1.27(0.06 ) & 3.05(0.08 ) & 2.67(0.19 ) & bzq & c & & ? & no + j1816.7 - 4942 & j181655.99 - 494344.7 & sumssj181655 - 494344 & 0.99(0.05 ) & 2.86(0.07 ) & 2.21(0.19 ) & und & c & & ? & yes + j1818.7 + 2138 & j181905.22 + 213234.0 & nvssj181905 + 213235 & 0.90(0.04 ) & 2.50(0.05 ) & 1.99(0.17 ) & bzb & c & m & ? & yes + j1820.6 + 3625 & j182023.61 + 362914.4 & & 1.04(0.04 ) & 2.61(0.05 ) & 2.36(0.17 ) & und & c & & ? & no + j1823.6 - 3453 & j182338.59 - 345412.0 & nvssj182338 - 345412 & 0.68(0.04 ) & 2.05(0.04 ) & 1.92(0.13 ) & bzb & b & m & ? & yes + j1823.8 + 4312 & j182352.33 + 431452.5 & & 1.12(0.04 ) & 2.96(0.05 ) & 2.50(0.13 ) & bzq & c & x & ? & no + j1825.1 - 5231 & j182513.79 - 523058.1 & sumssj182513 - 523057 & 1.65(0.03 ) & 2.86(0.03 ) & 2.26(0.07 ) & und & b & m & ? & yes + j1830.0 + 1325 & j183000.76 + 132414.4 & nvssj183000 + 132414 & 0.95(0.04 ) & 2.60(0.04 ) & 2.23(0.11 ) & und & c & m & ? & yes + j1830.2 - 4441 & j183000.86 - 444111.4 & sumssj183000 - 444112 & 1.17(0.04 ) & 3.10(0.05 ) & 2.37(0.10 ) & bzq & c & m & ? & yes + j1844.7 + 5716 & j184450.96 + 570938.6 & nvssj184451 + 570940 & 0.86(0.03 ) & 2.38(0.04 ) & 2.16(0.10 ) & bzb & b & m & ? & yes + j1936.9 + 8402 & j193930.23 + 835925.8 & & 1.08(0.04 ) & 2.97(0.05 ) & 2.43(0.14 ) & bzq & c & & ? & no + j1940.8 - 6213 & j194121.76 - 621120.8 & sumssj194121 - 621120 & 1.30(0.04 ) & 2.88(0.04 ) & 2.36(0.09 ) & bzq & c & & ? & yes + j1942.8 + 1033 & j194247.48 + 103327.2 & nvssj194247 + 103327 & 0.72(0.03 ) & 2.06(0.03 ) & 1.85(0.07 ) & bzb & b & m & ? & yes + j1959.9 - 4727 & j195945.66 - 472519.2 & sumssj195945 - 472519 & 0.80(0.03 ) & 2.15(0.04 ) & 1.66(0.17 ) & bzb & c & m & ? & yes + j2040.2 - 7109 & j203931.44 - 711033.0 & & 1.12(0.04 ) & 3.01(0.05 ) & 2.71(0.12 ) & bzq & c & & ? & no + j2049.8 + 1001 & j204932.28 + 095911.8 & & 1.23(0.06 ) & 3.13(0.10 ) & 2.68(0.23 ) & bzq & c & & ? & no + j2103.6 - 6236 & j210338.38 - 623225.8 & sumssj210338 - 623226 & 0.81(0.03 ) & 2.30(0.03 ) & 1.80(0.12 ) & bzb & b & m & ? & yes + j2250.2 - 4205 & j225014.94 - 420218.6 & & 1.05(0.04 ) & 2.64(0.07 ) & 2.39(0.20 ) & und & c & & ? & no + & j225022.20 - 420613.2 & sumssj225022 - 420613 & 0.94(0.03 ) & 2.47(0.04 ) & 1.88(0.13 ) & bzb & c & m & 0.1187 ? & yes + j2317.3 - 4534 & j231731.97 - 453359.6 & sumssj231731 - 453400 & 0.70(0.04 ) & 2.19(0.08 ) & 1.92(0.34 ) & bzb & c & s , m,6 & ? & no + j2323.0 - 4918 & j232255.30 - 491942.0 & & 1.08(0.06 ) & 3.21(0.10 ) & 2.34(0.30 ) & bzq & c & & ? & no + j2324.6 + 0801 & j232445.31 + 080206.3 & nvssj232445 + 080205 & 0.83(0.04 ) & 2.30(0.06 ) & 2.05(0.21 ) & bzb & c & m & ? & yes + j2325.4 + 1650 & j232526.62 + 164941.1 & & 1.34(0.04 ) & 3.04(0.05 ) & 2.42(0.12 ) & bzq & c & & ? & no + & j232538.11 + 164642.8 & nvssj232538 + 164641 & 0.78(0.04 ) & 2.15(0.10 ) & 2.27(0.33 ) & bzb & c & m & ? & yes + + col . ( 1 ) 2fgl name . ( 2 ) _ wise _ name . ( 3 ) other name if present in literature and in the following order : roma - bzcat , nvss , sdss , at20 g , ned . ( 4,5,6 ) infrared colors from the _ wise _ all sky catalog corrected for galactic extinciton . values in parentheses are 1@xmath12 uncertainties ( 7 ) type of candidate according to our method : bzb - bzq - und ( undetermined ) . ( 8) class of candidate according to our method . ( 9 ) notes : n = nvss , f = first , s = sumss , a = at20 g , m = 2mass , s = sdss dr9 , 6 = 6dfgs , x = _ xmm - newton _ or _ chandra _ , x = rosat ; qso = quasar , sy = seyfert , lnr = liner , bl = bl lac ; v = variability in _ wise _ ( var_flag @xmath11 5 in at least one band ) ( 10 ) redshift : ( ? ) = unknown , ( number ? ) = uncertain . ( 11 ) re - association flag : `` yes '' if the association of our method corresponds to the one provided in the 2fgl , `` no '' otherwise . [ tab : agu2 ] 2fglj0032.7 - 5521 & 2fglj1117.2 - 5341 & 2fglj1722.5 - 0420 + 2fglj0048.8 - 6347 & 2fglj1120.0 - 2204 & 2fglj1730.6 - 2409 + 2fglj0212.1 + 5318 & 2fglj1129.0 - 0532 & 2fglj1744.1 - 7620 + 2fglj0224.0 + 6204 & 2fglj1208.5 - 6240 & 2fglj1747.6 + 0324 + 2fglj0239.5 + 1324 & 2fglj1221.4 - 0633 & 2fglj1748.8 + 3418 + 2fglj0248.5 + 5131 & 2fglj1231.3 - 5112 & 2fglj1748.9 - 3923 + 2fglj0305.0 - 1602 & 2fglj1240.6 - 7151 & 2fglj1753.8 - 4446 + 2fglj0307.4 + 4915 & 2fglj1306.2 - 6044 & 2fglj1757.5 - 6028 + 2fglj0318.0 + 0255 & 2fglj1312.9 - 2351 & 2fglj1759.4 - 2954 + 2fglj0336.0 + 7504 & 2fglj1335.3 - 4058 & 2fglj1820.6 - 3219 + 2fglj0338.2 + 1306 & 2fglj1353.5 - 6640 & 2fglj1821.8 + 0830 + 2fglj0353.2 + 5653 & 2fglj1400.7 - 1438 & 2fglj1828.7 + 3231 + 2fglj0359.5 + 5410 & 2fglj1410.4 + 7411 & 2fglj1830.9 - 3132 + 2fglj0409.5 + 0509 & 2fglj1417.7 - 5028 & 2fglj1902.7 - 7053 + 2fglj0418.9 + 6636 & 2fglj1423.9 - 7842 & 2fglj1906.5 + 0720 + 2fglj0420.9 - 3743 & 2fglj1424.2 - 1752 & 2fglj1919.5 - 7324 + 2fglj0426.7 + 5434 & 2fglj1458.5 - 2121 & 2fglj1946.7 - 1118 + 2fglj0438.0 - 7331 & 2fglj1507.0 - 6223 & 2fglj1947.8 - 0739 + 2fglj0439.8 - 1858 & 2fglj1513.5 - 2546 & 2fglj2002.8 - 2150 + 2fglj0516.7 + 2634 & 2fglj1513.9 - 2256 & 2fglj2017.5 - 1618 + 2fglj0523.3 - 2530 & 2fglj1518.4 - 5233 & 2fglj2018.0 + 3626 + 2fglj0524.1 + 2843 & 2fglj1536.4 - 4949 & 2fglj2034.7 - 4201 + 2fglj0533.9 + 6759 & 2fglj1539.2 - 3325 & 2fglj2034.9 + 3632 + 2fglj0539.3 - 0323 & 2fglj1544.5 - 1126 & 2fglj2041.2 + 4735 + 2fglj0605.3 + 3758 & 2fglj1548.3 + 1453 & 2fglj2042.8 - 7317 + 2fglj0658.4 + 0633 & 2fglj1601.1 - 4220 & 2fglj2046.2 - 4259 + 2fglj0719.2 - 5000 & 2fglj1617.3 - 5336 & 2fglj2103.5 - 1112 + 2fglj0758.8 - 1448 & 2fglj1617.5 - 2657 & 2fglj2107.8 + 3652 + 2fglj0803.2 - 0339 & 2fglj1622.8 - 5006 & 2fglj2110.3 + 3822 + 2fglj0843.6 + 6715 & 2fglj1624.1 - 4040 & 2fglj2112.5 - 3042 + 2fglj0854.7 - 4501 & 2fglj1626.4 - 4408 & 2fglj2115.4 + 1213 + 2fglj0859.4 - 2532 & 2fglj1631.6 - 2819 & 2fglj2117.5 + 3730 + 2fglj0934.0 - 6231 & 2fglj1646.7 - 1333 & 2fglj2212.6 + 0702 + 2fglj0952.7 - 3717 & 2fglj1649.2 - 3004 & 2fglj2246.3 + 1549 + 2fglj1016.4 - 4244 & 2fglj1704.3 + 1235 & 2fglj2249.1 + 5758 + 2fglj1033.5 - 5032 & 2fglj1704.6 - 0529 & 2fglj2339.6 - 0532 + 2fglj1036.1 - 6722 & 2fglj1709.0 - 0821 & 2fglj2351.6 - 7558 + [ tab : dm ] among the whole sample of 590 ugss analyzed , 299 without and 291 with @xmath1-ray analysis flag there are 28 sources having at least one @xmath1-ray blazar candidate that were also unidentified in the first fermi @xmath1-ray lat catalog ( 1fgl ; * ? ? ? * ) and they were analyzed using two different statistical approaches : the classification tree and the logistic regression analyses ( see * ? ? ? * and references therein ) . for these 28 ugss , analyzed on the basis of the above statistical approaches , we performed a comparison with our results to verify if the 2fgl sources that we associated with a @xmath1-ray blazar candidates have been also classified as agns . by comparing the results of our association method with those in ackermann et al . ( 2012 ) , we found that 23 out of 28 ugss that we associate with @xmath1-ray blazar candidates are classified as agns , all of them with a probability higher than 66% and 12 of them higher than 80% ( see * ? ? ? * ) . among the remaining 5 sources , 4 have been classified as pulsars , with a very low probability with respect to the whole sample , systematically lower than 56% . in addition , there is one with an ambiguous classification . consequently , we emphasize that for the subamples where we overlap our results are in good agreement with the classification suggested by ackermann et al . ( 2012 ) , consistent with the @xmath1-ray blazar nature of the _ wise _ candidates proposed in our analysis . a new association method has been recently developed on the basis of the striking discovery that @xmath1-ray emitting blazars occupy a distinct region in the _ wise _ 3-dimensional color space , well separated from that occupied by other extragalactic and galactic sources @xcite . according to dabrusco et al . ( 2013 ) the 3-dimensional region occupied by @xmath1-ray emitting blazars is the @xmath2 ; its 2-dimensional projection in the [ 3.4]-[4.6]-[12 ] @xmath3 m parameter space , retains its historical definition of _ wise _ gamma - ray strip @xcite . additional improvements , mostly based on the _ wise _ all - sky data release , available since march 2012 ( e.g. , * ? ? ? * ) , and on a new parametrization of the @xmath2 in the parameter space of its principal components have been subsequently developed @xcite . in this work we describe the results obtained by applying our new association procedure to the search for new @xmath1-ray blazar candidates in the two samples : the unidentified gamma - ray sources ( ugss ) , and the active galaxies of uncertain type ( agus ) , as listed in the 2fgl @xcite . we present the complete list of @xmath1-ray blazar candidates found using the _ wise _ observations . we also perform an extensive archival search to see if the sources associated with our method , show additional blazar - like characteristics ; as for example the presence of a radio counterpart and/or of a spectrum that could be featureless as for bzbs or similar to those of broad - line quasars as generally occurs in bzqs . we found 62 @xmath1-ray blazar candidates for the ugs without any @xmath1-ray analysis flag and 49 for those with @xmath1-ray analysis flag , out of a total of 590 sources investigated . for the agus sample , we confirmed the blazar - like nature of 87 out 210 of agus analyzed on the basis of their ir colors . a significant fraction ( i.e. , @xmath0 36% ) of the _ wise _ sources associated with our method with ugss have a radio counterpart , more than 50% are also detected in the 2mass catalog as generally occurs for blazars , and more than @xmath010% appear to be variable according to the _ wise _ analysis flags @xcite . notably , all the sources for which an optical spectrum was available in literature clearly show blazar - like features , being either featureless or having broad emission lines typical of quasars , the only exception being sdss j015910.05 + 010514.5 , one of the counterparts associated with 2fglj0158.4 + 0107 . as generally expected for @xmath1-ray blazars a handful of the selected candidates are also detected in the x - rays . a deeper investigation of their x - ray counterparts will be addressed in a forthcoming paper @xcite . all the @xmath1-ray blazar candidates selected with our association procedure appear to be extragalactic in nature ; moreover our selection seems not to be highly contaminated by any class of non - blazar - like sources , as for example obscured quasars or seyfert galaxies . our results are in good agreement with those based on different statistical approaches like the classification tree and the logistic regression analyses @xcite . in particular , 23 out of 28 ugss that we associate to a @xmath1-ray blazar candidate are also classified as active galaxies by the above methods at high level of confidence . for ugss associated with a pulsar in the 2fgl analysis as reported in the public list of lat - detected gamma - ray pulsars ( see section [ sec : ugs ] ) , we did not find any _ wise _ @xmath1-ray blazar candidate , confirming the reliability of our selection procedure . we provide a list of the ugss for which we did not find any @xmath1-ray blazar candidates using either the new improved method or the old parametrization ( i.e. , less conservative ) , within their positional uncertainty regions at 95% level of confidence . this list of _ fermi _ sources reported in table [ tab : dm ] could be useful for follow up observations aiming at discover new pulsars or to constrain exotic high - energy physics phenomena such as dark matter signatures , or new classes of sources ( e.g. , * ? ? ? * ; * ? ? ? finally , we emphasize that additional investigations of different samples of active galactic nuclei , such as seyfert galaxies , are necessary to study the problem of the contamination of our association method by extragalactic sources with infrared colors similar to those of @xmath1-ray blazars . moreover extensive ground - based spectroscopic follow up observations in the optical and in the near ir would be ideal to verify the nature of the selected _ wise _ sources and to estimate the fraction of non - blazar objects , similar to the recent studies performed for the unidentified integral sources ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . _ note added to the proofs : _ the infrared source wise j182352.33 + 431452.5 , potential counterpart of 2fgl j1823.8 + 4312 , is a possible contaminant of our selections given its optical spectrum typical of an obscured red quasar ( d. stern priv . comm . ) . we thank the anonymous referee for useful comments that led to improvements in the paper . f. massaro is grateful to s. digel and d. thompson for their helpful discussions and to m. ajello , e. ferrara and j. ballet for their support . the work is supported by the nasa grants nnx12ao97 g . r. dabrusco gratefully acknowledges the financial support of the us virtual astronomical observatory , which is sponsored by the national science foundation and the national aeronautics and space administration . the work by g. tosti is supported by the asi / inaf contract i/005/12/0 . h. a. smith acknowledges partial support from nasa / jpl grant rsa 1369566 . topcat[multiblock footnote omitted ] @xcite and saoimage ds9 were used extensively in this work for the preparation and manipulation of the tabular data and the images . part of this work is based on archival data , software or on - line services provided by the asi science data center . this research has made use of data obtained from the high energy astrophysics science archive research center ( heasarc ) provided by nasa s goddard space flight center ; the simbad database operated at cds , strasbourg , france ; the nasa / ipac extragalactic database ( ned ) operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . part of this work is based on the nvss ( nrao vla sky survey ) ; the national radio astronomy observatory is operated by associated universities , inc . , under contract with the national science foundation . this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation . this publication makes use of data products from the wide - field infrared survey explorer , which is a joint project of the university of california , los angeles , and the jet propulsion laboratory / california institute of technology , funded by the national aeronautics and space administration . funding for the sdss and sdss - ii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the u.s . department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england . the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium for the participating institutions . the participating institutions are the american museum of natural history , astrophysical institute potsdam , university of basel , university of cambridge , case western reserve university , university of chicago , drexel university , fermilab , the institute for advanced study , the japan participation group , johns hopkins university , the joint institute for nuclear astrophysics , the kavli institute for particle astrophysics and cosmology , the korean scientist group , the chinese academy of sciences ( lamost ) , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , ohio state university , university of pittsburgh , university of portsmouth , princeton university , the united states naval observatory , and the university of washington . abdo , a. a. , et al . 2009 , astroparticle physics , 32 , 193 abdo , a. a. et al . 2010 apjs 188 405 ackermann , m. et al . 2011 apj , 743 , 171 ackermann , m. et al . 2012 apj , 753 , 83 adelman - mccarthy , j. , agueros , m.a . , allam , s.s . , et al . 2008 , apjs , 175 , 297 becker , r. h. , white , r. l. , helfand , d. j.1995 apj , 450 , 559 bauer , a. et al . 2009 apj , 705 , 46 chen , p. s. , fu , h. w. & gao , y. f. 2005 newa , 11 , 27 comastri , a. , mignoli , m. , ciliegi , p. , et al . 2002 , apj , 571 , 771 condon , j. j. , cotton , w. d. , greisen , e. w. , yin , q. f. , perley , r. a. , taylor , g. b. , & broderick , j. j. 1998 , aj , 115 , 1693 cutri et al . 2012 wise.rept , 1c dabrusco , r. , massaro , f. , ajello , m. , grindlay , j. e. , smith , howard a. & tosti , g. 2012 apj , 748 , 68 dabrusco , r. , massaro , f. , paggi , a. , masetti , n. , giroletti , m. , tosti , g. , smith , howard , a. 2013 apjs submitted draine , b. t. 2003 , ara&a , 41 , 241 evans , i. n. et al . 2010 apjs , 189 , 37 falco , e. e. et al . 1998 apj , 494 , 47 giommi , p. et al . 2005 , a&a , 434 , 385 hartman , r.c . et al . , 1999 apjs 123 healey , s. e. et al . 2007 apjs , 171 , 61 hu , c. et al . 2008 , 687 , 78 jolliffe i.t . principal component analysis , series : springer series in statistics " , 2nd ed . , springer , ny , 2002 , xxix , 487 , 28 jones , h. d. et al . 2004 mnras , 355 , 747 jones , h. d. et al . 2009 mnras , 399 , 683 mirabal , n. 2009 [ arxiv.org/abs/0908.1389v2 ] mirabal , n. 2009 apj , 701 , 129 mirabal , nieto , d. & pardo , s. 2010 a&a submitted , [ arxiv.org/abs/1007.2644v2 ] maselli , a. , massaro , e. , nesci , r. , sclavi , s. , rossi , c. , giommi , p. 2010 a&a , 512a , 74 maselli , a. , cusumano , g. , massaro , e. , la parola , v. , segreto , a. , sbarufatti , b. 2010 a&a , 520a , 47 masetti , n. et al . 2008 a&a , 482 , 113 masetti , n. et al . 2009 a&a , 495 , 121 masetti , n. et al . 2010 a&a , 519a , 96 masetti , n. et al . 2012 a&a , 538a , 123 massaro , e. , giommi , p. , leto , c. , marchegiani , p. , maselli , a. , perri , m. , piranomonte , s. , sclavi , s. 2009 a&a , 495 , 691 massaro , e. , giommi , p. , leto , c. , marchegiani , p. , maselli , a. , perri , m. , piranomonte , s. , sclavi , s. 2010 http://arxiv.org/abs/1006.0922 massaro , e. , giommi , p. , leto , c. , marchegiani , p. , maselli , a. , perri , m. , piranomonte , s. , 2011 multifrequency catalogue of blazars ( 3rd edition ) " , aracne editrice , rome , italy massaro , f. , dabrusco , r. , ajello , m. , grindlay , j. e. & smith , h. a. 2011b apj , 740l , 48 massaro , f. , dabrusco , r. , tosti , g. , ajello , m. , gasparrini , d. , grindlay , j. e. & smith , howard a. 2012a apj , 750 , 138 massaro , f. , dabrusco , r. , tosti , g. , ajello , m. , paggi , a. , gasparrini , 2012b apj , 752 , 61 massaro , f. , dabrusco , r. , paggi , a. , tosti , g. , gasparrini , d. 2012c apj , 750l , 35 massaro , f. , paggi , a. , dabrusco , r. , tosti , g. , grindlay , j. e. , smith , howard a. , digel , s. w. , funk , s. 2012d apj , 757l , 27 mauch , t. , murphy , t. , buttery , h. j. , curran , j. , hunstead , r. w. , piestrzynski , b. , robertson , j. g. , sadler , e. m. 2003 mnras , 342 , 1117 monet , d. g. et al . 2003 aj , 125 , 984 murphy , t. et al . 2010 mnras , 402 , 2403 mukherjee , r. et al . , 1997 apj , 490 , 116 nolan et al . 2012 apjs , 199 , 31 paggi , a. , massaro , f. , dabrusco , r. et al . 2013 apjs in prep . paris , i. et al . 2012 a&a , 548a , 66 pearson , k. 1901 philosophical magazine 2 , 559 . schneider et al . 2007 , aj , 134 , 102 stickel , m. , padovani , p. , urry , c. m. , fried , j. w. , kuehr , h. 1991 apj , 374 , 431 stocke et al . 1991 , apjs , 76 , 813 skrutskie , m. f. et al . 2006 , aj , 131 , 1163 su , m. & finkbeiner , d. p. 2012 apj submitted http://arxiv.org/abs/1207.7060v1 urry , c. m. , & padovani , p. 1995 , pasp , 107 , 803 thompson , d. j. 2008 rpph , 71k6901 taylor , m. b. 2005 , asp conf . ser . , 347 , 29 voges , w. et al . 1999 a&a , 349 , 389 white , r. l. , becker , r. h. helfand , d. j. , gregg , m. d. et al . 1997 apj , 475 , 479 white , g. l. et al . 1988 apj , 327 , 561 wright , e. l. , et al . 2010 aj , 140 , 1868 zechlin , h .- s . , fernandes , m. v. , elsasser , d. , horns , d. 2012 a&a , 538a , 93 in tables [ tab : ugs_usno ] , [ tab : ugf_usno ] , [ tab : agu_usno1 ] and [ tab : agu_usno2 ] , we report the magnitudes of the optical counterpart uniquely found within 3.3 , for all the @xmath1-ray blazar candidates , selected according to our association procedure . this information permits us to optimize the strategy for the future follow up optical observations needed to clarify the nature of the selected sources and to determine their redshifts via spectroscopy . name & name & mag & mag & mag & mag & mag + j0039.1 + 4331 & j003858.27 + 432947.0 & 19.29 & 19.26 & 19.03 & 18.84 & 18.32 + j0116.6 - 6153 & j011619.59 - 615343.5 & & 17.72 & 18.22 & 17.78 & 17.91 + j0133.4 - 4408 & j013306.35 - 441421.3 & & 18.38 & 19.70 & 18.12 & 18.76 + j0133.4 - 4408 & j013321.36 - 441319.4 & & 18.63 & 19.39 & 18.32 & 18.80 + j0143.6 - 5844 & j014347.39 - 584551.3 & & 16.70 & 18.48 & 16.64 & 17.04 + j0158.4 + 0107 & j015910.05 + 010514.7 & 17.94 & 18.27 & 17.71 & 18.43 & 16.33 + j0158.4 + 0107 & j015757.45 + 011547.8 & 21.13 & 19.39 & & 19.31 & 18.58 + j0158.4 + 0107 & j015836.25 + 010632.1 & 19.22 & 18.90 & 19.02 & 19.07 & 17.62 + j0158.6 + 8558 & j014935.30 + 860115.3 & 18.79 & 17.14 & 18.60 & 16.96 & 15.92 + j0158.6 + 8558 & j015550.16 + 854745.1 & 19.83 & 19.27 & 19.95 & 18.90 & 17.78 + j0227.7 + 2249 & j022744.35 + 224834.3 & & & 20.82 & 20.22 & 19.28 + j0316.1 - 6434 & j031614.31 - 643731.4 & & 16.59 & 18.19 & 16.57 & 16.82 + j0332.1 + 6309 & j033153.90 + 630814.1 & & & 20.66 & 19.92 & 18.35 + j0409.8 - 0357 & j040946.57 - 040003.4 & 19.45 & 19.18 & 17.53 & 16.98 & 16.86 + j0414.9 - 0855 & j041457.01 - 085652.0 & 20.48 & 18.15 & 20.21 & 17.31 & 17.69 + j0416.0 - 4355 & j041605.81 - 435514.6 & & 18.49 & 18.70 & 18.17 & 18.00 + j0555.9 - 4348 & j055618.74 - 435146.1 & & 19.23 & 18.88 & 19.08 & 18.08 + j0555.9 - 4348 & j055531.59 - 435030.7 & & 19.16 & 19.22 & 18.93 & 18.19 + j0602.7 - 4011 & j060237.10 - 401453.2 & & 17.61 & 17.76 & 17.74 & 17.62 + j0644.6 + 6034 & j064459.38 + 603131.7 & 19.44 & 19.03 & 19.33 & 18.23 & 18.28 + j0713.5 - 0952 & j071223.28 - 094536.3 & & & 19.48 & 17.88 & 17.24 + j0723.9 + 2901 & j072354.83 + 285929.9 & 19.78 & 19.05 & 19.97 & 18.72 & + j0744.1 - 2523 & j074402.19 - 252146.0 & & 13.33 & & 9.510 & + j0746.0 - 0222 & j074627.03 - 022549.3 & 19.03 & & 18.59 & 18.43 & 16.53 + j0756.3 - 6433 & j075624.60 - 643030.6 & & 18.80 & 19.13 & 17.26 & 18.56 + j0807.0 - 6511 & j080729.66 - 650910.3 & & 18.97 & 19.15 & 19.89 & + j0838.8 - 2828 & j083842.77 - 282830.9 & 19.66 & & 18.49 & 19.02 & 18.01 + j0841.3 - 3556 & j084121.63 - 355505.9 & & 17.20 & 17.57 & 16.72 & 16.64 + j0844.9 + 6214 & j084406.81 + 621458.6 & 19.39 & 16.38 & 18.75 & 16.62 & 15.85 + j0858.3 - 4333 & j085839.22 - 432642.7 & & 20.10 & & 20.12 & 17.83 + j0900.9 + 6736 & j090121.65 + 673955.8 & 19.09 & 18.52 & 19.61 & 18.25 & 18.13 + j0955.0 - 3949 & j095458.30 - 394655.0 & & 16.66 & 18.15 & 17.21 & 17.11 + j1013.6 + 3434 & j101256.54 + 343648.8 & 20.22 & 18.60 & 20.44 & 17.99 & 17.39 + j1016.1 + 5600 & j101544.44 + 555100.7 & 19.69 & 19.42 & 20.61 & 19.35 & + j1029.5 - 2022 & j102946.66 - 201812.6 & 18.01 & 18.22 & 18.41 & 18.25 & 18.37 + j1032.9 - 8401 & j103015.35 - 840308.7 & & 19.36 & 19.26 & 18.84 & 18.03 + j1038.2 - 2423 & j103754.92 - 242544.5 & 20.58 & 18.21 & 20.56 & 18.53 & 18.09 + j1207.3 - 5055 & j120746.43 - 505948.6 & & & 20.50 & 20.07 & + & j120750.50 - 510314.9 & & 19.12 & 19.71 & 20.27 & + j1254.2 - 2203 & j125422.47 - 220413.6 & & 19.88 & 18.67 & 19.11 & 18.22 + j1259.8 - 3749 & j125949.80 - 374858.1 & & 17.44 & 18.07 & 16.78 & 17.35 + j1340.5 - 0412 & j134042.02 - 041006.8 & 18.21 & 17.21 & 17.59 & 16.46 & 17.08 + j1346.0 - 2605 & j134621.08 - 255642.3 & 19.82 & 19.25 & 19.13 & 18.77 & 18.39 + j1347.0 - 2956 & j134706.89 - 295842.3 & 17.85 & 17.09 & 18.80 & 17.14 & 17.09 + j1404.0 - 5244 & j140313.11 - 524839.5 & & 17.49 & 18.92 & 18.57 & + j1517.2 + 3645 & j151649.26 + 365022.9 & 20.90 & & 21.49 & 20.07 & 19.16 + j1612.0 + 1403 & j161118.10 + 140328.9 & 18.39 & 18.39 & 19.06 & 19.16 & 18.43 + j1614.8 + 4703 & j161541.22 + 471111.8 & 17.55 & 16.03 & 16.90 & 15.39 & 15.51 + & j161434.67 + 470420.1 & 15.62 & 15.76 & 16.28 & 16.13 & 15.14 + & j161513.04 + 471355.2 & & 20.02 & 21.44 & 20.29 & 19.09 + & j161450.96 + 465953.7 & & 19.19 & 21.66 & 19.60 & 19.10 + j1622.8 - 0314 & j162225.35 - 031439.6 & & 19.85 & & 19.41 & + j1627.8 + 3219 & j162800.40 + 322414.0 & 20.65 & 19.50 & 19.02 & 18.88 & 19.01 + j1647.0 + 4351 & j164619.95 + 435631.0 & 20.43 & 19.73 & 20.42 & 19.67 & + j1730.6 - 0353 & j173052.86 - 035247.1 & 18.31 & 17.40 & 19.30 & 17.33 & 16.73 + j1745.6 + 0203 & j174526.95 + 020532.7 & 18.71 & 17.12 & 18.06 & 17.28 & 17.16 + & j174507.82 + 015442.5 & 19.21 & 16.40 & 18.11 & 16.30 & 15.98 + j1759.2 - 3853 & j175903.29 - 384739.5 & & 17.95 & 18.95 & & + j1842.3 + 2740 & j184201.25 + 274239.2 & 20.18 & 19.04 & 19.34 & 19.18 & 18.75 + j1904.8 - 0705 & j190444.57 - 070740.1 & & 19.73 & 19.87 & 18.45 & + j1924.9 - 1036 & j192501.63 - 104316.3 & & 18.63 & 19.42 & 18.04 & 17.75 + j2004.6 + 7004 & j200506.02 + 700439.3 & 20.73 & 19.25 & 19.24 & 18.65 & + j2021.5 + 0632 & j202155.45 + 062913.7 & 17.27 & 16.13 & 17.01 & 16.67 & 16.03 + & j202154.66 + 062908.7 & 19.15 & 17.44 & 17.81 & 17.24 & 17.35 + j2133.9 + 6645 & j213349.21 + 664704.3 & & & & 19.37 & 18.80 + j2134.6 - 2130 & j213430.18 - 213032.6 & 19.77 & 18.65 & 18.96 & 16.80 & 17.70 + j2300.0 - 3553 & j230010.16 - 360159.9 & & 18.43 & 19.23 & 18.17 & 17.63 + j2319.3 - 3830 & j232000.11 - 383511.4 & & 19.24 & 19.86 & 18.92 & 18.48 + j2358.4 - 1811 & j235828.61 - 181526.6 & 18.73 & 18.83 & 18.54 & 18.33 & 18.34 + name & name & mag & mag & mag & mag & mag + j0233.9 + 6238c & j023238.07 + 623651.9 & & & 20.95 & & 18.70 + & j023418.09 + 624207.8 & & & & 19.34 & 17.89 + j0341.8 + 3148c & j034158.52 + 314855.7 & 18.76 & 14.98 & 17.64 & 14.86 & 13.62 + & j034204.35 + 314711.4 & & 19.84 & & 19.86 & 17.33 + j0440.5 + 2554c & j043947.48 + 260140.5 & & 19.09 & & 19.11 & 15.70 + j0620.8 - 2556 & j062108.68 - 255757.9 & 19.78 & & 19.94 & 19.22 & + j0631.7 + 0428 & j063104.12 + 042012.6 & & 20.31 & & 20.36 & + j0637.0 + 0416c & j063647.19 + 042058.7 & 20.25 & 17.34 & 20.42 & 17.61 & 16.47 + & j063703.09 + 042146.1 & 18.95 & 16.95 & 19.88 & 17.23 & 16.04 + & j063705.96 + 042537.2 & 21.00 & 18.79 & 20.37 & 17.45 & 16.87 + & j063701.93 + 042037.2 & 21.00 & & 18.57 & & + j0922.2 - 5214c & j092154.24 - 521236.1 & & & 19.57 & 18.86 & 17.49 + j1059.9 - 2051 & j110025.72 - 205333.4 & 18.44 & 16.57 & 18.08 & 16.62 & 16.64 + j1208.6 - 2257 & j120816.33 - 224921.9 & 18.29 & 18.04 & 21.52 & 18.12 & 18.43 + j1255.8 - 5828 & j125459.44 - 582009.5 & & 16.56 & 18.67 & 16.71 & 16.33 + j1315.6 - 0730 & j131543.62 - 073659.0 & 19.88 & 19.09 & 18.21 & 17.95 & 18.06 + j1315.6 - 0730 & j131552.98 - 073301.9 & 19.78 & 18.68 & 18.75 & 17.75 & 17.56 + j1324.4 - 5411 & j132415.49 - 541104.4 & & 18.15 & & 18.86 & + j1345.8 - 3356 & j134543.05 - 335643.3 & & 17.98 & 19.58 & 18.65 & 18.12 + j1407.4 - 2948 & j140818.86 - 294203.2 & & & 20.80 & 19.36 & 18.75 + j1624.2 - 2124 & j162343.89 - 210707.0 & 19.70 & 18.57 & 19.35 & 18.74 & 18.51 + j1835.4 + 1036 & j183551.92 + 103056.8 & 18.37 & 16.97 & 17.97 & 16.73 & 16.30 + j1835.4 + 1349 & j183522.00 + 135733.9 & 19.69 & 18.29 & 19.36 & 18.15 & 17.24 + & j183535.34 + 134848.8 & 19.57 & 17.15 & 19.01 & 16.66 & 16.84 + j1837.9 + 3821 & j183656.31 + 382232.8 & 18.37 & 18.01 & 19.53 & 18.45 & 18.34 + & j183828.80 + 382704.3 & 20.75 & & 20.96 & 20.60 & + & j183837.16 + 381900.5 & 21.01 & 19.40 & 19.23 & 19.01 & 18.76 + j1844.3 + 1548 & j184425.36 + 154645.9 & 18.90 & 18.17 & 18.45 & 17.15 & 16.00 + j1844.9 - 1116 & j184456.29 - 111352.1 & & 18.29 & & & 13.34 + j1958.6 + 4020 & j195842.28 + 401125.8 & 18.81 & & 19.05 & & + j2124.0 - 1513 & j212423.63 - 152558.2 & 20.18 & 19.41 & 20.33 & 18.87 & 18.59 + j2128.7 + 5824 & j212900.37 + 583128.0 & 20.28 & 18.42 & 19.96 & 18.28 & 17.71 + name & name & mag & mag & mag & mag & mag + j0009.1 + 5030 & j000922.76 + 503028.8 & & & 19.74 & 19.35 & 17.32 + j0018.8 - 8154 & j001920.58 - 815251.3 & & 15.86 & 16.62 & 16.13 & 15.33 + j0022.2 - 1853 & j002209.25 - 185334.7 & 19.05 & 18.63 & 18.07 & 16.95 & 17.29 + j0022.3 - 5141 & j002200.08 - 514024.2 & & 15.65 & 17.38 & 15.94 & 16.65 + j0045.5 + 1218 & j004543.33 + 121712.0 & 18.22 & 17.40 & 18.75 & 16.91 & 15.66 + j0051.4 - 6241 & j005116.62 - 624204.3 & & 16.95 & 16.83 & 16.69 & 15.78 + j0059.7 - 5700 & j005846.56 - 565911.4 & & 17.45 & 17.39 & 16.95 & 17.07 + j0110.3 + 6805 & j011012.84 + 680541.1 & 18.66 & 16.51 & 18.14 & 16.33 & 15.37 + j0134.4 + 2636 & j013428.19 + 263843.0 & 16.87 & 16.74 & 17.09 & 15.91 & 15.55 + j0156.4 + 3909 & j015631.40 + 391430.9 & 17.99 & 18.05 & 18.71 & 19.03 & 18.25 + j0156.5 - 2419 & j015606.46 - 241754.3 & 21.01 & & 20.86 & 21.40 & + j0207.9 - 6832 & j020750.91 - 683755.1 & & 17.62 & 20.02 & 19.05 & 18.21 + j0210.7 - 5102 & j021046.19 - 510101.8 & & 14.85 & 17.39 & 14.82 & 15.13 + j0238.2 - 3905 & j023749.42 - 390050.3 & & 16.86 & 18.04 & 16.94 & 17.81 + & j023800.62 - 390504.6 & & 17.52 & 17.63 & 16.59 & 16.96 + j0248.6 + 8440 & j024948.30 + 843556.9 & 19.50 & 18.43 & 19.42 & 17.62 & 17.01 + j0253.4 + 3218 & j025333.64 + 321720.8 & & 19.83 & 20.39 & 19.69 & 19.46 + j0309.3 - 0743 & j030943.23 - 074427.4 & 18.33 & 15.96 & 18.16 & 16.01 & 16.71 + j0332.5 - 1118 & j033200.72 - 111456.1 & 17.78 & 17.08 & 18.85 & 17.75 & 18.30 + & j033223.25 - 111950.6 & 19.41 & 17.66 & 18.90 & 16.71 & 18.00 + j0333.7 + 2918 & j033349.00 + 291631.6 & 17.49 & 17.24 & 19.11 & 16.44 & 15.73 + j0334.3 + 6538 & j033356.74 + 653656.0 & & 18.93 & 19.70 & 17.66 & 17.01 + j0424.3 - 5332 & j042347.22 - 533026.6 & & 17.22 & 18.21 & 16.50 & 17.42 + & j042504.26 - 533158.3 & & 15.54 & 17.41 & 16.14 & 16.18 + j0433.9 - 5726 & j043344.12 - 572613.3 & & 17.73 & 18.90 & 17.78 & 18.67 + j0438.8 - 4521 & j043900.84 - 452222.6 & & 18.37 & 20.85 & 19.48 & + j0456.5 + 2658 & j045617.36 + 270221.1 & 20.42 & & 21.50 & & 18.68 + j0505.9 + 6116 & j050558.78 + 611335.9 & & 18.71 & 20.73 & 18.67 & 17.30 + j0506.7 - 5435 & j050657.80 - 543503.9 & & 15.95 & 16.73 & 16.91 & 16.25 + j0508.1 - 1936 & j050805.75 - 194721.6 & 17.57 & 17.51 & 19.44 & 18.22 & 18.09 + j0512.9 + 4040 & j051252.53 + 404143.7 & 16.46 & 15.11 & 16.39 & 15.35 & 14.83 + j0525.5 - 6011 & j052537.74 - 601732.0 & & 20.58 & & 20.86 & + j0532.0 - 4826 & j053158.61 - 482736.0 & & & 20.61 & & 18.98 + j0537.7 - 5716 & j053748.95 - 571830.2 & & 17.23 & 18.07 & 17.17 & 17.83 + j0609.4 - 0248 & j060915.06 - 024754.6 & 18.05 & 17.47 & 18.23 & 16.92 & 16.47 + j0621.9 + 3750 & j062157.63 + 375057.0 & 20.30 & & 20.29 & 19.96 & 18.63 + j0644.2 - 6713 & j064428.06 - 671257.3 & & & 20.47 & 20.87 & + j0647.8 - 6102 & j064806.55 - 610507.4 & & 18.75 & 18.89 & 19.26 & 18.05 + j0653.7 + 2818 & j065344.26 + 281547.5 & 19.15 & 18.26 & 18.12 & 17.54 & 17.72 + j0700.3 + 1710 & j070001.50 + 170921.8 & 18.29 & 18.08 & 18.55 & 16.02 & 17.10 + j0700.3 + 1710 & j070046.29 + 171019.8 & 18.32 & 17.18 & 17.29 & 17.23 & 16.70 + j0703.1 - 3912 & j070312.64 - 391418.9 & & 16.41 & 17.06 & 17.18 & 17.93 + j0706.5 + 7741 & j070651.32 + 774137.0 & 17.44 & 17.53 & 17.44 & 18.00 & 16.36 + j0706.7 - 4845 & j070549.12 - 483911.4 & & 18.96 & 19.14 & 18.53 & 18.10 + j0709.3 - 0256 & j070945.05 - 025517.4 & & & 19.61 & 19.29 & + j0726.0 - 0053 & j072550.63 - 005456.4 & 17.63 & 16.58 & 17.41 & 15.82 & 15.91 + j0734.2 - 7706 & j073443.44 - 771113.4 & & 19.76 & & 20.77 & + j0746.5 - 0713 & j074627.48 - 070949.7 & 19.73 & 18.81 & 19.64 & 19.52 & 17.37 + j0746.5 - 4758 & j074642.30 - 475455.2 & & 16.58 & 18.18 & 16.99 & 15.85 + j0816.7 - 2420 & j081639.46 - 242635.4 & 18.00 & & 18.26 & 18.06 & 16.14 + & j081640.41 - 242106.6 & 19.42 & & 20.70 & & + j0823.0 + 4041 & j082257.55 + 404149.8 & 19.00 & 18.7 & 19.34 & 19.47 & 18.16 + j0844.8 - 5459 & j084502.47 - 545808.5 & & & 19.35 & 16.80 & 17.65 + j0849.9 - 3540 & j084945.61 - 354101.2 & & 19.03 & 20.27 & 18.45 & + j0852.4 - 5756 & j085238.73 - 575529.4 & & 18.99 & 18.70 & 19.08 & 18.22 + j0855.1 - 0712 & j085435.20 - 071837.5 & 16.39 & 16.26 & 16.07 & 15.93 & 15.29 + j0856.0 + 7136 & j085654.85 + 714623.8 & 19.79 & 19.05 & & 19.75 & 17.29 + j0858.1 - 1952 & j085805.36 - 195036.8 & 19.19 & 18.54 & 18.63 & 18.93 & 17.76 + j0904.8 - 3513 & j090423.42 - 351203.0 & & 18.27 & 20.10 & 18.26 & + j0906.2 - 0906 & j090618.05 - 090544.9 & 19.06 & 19.13 & 19.07 & 18.59 & 18.04 + j0919.3 - 2203 & j092002.74 - 215835.0 & 18.55 & 17.99 & 19.26 & 19.24 & 18.12 + j0940.8 - 6105 & j094047.33 - 610728.5 & & 16.51 & 18.03 & 16.68 & 16.29 + j0941.9 - 0755 & j094221.46 - 075953.1 & 19.82 & 18.84 & 17.66 & 17.91 & 18.08 + j0946.9 - 2541 & j094709.52 - 254100.0 & 16.68 & 16.80 & 18.16 & 16.71 & 16.62 + j1016.2 - 0638 & j101542.96 - 063055.1 & 19.24 & & 19.94 & 19.70 & + j1016.2 - 0638 & j101626.98 - 063625.2 & 19.98 & & 19.81 & 18.86 & 18.07 + j1045.5 - 2931 & j104540.62 - 292726.4 & 19.30 & 19.06 & 18.77 & 19.17 & 18.64 + j1103.9 - 5356 & j110352.22 - 535700.7 & & 16.12 & 17.82 & 16.42 & 16.02 + j1106.3 - 3643 & j110624.04 - 364659.0 & & 18.71 & 19.65 & 19.20 & 18.37 + j1154.1 - 3242 & j115406.16 - 324243.0 & & 17.96 & 19.06 & 19.00 & 17.94 + j1230.2 - 5258 & j122939.88 - 530332.1 & & 16.62 & 18.03 & 17.41 & 16.72 + j1238.1 - 1953 & j123824.40 - 195913.4 & 18.03 & 17.48 & 17.66 & 17.01 & 17.97 + j1239.5 + 0728 & j123924.58 + 073017.2 & 19.07 & 17.76 & 19.34 & 17.92 & 17.96 + name & name & mag & mag & mag & mag & mag + j1301.6 + 3331 & j130147.03 + 332236.3 & 19.82 & 20.04 & 20.33 & 19.38 & 18.92 + j1303.8 - 5537 & j130349.23 - 554031.6 & & 16.82 & & 17.92 & 17.63 + j1304.1 - 2415 & j130416.70 - 241216.6 & 17.10 & 16.56 & 19.94 & 18.16 & 16.60 + j1304.3 - 4353 & j130421.01 - 435310.2 & & 15.75 & 17.48 & 16.10 & 15.48 + j1307.5 - 4300 & j130737.98 - 425938.9 & & 15.75 & 16.26 & 15.58 & 14.98 + j1307.6 - 6704 & j130817.51 - 670705.8 & & & 20.40 & 17.68 & + j1329.2 - 5608 & j132901.16 - 560802.5 & & & 17.38 & & 17.11 + j1330.1 - 7002 & j133011.34 - 700312.7 & & & 16.45 & 17.37 & + j1351.3 - 2909 & j135146.85 - 291217.4 & & & 20.22 & 19.44 & + j1406.2 - 2510 & j140609.60 - 250809.2 & 16.79 & 16.25 & 16.20 & 16.51 & 15.95 + j1416.3 - 2415 & j141642.23 - 241021.2 & 18.39 & 18.16 & 19.02 & 18.83 & 18.02 + j1419.4 - 0835 & j141922.56 - 083831.9 & 20.95 & 19.28 & 18.75 & 19.55 & 18.60 + j1514.6 - 4751 & j151440.03 - 474829.7 & & 17.61 & & 18.44 & 17.17 + j1518.2 - 2733 & j151803.60 - 273131.0 & 18.40 & 15.97 & 16.62 & 14.21 & 15.18 + j1553.2 - 2424 & j155331.62 - 242206.0 & 19.67 & 18.74 & 20.43 & 18.70 & 16.94 + j1558.3 + 8513 & j160031.76 + 850949.2 & 19.66 & 18.78 & 19.50 & 18.79 & 17.90 + j1604.5 - 4442 & j160431.03 - 444131.9 & & 17.50 & 20.00 & & + j1626.0 - 7636 & j162638.17 - 763855.4 & & 14.89 & 16.12 & 14.94 & 15.05 + j1725.1 - 7714 & j172350.86 - 771350.3 & & 19.12 & 19.71 & 18.94 & 18.31 + j1811.0 + 5340 & j181037.99 + 533501.5 & 18.88 & 18.77 & 19.34 & 18.45 & 16.71 + j1815.6 - 6407 & j181425.96 - 641008.8 & & 18.87 & 18.96 & 19.02 & + j1816.7 - 4942 & j181655.99 - 494344.7 & & 18.23 & 17.88 & 18.31 & 17.92 + j1818.7 + 2138 & j181905.22 + 213234.0 & 17.64 & 16.75 & 17.52 & 17.22 & 16.49 + j1820.6 + 3625 & j182023.61 + 362914.4 & 17.93 & 18.20 & & 17.94 & 17.43 + j1825.1 - 5231 & j182513.79 - 523058.1 & & 19.05 & 18.80 & 18.83 & 16.74 + j1830.0 + 1325 & j183000.76 + 132414.4 & 19.91 & 17.77 & 18.57 & 17.66 & 17.51 + j1830.2 - 4441 & j183000.86 - 444111.4 & & 16.54 & 18.17 & 17.47 & 16.88 + j1844.7 + 5716 & j184450.96 + 570938.6 & 17.69 & 17.79 & 18.52 & 17.49 & 17.55 + j1936.9 + 8402 & j193930.23 + 835925.8 & 19.59 & 18.55 & 18.89 & 19.11 & 18.45 + j1940.8 - 6213 & j194121.76 - 621120.8 & & & 21.07 & 20.04 & 18.53 + j1942.8 + 1033 & j194247.48 + 103327.2 & 18.59 & 16.82 & 16.69 & 15.37 & 15.24 + j1959.9 - 4727 & j195945.66 - 472519.2 & & 16.61 & 16.66 & 16.70 & 16.48 + j2103.6 - 6236 & j210338.38 - 623225.8 & & 16.00 & 17.83 & 16.24 & 16.13 + j2250.2 - 4205 & j225014.94 - 420218.6 & & 19.52 & 20.53 & 19.44 & 18.46 + & j225022.20 - 420613.2 & & 16.21 & 17.41 & 17.23 & 16.49 + j2317.3 - 4534 & j231731.97 - 453359.6 & & 17.36 & 18.24 & 18.80 & 17.44 + j2323.0 - 4918 & j232255.30 - 491942.0 & & 20.55 & 20.19 & 20.13 & 17.67 + j2324.6 + 0801 & j232445.31 + 080206.3 & 18.75 & 17.97 & 18.63 & 17.37 & 17.92 + j2325.4 + 1650 & j232526.62 + 164941.1 & 20.45 & & & 20.50 & + & j232538.11 + 164642.8 & 18.56 & 18.29 & 17.37 & 17.09 & 17.27 +
recent observations of the cosmic microwave background ( @xmath0 ) and of the large scale structure of the universe ( @xmath3 ) have led to the definition of a concordance cosmological model . the analysis of the @xmath0 temperature data over the whole celestial sphere from the wilkinson microwave anisotropy probe ( @xmath4 ) satellite experiment has been playing a dominant role in designing this precise picture of the universe @xcite . according to this model , the cosmic structures and the @xmath0 originate from gaussian adiabatic perturbations seeded in the early phase of inflation of the universe . however , cosmological scenarios motivated by theories for the unification of the fundamental interactions predict the existence of topological defects , resulting from phase transitions at the end of inflation @xcite . these defects would have participated to the formation of the cosmic structures , also imprinting the @xmath0 . cosmic strings are a line - like version of such defects which are also predicted in the framework of fundamental string theory @xcite . as a consequence , the issue of their existence is a central question in cosmology today . cosmic strings are parametrized by a string tension @xmath5 , i.e. a mass per unit length , which sets the overall strength of a string network . their main signature in the @xmath0 is characterized by temperature steps along the string positions . this localized effect , known as the kaiser - stebbins ( @xmath1 ) effect @xcite , hence implies a non - gaussian imprint of the string network in the @xmath0 . the most numerous strings appear at an angular size around @xmath6 @xmath7 on the celestial sphere . @xmath0 experiments with an angular resolution much below @xmath6 @xmath7 are thus required in order to resolve the width of cosmic strings . experimental constraints have been obtained on a possible string contribution in terms of upper bounds on the string tension @xmath5 @xcite . in this context , even though observations largely fit with an origin of the cosmic structures in terms of adiabatic perturbations , room is still available for the existence of cosmic strings . the observed @xmath0 signal can be modeled as a linear superposition of a statistically isotropic but non - gaussian string signal @xmath8 proportional to an unknown string tension , with statistically isotropic gaussian noise @xmath9 comprising the standard components of the @xmath0 induced by adiabatic perturbations , as well as instrumental noise . in a first recent work @xcite we developed an effective method for mapping the string network potentially imprinted in @xmath0 temperature data , in the perspective of forthcoming arcminute resolution experiments performing a mapping of planar patches of the sky in real space , with standard radiometers . we took a bayesian approach to this denoising problem , based on statistical models for both the string signal and the noise . the denoising is done in the wavelet domain , using a steerable wavelet transform well adapted for representing the strongly oriented features present in the string signal . we showed that the string signal coefficients are well modeled by generalized gaussian distributions ( @xmath10 s ) , which are fitted at each wavelet scale using a training simulation borrowed from the set of realistic string simulations recently produced by @xcite . in the present work we aim at extending the approach described above to the reconstruction of a string signal from radio - interferometric observations of the @xmath0 probing small patches of the celestial sphere at the same arcminute resolution . under the most standard assumptions , radio interferometers provide a noisy and incomplete fourier coverage of the planar signal . in these terms , the signal reconstruction amounts to solving an ill - posed inverse problem encompassing both denoising and deconvolution . relying on the @xmath1 effect in the @xmath0 , the string network itself can be mapped by the magnitude of the gradient @xmath11 of the string signal , which is very sparse or compressible in nature , with significant amplitudes only along the strings themselves . the theory of compressed sensing offers a new framework for solving ill - posed inverse problem when the signals are sparse or compressible @xcite . a band - limited signal may be expressed as the @xmath12-dimensional vector of its values sampled at the nyquist - shannon rate . by definition , a signal is sparse in some basis if its expansion contains only a small number @xmath13 of non - zero coefficients . more generally it is compressible if its expansion only contains a small number of significant coefficients , i.e. if a large number of its coefficients bear a negligible value . the theory of compressed sensing demonstrates that , for sparse or compressible signals , a small number @xmath14 of random measurements , in a sensing basis incoherent with the sparsity basis , will suffice for an accurate and stable reconstruction of the signals . the mutual coherence between two bases may be defined as the maximum complex modulus of the scalar product between unit - norm vectors of the two bases . a random selection of fourier measurements of a signal sparse in real space or in its gradient are particular examples of good sensing procedures . in this framework , the reconstruction of the signals may be approached in different ways , but all of them encompass a regularization of the ill - posed inverse problem by the introduction of a sparsity or compressibility constraint . in a second recent work @xcite , we presented results showing that compressed sensing offers powerful image reconstruction techniques for radio - interferometric data , which are stable relative to noise , as well as stable relative to non - exact sparsity , i.e. compressibility of the signals . these techniques are based on global minimization problems , which are solved by convex optimization algorithms @xcite . we particularly illustrated the versatility of the scheme relative to the inclusion of specific prior information on the signal in the minimization problems . we here wish to apply these results for the reconstruction of a string signal in the @xmath0 . as for the denoising algorithm proposed in @xcite , the reconstruction algorithm that we present may be considered as a modular component of a larger data analysis . it may notably find use as a pre - processing step for other methods for cosmic string detection based on explicit edge detection @xcite . our reconstruction algorithm is tested under different conditions with astrophysical noise components including various contributions to the standard components of the @xmath0 , i.e. primary and secondary anisotropies . our analyses rely on the set of realistic string simulations recently produced by @xcite . the remainder of this paper is organized as follows . in section [ sec : string signal and noise ] , we discuss the string signal and noise in the @xmath0 . in section [ sec : interferometric inverse problem ] , we describe the inverse problem posed in the context of radio interferometry . in section [ sec : compressed sensing reconstruction ] , we define the minimization problems to be solved for signal reconstruction in the context of compressed sensing , notably relying on a prior statistical model of the string signal fitted from simulations . in section [ sec : analyses and results ] , we describe our numerical analyses and results . we finally conclude in section [ sec : conclusion ] . in this section , we describe the string signal in the @xmath0 , as well as the astrophysical noise made up of the standard primary and secondary anisotropies . we also describe the numerical simulations used for our subsequent analyses . in an inflationary cosmological model , the phase transitions responsible for the formation of a cosmic string network occur after the end of inflation , so as to produce observable defects . from the epoch of last scattering until today , the cosmic string network continuously imprints the @xmath0 . the so - called scaling solution for the string network implies that the most numerous strings are imprinted just after last scattering and have a typical angular size around @xmath6 @xmath7 , of the order of the horizon size at that time @xcite . longer strings are also imprinted in the later stages of the universe evolution , but in smaller number , according to the number of corresponding horizon volumes required to fill the sky . the main signature of a cosmic string in the @xmath0 is described by the @xmath1 effect according to which a temperature step is induced along the string position . the relative amplitude of this step is given by @xmath15 where @xmath16 and @xmath17 with @xmath18 standing for the string velocity transverse to the line of sight and @xmath19 for the speed of light , and where @xmath20 is a dimensionless parameter uniquely associated with the string tension @xmath5 through @xmath21 with @xmath22 standing for the gravitational constant . in the following we call @xmath20 the string tension . on small angular scales , realistic simulations can be produced by stacking @xmath0 maps induced in different redshift ranges between last scattering and today . the simulations that we use in this work have been produced by this technique @xcite . the string signal is understood as a realization of a statistically isotropic but non - gaussian process on the celestial sphere with an overall amplitude rescaled by the string tension @xmath20 , and characterized by a nearly scale - free angular power spectrum : @xmath23 , where the positive integer index @xmath24 stands for the angular frequency index on the sphere . an analytical expression of this spectrum was provided for @xmath24 larger than a few hundreds by @xcite , on the basis of their simulations . we consider radio - interferometric @xmath0 observations with a small field of view corresponding to an angular opening @xmath25 on the celestial sphere . in this context , the small portion of the celestial sphere accessible is identified to a planar patch of size @xmath26 , and we may consider planar signals functions of a two - dimensional position vector @xmath27 . the corresponding spatial frequencies may be denoted as two - dimensional vectors @xmath28 with a radial component given by the norm @xmath29 of the vector . in this euclidean limit , the radial component identifies with the angular frequency on the celestial sphere , below some band limit @xmath30 set by the resolution of the interferometer under consideration : @xmath31 . for @xmath29 larger than a few hundreds , the nearly scale - free planar power spectrum of the string signal @xmath32 reads as : @xmath33 with @xmath34 for @xmath35 . the observed @xmath0 signal can be understood as a linear superposition of the string signal and statistically isotropic noise of astrophysical and instrumental origin @xmath36 with some angular power spectrum @xmath37 . in the euclidean limit , the corresponding planar power spectrum for the noise @xmath36 may be written as @xmath38 for @xmath35 . let us note that we consider zero mean signals , identifying perturbations around statistical means . angular power spectra of the string signal and noise ( borrowed from @xcite ) as functions of the angular frequency in the range @xmath39 $ ] in @xmath40 axes scaling . the spectrum of the string signal is represented for a string tension @xmath41 in terms of its analytical expression valid at high angular frequencies ( red straight line ) . the noise spectra ( ordered by decreasing amplitude at low angular frequencies ) are : the primary @xmath0 anisotropies ( black solid line ) and the gravitational lensing correction ( black dashed line ) , the thermal @xmath42 effect in the rayleigh - jeans limit ( blue solid line ) , the non - linear kinetic @xmath42 effect ( green solid line ) , and the ostriker - vishniac effect ( magenta solid line).,width=302 ] current @xmath0 experiments achieve an angular resolution on the celestial sphere of the order of @xmath43 arcminutes , corresponding to a limit angular frequency not far above @xmath44 . at such resolutions , the standard components of the @xmath0 primarily contain the statistically isotropic and gaussian primary anisotropies induced by adiabatic perturbations at last scattering . in this context , any possible string signal is confined to amplitudes largely dominated by these primary anisotropies . the constraints mainly come from a best fit analysis of the angular power spectrum of the overall @xmath0 signal in the @xmath4 temperature data @xcite . the tightest of these constraints @xcite gives the following upper bound at @xmath45 per cent confidence level : @xmath46 forthcoming experiments will provide access to higher angular resolution . among other instruments , radio interferometers such as the arcminute microkelvin imager ( ami ) will probe the @xmath0 at angular resolutions below @xmath6 @xmath47 @xcite . at these resolutions , the secondary @xmath0 anisotropies induced by interaction of @xmath0 photons with the evolving universe after last scattering will dominate the primary anisotropies , and must be accounted for . the primary anisotropies exhibit exponential damping at high angular frequencies . this contrasts with the slow decay of the nearly scale - free angular power spectrum of the string signal , which thus dominates over the primary anisotropies at high enough angular frequencies . the secondary anisotropies include gravitational effects such as the integrated sachs - wolfe ( @xmath48 ) effect , the rees - sciama ( @xmath49 ) effect , and gravitational lensing , as well as re - scattering effects such as the thermal and kinetic sunyaev - zeldovich ( @xmath42 ) effects . the @xmath42 effects dominate these secondary anisotropies @xcite . the @xmath48 and @xmath49 effects associated with the time evolution of the standard gravitational potentials can be neglected at these angular frequencies . one may thus restrict the secondary anisotropies considered in the noise to the linear ( ostriker - vishniac ) and non - linear kinetic @xmath42 effects , as well as the thermal @xmath42 effect . the @xmath42 effects are actually non - gaussian , spatially dependent , and the kinetic and thermal effects are correlated . as a simplifying assumption , we treat these two effects as two independent statistically isotropic gaussian noise components . the effect of gravitational lensing is very small relative to the @xmath42 effects , but we still take it into account as a correction to the angular power spectrum of the primary anisotropies . at arcminute resolution , the thermal and kinetic @xmath42 effects have standard deviations around @xmath43 and @xmath50 @xmath51 respectively . they also have a slow decay at high angular frequencies and will dominate the string signal for string tension values below the current experimental upper bound . arcminute @xmath0 experiments are in fact primarily dedicated to the detection of these secondary anisotropies . unlike the other effects considered , which have the same black body spectrum as the primary anisotropies , the thermal @xmath42 effect on the @xmath0 temperature depends on the frequency of observation . its amplitude decreases between the rayleigh - jeans limit ( null frequency ) and @xmath52 @xmath53 where it is expected to vanish , before increasing again at higher frequencies . figure [ fig : spectra ] represents the angular power spectra as functions of the angular frequency @xmath24 , for a string signal with string tension @xmath41 , for the primary @xmath0 anisotropies and the correction due to gravitational lensing , for the ostriker - vishniac and the non - linear kinetic @xmath42 effect , and for the thermal @xmath42 effect in the rayleigh - jeans limit . these spectra are explicitly borrowed from @xcite . in this context , the performance of the reconstruction algorithm to be defined will be studied in the following limits . in a first noise condition , we will consider the secondary anisotropies as a statistically isotropic gaussian noise , with power spectrum given by the rayleigh - jeans limit , that is added to the primary anisotropies . in a second noise condition , we will also assume an observation frequency around @xmath52 @xmath53 taking advantage of the frequency dependence of the thermal @xmath42 effect , and include in the noise secondary anisotropies in absence of this effect . this is equivalent to including only the kinetic @xmath42 effect and gravitational lensing in the secondary anisotropies . instrumental noise is considered to be negligible and simply discarded . these two different noise conditions are respectively denoted as @xmath54 ( secondary anisotropies with thermal @xmath42 effect ) and @xmath55 ( secondary anisotropies without thermal @xmath42 effect ) in the following . analyzing these limits can reveal to what extent the kinetic and thermal @xmath42 effects hamper the reconstruction of the string signal , as a function of the string tension . in a third noise condition , the reconstruction performance will also be examined in the limit where the noise only includes primary anisotropies , neglecting even instrumental noise , and assuming secondary anisotropies have been correctly separated . this case is denoted as @xmath56 ( primary anisotropies without instrumental noise ) and will be studied in order to understand the behaviour of the reconstruction algorithm in ideal conditions . for the sake of our analyses , foreground emissions such as galactic dust or point sources @xcite are disregarded . simulated temperature maps ( in @xmath57 ) of the string signal in the @xmath0 and of its magnitude of gradient for a tension @xmath58 are represented in the top and bottom left panels respectively . temperature maps of the primary @xmath0 anisotropies , i.e. the noise in the @xmath56 condition , and of its magnitude of gradient are represented in the top and bottom right panels respectively.,title="fig:",width=151 ] simulated temperature maps ( in @xmath57 ) of the string signal in the @xmath0 and of its magnitude of gradient for a tension @xmath58 are represented in the top and bottom left panels respectively . temperature maps of the primary @xmath0 anisotropies , i.e. the noise in the @xmath56 condition , and of its magnitude of gradient are represented in the top and bottom right panels respectively.,title="fig:",width=151 ] + simulated temperature maps ( in @xmath57 ) of the string signal in the @xmath0 and of its magnitude of gradient for a tension @xmath58 are represented in the top and bottom left panels respectively . temperature maps of the primary @xmath0 anisotropies , i.e. the noise in the @xmath56 condition , and of its magnitude of gradient are represented in the top and bottom right panels respectively.,title="fig:",width=151 ] simulated temperature maps ( in @xmath57 ) of the string signal in the @xmath0 and of its magnitude of gradient for a tension @xmath58 are represented in the top and bottom left panels respectively . temperature maps of the primary @xmath0 anisotropies , i.e. the noise in the @xmath56 condition , and of its magnitude of gradient are represented in the top and bottom right panels respectively.,title="fig:",width=151 ] we use simulations of the string signal borrowed from the set of realistic simulations produced by @xcite . the specific simulations considered are defined on planar patches of size @xmath26 for a field of view defined by an angular opening @xmath59 @xmath7 . the finite size of the patch induces a discretization of the spatial frequencies which may be defined on a grid with @xmath60 uniformly sampled points @xmath61 with @xmath62 , below some band limit @xmath30 . the original maps are sampled at nyquist - shannon rate on grids with @xmath60 uniformly sampled points @xmath63 with @xmath62 . we consider here samples with @xmath64 , so that the corresponding pixels have an angular size around @xmath65 , with a band limit @xmath66 on each each component of the spatial frequencies . for each noise component , a simulation may easily be produced by taking the fourier transform of gaussian white noise , renormalizing each fourier coefficient by the square root of the corresponding power spectrum , and inverting the fourier transform @xcite . in each noise condition considered , an overall noise simulation is obtained by simple superposition of the required independent components simulated . the power spectrum of the noise @xmath67 is the sum of the individual spectra . our approach for signal reconstruction notably relies on the introduction of specific statistical prior information on the signal in the regularization of the ill - posed inverse problem . a number of @xmath68 simulations of the string signal @xmath8 are used as training data for fitting a prior statistical model for the gradient of the signal . only @xmath6 additional simulation is reserved for testing the algorithm . in all noise conditions considered and for each string tension , this test string signal simulation is combined with @xmath69 independent realizations of the noise in order to produce multiple test simulations . for illustration , figure [ fig : maps - gradients ] represents the test simulation of the string signal and one simulation of the noise , as well as corresponding maps of the magnitude of gradient . note that due to the finite size of the simulations , different string signal simulations exhibit slightly different estimated power spectra . for simplicity of our argument though both training and test simulations have been normalized so as to exhibit the same total power for the same string tension ( see @xcite for more details ) . in this section , we describe the ill - posed inverse problem posed for reconstruction of the string signal in the context of radio - interferometric observations of the @xmath0 . we also describe a modified inverse problem including a whitening operation on the measured visibilities and discuss basic ideas for signal reconstruction . a map of the primary beam considered is represented in the left panel , with darkness increasing with the value of the function @xmath70 in the range [ @xmath71 . a map of the mask made up of a random selection of @xmath72 per cent of the spatial frequencies up to the band limit in the fourier plane is represented in the right panel , with a unit value at the position of the selected frequencies identified by black points , and zero otherwise.,title="fig:",width=151 ] a map of the primary beam considered is represented in the left panel , with darkness increasing with the value of the function @xmath70 in the range [ @xmath71 . a map of the mask made up of a random selection of @xmath72 per cent of the spatial frequencies up to the band limit in the fourier plane is represented in the right panel , with a unit value at the position of the selected frequencies identified by black points , and zero otherwise.,title="fig:",width=151 ] aperture synthesis in radio interferometry is a powerful technique in radio astronomy dating back to more than sixty years ago @xcite . in this context , the small portion of the celestial sphere around the pointing direction tracked by a radio telescope array during observation defines the original real planar signal or image probed @xmath8 . the field of view observed is limited by a primary beam represented by a function @xmath70 with values in the range [ @xmath71 . each telescope pair at one instant of observation identifies a baseline defined as the relative position between the two telescopes . to each baseline is associated one measurement called visibility . in the simplest setting one also considers baselines with negligible component in the pointing direction of the instrument . under this additional assumption , if the signal is made up of incoherent sources , each visibility corresponds to the value of the fourier transform of the signal multiplied by the primary beam at a spatial frequency @xmath28 identified by the components of the baseline in the plane of the signal . we assume that the spatial frequencies @xmath28 probed by all telescope pairs during the observation belong to the discrete grid of points @xmath73 ( see @xcite for more details ) . the fourier coverage provided by the @xmath74 spatial frequencies probed @xmath75 , with @xmath76 , can simply be identified by a binary mask in the fourier plane equal to @xmath6 for each spatial frequency probed and @xmath77 otherwise . the visibilities measured may be denoted by a vector of @xmath74 complex fourier coefficients @xmath78 . formally , the visibilities may equivalently be denoted by a vector of @xmath79 real measures @xmath80 consisting of the real and imaginary parts of the complex measures . in this discrete setting , the fourier coverage is in general incomplete in the sense that the number of real constraints @xmath79 is smaller than the number of unknowns @xmath12 : @xmath81 . an ill - posed inverse problem is thus defined for the reconstruction of the signal , seen as a vector of real sampled values @xmath82 , from the measured visibilities @xmath83 . in the context of our analysis , the noise affecting the signal consists in the vector of real sampled values for the standard @xmath0 components @xmath84 . the inverse problem therefore reads as : @xmath85 where the measurement matrix @xmath86 identifies the complete linear relation between the signal and the visibilities . the matrix @xmath87 is the diagonal matrix implementing the primary beam . the unitary matrix @xmath88 implements the discrete fourier transform providing the fourier coefficients . the matrix @xmath89 is the rectangular binary matrix implementing the mask characterizing the interferometer . it contains only one non - zero value on each line , at the index of the fourier coefficient corresponding to each of the spatial frequencies probed @xmath75 . neglecting small correlations introduced between spatial frequencies by the primary beam , the power spectrum of the noise @xmath90 is nonetheless modified as @xmath91 . the noise @xmath92 affecting the visibilities is gaussian noise with different variances @xmath93 for different components @xmath94 . let us recall that @xmath90 is real so that the components of @xmath95 are independent only when assuming that the mask @xmath96 contains no pair of antipodal spatial frequencies @xmath97 . for each such pair , the two corresponding noise components are complex conjugate of one another . in this work @xmath74 denotes the number of independent complex visibilities , while @xmath98 , with @xmath99 , will stand for the total number of complex visibilities including possible pairs of antipodal spatial frequencies . the measurement matrix may be augmented by a whitening operation which will be useful in a reconstruction perspective . it consists in dividing each measured visibility @xmath100 by the standard deviation of the real and imaginary parts of the corresponding noise component @xmath94 . this operation is simply implemented in terms of the diagonal matrix @xmath101^{1/2}\delta_{bb'}\}_{1\leq b , b'\leq ( m/2)}$ ] . this gives rise to the modified inverse problem @xmath102 where the whitened visibilities @xmath103 are now affected by identically distributed independent ( i.i.d . ) gaussian noise with unit variance of the real and imaginary part of each component @xmath94 , i.e. white gaussian noise . for each test simulation , the visibilities are computed in accordance with relation ( [ eq : iip1 ] ) . the primary beam @xmath70 considered for our observational set up is a gaussian function with a full width at half maximum ( @xmath104 ) equal to the angular opening of the field of view @xmath105 . the function has a maximum value equal to unity . two fourier coverages are considered independently , defined by masks made up of a random selection of @xmath106 spatial frequencies ( see @xcite for more details ) with either @xmath72 per cent or @xmath107 per cent of the fourier plane probed up to the band limit @xmath30 . note that , after discarding redundant visibilities associated with pairs of antipodal spatial frequencies in the two masks considered , these two percentages respectively correspond to selections of @xmath79 spatial frequencies with @xmath108 and @xmath109 per cent . the primary beam as well as the mask for a fourier coverage of @xmath72 per cent are illustrated in figure [ fig : obssetup ] . the visibilities are also whitened in accordance with relation ( [ eq : iip1 ] ) by the introduction of the suitable matrix @xmath110 in each noise condition . in the perspective of the reconstruction of the signal @xmath111 , relation ( [ eq : iip1 ] ) represents the measurement constraint . considering a candidate reconstruction @xmath112 , the residual noise reads as @xmath113 . the residual noise level estimator , defined as twice the negative logarithm of the likelihood associated with @xmath112 , simply reads as @xmath114 the notation @xmath115 for a scalar stands for the complex modulus when applied to a complex number and for the absolute value when applied to a real number . the @xmath116 norm of the residual noise is the standard norm of the corresponding vector : @xmath117 . this noise level estimator follows a chi - square distribution with @xmath79 degrees of freedom . typically , this estimator should be minimized by the good candidate reconstruction . the measurement constraint on the reconstruction may be defined as a bound @xmath118 , with @xmath119 corresponding to some suitable large percentile of the @xmath120 distribution in order to avoid noise over - fitting . in this context , many signals may formally satisfy the measurement constraint . a regularization scheme that encompasses enough prior information on the original signal is needed in order to find a unique solution . all image reconstruction algorithms will differ through the kind of regularization considered . firstly , the simplest possible reconstruction procedure would consist in assuming that the non - probed spatial frequencies have a null value , and simply inverting the fourier transform after dividing each visibility by the variance of the noise in order to enhance the signal - to - noise ratio . this reconstruction can by written as @xmath121 , where the notation @xmath122 stands for the conjugate transpose operation . this map simply identifies with the convolution of the original signal and noise @xmath123 by a beam @xmath124 , where @xmath125 stands for a spike at the center of the field of view : @xmath126 , with @xmath127 identifying the central pixel . in the standard vocabulary of radio interferometry , the beam is called the dirty beam , and the convolved map is called the dirty map in recognition of its visual appearance contrasting with the original signal . a normalization constant needs to be chosen so as to get a unit central value of the beam . in other words , no denoising or deconvolution is performed by this simple procedure that we call @xmath128 . the only important operation in this reconstruction is the enhancement of the signal - to - noise ratio by the matrix @xmath129 . secondly , the standard @xmath2 algorithm is essentially a matching pursuit ( @xmath130 ) algorithm @xcite which performs the deconvolution by iterative removal of the dirty beam , thus leading from the dirty map to a clean map @xcite . formally , the algorithm should be stopped when the @xmath120 constraint above is satisfied . in this section , we describe the standard global minimization problem @xmath131 for the reconstruction of the string signal in the context of compressed sensing . we model the prior statistical distribution of the gradient of the string signal with a @xmath10 , fitted on the basis of the training simulations . we also propose to enhance the @xmath131 problem on the basis of this statistical prior information , hence defining a @xmath132 minimization problem . we finally propose a power spectral model ( psm ) used for independent estimation of the string tension , which may notably be useful for accelerating the convergence of the iterative algorithms solving the minimization problems . for the reconstruction of a signal of sparsity @xmath133 , the theory of compressed sensing proves that the number @xmath79 of random measurements required , in a sensing basis incoherent with the sparsity basis , roughly scales with the sparsity , well below the number of samples @xmath12 of the signal at nyquist - shannon rate : @xmath134 @xcite . in the case of radio - interferometric data , the measurements are assumed to be random fourier samples which represent a good sensing procedure for a string signal @xmath8 exhibiting sparse magnitude of the gradient @xmath11 . in this case , in the presence of i.i.d . gaussian noise , the so - called total variation minimization problem ( @xmath131 ) applies @xcite , which corresponds to the minimization of the @xmath135 norm of the real signal under a @xmath136 constraint on the candidate reconstruction @xmath112 : @xmath137 the vector of sample values of the magnitude of the gradient associated with the vector @xmath111 is denoted as @xmath138 . by definition , the @xmath135 norm of a signal is the @xmath139 norm of the magnitude of its gradient , i.e. @xmath140 . we set @xmath119 to be the @xmath141 percentile of the @xmath120 distribution . the iterative algorithm considered for solving this convex optimization problem is based on the douglas - rachford splitting method @xcite in the framework of proximal operator theory @xcite . [ c ] [ c ] positive tail of modeled prior @xmath10 distribution @xmath142 for a string tension @xmath143 . the @xmath10 is superimposed on the histogram of the aggregate of the magnitude of the gradient of the @xmath68 test simulations.,title="fig:",width=226 ] in a bayesian perspective , the minimization of the @xmath135 norm of the signal amounts to assuming that the signal gradient follows a laplacian distribution with no spatial correlations , and to minimizing the negative logarithm of this prior distribution , i.e. to maximizing the distribution , subject to the @xmath120 constraint . this promotes compressibility in a generic sense . a laplacian distribution is indeed more peaked and more heavy - tailed than a gaussian prior for example . however , with realistic simulations of the string signal at hand one can study more precisely the signal distribution . in that perspective , the signal gradient can be modeled at each sample point by the positive tail of a generalized gaussian distribution ( @xmath10 ) : @xmath144 where @xmath145 is the gamma function , and where @xmath146 and @xmath147 are respectively called shape and scale parameters . the shape parameter @xmath146 can be considered as a continuous measure of the compressibility of the underlying distribution . setting @xmath148 recovers a gaussian distribution , and @xmath149 recovers a laplacian distribution . letting @xmath146 approach @xmath77 yields compressible distributions . this parameter determines the kurtosis @xmath150 , i.e. the ratio of the fourth central moment to the square of the variance ( second central moment ) , by @xmath151^{2}}.\label{eq : csr3}\ ] ] by nature this kurtosis is independent of the string tension @xmath20 . the scale parameter @xmath147 is linearly proportional to the standard deviation @xmath152 of the distribution . the corresponding variance reads as : @xmath153^{2}= \rho^{2}\frac{\gamma\left(3q^{-1}\right)}{\gamma\left(q^{-1}\right)}s^{2}.\label{eq : csr4}\ ] ] the shape and scale parameters are estimated by a moment method on the aggregate of the magnitude of the gradient of the @xmath68 test simulations . the values with two significant figures are @xmath154 and @xmath155 . the shape parameter value shows that the signal gradient exhibits a much more compressible gradient than a generic laplacian distribution . figure [ fig : histogram ] shows the positive tail of the modeled prior @xmath10 distribution @xmath142 for a string tension @xmath143 . the @xmath10 is superimposed on the histogram of the aggregate of the magnitude of the gradient of the @xmath68 test simulations . qualitatively , we see that the prior distribution is well modeled by a @xmath10 . accounting for the specific statistical prior information in the regularization of the inverse problem should enhance the reconstruction performance , beyond the standard @xmath131 minimization . in the bayesian perspective discussed , we define a statistical total variation @xmath132 minimization problem that consists in minimizing the logarithm of the @xmath10 prior distribution modeled , subject to the @xmath120 constraint : @xmath156 again , we set @xmath119 to be the @xmath141 percentile of the @xmath120 distribution . because @xmath157 the problem is non - convex in nature , but a re - weighted scheme allows convergence of the algorithm through iterations of simple @xmath131 minimizations @xcite . note that no prior knowledge of the string tension @xmath20 is formally required for the reconstruction . however we must acknowledge the fact that convergence appears to be faster when some prior estimation of @xmath20 is used to renormalize @xmath112 in the minimization term of relations ( [ eq : csr1 ] ) and ( [ eq : csr5 ] ) , as expected from relation ( [ eq : csr2 ] ) . as discussed in section [ sec : analyses and results ] , such an estimation of @xmath20 will also define a threshold below which one should not expect good reconstruction of the string network . our estimation procedure is described below . by bayes theorem , the posterior probability distribution for @xmath20 given the measured visibilities @xmath158 is simply obtained from the likelihood @xmath159 and the prior probability distribution @xmath160 on @xmath20 . for complete consistency , the likelihood should be calculated using the model established for the signal gradient and noise . however , while this model by construction accounts for the non - gaussianity , i.e. sparsity of the string signal , it ignores the spatial correlations . as in @xcite , we have observed that a likelihood yielding a precise localization of the string tension value can actually be obtained using a power spectral model . such a model assumes that both the string signal and the noise arise from statistically isotropic gaussian random processes , such that their fourier coefficients are independent gaussian variables . under this model , as the string signal and noise are independent , the observed signal @xmath123 has a power spectrum , as modified by the primary beam , of the form : @xmath161 with @xmath162 defined in section [ sub : reconstruction basics ] and similarly @xmath163 . in this setting the likelihood can be computed easily in terms of the measured visibilities @xmath83 as : @xmath164 the posterior probability distribution for @xmath20 given the measured visibilities thus reads as @xmath165 with normalization @xmath166 . we take the prior @xmath160 to be flat in an interval @xmath167 $ ] , with an upper bound @xmath168 large relative to the upper bound ( [ eq : ssn4 ] ) . we define an estimation @xmath169 of the string tension as the expectation value of this posterior probability distribution : @xmath170.\label{eq : csr9}\ ] ] in this section , we firstly describe our numerical analyses , including the performance criteria of reconstruction , and discuss their results . we also compare the corresponding eye visibility thresholds on the string tension with the detectability threshold obtained on the basis of the psm . we finally propose a small discussion of our results . the overall reconstruction is effective for mapping the string network if the magnitude of gradient of the reconstructed signal closely resembles the magnitude of gradient of the true string signal . a simple qualitative measure of the performance is given by whether the string network is visible in the magnitude of gradient of the reconstructed signal . we define the eye visibility threshold as the minimum string tension around which the overall denoising and mapping by the magnitude of the gradient begins to exhibit string features visible by eye . we will augment this qualitative assessment of the reconstruction performance with three quantitative measures , namely the signal - to - noise ratio , the correlation coefficient and the kurtosis of the magnitude of the gradient of the reconstructed string signal re - multiplied by the primary beam . the kurtosis is known to be a good statistic for discriminating between models with and without cosmic strings @xcite . the signal - to - noise ratio is defined in terms of the original and reconstructed string signal gradients as : @xmath171 where @xmath172 and @xmath173 respectively stand for the standard deviations of the discrepancy signal @xmath174 and of the original signal @xmath175 . the standard deviations are estimated from the sample variances on the basis of the signal realizations concerned . with this definition , the @xmath176 is measured in decibels ( @xmath177 ) . we will consider that the reconstruction is meaningful in terms of signal - to - noise ratio for the values of @xmath20 where this statistic is positive . the correlation coefficient is also defined in terms of the original and reconstructed string signal gradients as : @xmath178 where @xmath179 stands for the covariance between @xmath175 and @xmath180 , and @xmath173 and @xmath181 respectively stand for the standard deviations of the signals @xmath175 and @xmath180 . the covariance and variances are again estimated from the sample covariance and variances on the basis of the signal realizations concerned . we will consider that the reconstruction is meaningful in terms of correlation coefficient for the values of @xmath20 where this statistic is also positive . note that the variances and covariances discussed obviously depend on the string tension , but not their dimensionless ratios . for this reason we have omitted this dependence in the definitions ( [ eq : r1 ] ) and ( [ eq : r2 ] ) . analogously , kurtoses of the magnitude of the gradient of the reconstructed string signal re - multiplied by the primary beam are estimated from the sample kurtoses on the basis of the signal realizations concerned . the kurtosis of the magnitude of gradient of a string signal as expected on the basis of our test string signal simulation takes the value @xmath182 with two significant numbers . we compare the reconstruction results of the simple @xmath128 procedure and of the standard @xmath183 algorithm with those of the @xmath131 and @xmath132 algorithms . reconstructions are performed for string tensions equi - spaced in logarithmic scaling in the range @xmath184 $ ] , corresponding to ratio values for @xmath20 of @xmath185 , @xmath186 , @xmath187 , @xmath188 , and @xmath189 in each order of magnitude . for each string tension , each algorithm , each noise condition ( @xmath56 , @xmath55 , or @xmath54 ) and each coverage condition ( @xmath72 or @xmath107 per cent ) analysed , we consider that the quantitative measures described above indicate effective performance when this performance is significant over the @xmath69 corresponding simulations . .[tab : detectionthresholds ] psm detectability thresholds and eye visibility thresholds on the string tension , for each of the noise and coverage conditions considered . all values are given with two significant figures . [ cols="^,^,^,^",options="header " , ] magnitude of the gradient of reconstructed string signals re - multiplied by the primary beam , for a tension @xmath190 , in the noise condition @xmath56 and for a coverage at @xmath72 per cent . the reconstructions considered are those associated with the simple @xmath128 procedure ( top left panel ) , @xmath183 ( top right panel ) , @xmath131 ( bottom left panel ) , and @xmath132 ( bottom right panel).,title="fig:",width=151 ] magnitude of the gradient of reconstructed string signals re - multiplied by the primary beam , for a tension @xmath190 , in the noise condition @xmath56 and for a coverage at @xmath72 per cent . the reconstructions considered are those associated with the simple @xmath128 procedure ( top left panel ) , @xmath183 ( top right panel ) , @xmath131 ( bottom left panel ) , and @xmath132 ( bottom right panel).,title="fig:",width=151 ] + magnitude of the gradient of reconstructed string signals re - multiplied by the primary beam , for a tension @xmath190 , in the noise condition @xmath56 and for a coverage at @xmath72 per cent . the reconstructions considered are those associated with the simple @xmath128 procedure ( top left panel ) , @xmath183 ( top right panel ) , @xmath131 ( bottom left panel ) , and @xmath132 ( bottom right panel).,title="fig:",width=151 ] magnitude of the gradient of reconstructed string signals re - multiplied by the primary beam , for a tension @xmath190 , in the noise condition @xmath56 and for a coverage at @xmath72 per cent . the reconstructions considered are those associated with the simple @xmath128 procedure ( top left panel ) , @xmath183 ( top right panel ) , @xmath131 ( bottom left panel ) , and @xmath132 ( bottom right panel).,title="fig:",width=151 ] [ b ] [ b ] [ b ] [ ] [ b ] [ b ] [ b ] + + + in all noise and coverage conditions , the simple @xmath128 procedure is already effective at enhancing the visibility of strings by eye thanks to the multiplication by @xmath129 in fourier space , but the resulting maps are dirty . the string tensions representing the eye visibility thresholds are listed in table [ tab : detectionthresholds ] . these thresholds are well below the experimental upper bound in the noise condition @xmath56 , while for the noise conditions @xmath55 and @xmath54 they are respectively of the same order and larger . this evolution is naturally due to the larger amount of noise at high spatial frequencies , where the signal gradient lives , in the presence of secondary anisotropies . the eye visibility thresholds for @xmath183 , @xmath131 , and @xmath132 are the same as for @xmath128 . however it clearly appears that these more evolved algorithms offer much better reconstructions . this can already be acknowledged by simple eye inspection , as illustrated in figure [ fig : pa - in - rec ] for the magnitude of the gradient of the reconstructed signals re - multiplied by the primary beam in the noise condition @xmath56 at @xmath72 per cent and for a string tension @xmath191 . this qualitative assessment is confirmed by our quantitative measures for the reconstruction performance in all noise and coverage conditions and for all string tensions . figure [ fig : stat ] represents the @xmath176 , correlation coefficient , and kurtosis of the magnitude of the gradient of reconstructed string signals re - multiplied by the primary beam as functions of the string tension in all noise conditions and for a coverage at @xmath72 per cent , for @xmath183 , @xmath131 , and @xmath132 . the corresponding curves for @xmath128 are actually not shown for readability reasons as they are very far from optimal values . in this context , the major interest of the analysis actually resides in the comparison of the standard @xmath183 algorithm with the @xmath131 and @xmath132 algorithms . firstly , the reconstructions illustrated in figure [ fig : pa - in - rec ] seem very similar for @xmath183 , @xmath131 , and @xmath132 . note however that the magnitude of the gradient of the original signal ( not shown on the figure ) re - multiplied by the primary beam lies roughly in the range [ @xmath192 . the range of amplitudes reported on the maps therefore indicate that the dynamic range of the reconstructions is better for @xmath131 than for @xmath183 , and even better for @xmath132 . this is a generic behaviour that can be observed in all noise and coverage conditions and at all string tensions . secondly , as shown in figure [ fig : stat ] in all noise conditions and for a coverage at @xmath72 per cent , the eye visibility thresholds identify with the string tension where the @xmath176 and correlation coefficient get down to very small values , and the kurtosis exhibits a large variability over the @xmath69 simulations around the expected value . this allows us to identify rigorously the eye visibility threshold as the string tension below which the algorithms are not effective anymore at recovering the string network . thirdly , as also shown in figure [ fig : stat ] in all noise conditions and for a coverage at @xmath72 per cent , the @xmath176 and correlation coefficient are always significantly larger for @xmath131 than for @xmath183 . they are always at least as good for @xmath132 as for @xmath131 and often significantly larger . however , it appears that the kurtosis is less optimally recovered with @xmath131 than with @xmath183 . but the @xmath132 approach regularizes the kurtosis significantly better than @xmath183 in most cases . note that these last two conclusions also hold for a @xmath107 per cent coverage , hence suggesting their stability and robustness relative to the coverage . the reconstruction performance is always significantly better at @xmath107 per cent than at @xmath72 per cent , but the improvement is not drastic . this suggests that a small number of visibilities can already lead to good reconstructions of the string signal . in the perspective of the evaluation of the performance of the reconstruction algorithms , it is natural to ask whether the requirement for a complete reconstruction of the signal strongly limits the lowest string tension accessible in comparison with the lowest detectable string tension on the basis of the psm . in this regard , for any possible string tension @xmath20 , the probability distribution for the estimate @xmath169 on the basis of the psm may be obtained from simulations . one can identify the critical value @xmath193 such that , for null string tension , the probability that the estimate @xmath169 is larger than @xmath193 is small : @xmath194 , for some suitable positive value @xmath195 much smaller than unity . the test for the hypothesis of null string tension is then defined as follows . for estimated values @xmath196 , the hypothesis of null string tension may be rejected with a significance level @xmath195 . on the contrary for estimated values @xmath197 , the hypothesis of null string tension may not be rejected . we define the psm detectability threshold @xmath198 such that , for the string tension @xmath198 , the probability that the estimate @xmath169 is larger than @xmath193 is large : @xmath199 , for some suitable positive value @xmath200 much smaller than unity . consequently , for string tensions larger than @xmath198 , the probability of rejecting a null string tension on the basis of the hypothesis test defined is larger than @xmath201 , called the power of the test . the value @xmath198 is the smallest string tension that can be discriminated from the hypothesis of null string tension for given values of significance and power of the test . it may thus be understood as a detectability threshold determined on the basis of the psm . the psm detectability thresholds in the various noise and coverage conditions considered are reported in table [ tab : detectionthresholds ] for @xmath202 . in the noise condition @xmath56 , detectability thresholds are well below the experimental upper bound , while for the noise conditions @xmath55 and @xmath54 they are respectively of the same order and larger . in all cases , the psm detectability threshold @xmath203 is only slightly below the eye visibility threshold associated with the reconstruction algorithms discussed , by a factor smaller than @xmath204 . this discrepancy between these two thresholds confirms that the detection problem alone can be solved with the psm only down to slightly lower string tensions than the more difficult reconstruction problem . as already emphasized in @xcite , it is one thing to estimate a single global parameter such as the string tension on the basis of a psm , but quite another to explicitly reconstruct the string network itself . these comparison results are thus already very positive . our results strongly support the idea of the superiority of the compressed sensing approaches for a string signal reconstruction from radio - interferometric data , in particular when specific statistical prior information is introduced in the global minimization problem defined for reconstruction . however , at the high angular resolution considered and in the noise condition @xmath54 , the eye visibility thresholds , as well as the psm detectability thresholds , are above the current best experimental upper bound on the string tension . one first approach to circumvent this difficulty in the perspective of the reconstruction of a possible string network from forthcoming radio - interferometric observations is to consider an observation frequency around @xmath52 @xmath53 so as to eliminate the thermal @xmath42 effect and to fall in the noise condition @xmath55 . but even in this case the eye visibility thresholds remain slightly higher than the experimental upper bound . in this context , recent work by the authors might help to drastically enhance the signal reconstruction performance in the context of compressed sensing approaches @xcite . radio interferometers with small field of view and baselines with non - negligible and constant component in the pointing direction should be considered . in this context , the visibilities measured essentially identify with a noisy and incomplete fourier coverage of the product of the planar signal multiplied by the primary beam with a linear chirp modulation . theoretical and numerical results show the universality of the corresponding spread spectrum phenomenon relative to the sparsity basis , in terms of the achievable quality of reconstruction through global minimization problems such as @xmath131 and @xmath132 . preliminary numerical results regarding the application of such a framework to the reconstruction of a string signal in the @xmath0 look very promising . in another line of thoughts , global component separation techniques might be envisaged in order to simultaneously extract all non - gaussian components of the @xmath0 temperature data , including the @xmath42 effects and the string signal . finally , let us recall that the string signal simulations used for modelling the prior statistical distribution of the signal gradient as @xmath10 s and for computing its power spectrum are realistic . nevertheless , further analysis should be performed in order to study the stability of the @xmath132 reconstruction quality relative to a discrepancy between the statistical distribution of the true signal and our model . analyzing the difference between the @xmath131 and @xmath132 reconstruction qualities already provides useful information in that respect , as the @xmath131 minimization problem captures the essence of the compressiblity of the signal gradient , but with a non - exact @xmath10 shape parameter @xmath149 . in this context , one can argue in favour of a relative stability of the reconstruction quality in terms of the @xmath176 and correlation coefficient as these measures are indeed smaller for @xmath131 than for @xmath132 , but still larger than for @xmath183 . this stability is lost in terms of the kurtosis though . the @xmath132 algorithm proposed for the reconstruction of a string signal from arcminute resolution radio - interferometric data of the @xmath0 is designed in the framework of compressed sensing . it notably relies on a model of the prior statistical distribution of the signal , fitted on the basis of realistic simulations . the algorithm shows superior performance relative to the standard @xmath183 algorithm . secondary @xmath0 anisotropies strongly hamper the signal reconstruction though , and further work is still required in order to obtain good performance at string tensions below the current experimental upper bound . the authors wish to thank m. j. fadili for private communication of results on optimization by proximal methods . the authors also thank a. a. fraisse , c. ringeval , d. n. spergel , and f. r. bouchet for kindly providing simulations of a string signal . the authors also thank the reviewer for his valuable comments . y. w. is postdoctoral researcher of the belgian national science foundation ( f.r.s .- fnrs ) . jones m. e. , 2002 , in chen l .- w . , ma c .- p . , ng k .- w . , pen u .- , eds , asp conf . ser . 257 , amiba 2001 : high - z clusters , missing baryons , and @xmath0 polarization . , san francisco , p. 35
cygnus a ( @xmath2 ) is the best studied example and prototype of a powerful fr ii radio galaxy and is also one of the brightest nearby x - ray sources . imaging from _ einstein _ showed that cygnus a resides within @xmath3 of hot gas nearly @xmath4 mpc in extent @xcite . cygnus a itself has a very high x - ray luminosity , which classically , has been associated with a strong cooling - flow with a mass inflow rate possibly as high as @xmath5 yr@xmath6 @xcite . more recent observations of cooling - core clusters with _ xmm _ and _ chandra _ , however , have motivated both alternative explanations for the observed x - ray excess and imply substantially less cool gas than previously thought for these systems @xcite . while cygnus a itself has been studied extensively at all wavelengths , very little has been known of the optical properties of the galaxy cluster or group within which cygnus a is found on account of its low galactic latitude ( 5@xmath7 ) and attendant confusing star field . in paper 1 @xcite we presented 41 new velocity measurements within a 22 square region surrounding cygnus a. prior to that work only 4 concordant redshifts were known @xcite . the results indicated that cygnus a was located , but not centered , in a rich ( at least abell richness class 1 ) , high velocity dispersion cluster . while clearly showing substructure suggestive of a cluster merger , the number of velocities was insufficient to develop a dynamical model for the system . using _ asca _ , @xcite mapped the gas temperature across the cluster and showed the existence of a hot region between cygnus a and the secondary x - ray peak to the northwest ( in the direction of the dynamical centroid ) . the temperature structure could be explained by a fairly simple model involving the merger of two similar mass subclusters colliding head - on and developing a shock between them . in this paper , we analyze new imaging and spectroscopy of the cygnus a cluster . we compare the results of a detailed substructure analysis to the cluster merger model based on the x - ray gas temperature structure . we discuss the new observations in @xmath82 . in @xmath83.1 we look for evidence of bimodality in the spatial and velocity distribution of the cluster . a full 1-d , 2-d , and 3-d substructure analysis is presented in @xmath83.2 . in @xmath83.3 we calculate a new estimate for the cluster richness and mass , followed by a discussion of results and comparison to models in @xmath84 . conclusions are summarized in @xmath85 . spectroscopic observations were queue - scheduled on the wiyn 3.5 m telescope for semesters 1997b and 1998a with the hydra - bench spectrograph . we used the 400 line grating covering the spectral range 3880 - 7060 with a dispersion of 1.55 pixel@xmath6 and the blue simmons camera . in 1997b three fiber configurations were observed through clouds for 3x20 minute exposures on the nights of 8 and 30 september and 1 october 1997 . in 1998a five fiber configurations were observed in clear conditions on the nights of 29 may and 28 - 29 june 1998 . for each fiber configuration we typically assigned 7 - 8 fibers to alignment stars and 20 - 25 sky fibers , leaving 60 - 70 fibers available for galaxy candidates . in the first 1997b wiyn queue run we targeted 166 different objects . we used our original kpno 0.9 m 22 field centered on cygnus a from @xcite ( paper 1 ) supplemented by images from the digitized sky survey to take advantage of the larger field of view of hydra ( 1 degree ) . while these observations were strongly affected by clouds and mostly unsuccessful , we were able to cull 10 new galaxy velocities and confirmed another 4 of the lower s / n spectra from paper 1 . approximately 70% of the useful spectra turned out to be stars ; the poor resolution of the digitized sky survey made galaxy / star separation very difficult in this crowded field . even with good quality imaging , there is substantial confusion between blended , faint point - sources and the appearance of an extended object , making this one of the most challenging fields for galaxy selection . before the subsequent queue run in 1998a , we obtained new @xmath9-band images at the kpno 0.9 m telescope on 29 - 30 april 1998 using the 2048x2048 t2ka ccd at a scale of 068 pixel@xmath6 . to complement the @xmath10 image centered on cygnus a used in paper 1 , we imaged at two new positions ( 19:58:14 + 40:55:43 and 20:00:06 + 41:04:42 ) providing full spatial coverage over the extent of the x - ray emission . the two new images consisted of 4x10 min dithered exposures with a final image quality of 14 - 17 . candidate galaxies were identified using focas and inspected individually by eye . coupled with the focas results , we used the peak - flux as an additional cut on eliminating stars . priorities for spectroscopic observation were assigned primarily based on the optical magnitude . we targeted 254 galaxy candidates including many of the measured galaxies from paper 1 which were of low to medium s / n . having observed the brighter targets in 1997b ( suffering more stellar contamination ) and coupled with better weather , the 1998a observations were much more successful in finding new galaxy redshifts . from the 1998a data we measured 88 new velocities and re - confirmed at better signal - to - noise 15 of the more questionable objects from paper 1 . velocities were measured using the iraf cross - correlation task fxcor with the exception of a few emission - line objects . velocity errors were calculated from the tonry & davis r - value @xcite as determined by fxcor . for emission lines , we took the mean observed velocity over the lines and applied a heliocentric correction . we assume an error of 30 km sec@xmath6 on emission - line velocities . we imaged the cygnus a field again on 19 july 1999 in @xmath9-band at the kpno 0.9 m telescope , this time using the 8k x 8k mosaic camera giving a full one degree field of view with a scale of 0425 pixel@xmath6 . details of the data reduction , photometry , and astrometry are given in @xcite . the final image is a sum of 5@xmath115 min exposures under excellent conditions . the final image quality was 085 . we used this image in all subsequent analysis . we measured aperture magnitudes for all new cluster galaxies using a gunn - oke @xcite metric aperture of diameter 26.2 @xmath12 kpc . magnitudes for individual galaxies were calculated using a range of different procedures for handling of foreground stars , which indicated that the subtraction of such objects in the high stellar density field of cygnus a was the primary source of error in reported magnitudes . these tests indicated a global error of about 0.2 magnitudes for the gunn - oke aperture photometry , compared to an error of 0.05 magnitudes in the photometric solution @xcite . absolute magnitudes were calculated assuming a galactic absorption of @xmath13 = 0.81 @xcite and a k - correction of 0.08 magnitudes , using @xmath14 km sec@xmath6 mpc@xmath6 and @xmath15 . for non - cluster galaxies ( spectroscopically confirmed ) and newly identified galaxies we used a fixed aperture of 5 diameter . this smaller aperture was chosen in order to reduce the effects of foreground stars . an identical analysis to that performed for the gunn - oke photometry indicated a typical error of 0.1 magnitudes for galaxies brighter than @xmath16 . all magnitudes included in tables and figures have had foreground stars removed from their apertures . in table 1 we list all new velocities measured from wiyn . starting from all 139 velocities ( including those from paper 1 ) , we calculated the biweight estimators of location and scale @xcite . we then trimmed the list according to the first - pass @xmath17 rejection about the central velocity and re - computed the biweight estimators on the edited list . this 2nd iteration produced a list of 118 cluster members within a @xmath17 dispersion of the central biweight velocity @xmath18 km sec@xmath6 ( @xmath19 ) and biweight scale @xmath20 km sec@xmath6 , corrected to the rest frame of the cluster . these values compare quite well to those determined in @xcite ( @xmath21 km sec@xmath6 ) , while we have increased the number of confirmed cluster members by more than a factor of 2 . cygnus a ( @xmath22 km sec@xmath6 ) is therefore offset 2197 km sec@xmath6 from the mean cluster velocity . for cluster members , table 1 includes the right ascension and declination ( j2000.0 ) , the heliocentric velocity and error estimate , and gunn - oke aperture magnitudes . we re - list velocities for 19 galaxies re - observed from paper 1 for which we have adopted the new velocity measured with wiyn ( higher s / n ) . in all cases but one , the agreement between velocities was within a few hundred km sec@xmath6 ( typically 1 - 2@xmath23 ) . we also list velocities for 21 background / foreground galaxies and apparent magnitudes within a 5 aperture . we plot a velocity histogram for the 118 cluster members in figure [ fig1 ] . while qualitatively similar to that in paper 1 , the bimodality suggested by the previous work initially appears much less significant . there is , however , clearly a heavier tail in the distribution in the vicinity of cygnus a. we calculate the likelihood of a bimodal distribution using the kmm algorithm @xcite , which compares the goodness of fit between unimodal or multimodal gaussian functions and objectively partitions data into sub - populations accordingly . interestingly , we find that a single gaussian model is rejected at a significance of @xmath2499% . the estimated velocity locations of each group are 16648 ( group 1 ; just 163 km sec@xmath6 offset from cygnus a ) and 19428 km sec@xmath6 ( group 2 ) . the partitioning assigned 23 members to group 1 and 95 to group 2 . while this result is quite significant , there was some trouble in making the subgroup assignments with the estimates of correct allocation rates for each group being 76% and 96% , respectively . these results are for the homoscedastic case assuming a common covariance ( velocity dispersion squared ) where the estimated velocity dispersion for both groups was 1126 km sec@xmath6 . the results for the heteroscedastic hypothesis ( different velocity dispersions ) were quite different , suggesting both different mean velocities ( 18131 and 19849 km sec@xmath6 ) and mixing ratios ( 58 and 60 members ) for the groups . while still technically significant at the 99% level ( note that the probability distribution is not well determined in this case so the significance is not robust ) , the estimates for the correct overall allocation rate was 78% ( 73% and 84% for groups 1,2 respectively ) . the estimated dispersions were also quite different ; 1651 and 755 km sec@xmath6 for groups 1 and 2 , respectively . we also used a variant of the traditional kmm code which fits structures based on positions as well as velocities , although it assumes that the covariance among the three attributes is zero ( there is no correlation between position and velocity ; * ? ? ? * ) . as is the case for the standard kmm , the significance of the multicomponent fits are not easily derived . however , a monte carlo shuffling of the velocities is used to determine the frequency at which a randomized dataset fits as well as the observed one . the results , shown in figure [ fig2 ] , were quite similar to those of the heteroscedastic case given above ; two groups with mean velocities of 17648 ( n=61 ) and 20126 km sec@xmath6 ( n=57 ) and dispersions of 1134 and 600 km sec@xmath6 respectively . note that the velocity distribution of the higher velocity group appears much more gaussian - like than the more dispersed , lower - velocity group which includes cygnus a. in fact , statistical tests performed on this lower - velocity system indicate that its velocity distribution is non - gaussian . in summary , the kmm tests indicate that the cygnus a cluster is poorly fit by a single system . it may be represented by a pair of systems , with one being consistent with a relaxed cluster having a gaussian velocity distribution and the other ( which includes cygnus a ) more difficult to describe . the velocity dispersions of these two systems imply a virial mass ratio of over 3:1 . however , based on numbers of assigned galaxies and the observation that the grouping containing cygnus a is not well described as a virialized system , these two systems may be of comparable richness and hence mass . a model which fits these characteristics will be developed in the ensuing sections . we show the velocity - coded spatial distribution in figure [ fig3 ] . square symbols are the galaxies with velocity less than the cluster mean while triangles represent those at greater velocity . shading is used to indicate the magnitude of the difference between a galaxy s velocity and the cluster mean , with solid symbols for galaxies within @xmath25 and open symbols for galaxies with a larger velocity differences . for example , cygnus a has a velocity @xmath26 below the mean , and hence is coded as an open - square . note the significant grouping around cygnus a , below and slightly to the left of center , which includes seven satellite galaxies with a small dispersion in velocity . interestingly , the positions of the galaxies in this lower velocity bin ( the open squares in figure [ fig3 ] ) show a linear alignment along the principle axis of the cluster . we ran a suite of 1d , 2d , and 3d statistical tests in order to assess and quantify the amount and nature of substructure present in the cygnus a cluster . these tests are defined and explained in detail in ( * ? ? ? * hence referred to as p96 ) . briefly , the 1d tests are performed solely on the galaxy velocities , the 2d tests on galaxy positions , and the 3d tests on both velocities and positions . nearly all of the 1d tests produced a significant result , with the main finding being that the velocity distribution is skewed . averaging over the 3 skewness tests , the distribution is significantly skewed at better than the 98% level . the dip - stat @xcite tests for consistency with a unimodal population , and returned a 99% significance to deviations from a single gaussian distribution , consistent with the kmm results discussed above . the 2d substructure tests are useful for quantifying clumpiness , asymmetries , and elongations in the spatial distribution of galaxies in the cluster . in figure [ fig4 ] we show adaptively smoothed contours of the galaxy distribution overlaid onto the x - ray emission from the _ rosat _ pspc . while similar qualitatively to its counterpart in paper 1 , a much more obvious clump of galaxies is now seen near but not centered ( offset=72 @xmath12 kpc ) at the position of cygnus a. the two peaks in the spatial distribution are located at 19:59:30.6 + 40:45:02 and 19:59:03.1 + 40:49:36 with a separation of 458@xmath12 kpc . similarly , the galaxy distribution traces well the x - ray gas , although this is partly due to selection effects as our imaging data were centered at locations optimized to span the full extent of the intracluster medium ( we did target candidates outside this area based on the poss ) . note that the 2nd peak in the x - ray emission ( difficult to see with this stretch and in contrast to cygnus a ) , is located near 19:58:50.4 + 40:52:14 at a projected separation of @xmath27700@xmath12 kpc . from the 2d substructure tests , all but the ast test ( sensitive to non - central clumping ) returned significant results ( all @xmath28 ) , indicating that the spatial distribution is significantly elongated ( fourier elongation test ) , asymmetric ( @xmath29-test ) , and bimodal ( lee2d ) . the 3d substructure tests have the null hypothesis of constant mean velocity and dispersion with position . consistent with the 1d , 2d , and kmm results , we find evidence for significant substructure ( lee 3d , 99.4% ; @xmath30 , 99% ) . the remaining 3d tests were less significant than @xmath30 and lee-3d ( @xmath31 , 94% ; @xmath32 , 89% ; @xmath33 , 93% ) . we show in figure [ fig5 ] the standard @xmath30 bubble plot ; the symbol size is proportional to the actual @xmath34 for each galaxy as calculated by the dressler - shectman test , @xcite . from comparison to figures [ fig3 ] and [ fig4 ] , we confirm that the clumping responsible for the positive signal ( local velocity and/or dispersion are different from global values ) is located exactly on the galaxy clump associated with cygnus - a . we calculated global mass estimates using a projected mass estimator ( @xmath35 ) and the virial theorem ( @xmath36 ) , with values of @xmath37 and @xmath38 . the larger @xmath35 is consistent with the findings of p96 when there is a significant projected separation between two merging clusters at an epoch prior to core crossing . p96 found in general that these estimators over - predict the real cluster masses , which are additionally affected by the presence of merging subclusters . in figure [ fig6 ] we plot a histogram of the absolute magnitude distribution of cluster galaxies . the magnitudes ( @xmath39 ) were measured within a gunn - oke aperture of diameter 26.2@xmath12 kpc , on which system @xmath40 . from inspection of figure [ fig6 ] , one can see our incompleteness arises around @xmath41 due to sensitivity limits and our ability to pick out these fainter galaxies for target selection . in order to get a better estimate for the richness , we have taken advantage of the sub - arcsecond image quality on our @xmath9-band mosaic frame to locate galaxies not yet identified spectroscopically . we were fairly conservative in classifying objects as galaxies . over the entire field , we identified 362 additional galaxies which we list in table 2 . in figure [ fig7 ] we show the spatial distribution of these newly identified galaxies with adaptively smoothed contours as in figure [ fig4 ] but on a larger scale covering the full abell radius mapped by our mosaic image . note that the primary cluster concentration from figure [ fig4 ] coincides with the peak offset slightly to the east of center in the full - field plot of figure [ fig7 ] . there is also correspondence with the two lowest surface - density peaks in figure [ fig4 ] with the other two concentrations in figure [ fig7 ] . we indeed see that the missed galaxy distribution traces the same clustering pattern as the cluster members . there is an additional chain of less - dense concentrations from the southeast to southwest not sampled by our spectroscopy . these results suggest that overall , our spectroscopic sampling is quite good . we followed the general prescription of @xcite to estimate the cluster richness . the mosaic frame nearly exactly matches one abell radius ( @xmath42 , or 268 for @xmath19 ) . including the galaxies from table 1 , we found 394 galaxies within an abell radius of cygnus a ( 19:59:28.3 + 40:44:02 ) . of these , 133 lie within the range @xmath43 to @xmath44 defined by abell as the measure of richness . in the original paper , abell estimated contamination due to foreground and background galaxies using nearby field counts . lacking a direct analog , we simply estimated field galaxies using @xmath45 where @xmath46 galaxies steradian@xmath6 ( see * ? ? ? * ) . the corrected count is then 106 , placing the cygnus a cluster in the middle to high range of richness class ( rc ) 2 clusters . we note that this estimate for contamination by foreground / background objects is consistent with our spectroscopy results , where 118 of 139 galaxies were shown to be cluster members . an even higher richness is quite possible , given our fairly conservative cut on galaxy identification and the likelihood that additional galaxies are hidden by the numerous galactic stars . lastly , if the cygnus a system is an ongoing merger of two clusters of similar mass and richness , this would imply that they are each rich systems of rc=1 . the picture gleaned from the substructure analysis suggests that cygnus a is the dominant galaxy in a cluster which is currently merging with another subcluster . the fact that we get a consistent picture from each of the kmm , 1d , 2d , and 3d tests allows us to constrain somewhat the expected merger properties . depending on which variant of the kmm partitioning scheme is used we find that the velocity separation of the two subclusters is between 1600 and 2600 km sec@xmath6 ( corrected to the rest - frame of the cluster ) . we find different mixing ratios for the two cases ( homoscedastic or heteroscedastic assumption ) ; with equal velocity dispersions the ratio of galaxy members is of order 4 ( cygnus - a being the poorer group ) whereas they are nearly equal when the dispersions are fit independently . the projected separation of the subclusters is @xmath27700@xmath12 kpc from the x - rays or @xmath27460 @xmath12 kpc based on the galaxy surface density peak . we compare the properties of the two subclusters using figures [ fig2]- [ fig4 ] . the kmm fits suggest that the cygnus a subcluster is fairly dispersed in velocity whereas the higher velocity peak is more gaussian . as seen in projection , spatially the two subclusters appear to be well - mixed , although the immediate clumping centered on cygnus - a is quite significant with a small velocity - spread , well offset from the global mean . and as previously noted , the kmm group which includes cygnus a also shows an interesting linear alignment between the two subclusters , possibly coincident with the axis of the merger . now , we utilize the n - body results and statistics from the substructure tests in p96 to infer a best model for the merger state of this system . p96 note that the 1d , 2d , and 3d tests each have their own strengths and weaknesses so a collective interpretation of their results can reveal an estimate of the merger geometry . the 1d substructure tests are obviously most sensitive when viewed along the merger axis ( in our discussion , this is called a projection angle of 0@xmath7 ) , and fail to produce significant results for mergers viewed at projection angles greater than @xmath47 . the 2d tests require a larger projection angle to separate the components , although the size of this angle required to produce significant substructure tests is dependent on the epoch of the merger . projection angles smaller than @xmath48 only produce significant results well before or after core passage . the 3d tests are most sensitive to projection angles @xmath49 , similar to the 1d tests . thus , producing significant substructure results for all three sets of tests requires an intermediate projection angle of @xmath2730 - 45@xmath0 . furthermore , the simulations indicate that larger mass ratios decrease the significance of many tests , further implying a mass ratio of less than 3:1 for the two subclusters . specifically , the degree of significance of the 2d and 3d substructure tests we observe is only present in the simulations at times @xmath270.2 - 0.6 gyr prior to core passage , or else significantly after core passage ( @xmath242 gyr ) . the high velocity dispersion and 3d tests strongly favor the pre core - passage epoch . the 1d results also support a time quite near to core crossing , but do not discriminate between the pre or post - crossing event . the subcluster morphologies are interesting in this context ; the cygnus a subcluster appears dispersed ( figure [ fig2 ] ) as if the core of the subcluster were lagging behind its member galaxies in the merger . also note the compact core around cygnus a , which is responsible for the significant @xmath30-test results ( figure [ fig5 ] ) . post core - passage , it would seem unlikely that such a core would remain so distinct . the other subcluster appears more gaussian in velocity space , is quite well mixed spatially with the cygnus a subcluster , and lacks a well - defined core . it is not clear which of the subclusters is the _ primary _ in the merger , although the data do not suggest a large mass ratio so the distinction may be irrelevant . more details about the likely merger in this system can be found from properties of the x - ray gas . @xcite(msv99 ) studied x - ray gas temperature maps from _ asca _ for several nearby merging galaxy clusters , including cygnus a. msv99 mapped the x - ray temperature across the cluster from an annulus including the cygnus a group ( cygnus a was excised from the data before fitting the temperature ) in the direction of the 2nd spatial peak ( see figure [ fig4 ] ) . the results indicate a region between the two clusters with t@xmath50 kev whereas both subclusters are similar at @xmath51 kev . the hot region is consistent with a shocked region resulting from a head - on collision of similar mass clusters . under this fairly simple geometry and under the approximation of a 1-dimensional shock , msv99 estimate a subcluster collision velocity of @xmath52 km sec@xmath6 . this value is close to the expected free - fall velocity that two similar mass , @xmath53 kev , clusters would achieve by the time that they reached the observed separation . while this simple 1d shock model ignores the physics of the gas in the interaction , msv99 confirm that the velocity estimate is accurate to about 20% as compared to when hydrodynamical effects are included in the model . quite interestingly , this predicted velocity matches extremely well the velocity difference of the two subclusters inferred from the optical data ( 1600 - 2600 km sec@xmath6 ) . more recently , @xcite studied the icm surrounding cygnus a with _ chandra_. while the more diffuse x - ray emission extending to the northwest and the 2nd cluster is partially detected on two of the other chips , the sensitivity is insufficient to characterize its properties . for the southwestern half of the x - ray emission in figure [ fig2 ] , however , smith et al . confirm the _ asca _ temperature results and compute both the gas mass and total integrated cluster mass . within a 500 kpc radius they derive a gas mass of @xmath54 and total mass between @xmath55 ( depending on the central temperature profile ) . we estimate a factor of 4 - 5 higher total mass , based on the projected mass from the optical data . however , it is hard to directly compare the two estimates since smith et al assumed a lower hubble constant ( @xmath56 km s@xmath6mpc@xmath6 ) , and the _ chandra _ measurement includes about 1/3 of the full area of the galaxy distribution and extended x - ray emission . @xcite also show a complex temperature distribution for the cluster consistent with a unsettled dynamical situation in the hot gas . we looked for counterparts to the x - ray point sources detected in the _ chandra _ image of @xcite . with the exception of cygnus a itself , we find no identifications with our extended list of galaxies in tables 1 and 2 , which confirms their conclusion that these sources are most likely stars or background objects ( agns ) . one other observation to add to this picture is the presence of the cooling - core , a core with a cooling time much less than a hubble time , in cygnus a. while the interpretation of so - called cooling - flow clusters appears to be undergoing revision @xcite , some aspects of the more classical picture may survive . cygnus a is a central - dominant galaxy with a central x - ray peak , a negative or flat temperature gradient and positive metallicity gradient toward the core , and presumably sits at the bottom of the local gravitational potential well @xcite . thus one would expect significant cooling to occur in its core . however , the x - ray spectrum of the core shows a strong power - law continuum to very high energies as expected for a powerful agn / quasar @xcite . according to current models , it is expected that the powerful radio source may have a significant influence on the cooling properties of the central gas , with the result that the mass deposition rates inferred by the central x - ray excess alone may be over - estimations by more than an order of magnitude @xcite . current observations of cooling flows show that high metal abundances are common and thus require large enrichment times to create them @xcite . the implication is that cooling cores probably form very early in clusters , are very persistent phenomenon , and likely survive cluster mergers as well as the energy input over the lifetime of the central agn s . so most mergers must not destroy cooling cores . consistent with this hypothesis are the observed high frequency of cluster substructure and the frequency of cooling cores seen in large samples of clusters ( e.g. * ? ? ? disturbances may occur as a result of a merger , but may only take the form of a _ sloshing _ of the central gas . @xcite observed the signs of such an event in the perseus cluster . these authors also argue that the cold , dense , low - entropy core expected to be found in perseus ( and similarly in cygnus - a ) is probably well protected against the penetration of gas from infalling groups or clusters in a merger event . in an alternative hypothesis , @xcite suggest that cores of cool gas first form in subclusters which then merge to create rich clusters with cool , central x - ray excesses . in this model , cool cores actually form in hierarchical formation along with the clusters themselves . this model is attractive from the point of view of providing sufficient time to produce the observed central metallicities and is consistent with the presence of substantial substructure in these clusters . following the more classical picture of cluster cooling flows , @xcite studied the consequences of head - on 4:1 and 16:1 mass - ratio mergers on existing cooling - flows . their results indicated that the key parameters to whether the cooling - flow survives such an event were the gas dynamics in the interaction of secondary and primary cluster gas . those mergers with higher secondary gas densities were found to be the most disruptive . however , even in the cases of strong disruption of the cooling flow , their simulations indicate that the initial increase in the depth of the gravitational potential at core - crossing is insufficient to disrupt the flow . the demise of the cooling flow in these simulations occurred after a time - delay of 1 - 2 gyr following core - crossing . applied to the cygnus a cluster , the properties inferred from both optical and x - ray do not suggest a dense cluster core for the higher velocity subcluster . so even for a high mass density in the secondary core ( which is not supported by the data ) , the effects on the cooling flow are expected to be minor . thus the presence of the cooling core in cygnus a does not argue strongly either way for a pre or post core - crossing epoch . in any case , one would not expect significant disruption , even with the classical cooling flow model . the existence of the most powerful radio galaxy in the local universe in the midst of a massive , complex , cluster - cluster merger close to core passage , raises the question of whether there is a connection between the two phenomena . radio sources are believed to be transients . a soft lower limit to the age of @xmath57 can be set from synchrotron spectral aging analysis @xcite . a more complete physical argument yields an age of @xmath58 years @xcite . thus the current event we are observing seems to have an age shorter than the merger timescale . however , the dynamics and the morphology of the optical cluster we report in this paper suggest a complex interaction is underway . the x - ray total intensity and temperature imaging add to the already complicated picture . as can be seen in figure [ fig2 ] , cygnus a is not at the center of the velocity field of the subcluster to which it is assigned membership . furthermore , as can be seen in figure [ fig3 ] , cygnus a is not close to spatial center of its subcluster either . however , from figure [ fig4 ] , one can see that it is at the peak of the galaxy counts as well as the the x - ray peak . one does not see the symmetry one might expect based on what we see in other clusters with dense cores . thus while the timescales of the merger event do not clearly match the radio event , the large scale environment around cygnus a does seem unusually disturbed . this situation allows for a variety of physical phenomena which could disturb the cygnus a galaxy and produce the radio event . given the large velocities for the individual galaxies and the dense environment around cygnus a , a major merger of galaxies seems unlikely ( e.g. * ? ? ? . however , a galaxy in a stable orbit around the cygnus a might be perturbed by a close passage and fall into the central galaxy . recently , from adaptive optics imaging , @xcite have found a secondary point source located just 04 ( 400 pc in projection ) southwest of the radio core . the spectral energy distribution of the object is most consistent with the core of a gas - stripped galaxy , likely merging with the giant elliptical host . in comparison to models , these authors speculate that the core of such a galaxy ( 100 - 1000 times less massive than the host ) might survive a few pericenter passes or a radial encounter , the time - frame for such events being similar ( a few @xmath59 years ) to the estimated age of the radio source . also the cluster - cluster merger will produce a time- variable tidal field on galaxies or dense clouds near cygnus a @xcite . this could well disrupt a stable environment near cygnus a and cause material to fall into the core . infalling hi gas is seen against the nucleus @xcite and recently near - ir spectroscopy has revealed a giant molecular cloud also falling into the core of cygnus a. @xcite . thus indirectly the complex cluster environment could be responsible for the extreme agn we see in cygnus a. the picture which emerges from both the optical and x - rays is a cluster - cluster merger seen in projection at 30 - 45@xmath0 fairly near to core - crossing , the 2d and 3d substructure tests argue for a pre - crossing time - frame , with the level and type of substructure consistent with epochs between 0.2 - 0.6 gyr prior to core passage . the x - ray results from the literature , as discussed in section 4.3 , indicate the presence of a bow - shock located between the two subclusters with a temperature enhancement a factor of two higher than the subcluster icms . a shock with these conditions is expected to form by gas moving at supersonic speeds as two @xmath60 equal mass clusters come together and reach a separation of @xmath271 mpc . such a shock forms when the leading edge of the infalling cluster is still several hundreds of kpc from the primary cluster core . as the cores come closer together in time , the shock structure would actually become more difficult to discern because of the steep density profile near the core ( an issue of contrast ) . additionally , the shock will decelerate as the cores merge because of the increase in the ambient gas density . hence the x - ray results would also seem to argue for a pre core - crossing epoch . cygnus a resides near but offset from the center of a rc @xmath612 cluster which appears to be merging with another similar richness cluster . the implied mass ratio of the two subclusters is of order 2:1 at most , with the cygnus - a clump being somewhat more massive . we have spectroscopically identified 74 new members of the cygnus a cluster , bringing the total to 118 galaxies consistent with cluster membership based on 3@xmath23 clipping of the velocity distribution . the cluster has a biweight mean velocity @xmath62 km sec@xmath6 and biweight scale @xmath20 km sec@xmath6 ( corrected to the cluster rest frame ) . results from the optical 1d , 2d , and 3d substructure analysis indicate significant bimodality consistent with the merger of two subclusters . the level of substructure seen and the specific tests which detect it are consistent with a pre core - passage epoch , 200 - 600 myr from core - crossing . such a model is independently supported from the x - ray temperature variations and the presence of shock heated gas between the subclusters . the large velocity dispersion of the kmm substructure associated with cygnus a , the linear spatial distribution of galaxies along the likely merger axis , and the existence of shocked gas between the two cores suggests that we are viewing this system shortly before core passage . we thank dr . elizabeth rizza for her help with the 1998 imaging data . this research was supported by the gemini observatory , which is operated by the association of universities for research in astronomy , inc . , on behalf of the international gemini partnership of argentina , australia , brazil , canada , chile , the united kingdom and the united states of america . r r r r r 19:57:19.57 & 41:05:06.5 & 20986 & 51 & -20.6 + 19:57:20.93 & 40:58:29.6 & 17148 & 42 & -21.4 + 19:57:28.10 & 40:45:08.2 & 14330 & 88 & -19.5 + 19:57:38.06 & 40:53:18.8 & 19786 & 36 & -21.7 + 19:57:40.32 & 40:47:55.6 & 20146 & 132 & -19.2 + 19:57:41.86 & 41:04:57.8 & 14480 & 112 & -19.3 + 19:57:42.66 & 41:06:06.3 & 20806 & 49 & -21.3 + 19:57:46.87 & 40:46:13.1 & 18887 & 92 & -19.6 + 19:57:51.63 & 41:05:39.6 & 20746 & 59 & -20.9 + 19:57:58.25 & 40:52:54.4 & 20596 & 47 & -21.2 + 19:58:00.04 & 41:06:14.7 & 17598 & 64 & -19.3 + 19:58:06.83 & 40:46:44.6 & 20746 & 30 & -20.8 + 19:58:07.38 & 40:54:25.2 & 20746 & 88 & -20.2 + 19:58:08.75 & 40:47:32.7 & 19936 & 55 & -19.8 + 19:58:12.60 & 41:04:04.5 & 22245 & 80 & -18.9 + 19:58:14.08 & 40:56:36.6 & 18797 & 90 & -21.2 + 19:58:22.60 & 40:58:02.1 & 18197 & 80 & -20.8 + 19:58:25.88 & 40:51:00.6 & 17748 & 47 & -20.8 + 19:58:32.53 & 41:03:11.6 & 18257 & 44 & -21.7 + 19:58:38.55 & 40:47:33.4 & 19786 & 48 & -20.4 + 19:58:40.98 & 40:54:44.9 & 19427 & 65 & -20.5 + 19:58:43.44 & 40:57:47.1 & 14330 & 77 & -19.3 + 19:58:43.67 & 40:56:27.0 & 20236 & 86 & -19.5 + 19:58:44.36 & 40:53:27.5 & 20446 & 60 & -20.0 + 19:58:45.78 & 40:55:05.4 & 18557 & 82 & -20.5 + 19:58:50.38 & 40:49:54.8 & 19606 & 37 & -20.5 + 19:58:51.75 & 40:51:45.2 & 17658 & 30 & -18.6 + 19:58:52.75 & 41:06:22.0 & 20476 & 72 & -20.4 + 19:58:53.90 & 40:55:06.2 & 19007 & 70 & -20.9 + 19:58:55.20 & 40:52:29.9 & 20146 & 57 & -19.6 + 19:58:55.70 & 40:47:46.6 & 20056 & 54 & -20.8 + 19:58:58.78 & 41:01:18.3 & 19247 & 62 & -20.9 + 19:58:59.41 & 40:48:01.7 & 20476 & 66 & -20.0 + 19:59:04.63 & 40:48:55.9 & 20566 & 58 & -20.1 + 19:59:06.42 & 40:58:39.7 & 19996 & 92 & -19.6 + 19:59:07.57 & 40:47:12.1 & 19307 & 79 & -19.1 + 19:59:10.64 & 40:47:11.3 & 17148 & 73 & -19.0 + 19:59:11.72 & 40:42:26.3 & 19157 & 79 & -19.2 + 19:59:11.92 & 40:57:25.4 & 19457 & 84 & -21.2 + 19:59:13.34 & 40:45:32.4 & 17748 & 75 & -19.8 + 19:59:14.64 & 40:53:08.4 & 18018 & 96 & -20.1 + 19:59:20.00 & 40:58:32.3 & 19157 & 56 & -21.1 + 19:59:21.16 & 40:37:44.5 & 20446 & 83 & -19.1 + 19:59:21.57 & 40:52:25.4 & 21825 & 111 & -18.6 + 19:59:22.98 & 41:08:23.5 & 20626 & 69 & -21.3 + 19:59:24.07 & 40:43:20.6 & 16459 & 82 & -19.9 + 19:59:24.99 & 40:44:06.9 & 16908 & 62 & -20.0 + 19:59:25.64 & 40:45:08.3 & 18377 & 73 & -18.7 + 19:59:27.60 & 40:43:25.1 & 16608 & 64 & -20.4 + 19:59:28.68 & 40:43:53.0 & 16878 & 78 & -20.2 + 19:59:29.17 & 40:43:59.9 & 17388 & 134 & -19.6 + 19:59:30.18 & 40:51:11.2 & 18767 & 92 & -19.5 + 19:59:30.46 & 40:48:08.0 & 19786 & 72 & -19.8 + 19:59:30.64 & 40:45:07.9 & 21345 & 86 & -19.2 + 19:59:31.76 & 40:58:32.2 & 18947 & 56 & -19.6 + 19:59:33.58 & 40:44:03.9 & 16968 & 159 & -19.1 + 19:59:34.84 & 40:41:41.8 & 17928 & 116 & -19.6 + 19:59:36.04 & 41:07:13.3 & 18227 & 65 & -20.8 + 19:59:40.31 & 40:46:25.4 & 20206 & 61 & -19.9 + 19:59:43.14 & 40:38:43.0 & 17778 & 56 & -20.0 + 19:59:44.06 & 40:37:38.4 & 19277 & 30 & -18.7 + 19:59:50.48 & 40:38:14.8 & 18107 & 107 & -20.0 + 19:59:51.12 & 40:39:27.2 & 18677 & 133 & -19.5 + 20:00:04.51 & 40:44:02.6 & 19007 & 90 & -20.7 + 20:00:05.93 & 40:59:55.6 & 20056 & 69 & -20.6 + 20:00:08.44 & 40:59:37.2 & 17628 & 106 & -19.4 + 20:00:15.48 & 40:52:44.7 & 19666 & 156 & -19.0 + 20:00:18.54 & 40:35:13.5 & 17418 & 132 & -19.1 + 20:00:20.83 & 40:58:31.1 & 19936 & 65 & -21.0 + 20:00:21.14 & 40:57:00.9 & 18497 & 69 & -20.3 + 20:00:26.67 & 40:33:50.2 & 16638 & 150 & -18.6 + 20:00:42.78 & 41:01:24.2 & 18557 & 85 & -18.9 + 20:00:45.13 & 40:56:56.9 & 17688 & 30 & -19.9 + 20:00:58.56 & 41:02:56.4 & 18947 & 96 & -20.0 + 20:00:58.74 & 41:03:14.0 & 19487 & 63 & -20.6 + 20:01:03.35 & 40:56:45.8 & 20896 & 34 & -22.1 + 20:01:09.02 & 40:56:40.9 & 20836 & 68 & -21.3 + 19:58:33.11 & 40:42:47.5 & 17898 & 71 & -20.0 + 19:58:33.63 & 40:52:20.3 & 19457 & 48 & -20.6 + 19:58:40.20 & 40:50:01.4 & 16069 & 45 & -20.4 + 19:58:41.08 & 40:54:44.0 & 19337 & 65 & -20.5 + 19:58:45.43 & 40:45:38.2 & 19037 & 37 & -22.3 + 19:58:45.71 & 40:48:51.5 & 20086 & 49 & -21.1 + 19:58:56.74 & 40:48:52.2 & 16159 & 74 & -20.6 + 19:58:59.07 & 40:53:19.8 & 16818 & 87 & -20.4 + 19:59:00.92 & 40:51:08.0 & 19337 & 52 & -21.9 + 19:59:08.09 & 40:54:13.0 & 16668 & 51 & -20.7 + 19:59:13.30 & 40:50:05.7 & 19367 & 55 & -19.0 + 19:59:14.10 & 40:48:15.2 & 16219 & 53 & -19.3 + 19:59:16.19 & 40:50:07.3 & 15529 & 54 & -20.9 + 19:59:32.55 & 40:49:25.3 & 19636 & 59 & -20.2 + 19:59:33.83 & 40:52:17.0 & 19996 & 43 & -20.2 + 19:59:55.62 & 40:43:55.5 & 20266 & 98 & -20.3 + 20:00:02.10 & 40:45:57.8 & 19067 & 36 & -20.6 + 20:00:05.96 & 40:36:36.4 & 18437 & 41 & -21.9 + 20:00:07.67 & 40:46:02.9 & 17868 & 115 & -19.6 + 20:00:17.42 & 40:51:16.9 & 19187 & 50 & -20.6 + 19:57:21.93 & 41:03:38.0 & 24763 & 99 & 18.3 + 19:57:30.43 & 40:54:44.1 & 10373 & 72 & 20.4 + 19:57:38.38 & 41:01:01.0 & 41162 & 105 & 18.9 + 19:57:59.78 & 40:56:59.7 & 45209 & 80 & 20.4 + 19:57:59.94 & 40:49:01.0 & 38764 & 102 & 20.0 + 19:58:08.78 & 41:06:34.1 & 31748 & 90 & 18.2 + 19:58:09.35 & 40:49:36.0 & 31419 & 82 & 19.4 + 19:58:22.39 & 41:01:44.2 & 24613 & 77 & 18.8 + 19:58:22.85 & 40:58:54.1 & 5966 & 77 & 18.3 + 19:58:34.89 & 40:55:32.4 & 32318 & 76 & 18.4 + 19:58:44.73 & 40:34:41.7 & 104540 & 104 & 19.8 + 19:58:46.53 & 41:05:19.5 & 41792 & 70 & 19.7 + 19:58:53.89 & 40:33:19.5 & 31299 & 88 & 19.8 + 19:58:57.25 & 40:38:37.3 & 36006 & 82 & 18.8 + 19:58:57.74 & 40:59:22.0 & 26082 & 82 & 18.9 + 19:58:58.35 & 40:59:39.5 & 35796 & 66 & 19.3 + 19:59:04.29 & 40:49:24.8 & 68294 & 93 & 18.2 + 19:59:23.50 & 40:45:45.7 & 40592 & 111 & 19.1 + 20:00:38.42 & 40:59:06.7 & 43830 & 116 & 19.9 + 20:00:41.40 & 41:00:52.4 & 13431 & 94 & 19.7 + 20:00:50.58 & 41:12:14.8 & 48897 & 75 & 20.1 + r r r 19:56:48.41 & 40:27:52.9 & 18.1 + 19:56:48.99 & 40:56:40.6 & 18.1 + 19:56:49.27 & 40:51:04.6 & 19.0 + 19:56:49.28 & 40:48:48.6 & 19.4 + 19:56:51.47 & 40:34:01.2 & 19.3 + 19:56:52.03 & 40:36:19.0 & 19.0 + 19:56:52.57 & 41:07:43.7 & 15.8 + 19:56:53.77 & 40:27:53.2 & 19.6 + 19:56:55.56 & 40:59:23.2 & 17.6 + 19:56:55.82 & 41:03:51.3 & 18.4 + 19:56:58.25 & 40:20:38.5 & 19.8 + 19:57:00.04 & 40:50:02.7 & 17.9 + 19:57:00.74 & 41:06:17.8 & 20.3 + 19:57:00.75 & 40:48:43.6 & 18.7 + 19:57:03.27 & 40:40:18.2 & 17.0 + 19:57:04.00 & 40:17:04.0 & 18.7 + 19:57:04.72 & 40:14:38.1 & 19.5 + 19:57:04.73 & 41:00:43.2 & 19.9 + 19:57:05.45 & 40:38:30.0 & 18.5 + 19:57:05.75 & 40:15:10.4 & 17.5 + 19:57:06.06 & 41:07:22.3 & 20.7 + 19:57:06.12 & 40:49:53.8 & 18.3 + 19:57:06.14 & 40:57:09.6 & 17.6 + 19:57:07.12 & 40:54:41.8 & 19.3 + 19:57:08.16 & 40:56:11.6 & 18.8 + 19:57:08.61 & 41:11:40.4 & 18.9 + 19:57:08.61 & 41:08:00.8 & 19.0 + 19:57:09.11 & 40:31:13.6 & 20.0 + 19:57:09.96 & 40:52:10.1 & 16.4 + 19:57:10.62 & 41:01:21.6 & 20.8 + 19:57:10.95 & 40:38:37.0 & 20.5 + 19:57:11.72 & 41:08:15.2 & 18.5 + 19:57:12.71 & 40:38:38.4 & 20.1 + 19:57:16.18 & 40:39:13.5 & 20.1 + 19:57:25.40 & 40:35:19.2 & 18.8 + 19:57:26.37 & 41:02:01.7 & 20.2 + 19:57:27.03 & 41:04:25.5 & 19.7 + 19:57:28.11 & 40:56:26.1 & 21.2 + 19:57:28.45 & 40:55:00.0 & 20.7 + 19:57:30.19 & 40:51:16.4 & 20.7 + 19:57:30.72 & 40:39:00.2 & 20.3 + 19:57:32.29 & 40:14:33.2 & 19.4 + 19:57:32.38 & 40:21:20.0 & 17.4 + 19:57:33.30 & 40:40:48.7 & 19.9 + 19:57:33.42 & 40:30:51.4 & 20.6 + 19:57:33.75 & 41:13:02.7 & 19.8 + 19:57:37.40 & 40:54:05.2 & 15.3 + 19:57:37.96 & 41:09:26.8 & 18.4 + 19:57:37.97 & 40:15:31.6 & 20.5 + 19:57:39.52 & 40:46:39.9 & 20.2 + 19:57:39.67 & 40:36:10.8 & 20.1 + 19:57:39.82 & 41:13:02.1 & 19.9 + 19:57:40.61 & 41:08:46.1 & 18.8 + 19:57:41.86 & 41:10:18.4 & 18.6 + 19:57:42.48 & 40:20:06.9 & 18.2 + 19:57:43.72 & 41:03:27.3 & 18.7 + 19:57:43.75 & 40:19:11.8 & 16.7 + 19:57:44.16 & 41:02:11.5 & 19.9 + 19:57:44.93 & 40:18:09.3 & 19.8 + 19:57:45.95 & 40:21:53.2 & 20.2 + 19:57:46.25 & 40:34:03.4 & 18.9 + 19:57:46.70 & 40:53:12.1 & 20.3 + 19:57:47.94 & 40:54:38.1 & 20.7 + 19:57:48.57 & 40:41:00.5 & 19.2 + 19:57:48.73 & 41:13:02.9 & 17.2 + 19:57:49.52 & 40:33:39.1 & 20.9 + 19:57:49.59 & 40:22:33.7 & 17.2 + 19:57:49.67 & 40:36:17.3 & 18.7 + 19:57:49.94 & 41:04:54.3 & 20.1 + 19:57:50.38 & 40:25:37.1 & 18.8 + 19:57:51.15 & 40:38:12.2 & 18.9 + 19:57:53.82 & 40:55:10.6 & 20.9 + 19:57:53.99 & 40:14:53.0 & 19.9 + 19:57:55.18 & 40:57:14.5 & 19.6 + 19:57:55.70 & 40:13:39.7 & 18.9 + 19:57:57.06 & 40:42:07.6 & 20.1 + 19:57:57.26 & 41:00:36.1 & 19.1 + 19:57:57.59 & 40:56:17.3 & 18.0 + 19:57:59.45 & 40:57:05.5 & 18.8 + 19:58:00.83 & 40:56:54.9 & 19.5 + 19:58:01.10 & 40:52:51.9 & 19.0 + 19:58:01.84 & 40:38:45.7 & 17.9 + 19:58:02.25 & 40:25:04.6 & 18.0 + 19:58:02.25 & 40:14:48.6 & 18.4 + 19:58:03.34 & 41:06:07.6 & 20.4 + 19:58:04.07 & 40:19:43.1 & 20.3 + 19:58:04.97 & 40:20:11.0 & 20.3 + 19:58:04.99 & 40:55:09.7 & 17.0 + 19:58:05.01 & 40:37:24.4 & 18.9 + 19:58:05.59 & 40:37:01.9 & 18.7 + 19:58:05.70 & 40:48:52.9 & 20.6 + 19:58:06.62 & 40:55:44.8 & 19.3 + 19:58:07.92 & 40:38:23.9 & 19.9 + 19:58:08.38 & 41:05:14.2 & 19.6 + 19:58:08.40 & 40:54:55.9 & 18.7 + 19:58:08.73 & 40:57:45.4 & 19.3 + 19:58:09.38 & 40:49:35.4 & 19.4 + 19:58:09.98 & 41:12:54.0 & 19.0 + 19:58:10.92 & 40:52:30.4 & 20.0 + 19:58:13.98 & 40:24:38.4 & 20.7 + 19:58:14.01 & 40:54:44.0 & 19.9 + 19:58:14.77 & 40:26:58.6 & 19.6 + 19:58:16.27 & 41:11:59.7 & 19.9 + 19:58:16.55 & 41:00:39.4 & 20.0 + 19:58:16.66 & 40:56:32.9 & 19.0 + 19:58:18.88 & 40:45:22.1 & 20.3 + 19:58:19.19 & 40:43:21.6 & 18.3 + 19:58:19.25 & 40:40:02.9 & 17.8 + 19:58:20.32 & 40:45:22.1 & 20.8 + 19:58:20.71 & 40:57:30.2 & 20.0 + 19:58:21.62 & 40:54:38.7 & 20.8 + 19:58:21.96 & 41:09:49.9 & 19.0 + 19:58:22.41 & 40:57:44.7 & 19.8 + 19:58:22.48 & 41:01:42.6 & 18.8 + 19:58:25.57 & 40:43:30.3 & 19.8 + 19:58:26.40 & 41:02:54.7 & 20.2 + 19:58:26.78 & 41:00:26.4 & 20.7 + 19:58:26.84 & 40:58:47.1 & 19.9 + 19:58:27.01 & 41:13:30.9 & 18.5 + 19:58:27.35 & 40:37:08.0 & 17.9 + 19:58:28.23 & 40:32:39.7 & 19.7 + 19:58:28.85 & 40:40:04.1 & 17.7 + 19:58:29.18 & 40:34:14.6 & 19.2 + 19:58:30.05 & 41:07:59.0 & 19.8 + 19:58:30.64 & 41:14:13.9 & 19.5 + 19:58:32.49 & 41:05:16.7 & 17.8 + 19:58:34.05 & 40:47:32.6 & 20.4 + 19:58:36.32 & 40:44:58.3 & 19.7 + 19:58:38.61 & 41:10:49.1 & 20.2 + 19:58:39.21 & 40:58:43.1 & 20.5 + 19:58:39.47 & 40:57:08.0 & 19.2 + 19:58:39.76 & 40:57:32.2 & 18.6 + 19:58:41.46 & 40:26:08.0 & 20.6 + 19:58:41.61 & 40:56:56.3 & 19.4 + 19:58:42.81 & 40:22:17.9 & 18.7 + 19:58:42.90 & 40:45:46.1 & 20.0 + 19:58:43.76 & 40:20:08.2 & 18.7 + 19:58:43.99 & 41:06:14.8 & 21.1 + 19:58:47.41 & 40:34:04.4 & 17.8 + 19:58:48.43 & 41:02:25.5 & 17.6 + 19:58:48.48 & 40:16:32.6 & 18.5 + 19:58:48.64 & 40:16:13.2 & 17.6 + 19:58:48.66 & 40:38:53.9 & 18.7 + 19:58:49.84 & 40:25:41.4 & 20.9 + 19:58:50.48 & 40:52:41.7 & 17.1 + 19:58:51.28 & 40:45:48.1 & 19.7 + 19:58:51.47 & 40:27:12.5 & 17.9 + 19:58:51.64 & 40:50:17.8 & 18.3 + 19:58:53.22 & 40:56:16.8 & 19.5 + 19:58:53.43 & 40:59:37.4 & 19.2 + 19:58:53.92 & 40:33:19.0 & 19.8 + 19:58:53.94 & 40:44:34.1 & 19.9 + 19:58:56.09 & 41:12:50.4 & 18.7 + 19:58:56.31 & 40:51:58.8 & 19.7 + 19:58:57.00 & 40:29:53.1 & 16.5 + 19:58:58.08 & 40:13:42.3 & 17.7 + 19:58:58.68 & 40:41:30.1 & 20.5 + 19:58:59.57 & 41:07:22.3 & 20.9 + 19:59:01.38 & 41:11:18.1 & 19.6 + 19:59:03.46 & 40:56:08.3 & 20.2 + 19:59:03.78 & 40:22:22.2 & 17.8 + 19:59:04.23 & 40:46:56.6 & 17.8 + 19:59:04.55 & 40:51:30.6 & 19.5 + 19:59:04.88 & 40:19:07.1 & 19.9 + 19:59:05.26 & 40:19:18.8 & 17.1 + 19:59:05.86 & 40:54:04.4 & 20.8 + 19:59:06.03 & 40:49:22.8 & 19.9 + 19:59:08.69 & 40:54:35.3 & 19.7 + 19:59:11.68 & 40:52:03.9 & 20.1 + 19:59:12.42 & 40:18:40.2 & 18.2 + 19:59:12.46 & 41:05:01.2 & 19.7 + 19:59:13.50 & 40:16:05.6 & 20.3 + 19:59:14.62 & 40:46:44.9 & 19.5 + 19:59:15.21 & 40:54:43.8 & 19.1 + 19:59:15.78 & 41:08:55.4 & 19.5 + 19:59:16.32 & 41:04:36.3 & 21.0 + 19:59:17.51 & 40:43:58.9 & 17.8 + 19:59:17.93 & 40:22:00.6 & 20.1 + 19:59:20.31 & 40:32:01.1 & 19.0 + 19:59:21.68 & 40:27:01.1 & 20.4 + 19:59:22.09 & 40:44:23.7 & 17.2 + 19:59:22.24 & 40:25:54.4 & 19.1 + 19:59:22.96 & 40:49:51.0 & 19.6 + 19:59:23.89 & 40:51:43.3 & 18.9 + 19:59:24.09 & 40:50:29.9 & 19.1 + 19:59:24.21 & 40:49:19.9 & 19.5 + 19:59:25.64 & 40:42:44.5 & 17.5 + 19:59:26.24 & 40:44:48.2 & 19.4 + 19:59:26.67 & 40:43:37.1 & 18.0 + 19:59:27.15 & 40:14:40.2 & 19.7 + 19:59:28.07 & 40:44:40.8 & 18.6 + 19:59:28.52 & 40:21:07.1 & 18.5 + 19:59:29.01 & 40:14:13.4 & 17.9 + 19:59:31.22 & 40:24:16.6 & 17.5 + 19:59:31.30 & 40:54:39.7 & 20.9 + 19:59:32.07 & 40:14:30.8 & 19.4 + 19:59:33.70 & 40:24:50.2 & 20.5 + 19:59:34.84 & 40:39:42.8 & 18.0 + 19:59:35.33 & 41:10:40.8 & 19.5 + 19:59:35.66 & 40:45:32.5 & 17.9 + 19:59:35.90 & 40:34:27.1 & 19.9 + 19:59:36.66 & 40:56:42.2 & 20.6 + 19:59:37.08 & 40:51:20.5 & 19.9 + 19:59:37.15 & 40:37:59.3 & 20.6 + 19:59:37.91 & 40:46:03.5 & 16.9 + 19:59:38.15 & 40:47:58.1 & 19.7 + 19:59:38.55 & 40:19:41.5 & 18.6 + 19:59:39.85 & 40:47:21.3 & 18.1 + 19:59:43.52 & 40:46:35.0 & 17.4 + 19:59:43.63 & 40:25:41.2 & 17.8 + 19:59:44.99 & 41:12:52.4 & 18.1 + 19:59:45.00 & 40:18:22.8 & 17.7 + 19:59:45.63 & 40:40:24.8 & 17.2 + 19:59:47.36 & 40:23:32.2 & 17.6 + 19:59:47.44 & 40:46:37.4 & 20.2 + 19:59:48.49 & 40:58:01.9 & 17.6 + 19:59:48.91 & 40:43:26.7 & 20.3 + 19:59:49.07 & 40:32:18.4 & 18.6 + 19:59:49.70 & 40:55:52.5 & 19.3 + 19:59:49.85 & 40:56:22.4 & 20.2 + 19:59:50.47 & 40:55:14.1 & 17.5 + 19:59:51.80 & 40:53:35.3 & 19.0 + 19:59:51.82 & 41:05:51.9 & 18.1 + 19:59:52.28 & 40:41:36.9 & 19.8 + 19:59:52.46 & 40:49:17.0 & 18.1 + 19:59:52.89 & 40:44:17.7 & 19.0 + 19:59:53.90 & 41:01:25.8 & 20.5 + 19:59:54.01 & 40:24:13.2 & 18.3 + 19:59:55.48 & 40:28:34.6 & 19.6 + 19:59:55.56 & 40:23:17.2 & 19.1 + 19:59:56.18 & 40:48:49.4 & 20.7 + 19:59:56.24 & 40:26:22.9 & 17.3 + 19:59:56.45 & 40:24:55.6 & 20.1 + 19:59:56.88 & 40:30:55.7 & 18.2 + 19:59:58.28 & 40:52:57.6 & 18.2 + 19:59:59.77 & 40:22:04.8 & 18.7 + 20:00:00.62 & 40:46:07.5 & 18.3 + 20:00:00.75 & 40:43:17.0 & 19.2 + 20:00:01.28 & 40:44:57.4 & 18.2 + 20:00:02.04 & 40:39:52.6 & 19.9 + 20:00:02.36 & 40:57:07.6 & 20.0 + 20:00:02.62 & 40:57:06.1 & 20.1 + 20:00:03.24 & 40:40:46.6 & 17.2 + 20:00:03.51 & 40:24:01.3 & 19.9 + 20:00:04.24 & 40:40:13.5 & 19.7 + 20:00:04.25 & 40:57:27.7 & 18.7 + 20:00:04.54 & 40:41:31.8 & 20.9 + 20:00:06.74 & 41:07:53.6 & 18.5 + 20:00:08.54 & 40:59:52.9 & 20.2 + 20:00:10.06 & 41:08:12.6 & 20.8 + 20:00:11.48 & 40:51:02.2 & 20.2 + 20:00:12.60 & 40:54:18.4 & 19.3 + 20:00:12.99 & 40:17:18.7 & 19.7 + 20:00:13.68 & 40:54:32.1 & 20.5 + 20:00:15.08 & 40:43:32.5 & 19.8 + 20:00:15.50 & 41:10:00.0 & 19.8 + 20:00:15.75 & 40:39:57.0 & 18.0 + 20:00:18.03 & 40:55:34.2 & 20.8 + 20:00:18.65 & 40:24:02.0 & 19.2 + 20:00:19.69 & 40:54:10.1 & 19.3 + 20:00:20.24 & 40:37:12.0 & 18.7 + 20:00:20.31 & 40:51:46.3 & 17.3 + 20:00:20.58 & 40:26:45.1 & 18.7 + 20:00:21.94 & 40:53:57.3 & 20.8 + 20:00:23.32 & 40:31:25.3 & 19.9 + 20:00:23.64 & 40:45:52.1 & 20.4 + 20:00:23.69 & 40:31:21.4 & 19.6 + 20:00:23.91 & 40:32:48.8 & 19.5 + 20:00:24.28 & 41:08:15.0 & 20.4 + 20:00:24.37 & 40:56:26.9 & 19.7 + 20:00:25.24 & 40:38:26.4 & 20.6 + 20:00:27.57 & 40:40:03.1 & 20.2 + 20:00:27.63 & 41:02:18.7 & 20.2 + 20:00:27.82 & 40:27:55.9 & 18.7 + 20:00:28.02 & 41:02:37.1 & 19.7 + 20:00:29.86 & 40:25:51.4 & 20.0 + 20:00:29.99 & 41:05:54.7 & 18.8 + 20:00:30.30 & 40:46:24.6 & 17.1 + 20:00:31.76 & 40:51:31.6 & 20.3 + 20:00:32.87 & 40:37:46.6 & 19.9 + 20:00:33.39 & 40:52:42.7 & 18.6 + 20:00:33.62 & 40:34:35.6 & 19.2 + 20:00:34.51 & 40:55:49.6 & 19.8 + 20:00:34.83 & 40:34:48.8 & 17.7 + 20:00:34.84 & 40:52:20.7 & 18.0 + 20:00:37.51 & 40:46:45.8 & 20.1 + 20:00:39.16 & 40:50:30.4 & 18.5 + 20:00:39.20 & 41:02:38.4 & 20.2 + 20:00:39.51 & 40:25:43.0 & 19.6 + 20:00:40.81 & 40:57:18.7 & 18.5 + 20:00:40.86 & 40:20:05.3 & 19.2 + 20:00:40.92 & 40:40:48.6 & 17.6 + 20:00:43.29 & 40:18:45.5 & 19.4 + 20:00:43.47 & 40:30:50.6 & 21.1 + 20:00:43.81 & 40:24:46.2 & 19.4 + 20:00:45.03 & 40:51:31.2 & 20.6 + 20:00:46.40 & 40:29:07.7 & 18.9 + 20:00:47.43 & 40:39:55.5 & 17.6 + 20:00:47.94 & 40:33:37.9 & 19.4 + 20:00:49.02 & 40:53:40.3 & 19.6 + 20:00:50.18 & 40:53:21.1 & 17.7 + 20:00:51.39 & 40:28:38.7 & 18.0 + 20:00:51.68 & 40:46:28.3 & 18.9 + 20:00:51.97 & 40:32:56.1 & 19.2 + 20:00:52.07 & 40:33:28.7 & 19.2 + 20:00:52.51 & 41:01:36.0 & 19.2 + 20:00:52.57 & 40:31:16.5 & 16.8 + 20:00:52.67 & 40:50:32.7 & 18.0 + 20:00:53.18 & 40:54:38.1 & 20.3 + 20:00:53.96 & 40:58:57.2 & 19.6 + 20:00:55.08 & 40:41:14.6 & 17.9 + 20:00:55.21 & 40:44:30.5 & 18.8 + 20:00:55.98 & 40:52:16.3 & 18.5 + 20:00:56.02 & 41:01:54.5 & 18.5 + 20:00:56.19 & 40:32:05.0 & 21.1 + 20:00:56.75 & 40:54:01.7 & 18.9 + 20:00:57.56 & 40:25:33.1 & 18.3 + 20:00:59.44 & 40:23:18.9 & 19.5 + 20:01:00.80 & 40:59:47.8 & 18.9 + 20:01:01.44 & 40:47:55.3 & 20.1 + 20:01:01.50 & 41:10:06.8 & 16.6 + 20:01:03.32 & 41:00:54.0 & 18.8 + 20:01:03.95 & 40:50:49.8 & 19.9 + 20:01:04.68 & 40:40:19.9 & 18.8 + 20:01:05.19 & 40:55:56.0 & 20.4 + 20:01:06.35 & 40:56:35.0 & 19.7 + 20:01:06.96 & 41:09:26.8 & 19.7 + 20:01:07.21 & 40:50:47.8 & 17.9 + 20:01:07.21 & 40:55:01.2 & 19.5 + 20:01:07.30 & 40:50:16.4 & 18.7 + 20:01:08.77 & 40:35:22.7 & 19.6 + 20:01:08.99 & 40:32:07.0 & 20.2 + 20:01:13.39 & 40:17:05.6 & 20.2 + 20:01:14.79 & 40:33:10.9 & 19.5 + 20:01:16.07 & 40:56:29.7 & 20.9 + 20:01:18.09 & 40:54:48.6 & 19.9 + 20:01:18.96 & 40:52:34.7 & 20.1 + 20:01:19.99 & 40:41:43.0 & 20.3 + 20:01:22.90 & 40:27:21.3 & 18.5 + 20:01:23.76 & 40:26:01.8 & 17.2 + 20:01:25.88 & 41:13:04.0 & 20.3 + 20:01:29.64 & 40:53:22.4 & 19.1 + 20:01:30.76 & 40:29:11.0 & 19.0 + 20:01:31.03 & 40:50:44.8 & 19.4 + 20:01:32.02 & 40:59:55.3 & 19.2 + 20:01:33.20 & 40:37:31.6 & 18.2 + 20:01:33.56 & 41:01:05.3 & 19.0 + 20:01:33.57 & 40:45:14.4 & 17.6 + 20:01:35.82 & 40:59:04.0 & 19.4 + 20:01:38.28 & 40:31:55.1 & 18.8 + 20:01:42.30 & 41:09:25.8 & 18.1 + 20:01:42.69 & 40:38:42.1 & 19.9 + 20:01:46.20 & 40:26:01.3 & 18.4 + 20:01:46.75 & 40:36:00.3 & 19.5 + 20:01:47.48 & 40:30:38.8 & 20.0 + 20:01:47.74 & 41:11:08.2 & 20.1 + 20:01:48.00 & 41:11:05.1 & 18.3 + 20:01:48.31 & 40:35:57.6 & 20.2 + 20:01:50.22 & 40:35:26.7 & 19.8 + 20:01:51.72 & 40:25:22.5 & 17.4 + 20:01:51.83 & 40:40:04.0 & 18.6 + 20:02:04.97 & 41:07:32.6 & 18.5 +
linear regression is one of the most common statistical techniques used in astronomical data analysis . in general , linear regression in astronomy is characterized by intrinsic scatter about the regression line , and measurement errors in both the independent and dependent variables . the source of intrinsic scatter is variations in the physical properties of astronomical sources that are not completely captured by the variables included in the regression . it is important to correctly account for both measurement error and intrinsic scatter , as both aspects can have a non - negligible effect on the regression results . in particular , ignoring the intrinsic scatter and weighting the data points solely by the measurement errors can result in the higher - precision measurements being given disproportionate influence on the regression results . furthermore , when the independent variable is measured with error , the ordinary least squares ( ols ) estimate of the regression slope is biased toward zero ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? when there are multiple independent variables , measurement error can have an even stronger and more unpredictable effect @xcite . in addition , the existence of non - detections , referred to as ` censored data ' , in the data set will result in additional complications ( e.g. , * ? ? ? * ) . therefore , when performing regression , it is essential to correctly account for the measurement errors and intrinsic scatter in order to ensure that the data analysis , and thus the scientific conclusions based on it , are trustworthy . many methods have been proposed for performing linear regression when intrinsic scatter is present and both variables are measured with error . these include methods that correct the observed moments of the data ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , minimize an ` effective ' @xmath1 statistic ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , assume a probability distribution for the true independent variable values ( so - called structural equation models , e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) ; bayesian approaches to these models have also been developed ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? in addition , methods have been proposed to account for measurement error in censored regression ( e.g. , * ? ? ? * ; * ? ? ? the most commonly used methods in astronomy are the bces estimator @xcite and the ` fitexy ' estimator @xcite . both methods have their advantages and disadvantages , some of which have been pointed out by @xcite . however , neither method is applicable when the data contain non - detections . in this work i describe a bayesian method for handling measurement errors in astronomical data analysis . my approach starts by computing the likelihood function of the complete data , i.e. , the likelihood function of both the unobserved true values of the data and the measured values of the data . the measured data likelihood is then found by integrating the likelihood function for the complete data over the unobserved true values ( e.g. , * ? ? ? * ; * ? ? ? this approach is known as ` structural equation modelling ' of measurement error problems , and has been studied from both a frequentist approach ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and a bayesian approach ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in this work , i extend the statistical model of @xcite to allow for measurement errors of different magnitudes ( i.e. , ` heteroscedastic ' errors ) , non - detections , and selection effects , so long as the selection function can be modelled mathematically . our method models the distribution of independent variables as a weighted sum of gaussians . the mixture of gaussians model allows flexibility when estimating the distribution of the true values of the independent variable , thus increasing its robustness against model mispecification ( e.g. , * ? ? ? the basic idea is that one can use a suitably large enough number of gaussians to accurately approximate the true distribution of independent variables , even though in general the individual gaussians have no physical meaning . the paper is organized as follows . in [ s - notation ] we summarize some notation , and in [ s - measerr ] i review the effects of measurement error on the estimates for the regression slope and correlation coefficient . in [ s - statmod ] i describe the statistical model and derive the likelihood functions , and in [ s - dcissues ] i describe how to incorporate knowledge of the selection effects and account for non - detections . in [ s - prior ] i describe the prior distribution for this model , and in [ s - markov ] i describe a gibbs sampler for sampling from the posterior distributions . in [ s - sims ] i use simulation to illustrate the effectiveness of this structural model and compare with the ols , bces(@xmath2 ) , and fitexy estimators . finally , in [ s - data ] i illustrate the method using astronomical data by performing a regression of the x - ray photon index , @xmath3 , on the eddington ratio using a sample of 39 @xmath4 radio - quiet quasars . sections [ s - statmod ] , [ s - dcissues ] , and [ s - compmeth ] are somewhat technical , and the reader who is uninterested in the mathematical and computational details may skip to them . i will use the common statistical notation that an estimate of a quantity is denoted by placing a ` hat ' above it ; e.g. , @xmath5 is an estimate of the true value of the parameter @xmath6 . in general , greek letters will denote the true value of a quantity , while roman letters will denote the contaminated measured value . i will frequently refer to the ` bias ' of an estimator . the bias of an estimator is @xmath7 , where @xmath8 is the expectation value of the estimator @xmath5 , and @xmath9 is the true value of @xmath6 . an unbiased estimator is one such that @xmath10 . i will denote a normal density with mean @xmath11 and variance @xmath12 as @xmath13 , and i will denote as @xmath14 a multivariate normal density with @xmath15-element mean vector @xmath11 and @xmath16 covariance matrix @xmath17 . if i want to explicitly identify the argument of the gaussian function , i will use the notation @xmath18 , which should be understood to be a gaussian with mean @xmath11 and variance @xmath12 as a function of @xmath19 . following @xcite , i denote the scaled inverse-@xmath1 density as inv-@xmath20 , where @xmath21 is the degrees of freedom and @xmath22 is the scale parameter , and we denote the inverse - wishart as inv-@xmath23 , where @xmath21 is the degrees of freedom and @xmath24 is the scale matrix . the inverse - wishart distribution can be thought of as a multivariate generalization of the scaled inverse-@xmath1 distribution . i will often use the common statistical notation where `` @xmath25 '' means `` is drawn from '' or `` is distributed as '' . for example , @xmath26 states that @xmath19 is drawn from a normal density with mean @xmath11 and variance @xmath12 . it is well known that measurement error can attenuate the estimate of the regression slope and correlation coefficient ( e.g. , * ? ? ? * ; * ? ? ? * ) . for completeness , i give a brief review of the effect of measurement error on correlation and regression analysis for the case of one independent variable . denote the independent variable as @xmath27 and the dependent variable as @xmath28 ; @xmath27 and @xmath28 are also referred to as the ` covariate ' and the ` response ' , respectively . i assume that @xmath27 is a random vector of @xmath29 data points drawn from some probability distribution . the dependent variable , @xmath28 , depends on @xmath27 according to the usual additive model : @xmath30 here , @xmath31 is a random variable representing the intrinsic scatter in @xmath32 about the regression relationship , and @xmath33 are the regression coefficients . the mean of @xmath34 is assumed to be zero , and the variance of @xmath34 is assumed to be constant and is denoted as @xmath12 . we do not observe the actual values of @xmath35 , but instead observe values @xmath36 which are measured with error . the measured values are assumed to be related to the actual values as @xmath37 where @xmath38 and @xmath39 are the random measurement errors on @xmath40 and @xmath41 , respectively . in general , the errors are normally distributed with known variances @xmath42 and @xmath43 , and covariance @xmath44 . for simplicity , throughout the rest of this section i assume that @xmath45 and @xmath46 are the same for each data point . when the data are measured without error , the least - squares estimate of the regression slope , @xmath47 , and the estimated correlation coefficient , @xmath48 , are @xmath49 here , @xmath50 is the sample covariance between @xmath27 and @xmath28 , and @xmath51 is the sample variance of @xmath27 . when the data are measured with error , the least - squares estimate of the regression slope , @xmath52 , and the estimated correlation coefficient , @xmath53 , become @xmath54 from these equations it is apparent that the estimated slope and correlation are biased when the data are measured with error . it is informative to assess the effect of measurement error in terms of the ratios @xmath55 , as these quantities can be calculated from the data . the fractional bias in the estimated slope and correlation may then be expressed as @xmath56 from equations ( [ eq - frac_bbias ] ) and ( [ eq - frac_rbias ] ) it is apparent that measurement errors have the following effects . first , covariate measurement error reduces the magnitude of the observed correlation between the independent variable and the response , as well as biasing the estimate of the slope towards zero . second , measurement error in the response also reduces the magnitude of the observed correlation between the variables . third , if the measurement errors are correlated the effects depend on the sign of this correlation . if the measurement error correlation has the same sign as the intrinsic correlation between @xmath27 and @xmath28 , then the measurement errors cause a spurious increase in the observed correlation ; otherwise the measurement errors cause a spurious decrease in the observed correlation . the magnitude of these effects depend on how large the measurement errors are compared to the observed variance in @xmath19 and @xmath57 . in figure [ f - bias ] i plot the fractional bias in the correlation coefficient , @xmath58 , as a function of @xmath59 and @xmath60 when the errors are uncorrelated . as can be seen , measurement error can have a significant effect on the estimation of the linear correlation coefficient . for example , when @xmath61 and @xmath62 , the estimated correlation is @xmath63% lower than the true correlation . therefore , interpretation of correlation coefficients and regression slopes must be approached with caution when the data have been contaminated by measurement error . to ensure accurate results , it is necessary to employ statistical methods that correct for the measurement errors . i assume that the independent variable , @xmath27 , is drawn from a probability distribution @xmath64 , where @xmath65 denotes the parameters for this distribution . the dependent variable is then drawn from the conditional distribution of @xmath28 given @xmath27 , denoted as @xmath66 ; @xmath6 denotes the parameters for this distribution . the joint distribution of @xmath27 and @xmath28 is then @xmath67 . in this work i assume the normal linear regression model given by equation ( [ eq - adderr ] ) , and thus @xmath66 is a normal density with mean @xmath68 and variance @xmath12 , and @xmath69 . since the data are a randomly observed sample , we can derive the likelihood function for the measured data . the likelihood function of the measured data , @xmath70 , is obtained by integrating the complete data likelihood over the missing data , @xmath27 and @xmath28 ( e.g. , * ? ? ? * ; * ? ? ? * ) : @xmath71 here , @xmath72 is the complete data likelihood function . because of the hierarchical structure inherent in the measurement error model , it is helpful to decompose the complete data likelihood into conditional probability densities : @xmath73 the density @xmath74 describes the joint distribution of the measured values @xmath19 and @xmath57 at a given @xmath27 and @xmath28 , and depends on the assumed distribution of the measurement errors , @xmath75 and @xmath76 . in this work i assume gaussian measurement error , and thus @xmath77 is a multivariate normal density with mean @xmath78 and covariance matrix @xmath79 , where @xmath80 and @xmath81 . the statistical model may then be conveniently expressed hierarchically as @xmath82,\sigma_i ) \label{eq - hierxy } \end{aligned}\ ] ] note that if @xmath40 is measured without error , then @xmath83 is a dirac delta function , and @xmath84 . an equivalent result holds if @xmath41 is measured without error . equation ( [ eq - obslik2 ] ) may be used to obtain the observed data likelihood function for any assumed distribution of @xmath27 . in this work , i model @xmath64 as a mixture of @xmath85 gaussians , @xmath86 where @xmath87 . note that , @xmath88 may be interpreted as the probability of drawing a data point from the @xmath89 gaussian . i will use the convenient notation @xmath90 and @xmath91 ; note that @xmath92 . it is useful to model @xmath64 using this form because it is flexible enough to adapt to a wide variety of distributions , but is also conjugate for the regression relationship ( eq.[[eq - adderr ] ] ) and the measurement error distribution , thus simplifying the mathematics . assuming the gaussian mixture model for @xmath64 , the measured data likelihood for the @xmath93 data point can be directly calculated using equation ( [ eq - obslik2 ] ) . denoting the measured data as @xmath94 , the measured data likelihood function for the @xmath93 data point is then a mixture of bivariate normal distributions with weights @xmath95 , means @xmath96 , and covariance matrices @xmath97 . because the data points are statistically independent , the full measured data likelihood is then the product of the likelihood functions for the individual data points : @xmath98 here , @xmath99 denotes the transpose of @xmath100 . equation ( [ eq - bobslik ] ) may be maximized to compute the maximum - likelihood estimate ( mle ) . when @xmath101 , the expectation - maximization ( em ) algorithm @xcite is probably the most efficient tool for calculating the mle . @xcite describe an em algorithm when @xmath102 is assumed to be a mixture of normals and the measurement error distribution is multivariate @xmath103 , and their results can be extended to the statistical model described in this work . it is informative to decompose the measured data likelihood , @xmath104 , as this representation is useful when the data contain non - detections ( cf . , [ s - nondet ] ) . the marginal distribution of @xmath40 is @xmath105 and the conditional distribution of @xmath41 given @xmath40 is @xmath106 ^ 2}{var(y_i\|x_i , k ) } \right\ } \label{eq - cdensyx } \\ \gamma_k & = & \frac{\pi_k n(x_i\|\mu_k , \tau^2_k + \sigma^2_{x , i } ) } { \sum_{j=1}^k \pi_j n(x_i\|\mu_j , \tau^2_j + \sigma^2_{x , i } ) } \label{eq - cdenskx } \\ e(y_i\|x_i , k ) & = & \alpha + \left ( \frac{\beta \tau^2_k + \sigma_{xy , i } } { \tau^2_k + \sigma^2_{x , i } } \right ) x_i + \left ( \frac{\beta \sigma^2_{x , i } - \sigma_{xy , i } } { \tau_k^2 + \sigma^2_{x , i } } \right ) \mu_k \label{eq - cmeanyx } \\ var(y_i\|x_i , k ) & = & \beta^2 \tau^2_k + \sigma^2 + \sigma^2_{y , i } - \frac{(\beta \tau^2_k - \sigma_{xy , i})^2}{\tau^2_k + \sigma^2_{x , i } } \label{eq - cvary}. \end{aligned}\ ] ] here , @xmath107 can be interpreted as the probability that the @xmath93 data point was drawn from the @xmath89 gaussian given @xmath40 , @xmath108 gives the expectation value of @xmath41 at @xmath40 , given that the data point was drawn from the @xmath89 gaussian , and @xmath109 gives the variance in @xmath41 at @xmath40 , given that the data point was drawn from the @xmath89 gaussian . it is informative to investigate the case where the distribution of @xmath27 is assumed to be uniform , @xmath110 . interpreting @xmath102 as a ` prior ' on @xmath27 , one may be tempted to consider assuming @xmath110 as a more objective alternative to the normal distribution . a uniform distribution for @xmath27 may be obtained as the limit @xmath111 , and thus the likelihood function for @xmath110 can be calculated from equation ( [ eq - cdensyx ] ) by taking @xmath111 and @xmath112 . when the measurement errors are uncorrelated , the likelihood for uniform @xmath102 is @xmath113 the argument of the exponential is the fitexy goodness of fit statistic , @xmath114 , as modified by @xcite to account for intrinsic scatter ; this fact has also been recognized by @xcite . despite this connection , minimizing @xmath114 is not the same as maximizing the conditional likelihood of @xmath57 given @xmath19 , as both @xmath115 and @xmath12 appear in the normalization of the likelihood function as well . for a given value of @xmath12 , minimizing @xmath114 can be interpreted as minimizing a weighted sum of squared errors , where the weights are given by the variances in @xmath41 at a given @xmath40 , and one assumes a uniform distribution for @xmath27 . unfortunately , this is only valid for a fixed value of @xmath12 . moreover , little is known about the statistical properties of the fitexy estimator , such as its bias and variance , although bootstrapping ( e.g. , * ? ? ? * ; * ? ? ? * ) may be used to estimate them . furthermore , it is ambiguous how to calculate the fitexy estimates when there is an intrinsic scatter term . the fitexy goodness - of - fit statistic , @xmath114 , can not be simultaneously minimized with respect to @xmath116 and @xmath12 , as @xmath114 is a strictly decreasing function of @xmath12 . as such , it is unclear how to proceed in the optimization beyond an _ ad hoc _ approach . many authors have followed the approach adopted by @xcite and increase @xmath12 until @xmath117 , or assume @xmath118 if @xmath119 . despite the fact that minimizing @xmath120 is not the same as maximizing equation ( [ eq - uniflik ] ) , one may still be tempted to calculate a mle based on equation ( [ eq - uniflik ] ) . however , it can be shown that if one assumes @xmath110 , and if all of the @xmath19 and @xmath57 have the same respective measurement error variances , @xmath121 and @xmath122 , the mle estimates for @xmath123 and @xmath115 are just the ordinary least squares estimates @xcite . while this is not necessarily true when the magnitudes of the measurement errors vary between data points , one might expect that the mle will behave similarly to the ols estimate . i confirm this fact using simulation in [ s - sims1 ] . unfortunately , this implies that the mle for @xmath110 inherits the bias in the ols estimate , and thus nothing is gained . furthermore , as argued by @xcite , one can easily be convinced that assuming @xmath110 is incorrect by examining a histogram of @xmath19 . the formalism developed in [ s - normreg ] can easily be generalized to multiple independent variables . in this case equation ( [ eq - adderr ] ) becomes @xmath124 where @xmath115 is now a @xmath15-element vector and @xmath125 is a @xmath15-element vector containing the values of the independent variables for the @xmath93 data point . similar to before , we assume that the distribution of @xmath125 can be approximated using a mixture of @xmath85 multivariate normal densities with @xmath15-element mean vectors @xmath126 , @xmath16 covariance matrices @xmath127 , and weights @xmath128 . the measured value of @xmath125 is the @xmath15-element vector @xmath129 , and the gaussian measurement errors on @xmath130 have @xmath131 covariance matrix @xmath79 . the statistical model is then @xmath132 , \sigma_i ) . \label{eq - mhierxy } \\ \end{aligned}\ ] ] denoting @xmath133 , the measured data likelihood is @xmath134 here , @xmath135 is the @xmath136-element mean vector of @xmath137 for gaussian @xmath138 , @xmath139 is the @xmath140 covariance matrix of @xmath137 for gaussian @xmath138 , @xmath43 is the variance in the measurement error on @xmath41 , @xmath44 is the @xmath15-element vector of covariances between the measurement errors on @xmath41 and @xmath129 , and @xmath141 is the @xmath16 covariance matrix of the measurement errors on @xmath129 . similar to the case for one independent variable , the measured data likelihood can be decomposed as @xmath142 , where @xmath143 and @xmath144 ^ 2}{var(y_i\|{\bf x}_i , k ) } \right\ } \label{eq - mcdensyx } \\ \gamma_k & = & \frac{\pi_k n({\bf x}_i\|\mu_k , t_k + \sigma_{x , i } ) } { \sum_{j=1}^k \pi_j n({\bf x}_i\|\mu_j , t_j + \sigma_{x , i } ) } \label{eq - mcdenskx } \\ e(y_i\|{\bf x}_i , k ) & = & \alpha + \beta^t \mu_k + ( \beta^t t_k + \sigma^t_{xy , i } ) ( t_k + \sigma_{x , i})^{-1 } ( { \bf x}_i - \mu_k ) \label{eq - mxihati } \\ var(y_i\|{\bf x}_i , k ) & = & \beta^t t_k \beta + \sigma^2 + \sigma^2_{y , i } - ( \beta^t t_k + \sigma^t_{xy , i } ) ( t_k + \sigma_{x , i})^{-1 } ( t_k \beta + \sigma_{xy , i } ) \label{eq - mcvary}. \end{aligned}\ ] ] there are several issues common in the collection of astronomical data that violate the simple assumptions made in [ s - statmod ] . astronomical data collection consists almost entirely of passive observations , and thus selection effects are a common concern . instrumental detection limits often result in the placement of upper or lower limits on quantities , and astronomical surveys are frequently flux - limited . in this section i modify the likelihood functions described in [ s - statmod ] to include the effects of data collection . general methods for dealing with missing data are described in @xcite and @xcite , and i apply the methodology described in these references to the measurement error model developed here . although in this work i focus on linear regression , many of these results can be applied to more general statistical models , such as estimating luminosity functions . suppose that one collects a sample of @xmath29 sources out of a possible @xmath145 sources . one is interested in understanding how the observable properties of these sources are related , but is concerned about the effects of the selection procedure on the data analysis . for example , one may perform a survey that probes some area of the sky . there are @xmath145 sources located within this solid angle , where @xmath145 is unknown . because of the survey s selection method , the sample only includes @xmath29 sources . in this case the astronomer is interested in how measurement error and the survey s selection method affect statistical inference . i investigate selection effects within the framework of our statistical model by introducing an indicator variable , @xmath146 , which denotes whether a source is included in the sample . if the @xmath147 source is included in the sample , then @xmath148 , otherwise @xmath149 . in addition , i assume that the selection function only depends on the measured values , @xmath19 and @xmath57 . under this assumption , the selection function of the sample is the probability of including a source with a given @xmath19 and @xmath57 , @xmath150 . this is commonly the case in astronomy , where sources are collected based on their measured properties . for example , one may select sources for a sample based on their measured properties as reported in the literature . in addition , if one performs a flux - limited survey then a source will only be considered detected if its measured flux falls above some set flux limit . if a sample is from a survey with a simple flux limit , then @xmath151 if the measured source flux @xmath41 is above the flux limit , and @xmath152 if the measured source flux is below the flux limit . since the selection function depends on the measured flux value , and not the true flux value , sources with true flux values above the flux limit can be missed by the survey , and sources with true flux below the limit can be detected by the survey . this effect is well - known in astronomy and is commonly referred to as malmquist bias ( e.g. , * ? ? ? including the variable @xmath146 , the complete data likelihood can be written as @xmath153 equation ( [ eq - complik ] ) is valid for any number of independent variables , and thus @xmath40 and @xmath125 may be either scalar or vector . integrating equation ( [ eq - complik ] ) over the missing data , the observed data likelihood is @xmath154 here , @xmath155 is the binomial coefficient , @xmath156 denotes the set of @xmath29 included sources , @xmath157 and @xmath158 denote the values of @xmath19 and @xmath57 for the included sources , and @xmath159 denotes the set of @xmath160 missing sources . in addition , i have omitted terms that do not depend on @xmath161 , or @xmath145 . note that @xmath145 is unknown and is thus also a parameter of the statistical model . the binomial coefficient is necessary because it gives the number of possible ways to select a sample of @xmath29 sources from a set of @xmath145 sources . it is apparent from equation ( [ eq - mislik ] ) that statistical inference on the regression parameters is unaffected if the selection function is independent of @xmath57 and @xmath19 . ( e.g. , * ? ? ? * ; * ? ? ? * ) . in this case the selection function may be ignored . it is commonly the case that a sample is selected based only on the measured independent variables . for example , suppose one performs a survey in which all sources with measured optical flux greater than some threshold are included . then , these optically selected sources are used to fit a regression in order to understand how the x - ray luminosity of these objects depends on their optical luminosity and redshift . in this case , the probability of including a source only depends on the measured values of the optical luminosity and redshift , and is thus independent of the x - ray luminosity . when the sample selection function is independent of @xmath57 , given @xmath19 , then @xmath162 . because we are primarily interested in the regression parameters , @xmath6 , i model the distributions of @xmath27 for the included and missing sources seperately , with the parameters for the distribution of included sources denoted as @xmath163 . in addition , i assume that the measurement errors between @xmath57 and @xmath19 are statistically independent . then the @xmath160 integrals over @xmath57 and @xmath28 for the missing sources in equation ( [ eq - mislik ] ) are equal to unity , and we can write the observed data likelihood as @xmath164 where @xmath165 is the distribution of those @xmath27 included in one s sample . here i have omitted terms depending on @xmath145 because one is primarily interested in inference on the regression parameters , @xmath6 . equation ( [ eq - indlik ] ) is identical to equation ( [ eq - obslik2 ] ) , with the exception that @xmath64 now only models the distribution of those @xmath27 that have been included in one s sample , and i have now assumed that the measurement errors on @xmath57 and @xmath19 are independent . in particular , for the gaussian mixture models described in [ s - normreg ] and [ s - multreg ] , the observed data likelihood is given by equations ( [ eq - bobslik ] ) and ( [ eq - mobslik ] ) , where @xmath95 , @xmath11 , and @xmath166 ( or @xmath167 ) should be understood as referring to the parameters for the distribution of the observed @xmath27 . as is evident from the similarity between equations ( [ eq - indlik ] ) and ( [ eq - obslik2 ] ) , _ if the sample is selected based on the measured independent variables , and if the measurement errors on the dependent and independent variables are statistically independent , then inference on the regression parameters , @xmath6 , is unaffected by selection effects_. if the method in which a sample is selected depends on the measured dependent variable , @xmath57 , or if the measurement error in @xmath19 and @xmath57 are correlated , the observed data likelihood becomes more complicated . as an example , one might encounter this situation if one uses an x - ray selected sample to investigate the dependence of x - ray luminosity on optical luminosity and redshift . in this case , the selection function of the sample depends on both the x - ray luminosity and redshift , and is thus no longer independent of the dependent variable . such data sets are said to be ` truncated ' . if the selection function depends on @xmath57 , or if the measurement errors on @xmath57 and @xmath19 are not independent , one can not simply ignore the terms depending on @xmath145 , since the @xmath168 integrals in equation ( [ eq - mislik ] ) depend on @xmath6 . however , we can eliminate the dependence of equation ( [ eq - mislik ] ) on the unknown @xmath145 by applying a bayesian approach . the posterior distribution of @xmath169 and @xmath145 is related to the observed data likelihood function as @xmath170 , where @xmath171 is the prior distribution of @xmath172 . if we assume a uniform prior on @xmath173 and @xmath174 , then one can show ( e.g. , * ? ? ? * ) that the posterior distribution of @xmath6 and @xmath65 is @xmath175^{-n } \prod_{i=1}^n p(x_i , y_i\|\theta , \psi ) . \label{eq - trunclik}\ ] ] here , @xmath176 is given by equation ( [ eq - obslik2 ] ) , and @xmath177 is the probability of including a source in one s sample , given the model parameters , @xmath6 and @xmath65 : @xmath178 i have left off the subscripts for the data points in equation ( [ eq - detprob ] ) because the integrals are the same for each @xmath179 . if one assumes the gaussian mixture model of sections [ s - normreg ] and [ s - multreg ] , then @xmath176 is given by equations ( [ eq - bobslik ] ) or ( [ eq - mobslik ] ) . the posterior mode can then be used as an estimate of @xmath6 and @xmath65 , which is found by maximizing equation ( [ eq - trunclik ] ) . in addition to issues related to the sample selection method , it is common in astronomical data to have non - detections . such non - detections are referred to as ` censored ' data , and the standard procedure is to place an upper and/or lower limit on the censored data point . methods of data analysis for censored data have been reviewed and proposed in the astronomical literature , ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and @xcite describe censored regression when the variables are measured without error . see @xcite for a review of censored data in astronomy . to facilitate the inclusion of censored data , i introduce an additional indicator variable , @xmath180 , indicating whether a data point is censored or not on the dependent variable . if @xmath41 is detected , then @xmath181 , else if @xmath41 is censored then @xmath182 . it is commonly the case that a source is considered ` detected ' if its measured flux falls above some multiple of the background noise level , say @xmath183 . then , in this case , the probability of detecting the source given the measured source flux @xmath41 is @xmath184 if @xmath185 , and @xmath186 if @xmath187 . since source detection depends on the measured flux , some sources with intrinsic flux @xmath28 above the flux limit will have a measured flux @xmath57 that falls below the flux limit . similarly , some sources with intrinsic flux below the flux limit will have a measured flux above the flux limit . i assume that a sample is selected based on the independent variables , i.e. , @xmath162 . it is difficult to imagine obtaining a censored sample if the sample is selected based on its dependent variable , as some of the values of @xmath57 are censored and thus unknown . therefore , i only investigate the effects of censoring on @xmath57 when the probability that a source is included in the sample is independent of @xmath57 , given @xmath19 . in addition , i do not address the issue of censoring on the independent variable . although such methods can be developed , it is probably simpler to just omit such data as inference on the regression parameters is unaffected when a sample is selected based only on the independent variables ( cf . , [ s - seffects_x ] ) . the observed data likelihood for an @xmath19-selected sample is given by equation ( [ eq - indlik ] ) . we can modify this likelihood to account for censored @xmath57 by including the indicator variable @xmath180 and again integrating over the missing data : @xmath188 here , the first product is over the set of data points with detections , @xmath189 , and the second product is over the set of data points with non - detections , @xmath190 . the conditional distribution @xmath191 and the marginal distribution @xmath192 for the gaussian mixture model are both given in [ s - normreg ] and [ s - multreg ] . if the data points are measured without error and one assumes the normal regression model , @xmath193 , then equation [ eq - censlik ] becomes the censored data likelihood function described in @xcite . a mle for censored regression with measurement errors is then obtained by maximizing equation ( [ eq - censlik ] ) . in this section i describe a bayesian method for computing estimates of the regression parameters , @xmath6 , and their uncertainties . the bayesian approach calculates the posterior probability distribution of the model parameters , given the observed data , and therefore is accurate for both small and large sample sizes . the posterior distribution follows from baye s formula as @xmath194 , where @xmath195 is the prior distribution of the parameters . i describe some markov chain methods for drawing random variables from the posterior , which can then be used to estimate quantities such as standard errors and confidence intervals on @xmath6 and @xmath65 . @xcite is a good reference on bayesian methods , and @xcite gives a review of bayesian methods intended for astronomers . further details of markov chain simulation , including methods for making the simulations more efficient , can be found in @xcite . in order to ensure a proper posterior for the gaussian mixture model , it is necessary to invoke a proper prior density on the mixture parameters @xcite . i adopt a uniform prior on the regression parameters @xmath196 , and take @xmath197 . the dirichlet density is a multivariate extension of the beta density , and the @xmath198 prior adopted in this work is equivalent to a uniform prior on @xmath95 , under the constraint @xmath199 . the prior on @xmath11 and @xmath166 ( or @xmath167 ) adopted in this work is very similar to that advocated by @xcite and @xcite . i adopt a normal prior on the individual @xmath200 with mean @xmath201 and variance @xmath202 ( or covariance matrix @xmath203 ) . this reflects our prior belief that the distribution of @xmath27 is more likely to be fairly unimodal , and thus that we expect it to be more likely that the individual gaussians will be close together than far apart . if there is only one covariate , then i adopt a scaled inverse-@xmath1 prior on the individual @xmath204 with scale parameter @xmath205 and one degree of freedom , otherwise if there are @xmath206 covariates i adopt an inverse - wishart prior on the individual @xmath207 with scale matrix @xmath208 and @xmath15 degrees of freedom . this reflects our prior expectation that the variances for the individual gaussian components should be similar , but the low number of degrees of freedom accomodates a large range of scales . both the gaussian means and variances are assumed to be independent in their prior distribution , and the ` hyper - parameters ' @xmath209 ( or @xmath203 ) , and @xmath205 ( or @xmath208 ) are left unspecified . by leaving the parameters for the prior distribution unspecified , they becomes additional parameters in the statistical model , and therefore are able to adapt to the data . since the hyper - parameters are left as free parameters they also require a prior density . i assume a uniform prior on @xmath201 and @xmath205 ( or @xmath208 ) . if there is one covariate , then i assume a scaled inverse-@xmath1 prior for @xmath202 with scale parameter @xmath205 and one degree of freedom , otherwise if there are multiple covariate we assume a inverse - wishart prior for @xmath203 with scale matrix @xmath208 and @xmath15 degrees of freedom . the prior on @xmath202 ( or @xmath203 ) reflects the prior expectation that the dispersion of the gaussian components about their mean @xmath201 should be on the order of the typical dispersion of each individual gaussian . the prior density for one covariate is then @xmath210 and is summarized hierarchically as @xmath211 the prior density for multiple covariates is just the multivariate extension of equations ( [ eq - coefprior])([eq - wsqrprior ] ) . the posterior distribution summarizes our knowledge about the parameters in the statistical model , given the observed data and the priors . direct computation of the posterior distribution is too computationally intensive for the model described in this work . however , we can obtain any number of random draws from the posterior using markov chain monte carlo ( mcmc ) methods . in mcmc methods , we simulate a markov chain that performs a random walk through the parameter space , saving the locations of the walk at each iteration . eventually , the markov chain converges to the posterior distribution , and the saved parameter values can be treated as a random draw from the posterior . the random draws can then be used to estimate posterior medians , standard errors , of plot histogram estimates of the posterior . the easiest method for sampling from the posterior is to construct a gibbs sampler . the basic idea behind the gibbs sampler is to construct a markov chain , where new values of the model parameters and missing data are simulated at each iteration , conditional on the values of the observed data and the current values of the model parameters and the missing data . within the context of the measurement error model considered in this work , the gibbs sampler undergoes four different stages . the first stage of the gibbs sampler simulates values of the missing data , given the measured data and current parameter values , a process known as data augmentation . in this work the missing data are @xmath212 and any non - detections . in addition , i introduce an additional latent variable , @xmath213 , which gives the class membership for the @xmath93 data point . the vector @xmath213 has @xmath85 elements , where @xmath214 if the @xmath93 data point comes from the @xmath89 gaussian , and @xmath215 if @xmath216 . i will use @xmath217 to refer to the set of @xmath29 vectors @xmath218 . noting that @xmath88 gives the probability of drawing a data point from the @xmath89 gaussian , the mixture model for @xmath27 may then be expressed hierarchically as @xmath219 here , @xmath220 is a multinomial distribution with @xmath221 trials , where @xmath222 is the probability of success for the @xmath89 class on any particular trial . the vector @xmath213 is also considered to be missing data , and is introduced to simplify construction of the gibbs sampler . the new values of the missing data simulated in the data augmentation step are then used to simulate new values of the regression and gaussian mixture parameters . the second stage of the gibbs sampler simulates values of the regression parameters , @xmath6 , given the current values of @xmath223 and @xmath28 . the third stage simulates values of the mixture parameters , @xmath65 , given the current values of @xmath27 and @xmath28 . the fourth stage uses the new values of @xmath6 and @xmath65 to update the parameters of the prior density . the values of the parameters are saved , and the process is repeated , creating a markov chain . after a large number of iterations , the markov chain converges , and the saved values of @xmath6 and @xmath65 from the latter part of the algorithm may then be treated as a random draw from the posterior distribution , @xmath224 . methods for simulating random variables from the distributions used for this gibbs sampler are described in various works ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? a gibbs sampler for the gaussian mixture model is 1 . [ i - gstart ] start with initial guesses for @xmath225 and the prior parameters [ i - ycens ] if there are any non - detections , then draw @xmath41 for the censored data points from @xmath226 . this may be done by first drawing @xmath41 from @xmath227 : @xmath228 one then draws a random variable @xmath229 , uniformly - distributed on @xmath230 $ ] . if @xmath231 then the value of @xmath41 is kept , otherwise one draws a new value of @xmath41 and @xmath229 until @xmath232 . [ i - xi ] draw values of @xmath27 from @xmath233 . the distribution @xmath233 can be derived from equations ( [ eq - hierxi])([eq - hierxy ] ) or ( [ eq - mhierxi])([eq - mhierxy ] ) and the properties of the multivariate normal distribution : 1 . [ i - xi1 ] if there is only one independent variable then @xmath125 is updated as : @xmath234 \label{eq - gxihatk } \\ \hat{\xi}_{xy , i } & = & x_i + \frac{\sigma_{xy , i}}{\sigma^2_{y , i } } ( \eta_i - y_i ) \label{eq - gxixy } \\ \sigma^2_{\hat{\xi},i } & = & \sum_{k=1}^{k } g_{ik } \sigma^2_{\hat{\xi},ik } \label{eq - gxihvark } \\ \sigma^2_{\hat{\xi},ik } & = & \left [ \frac{1}{\sigma^2_{x , i}(1 - \rho^2_{xy , i } ) } + \frac{\beta^2}{\sigma^2 } + \frac{1}{\tau_k^2 } \right]^{-1}. \label{eq - gxihvar } \end{aligned}\ ] ] here , @xmath235 is the correlation between the measurement errors on @xmath40 and @xmath41 . note that @xmath125 is updated using only information from the @xmath89 gaussian , since @xmath236 only for @xmath237 and @xmath215 otherwise . [ i - xi2 ] if there are multiple independent variables , i have found it easier and computationally faster to update the values of @xmath125 using a scalar gibbs sampler . in this case , the @xmath15 elements of @xmath125 are updated individually . i denote @xmath238 to be the value of the @xmath239 independent variable for the @xmath93 data point , and @xmath240 to be the measured value of @xmath238 . in addition , i denote @xmath241 to be the @xmath242-element vector obtained by removing @xmath238 from @xmath125 , i.e. , @xmath243 . similarly , @xmath244 denotes the @xmath242-element vector of regression coefficients obtained after removing @xmath245 from @xmath115 . then , @xmath238 is updated as @xmath246^{-1}. \label{eq - gmxihvark } \end{aligned}\ ] ] here , @xmath247 is a @xmath136-element vector obtained by subtracting @xmath248 from @xmath249 , with the exception of the @xmath239 element of @xmath125 ; instead , the @xmath250 element of @xmath251 is just @xmath240 . the @xmath15-element vector @xmath252 is obtained in an equivalent manner . the @xmath140 matrix @xmath79 is the covariance matrix of the measurement errors on @xmath253 . the term @xmath254 denotes the @xmath250 element of the vector @xmath255 , and likewise for @xmath256 . the terms @xmath257 and @xmath258 denote the @xmath250 and @xmath239 elements of the diagonals of @xmath259 and @xmath260 , respectively . this step is repeated until all @xmath15 independent variables have been updated for each data point . + if any of the @xmath125 are measured without error , then one simply sets @xmath261 for those data points . 4 . [ i - eta ] draw values of @xmath28 from @xmath262 . similar to @xmath27 , the distribution @xmath262 can be derived from equations ( [ eq - hierxi])([eq - hierxy ] ) or ( [ eq - mhierxi])([eq - mhierxy ] ) and the properties of the multivariate normal distribution . 1 . [ i - eta1 ] if there is only one covariate then @xmath28 is updated as @xmath263 \label{eq - gmetahat } \\ \sigma^2_{\hat{\eta},i } & = & \left [ \frac{1}{\sigma^2_{y , i}(1 - \rho^2_{xy , i } ) } + \frac{1}{\sigma^2 } \right]^{-1}. \label{eq - gmetahvar } \end{aligned}\ ] ] 2 . [ i - eta2 ] if there are multiple covariates then @xmath28 is updated as @xmath264^{-1 } \label{eq - gmetahvar2 } \\ { \bf z}^_i & = & ( y_i , { \bf x}_i - \xi_i ) . \label{eq - zstar } \end{aligned}\ ] ] here , @xmath265 is the first element of the vector @xmath266 , @xmath247 is a @xmath136-element vector whose first element is @xmath41 and remaining elements are @xmath267 , and @xmath268 is the first diagonal element of @xmath259 . + if any of the @xmath28 are measured without error , then one sets @xmath269 for those data points . 5 . [ i - glabel ] draw new values of the gaussian labels , @xmath217 . the conditional distribution of @xmath213 is multinomial with number of trials @xmath270 and group probabilities @xmath271 : @xmath272 note that if there is only one covariate then @xmath273 and @xmath274 . [ i - coef ] draw @xmath275 from @xmath276 . given @xmath277 and @xmath12 , the distribution of @xmath123 and @xmath115 is obtained by ordinary regression : @xmath278 here , @xmath279 is a @xmath280 matrix , where the first column is a column of ones , the second column contains the @xmath29 values of @xmath125 for the first independent variable , the third column contains the @xmath29 values of @xmath125 for the second independent variable , etc . 7 . [ i - sigsqr ] draw a new value of @xmath12 from @xmath281 . the distribution @xmath281 is derived by noting that given @xmath282 and @xmath125 , @xmath32 is normally distributed with mean @xmath283 and variance @xmath12 . re - expressing this distribution in terms of @xmath12 instead of @xmath28 , and taking the product of the distributions for each data point , it follows that @xmath12 has a scaled inverse-@xmath1 distribution : @xmath284 8 . [ i - pi ] draw new values of the group proportions , @xmath95 . given @xmath217 , @xmath95 follows a dirichlet distribution : @xmath285 note that @xmath286 is the number of data points that belong to the @xmath89 gaussian . [ i - mu ] draw a new value of @xmath200 from @xmath287 . if there is only one independent variable , then @xmath288 and @xmath289 . the new value of @xmath200 is simulated as @xmath290 10 . [ i - tausqr ] draw a new value of @xmath204 or @xmath207 . the distribution of @xmath291 or @xmath292 is derived in a manner similar to @xmath293 , and noting that the prior is conjugate for this likelihood . the distribution of @xmath294 is a scaled inverse-@xmath1 distribution , and the distribution of @xmath292 is an inverse - wishart distribution : 1 . [ i - tsqr ] if there is only one independent variable then draw @xmath295 . \label{eq - tsqr } \end{aligned}\ ] ] 2 . [ i - tcovar ] if there are multiple independent variables then draw @xmath296 11 . [ i - mu0 ] draw a new value for @xmath297 . noting that conditional on @xmath201 and @xmath203 , @xmath298 are independently distributed as @xmath299 , it is straight - forward to show that @xmath300 if there is only one covariate then @xmath301 and @xmath289 . 12 . [ i - usqr0 ] draw a new value for @xmath202 or @xmath203 , given @xmath302 and @xmath205 ( or @xmath208 ) . similar to the case for @xmath303 or @xmath207 , the conditional distribution of @xmath202 or @xmath203 is scaled inverse-@xmath1 or inverse - wishart [ i - usqr ] if there is only one covariate then @xmath304 . \label{eq - usqrhat } \end{aligned}\ ] ] 2 . [ i - umat ] if there are multiple covariates then @xmath305 13 . [ i - wsqr0 ] finally , draw a new value of @xmath306 or @xmath307 : 1 . [ i - wsqr ] if there is only one covariate then @xmath306 is drawn from a gamma distribution . this can be derived by noting that @xmath308 has the form of a gamma distribution as a function of @xmath205 . the new value of @xmath205 is then simulated as @xmath309 . \label{eq - b } \end{aligned}\ ] ] 2 . [ i - wmat ] if there are multiple covariates then @xmath307 is drawn from a wishart distribution . this can be derived by noting that @xmath310 has the form of a wishart distribution as a function of @xmath208 . the new value of @xmath208 is then simulated as @xmath311 after completing steps [ i - ycens][i - wsqr0 ] above , an iteration of the gibbs sampler is complete . one then uses the new simulated values of @xmath312 and the prior parameters , and repeats steps [ i - ycens][i - wsqr0 ] . the algorithm is repeated until convergence , and the values of @xmath6 and @xmath65 at each iteration are saved . upon reaching convergence , one discards the values of @xmath6 and @xmath65 from the beginning of the simulation , and the remaining values of @xmath313 and @xmath166 ( or @xmath167 ) may be treated as a random draw from the posterior distribution , @xmath224 . one can then use these values to calculate estimates of the parameters , and their corresponding variances and confidence intervals . the posterior distribution of the parameters can also be estimated from these values of @xmath6 and @xmath65 using histogram techniques . techniques for monitering convergence of the markov chains can be found in @xcite . the output from the gibbs sampler may be used to perform bayesian inference on other quantities of interest . in particular , the pearson linear correlation coefficient , @xmath314 , is often used in assessing the strength of a relationship between the @xmath19 and @xmath57 . a random draw from the posterior distribution for the correlation between @xmath28 and @xmath315 , denoted as @xmath316 , can be calculated from equation ( [ eq - rho ] ) for each draw from the gibbs sampler . for the gaussian mixture model , the variance @xmath317 and covariance matrix @xmath318 are @xmath319 and @xmath320 is the @xmath239 diagonal element of @xmath321 . the simplification for one covariate is self - evident . if there is considerable posterior probability near @xmath322 or @xmath323 , then the gibbs sampler can get ` stuck ' . for example , if @xmath323 , then step [ i - xi1 ] of the gibbs sampler will draw values of @xmath324 . then , step [ i - mu ] will produce a new value of @xmath200 that is almost identical to the previous iteration , step [ i - tsqr ] will produce a new value of @xmath323 , and so on . the gibbs sampler will eventually get ` unstuck ' , but this can take a long time and result in very slow convergence . in particular , it is very easy for the gibbs sampler to get stuck if the measurement errors are large relative to @xmath12 or @xmath204 , or if the number of data points is small . in this situation i have found it useful to use the metropolis - hastings algorithm instead . if the selection function is not independent of @xmath57 , given the independent variables ( cf . eq.[[eq - trunclik ] ] ) , or if the selection function depends on @xmath19 and the measurement errors are correlated , then posterior simulation based on the gibbs sampler is more complicated . in addition , if the measurement errors are large compared to the intrinsic dispersion in the data , or if the sample size is small , then the gibbs sampler can become stuck and extremely inefficient . in both of these cases one can use the metropolis - hastings algorithm @xcite to sample from the posterior distribution , as the metropolis - hasting algorithm can avoid constructing markov chains for @xmath27 and @xmath28 . for a description of the metropolis - hastings algorithm , we refer the reader to @xcite or @xcite . in this section i perform simulations to illustrate the effectiveness of the gaussian structural model for estimating the regression parameters , even in the presence of severe measurement error and censoring . in addition , i compare the ols , bces(@xmath2 ) , and fitexy estimators with a maximum - likelihood estimator based on the gaussian mixture model with @xmath112 gaussian . the first simulation i performed is for a simple regression with one independent variable . i generated @xmath325 data sets by first drawing @xmath29 values of the independent variable , @xmath27 , from a distribution of the form @xmath326 the distribution of @xmath27 is shown in figure [ f - xidist ] , along with the best - fitting one and two gaussian approximations . in this case the two gaussian mixture is nearly indistinguishable from the actual distribution of @xmath27 , and thus should provide an excellent approximation to @xmath102 . the values for @xmath27 had a mean of @xmath327 and a dispersion of @xmath328 . i varied the number of data points in the simulated data sets as @xmath329 and @xmath330 . i then simulated values of @xmath28 according to equation ( [ eq - adderr ] ) , with @xmath331 and @xmath332 . the intrinsic scatter , @xmath34 , had a normal distribution with mean zero and standard deviation @xmath333 , and the correlation between @xmath28 and @xmath27 was @xmath334 . the joint distribution of @xmath27 and @xmath28 for one simulated data set with @xmath335 is shown in figure [ f - simdist1 ] . measured values for @xmath27 and @xmath28 were simulated according to equations ( [ eq - xerr ] ) and ( [ eq - yerr ] ) . the measurement errors had a zero mean normal distribution of varying dispersion and were independent for @xmath19 and @xmath57 . the variances in the measurement errors , @xmath336 and @xmath43 , were different for each data point and drawn from a scaled inverse-@xmath1 distribution . the degrees of freedom for the inverse-@xmath1 distribution was @xmath337 , and the scale parameters are denoted as @xmath103 and @xmath338 for the @xmath19 and @xmath57 measurement error variances , respectively . the scale parameters dictate the typical size of the measurements errors , and were varied as @xmath339 and @xmath340 . these values corresponded to values of @xmath341 and @xmath342 respectively . i simulated @xmath343 data sets for each grid point of @xmath344 and @xmath29 , giving a total of @xmath325 simulated data sets . the joint distributions of @xmath19 and @xmath57 for varying values of @xmath345 and @xmath346 are also shown in figure [ f - simdist1 ] . these values of @xmath19 and @xmath57 are the ` measured ' values of the simulated data set shown in the plot of @xmath28 as a function of @xmath27 . for each simulated data set , i calculated the maximum - likelihood estimate , found by maximizing equation ( [ eq - bobslik ] ) . for simplicity , i only use @xmath112 gaussian . i also calculated the ols , bces(@xmath2 ) , and fitexy estimates for comparison . i calculated a ols estimate of @xmath12 by subtracting the average @xmath122 from the variance in the regression residuals . if the ols estimate of @xmath12 was negative , i set @xmath347 . following @xcite , i estimate @xmath12 for a bces(@xmath2)-type estimator as @xmath348 , where @xmath349 is the average measurement error variance in @xmath57 , and @xmath350 is the bces(@xmath2 ) estimate of the slope . if @xmath351 is negative , i set @xmath352 . following @xcite , i compute a fitexy estimate of @xmath353 by increasing @xmath12 until @xmath117 , or assume @xmath118 if @xmath119 . the sampling distributions of the slope and intrinsic scatter estimators for @xmath354 are shown in figures [ f - sampdist1 ] and [ f - sampdist2 ] as a function of @xmath355 , and the results of the simulations are summarized in table [ t - univest ] . the bias of the ols estimate is apparent , becoming more severe as the measurement errors in the independent variable increase . in addition , the variance in the ols slope estimate decreases as the measurement errors in @xmath27 increase , giving one the false impression that one s estimate of the slope is more precise when the measurement errors are large . this has the effect of concentrating the ols estimate of @xmath115 around @xmath356 , thus effectively erasing any evidence of a relationship between the two variables . when the measurement errors are large , the ols estimate of the intrinsic scatter , @xmath357 , is occasionally zero . the bces(@xmath2 ) estimator performs better than the ols and fitexy estimators , being approximately unbiased when the measurement errors are @xmath358 . however , the bces estimate of the slope , @xmath359 , suffers some bias when the measurement errors are large and/or the sample size is small . in addition , the variance in @xmath360 is larger than the mle , and @xmath360 becomes considerably unstable when the measurement errors on @xmath27 are large . this instability results because the denominator in the equation for @xmath350 is @xmath361 . if @xmath362 , then the denominator is @xmath363 , and @xmath364 can become very large . similar to the ols and fitexy estimates , the estimate of the intrinsic variance for the bces - type estimator is often zero when the measurement errors are large , suggesting the false conclusion that there is no intrinsic scatter about the regression line . the fitexy estimator performed poorly in the simulations , being both biased and highly variable . the bias of the fitexy estimator is such that @xmath365 tends to overestimate @xmath115 , the severity of which tends to increase as @xmath60 decreases . this upward bias in @xmath365 has been noted by @xcite , who also performed simulations comparing @xmath365 with @xmath360 . they note that when one minimizes @xmath114 alternatively with respect to @xmath115 and @xmath12 , and iterates until convergence , then the bias in @xmath365 can be improved . i have tested this and also find that the bias in @xmath365 is reduced , but at the cost of a considerable increase in variance in @xmath365 . in general , our simulations imply that the variance of the fitexy estimator is comparable to that of the bces(@xmath2 ) estimator if one does not iterate the minimization of @xmath114 , and the variance of @xmath365 is larger if one does iterate . however , since @xmath360 is approximately unbiased when @xmath59 is not too large , @xmath360 should be preferred over @xmath365 . in addition , when the measurement errors are large the fitexy estimate of @xmath353 is commonly @xmath366 , similar to the bces - type estimate of the intrinsic dispersion . the maximum - likelihood estimator based on the gaussian structural model performs better than the ols , bces , and fitexy estimators , and gives fairly consistent estimates even in the presence of severe measurement error and low sample size . the mle is approximately unbiased , in spite of the fact that the mle incorrectly assumes that the independent variables are normally distributed . the variance in the mle of the slope , @xmath367 , is smaller than that of @xmath360 and @xmath365 , particularly when @xmath59 is large . in contrast to the ols estimate of the slope , the dispersion in @xmath367 increases as the measurement errors increases , reflecting the additional uncertainty in @xmath367 caused by the measurement errors . finally , in contrast to the other estimators , the mle of the intrinsic variance is always positive , and the probability of obtaining @xmath368 is negligible for these simulations . i argued in [ s - normreg ] that assuming a uniform distribution on @xmath27 does not lead to better estimates than the usual ols case . i also used these simulations to estimate the sampling density of the mle assuming @xmath110 . the results were nearly indistinguishable from the ols estimator , supporting our conjecture that assuming @xmath110 does not offer an improvement over ols . while it is informative to compare the sampling distribution of our proposed maximum - likelihood estimator with those of the ols , bces(@xmath2 ) , and fitexy estimators , i do not derive the uncertainties in the regression parameters from the sampling distribution of the mle . as described in [ s - markov ] , we derive the uncertainties in the regression parameters by simulating draws from the posterior distribution , @xmath224 . this allows a straight - forward method of interpreting the parameter uncertainties that does not rely on large - sample approximations , as the posterior distribution is the probability distribution of the parameters , given the observed data . the posterior distributions of @xmath369 and @xmath353 for a simulated data set with @xmath370 and @xmath371 is shown in figure [ f - post1d ] . when estimating these posteriors , i used @xmath372 gaussians in the mixture model . as can be seen from figure [ f - post1d ] , the true values of @xmath369 and @xmath353 are contained within the regions of non - negligible posterior probability . i have estimated posteriors for other simulated data sets , varying the number of data points and the degree of measurement error . as one would expect , the uncertainties in the regression parameters , represented by the widths of the posterior distributions , increase as the size of the measurement errors increase and the sample size decreases . a common frequentist approach is to compute the covariance matrix of the mle by inverting the estimated fisher information matrix , evaluated at the mle . then , under certain regularity conditions , the mle of the parameters is asympotically normally distributed with mean equal to the true value of the parameters and covariance matrix equal to the inverse of the fisher information matrix . furthermore , under these regularity conditions the posterior distribution and sampling distribution of the mle are asymptotically the same . figure [ f - fbcompare ] compares the posterior distribution of the slope for a simulated data set with that inferred from the mle . the posterior and mle was calculated assuming @xmath112 gaussian . as can be seen , the posterior distribution for @xmath115 is considerably different from the approximation based on the mle of @xmath115 , and thus the two have not converged for this sample . in particular , the posterior is more skewed and heavy - tailed , placing more probability on values of @xmath373 than does the distribution approximated by the mle . therefore , uncertainties in the mle should be interpreted with caution if using the asymptotic approximation to the sampling distribution of the mle . to assess the effectiveness of the gaussian structural model in dealing with censored data sets with measurement error , i introduced non - detections into the simulations . the simulations were performed in an identical manner as that described in [ s - sims1 ] , but now i only consider sources to be ` detected ' if @xmath374 . for those sources that were ` censored ' ( @xmath375 ) , i placed an upper limit on them of @xmath376 . i focus on the results for a simulated data set with @xmath377 data points and measurement errors similar to the intrinsic dispersion in the data , @xmath371 and @xmath378 . the detection threshold of @xmath374 resulted in a detection fraction of @xmath379 . this simulation represents a rather extreme case of large measurement errors and low detection fraction , and provides an interesting test of the method . in figure [ f - cens_reg ] i show the distribution of @xmath27 and @xmath28 , as well as the distribution of their measured values , for one of the simulated data sets . for this particular data set , there were 29 detections and 71 non - detections . as can be seen , the significant censoring and large measurement errors have effectively erased any visual evidence for a relationship between the two variables . i estimated the posterior distribution of the regression parameters for this data set using the gibbs sampler ( cf , [ s - gibbs ] ) with @xmath372 gaussians . the posterior median of the regression line , as well as the @xmath380 pointwise confidence intervals of the posterior probability . while the difference between confidence intervals and credibility intervals is not purely semantical , i do not find the difference to be significant within the context of my work , so i use the more familiar term ` confidence interval ' . ] on the regression line are shown in figure [ f - cens_reg ] . the posterior distributions for @xmath369 and @xmath353 are shown in figure [ f - censpost ] . as can be seen , the true value of the parameters is contained within the @xmath381 probability regions , although the uncertainty is large . for this particular data set , we can put limits on the value of the correlation coefficient as @xmath382 and the slope as @xmath383 . for comparison , the usual maximum - likelihood estimate that ignores the measurement error ( e.g. , ? ? ? * ) concludes @xmath384 . this estimate is biased and differs from the true value of @xmath115 at a level of @xmath385 . the posterior constraints on the regression parameters are broad , reflecting our considerable uncertainty in the slope , but they are sufficient for finding a positive correlation between the two variables , @xmath27 and @xmath28 . therefore , despite the high level of censoring and measurement error in this data set , we would still be able to conclude that @xmath28 increases as @xmath27 increases . to further illustrate the effectiveness of the method , i apply it to a data set drawn from my work on investigating the x - ray properties of radio - quiet quasars ( rqqs ) . recent work has suggested a correlation between quasar x - ray spectral slope , @xmath387 and quasar eddington ratio , @xmath386 ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in this section i apply the regression method to a sample of 39 @xmath4 rqqs and confirm the @xmath3@xmath386 correlation . because the purpose of this section is to illustrate the use of this regression method on real astronomical data , i defer a more in - depth analysis to a future paper . estimation of the eddington luminosity , @xmath388 , requires an estimate of the black hole mass , @xmath389 . black hole virial masses may be estimated as @xmath390 , where @xmath391 is the broad line region size , and @xmath392 is the velocity dispersion of the gas emitting the broad emission lines . a correlation has been found between the luminosity of a source and the size of it s broad line region ( the @xmath391@xmath393 relationship , e.g. , * ? ? ? one can then exploit this relationship , and use the broad line @xmath394 as an estimate for @xmath392 , obtaining virial mass estimates @xmath395 ( e.g. , * ? ? ? * ) , where the exponent is @xmath396 ( e.g. , * ? ? ? unfortunately , the uncertainty on the broad line estimates of @xmath389 can be considerable , having a standard deviation of @xmath397 dex ( e.g. , * ? ? * ; * ? ? ? * ; * ? ? ? * ) . for ease of comparison with previous work , i estimate @xmath389 using only the h@xmath115 emission line . the logarithm of the virial mass estimates were calculated using the h@xmath115 luminosity and @xmath394 according to the relationship given by @xcite . my sample consists of a subset of the sample of @xcite . these sources have measurements of the x - ray photon index , @xmath398 , obtained from _ chandra _ observations , and measurements of the optical / uv luminosity at @xmath399 , denoted as @xmath400 , obtained from sdss spectra . the h@xmath115 profile was modeled as a sum of gaussians and extracted from the sdss spectra according to the procedure described in @xcite . i estimated the h@xmath115 @xmath394 and luminosity from the line profile fits . i estimate the bolometric luminosity , @xmath401 , from the luminosity at @xmath399 , assuming a constant bolometric correction @xmath402 @xcite . the standard deviation in this bolometric correction reported by @xcite is 3.1 , implying an uncertainty in @xmath403 of @xmath404 dex . combining this with the @xmath405 dex uncertainty on @xmath406 , the total ` measurement error ' on @xmath407 becomes @xmath408 dex . the distribution of @xmath3 as a function of @xmath409 is shown in figure [ f - gamx_eddrat ] . as can be seen , the measurement errors on both @xmath3 and @xmath409 are large and make a considerable contribution to the observed scatter in both variables , where @xmath410 and @xmath411 . therefore , we expect the measurement errors to have a significant effect on the correlation and regression analysis . i performed the regression assuming the linear form @xmath412 , and modelleling the intrinsic distribution of @xmath413 using @xmath372 gaussians . draws from the posterior were obtained using the gibbs sampler . the marginal posterior distributions for @xmath414 , and the correlation between @xmath3 and @xmath413 , @xmath314 , are shown in figure [ f - posthb ] , and the posterior median and @xmath381 ( @xmath415 ) pointwise intervals on the regression line are shown in figure [ f - gamx_eddrat ] . the posterior median estimate of the parameters are @xmath416 for the constant , @xmath417 for the slope , @xmath418 for the intrinsic scatter about the regression line , @xmath419 for the mean of @xmath409 , and @xmath420 dex for the dispersion in @xmath413 . here , i have used a robust estimate of the posterior standard deviation as an ` error bar ' on the parameters . these results imply that the observed scatter in @xmath409 is dominated by measurement error , @xmath421 , as expected from the large value of @xmath59 . for comparison , the bces(@xmath2 ) estimate of the slope is @xmath422 , the fitexy estimate is @xmath423 , and the ols estimate is @xmath424 ; the standard error on @xmath365 was estimated using bootstrapping . figure [ f - gamx_eddrat ] also compares the ols , bces , and fitexy best - fit lines with the posterior median estimate . the @xmath381 confidence region on the slope implied by the posterior draws is @xmath425 , whereas the approximate @xmath381 confidence region implied by the bces , fitexy , and ols standard errors are @xmath426 , @xmath427 , and @xmath428 , respectively . the ols and fitexy estimates and the bayesian approach give ` statistically significant ' evidence for a correlation between @xmath409 and @xmath3 ; however the bces estimate is too variable to rule out the null hypothesis of no correlation . as noted before , the large measurement errors on @xmath429 bias the ols estimate of @xmath115 toward shallower values and the fitexy estimate of @xmath115 toward steeper values . because of this bias , confidence regions based on @xmath47 and @xmath365 are not valid because they are not centered on the true value of @xmath115 , and thus do not contain the true value with the stated probability ( e.g. , @xmath381 ) . on the other hand , confidence regions based on the bces estimate are likely to be approximately valid ; however , in this example the large measurement errors have caused @xmath360 to be too variable to give meaningful constraints on the regression slope . the bces - type estimate of the intrinsic dispersion was @xmath430 and the ols estimate of the intrinsic dispersion was @xmath431 , where both were calculated in the same manner as in [ s - sims1 ] . the fitexy estimate of the intrinsic dispersion was @xmath366 , as @xmath119 . the bces - type estimate of @xmath353 is similar to the bayesian posterior median estimate , while @xmath432 overestimates the scatter compared to the bayesian estimate by @xmath433 . in contrast , the fitexy estimator does not find any evidence for intrinsic scatter in the regression , which is inconsistent with the posterior distribution of @xmath353 . from the posterior distribution , we can constrain the correlation between @xmath3 and @xmath409 to be @xmath434 with @xmath435 probability , confirming the positive correlation between @xmath3 and eddington ratio seen previously . the posterior median estimate of the correlation is @xmath436 , compared with an estimate of @xmath437 if one naively calculates the correlation directly from the measured data . the large measurement errors significantly attenuate the observed correlation , making the observed correlation between @xmath3 and @xmath413 appear weaker than if one does not correct for the measurement errors . in this work i have derived a likelihood function for handling measurement errors in linear regression of astronomical data . our probability model assumes that the measurement errors are gaussian with zero mean and known variance , that the intrinsic scatter in the dependent variable about the regression line is gaussian , and that the intrinsic distribution of the independent variables can be well approximated as a mixture of gaussians . i extend this model to enable the inclusion of non - detections , and describe how to incorporate the data selection process . a gibbs sampler is described to enable simulating random draws from the posterior distribution . i illustrated the effectiveness of structural gaussian mixture model using simulation . for the specific simulations performed , a maximum - likelihood estimator based on the gaussian structural model performed better than the ols , bces(@xmath2 ) , and fitexy estimators , especially when the measurement errors were large . in addition , our method also performed well when the measurement errors were large and the detection fraction was small , with the posterior distributions giving reasonable bounds on the regression parameters . these results were in spite of the fact that the intrinsic distribution of the independent variable was not a sum of gaussians for the simulations , suggesting that approximating the distribution of the independent variable as a mixture of gaussians does not lead to a significant bias in the results . finally , i concluded by using the method to fit the radio - quiet quasar x - ray photon index as a function of @xmath409 , using a sample of 39 @xmath4 sources . the posterior distribution for this data set constrained the slope to be @xmath438 and the linear correlation coefficient to be @xmath439 , confirming the correlation between x - ray spectral slope and eddington ratio seen by other authors . although i have focused on linear regression in this work , the approach that i have taken is quite general and can be applied to other applications . in particular , equations ( [ eq - obslik2 ] ) , ( [ eq - trunclik ] ) , and ( [ eq - censlik ] ) are derived under general conditions and are not limited to regression . in this work , i assume forms for the respective probability densities that are appropriate for linear regression ; however , these equation provide a framework for constructing more general probability models of one s data , as in , for example , nonlinear fitting * ? ? ? * or estimation of luminosity functions . this work was supported in part by nsf grant ast-0307384 . the author would like to thank the referee for comments that contributed to the improvement of this paper , and for providing some of the references to the statistics literature . the author would also like to thank jill bechtold , eric feigelson , and aneta siemiginowska for looking over and offering helpful comments on an early version of this paper . aitken , m. , & rocci , r. , 2002 , statistics and computing , 12 , 163 akritas , m. g. , & bershady , m. a. 1996 , , 470 , 706 akritas , m. g. , & siebert , j. 1996 , , 278 , 919 barker , d. r. , & diana , l. m. 1974 , am . j. phys . , 42 , 224 carroll , r. j. , roeder , k. , & wasserman , l. , 1999 , biometrics , 55 , 44 carroll , r. j. , ruppert , d. , & stefanski , l. a. , 1995 , measurement error in nonlinear models ( london : chapman & hall ) chib , s. , & greenberg , e. 1995 , amer . stat . , 49 , 327 clutton - brock , m. 1967 , technometrics , 9 , 261 davison , a. c. , & hinkley , d. v. 1997 , bootstrap methods and their application ( cambridge : cambridge university press ) dellaportas , p. , stephens , d. a. , 1995 , biometrics , 51 , 1085 dempster , a. , laird , n. , & rubin , d. 1977 , j. r. statist . b. , 39 , 1 efron , b. 1979 , ann . , 7 , 1 elvis , m. , et al . 1994 , , 95 , 1 feigelson , e. d. , & nelson , p. i. 1985 , , 293 , 192 feigelson , e. d. , 1992 , in statistical challenges in modern astronomy , ed . e. feigelson & g. babu , ( new york : springer - verlag ) fox , j. 1997 , applied regression analysis , linear models , and related methods ( thousand oaks : sage publications , inc . ) freedman , l. s. , fainberg , v. , kipnis , v. , midthune , d. , & carrol , r. j. , 2004 , biometrics , 60 , 172 fuller , w. a. 1987 , measurement error models ( new york : john wiley & sons ) gelman , a. , carlin , j. b. , stern , h. s. , & rubin , d. b. 2004 , bayesian data analysis ( 2nd ed . ; boca raton : chapman & hall / crc ) gull , s. f. 1989 , in maximum entropy and bayesian methods , ed . j. skilling , ( dordrecht : kluwer academic publishers ) 511 hastings , w. k. 1970 , biometrika , 57 , 97 huang , x. , stefanski , l. a. , & davidian , m. 2006 , biometrika , 93 , 53 isobe , t. , feigelson , e. d. , & nelson , p. i. 1986 , , 306 , 490 isobe , t. , feigelson , e. d. , akritas , m. g. , & babu , g. j. 1990 , , 364 , 104 kaspi , s. , maoz , d. , netzer , h. , peterson , b. m. , vestergaard , m. , & jannuzi , b. t. 2005 , , 629 , 61 kelly , b. c. , bechtold , j. , siemiginowska , a. , aldcroft , t. , & sobolewska , m. 2007 , , 657 , 116 kelly , b. c. , & bechtold , j. 2007 , , 168 , 1 landy , s. d. , & szalay , a. s. 1992 , , 391 , 494 little , r. j. a. , & rubin , d. b. 2002 , statistical analysis with missing data ( 2nd ed . ; hoboken : john wiley & sons ) loredo , t. j. , 1992 , in statistical challenges in modern astronomy , ed . e. feigelson & g. babu , ( new york : springer - verlag ) marshall , h. l. , 1992 , in statistical challenges in modern astronomy , ed . e. feigelson & g. babu , ( new york : springer - verlag ) mclure , r. j. , & jarvis , m. j. 2002 , , 337 , 109 metropolis , n. , & ulam , s. 1949 , j. amer . assoc . , 44 , 335 metropolis , n. , rosenbluth , a. w. , rosenbluth , m. n. , teller , a. h. , & teller , e. 1953 , j. chem . phys . , 21 , 1087 mller , p. , & roeder , k. , 1997 , biometrika , 84 , 523 piconcelli , e. , jimenez - bailn , e. , guainazzi , m. , schartel , n. , rodrguez - pascual , p. m. , & santos - lle , m. 2005 , , 432 , 15 porquet , d. , reeves , j. n. , obrien , p. , & brinkmann , w. 2004 , , 422 , 85 press , w. h. , teukolsky , s. a. , vetterling , w. t. , & flannery , b. p. 1992 , numerical recipes ( second ed . ; cambridge : cambridge unv . press ) richardson , s. , & leblond , l. , 1997 , statistics in medicine , 16 , 203 richardson , s. , leblond , l. , jaussent , i. , & green , p. j. , 2002 , j. r. statist . a , 165 , 549 ripley , b. d. 1987 , stochastic simulation ( new york : john wiley & sons ) roeder , k. , & wasserman , l. 1997 , j. amer . assoc . , 92 , 894 roy , s. , banerjee , t. , 2006 , ann . instit . statist . math . , 58 , 153 schafer , d. w. , 1987 , 74 , 385 schafer , d. w. , 2001 , biometrics , 57 , 53 scheines , r. , hoijtink , h. , boomsma , a. , 1999 , psychometrika , 64 , 37 schmitt , j. h. m. m. 1985 , , 293 , 178 shemmer , o. , brandt , w. n. , netzer , h. , maiolino , r. , & kaspi , s. 2006 , , 646 , l29 stapleton , d. c. & young , d. j. 1984 , econometrica , 52 , 737 tremaine , s. , et al . 2002 , , 574 , 740 vestergaard , m. , & peterson , b. m. 2006 , , 641 , 689 wandel , a. , peterson , b. m. , & malkan , m. a. 1999 , , 526 , 579 weiner , b. j. , et al . 2006 , , 653 , 1049 weiss , a. a. , 1993 , j. econometrics . , 56 , 169 . zellner , a. 1971 , an introduction to bayesian inference in econometrics ( new york : john wiley & sons ) and @xmath27 ( left ) , and the measured values of @xmath57 and @xmath19 ( right ) , from a simulated censored data set of @xmath354 data points , @xmath378 , and @xmath371 ( cf . , [ s - sims2 ] ) . in the plot of @xmath28 and @xmath27 , the solid squares denote the values of @xmath27 and @xmath28 for the detected data points , and the hollow squares denote the values of @xmath27 and @xmath28 for the undetected data points . the solid line in both plots is the true regression line . in the plot of @xmath57 and @xmath19 , the squares denote the measured values of @xmath19 and @xmath57 for the detected data points , and the arrows denote the ` upper limits ' on @xmath57 for the undetected data points . the fictitious data point with error bars illustrates the median values of the error bars . the dashed - dotted line shows the best fit regression line , as calculated from the posterior median of @xmath123 and @xmath115 , and the filled region defines the approximate @xmath381 @xmath440 pointwise confidence intervals on the regression line . the true values of the regression line are contained within the @xmath381 confidence intervals.[f - cens_reg],title="fig : " ] and @xmath27 ( left ) , and the measured values of @xmath57 and @xmath19 ( right ) , from a simulated censored data set of @xmath354 data points , @xmath378 , and @xmath371 ( cf . , [ s - sims2 ] ) . in the plot of @xmath28 and @xmath27 , the solid squares denote the values of @xmath27 and @xmath28 for the detected data points , and the hollow squares denote the values of @xmath27 and @xmath28 for the undetected data points . the solid line in both plots is the true regression line . in the plot of @xmath57 and @xmath19 , the squares denote the measured values of @xmath19 and @xmath57 for the detected data points , and the arrows denote the ` upper limits ' on @xmath57 for the undetected data points . the fictitious data point with error bars illustrates the median values of the error bars . the dashed - dotted line shows the best fit regression line , as calculated from the posterior median of @xmath123 and @xmath115 , and the filled region defines the approximate @xmath381 @xmath440 pointwise confidence intervals on the regression line . the true values of the regression line are contained within the @xmath381 confidence intervals.[f - cens_reg],title="fig : " ] , as a function of @xmath409 for 39 @xmath0 radio - quiet quasars . in both plots the thick solid line shows the posterior median estimate ( pme ) of the regression line . in the left plot , the filled region denotes the @xmath381 @xmath440 pointwise confidence intervals on the regression line . in the right plot , the thin solid line shows the ols estimate , the dashed line shows the fitexy estimate , and the dot - dashed line shows the bces(@xmath2 ) estimate ; the error bars have been omitted for clarity . a significant positive trend is implied by the data.[f - gamx_eddrat],title="fig : " ] , as a function of @xmath409 for 39 @xmath0 radio - quiet quasars . in both plots the thick solid line shows the posterior median estimate ( pme ) of the regression line . in the left plot , the filled region denotes the @xmath381 @xmath440 pointwise confidence intervals on the regression line . in the right plot , the thin solid line shows the ols estimate , the dashed line shows the fitexy estimate , and the dot - dashed line shows the bces(@xmath2 ) estimate ; the error bars have been omitted for clarity . a significant positive trend is implied by the data.[f - gamx_eddrat],title="fig : " ] cccccccccc 0.5 & 25 & @xmath441 & @xmath442 & @xmath443 & @xmath444 & @xmath445 & @xmath446 & @xmath447 & @xmath448 + & 50 & @xmath449 & @xmath450 & @xmath451 & @xmath452 & @xmath453 & @xmath454 & @xmath455 & @xmath456 + & 100 & @xmath457 & @xmath458 & @xmath459 & @xmath460 & @xmath461 & @xmath462 & @xmath463 & @xmath464 + 1.0 & 25 & @xmath465 & @xmath466 & @xmath467 & @xmath468 & @xmath469 & @xmath470 & @xmath471 & @xmath472 + & 50 & @xmath473 & @xmath474 & @xmath475 & @xmath476 & @xmath477 & @xmath478 & @xmath479 & @xmath480 + & 100 & @xmath481 & @xmath482 & @xmath483 & @xmath484 & @xmath485 & @xmath486 & @xmath487 & @xmath488 + 2.0 & 25 & @xmath489 & @xmath490 & @xmath491 & @xmath492 & @xmath493 & @xmath494 & @xmath495 & @xmath496 + & 50 & @xmath497 & @xmath498 & @xmath499 & @xmath500 & @xmath501 & @xmath502 & @xmath503 & @xmath504 + & 100 & @xmath505 & @xmath506 & @xmath507 & @xmath508 & @xmath509 & @xmath510 & @xmath511 & @xmath512 +
relations between social actors can be studied by methods of social network analysis [ e.g. , wasserman and faust ( @xcite ) ; carrington , scott and wasserman ( @xcite ) ] . examples are friendship between pupils in a school class or alliances between firms . a basic data structure for social networks is the directed graph or digraph , where the actors are represented by the nodes , and the arcs between the nodes indicate the social ties . social network analysis traditionally has had a focus on rich description of network data , but the recent development of methods of statistical inference for network data [ e.g. , airoldi et al . ( @xcite ) ; hunter and handcock ( @xcite ) ] has the potential of moving this field toward a wider use of inferential approaches . longitudinal studies are especially important for obtaining insight into social networks , but statistical methods for longitudinal network data that are versatile enough for realistic modeling are only just beginning to be developed . this article considers repeated observations of a relation , or network , on a given set of actors @xmath0 , observed according to a panel design , and represented as a sequence of digraphs @xmath1 for @xmath2 , where @xmath3 are the observation moments . the nodes represent social actors ( which may be individuals , companies , etc . ) , and the node set @xmath4 is the same for all observation moments . a digraph is defined here as an irreflexive relation , that is , a subset @xmath5 of @xmath6 , and when @xmath7 we shall say that there is an arc , or a tie , from @xmath8 to @xmath9 . it often is realistic to assume that the social network has been developing between the observation moments , which leads to the assumption that the observations @xmath1 are realizations of stochastic digraphs @xmath10 embedded in a continuous - time stochastic process @xmath11 , @xmath12 . holland and leinhardt ( @xcite ) proposed to use continuous - time markov chains , defined on the space of all digraphs with a given node set , for modeling social network dynamics even if the observations are made at a few discrete time points and not continuously . continuous - time markov chains provide a natural starting point for modeling longitudinal network data . wasserman ( @xcite , @xcite ) and leenders ( @xcite ) elaborated _ dyad - independent models _ , where the ties between pairs of actors ( _ dyads _ ) develop according to processes that are mutually independent between dyads . this is not realistic for social processes , because dependence between the set of ties among three or more actors can be very strong , as was found already by davis ( @xcite ) who showed that a basic feature of many social networks is the tendency toward transitivity ( `` friends of my friends are my friends '' ) . snijders and van duijn ( @xcite ) and snijders ( @xcite ) proposed so - called _ actor - oriented models _ , explained in the next section , which do allow such higher - order dependencies . the actor - oriented models are too complicated for the calculation of likelihoods or estimators in closed form , but they represent stochastic processes which can be easily simulated . this was exploited by the estimation method proposed in the papers mentioned , which is a method of moments ( _ mom _ ) estimator implemented algorithmically by stochastic approximation [ robbins and monro ( @xcite ) ; also see , e.g. , kushner and yin ( @xcite ) ] . this estimator usually performs well [ some empirical applications are given in de federico ( @xcite ) , van de bunt , van duijn and snijders ( @xcite ) , and van duijn et al . ( @xcite ) ] . it is to be expected , however , that the statistical efficiency of the maximum likelihood ( _ ml _ ) estimator will be greater . ml estimation also paves the way for likelihood - based model selection , which will be a marked improvement on existing methods of model selection . this article presents an mcmc algorithm for approximating the ml estimator , combining and extending ideas in gu and kong ( @xcite ) , snijders ( @xcite ) , and koskinen and snijders ( @xcite ) , the latter of which proposed for this model a mcmc algorithm for bayesian inference . section [ sec2 ] presents the model definition . the algorithm for obtaining the ml estimator is described in section [ sec3 ] . section [ sec4 ] reports results of an empirical example and section [ sec5 ] presents a very small simulation study comparing the ml and mom estimators . the paper finishes with an algorithm for approximating the likelihood ratio test in section [ sec6 ] and a discussion in section [ sec7 ] . we assume that repeated observations @xmath13 on the network are available for some @xmath14 . the network , or digraph , @xmath5 will be identified with its @xmath15 adjacency matrix , of which the elements denote whether there is a tie from node @xmath8 to node @xmath9 ( @xmath16 ) or not ( @xmath17 ) . self - ties are not allowed , so that the diagonal is structurally zero . random variables are denoted by capitals . the stochastic process @xmath11 , in which the observations @xmath1 are embedded , is modeled as being right - continuous . various models have been proposed , most of them being markov processes of some kind . we focus on actor - oriented models [ snijders ( @xcite ) ] . tie - oriented models [ snijders ( @xcite ) ] can be treated similarly . the basic idea of actor - oriented models [ snijders ( @xcite ) ] is that the nodes of the graph represent social actors , who have control , albeit under constraints , of their outgoing ties ; and the graph develops as a continuous - time markov process ( even though it is observed only at @xmath18 discrete time points ) . the constraints are that ties may change only one by one , and actors do not coordinate their changes of ties . thus , at any given moment , one actor @xmath8 may create one new tie or delete one existing tie , where the probability distribution of such changes depends on the current digraph ; excluded are simultaneous changes such as swapping one tie for another , or bargaining between actors over ties . this constraint was proposed already by holland and leinhardt ( @xcite ) , and it has the virtue of splitting up the change process in its smallest possible constituents . the actor - oriented model is further specified as follows ; further discussion and motivation is given in snijders ( @xcite ) . 1 . _ opportunities for change _ each actor @xmath8 gets at stochastically determined moments the opportunity to change one of the outgoing tie variables @xmath19 ( @xmath20 , @xmath21 ) . since the process is assumed to be markovian , waiting times between opportunities have exponential distributions . each of the actors @xmath8 has a _ rate function _ @xmath22 which defines how quickly this actor gets an opportunity to change a tie variable , when the current value of the digraph is @xmath5 , and where @xmath23 is a parameter . at any time point @xmath24 with @xmath25 , the waiting time until the next opportunity for change by any actor is exponentially distributed with parameter @xmath26 given that an opportunity for change occurs , the probability that it is actor @xmath8 who gets the opportunity is given by @xmath27 the rate functions can be constant between observation moments , or they can depend on functions @xmath28 which may be covariates or positional characteristics of the actors such as outdegrees @xmath29 . a convenient assumption is to use an exponential link function , @xmath30 1 . _ options for change _ when actor @xmath8 gets the opportunity to make a change , this actor has a permitted set @xmath31 of values to which the digraph may be changed , where @xmath32 is the current value of the digraph . the assumption that the actor controls his or her outgoing ties , but can change only one tie variable at the time , is equivalent to [ ai ] @xmath33 where @xmath34 is the set of adjacency matrices differing from @xmath5 in exactly one element , @xmath35\\[-8pt ] & & \hspace{14pt}\mbox { and } x_{hk } = x^0_{hk } \mbox { for all other } ( h , k ) \ } .\nonumber\end{aligned}\ ] ] including @xmath32 in @xmath31 can be important for expressing the property that actors who are satisfied with the current network will prefer to keep it unchanged . therefore , the usual model is @xmath36 . this does not lead to identifiability problems because the ratio between not making a change and making a change is not a free parameter , but fixed by assumption ( [ choiceprob ] ) for the conditional probabilities of the new state of the network . some alternatives are models with structurally impossible ties , where the impossible digraphs are left out of @xmath31 , and models where the actor is required to make a change whenever there is the opportunity , obtained by leaving the current element @xmath32 out of @xmath31 . it is assumed that the network dynamics is driven by a so - called _ objective function _ @xmath37 that can be interpreted as the relative attractiveness for actor @xmath8 of moving from the network represented by @xmath32 to the network @xmath5 , and where @xmath38 is a parameter . under the condition that the current digraph is @xmath32 and actor @xmath8 gets the opportunity to make a change , the conditional probability that the next digraph is @xmath5 is modeled as @xmath39 where the summation extends over @xmath40 . this formula can be motivated by a random utility argument as used in econometrics [ see , e.g. , maddala ( @xcite ) ] , where it is assumed that the actor maximizes @xmath37 plus a random disturbance having a standard gumbel distribution . assumption ( [ ai ] ) implies that instead of @xmath41 we can also write @xmath42 where the correspondence between @xmath5 and @xmath9 is defined as follows : if @xmath43 , @xmath9 is the unique element of @xmath4 for which @xmath44 ; if @xmath45 , @xmath46 . this less redundant notation will be used in the sequel . thus , for @xmath21 , @xmath42 is the probability that , under the condition that actor @xmath8 has the opportunity to make a change and the current digraph is @xmath32 , the change will be to @xmath47 with the rest unchanged , while @xmath48 is the probability that , under the same condition , the digraph will not be changed . the most usual models are based on objective functions that depend on @xmath5 only . this has the interpretation that actors wish to maximize a function @xmath49 independently of `` where they come from . '' the greater generality of ( [ choiceprob ] ) , where the objective function can depend also on the previous state @xmath32 , makes it possible to model path - dependencies , or hysteresis , where the loss suffered from withdrawing a given tie differs from the gain from creating this tie , even if the rest of the network has remained unchanged . various ingredients for specifying the objective function were proposed in snijders ( @xcite ) . a linear form is convenient , @xmath50 where the functions @xmath51 are determined by subject - matter knowledge and available social scientific theory . these functions can represent essential aspects of the network structure , assessed from the point of view of actor @xmath8 , such as @xmath52 they can also depend on covariates such as resources and preferences of the actors , or costs of exchange between pairs of actors or combinations of network structure and covariates . for example , de federico ( @xcite ) found in a study of friendship between foreign exchange students that friendship formation tends to be reciprocal , with a negative parameter for forming indirect ties ( thus leading to relatively closed networks ) , and that friendships are more likely to be formed between individuals from the same region ( a covariate effect ) , but that reciprocation adds less to the tendency to form ties between persons from the same region than between arbitrary individuals ( negative interaction between covariate and reciprocity ) . 1 . _ intensity matrix ; time - homogeneity _ the model description given above defines @xmath11 as a continuous - time markov process with @xmath53-matrix or intensity matrix [ e.g. , norris ( @xcite ) ] for @xmath43 given by @xmath54 the assumptions do not imply that the distribution of @xmath11 is stationary . the intensity matrix is time - homogeneous , however , except for time dependence reflected by time - varying components in the functions @xmath51 . for the data - collection designs to which this paper is devoted , where observations are done at discrete time points @xmath55 , these time points can be used for marking time - heterogeneity of the transition distribution ( cf . the example in section [ sec6 ] ) . for example , covariates may be included with values allowed to change at the observation moments . a special role is played here by the time durations @xmath56 between successive observations . standard theory for continuous - time markov chains [ e.g. , norris ( @xcite ) ] shows that the matrix of transition probabilities from @xmath10 to @xmath57 is @xmath58 , where @xmath53 is the matrix with elements ( [ q ] ) . thus , changing the duration @xmath56 can be compensated by multiplication of the rate function @xmath59 by a constant . since the connection between an externally defined real - valued time variable @xmath60 and the rapidity of network change is tenuous at best , it usually is advisable [ e.g. , see snijders ( @xcite ) ] to include a multiplicative parameter in the rate function which is constant between observation moments @xmath60 but differs freely between periods @xmath61 . with the inclusion of such a parameter , the numerical values @xmath60 become unimportant because changes in their values can be compensated by the multiplicative rate parameters . popular models for analyzing cross - sectional network data are exponential families of graph distributions such as the markov model of frank and strauss ( @xcite ) , generalized to the @xmath62 exponential random graph ( erg ) model by frank ( @xcite ) and wasserman and pattison ( @xcite ) , and elaborated and further specified by hunter and handcock ( @xcite ) and snijders et al . ( @xcite ) . discrete - time markov chain models , as longitudinal models of this kind , were proposed by robins and pattison ( @xcite ) , krackhardt and handcock ( @xcite ) , and hanneke and xing ( @xcite ) . there are a number of essential differences between these models and the actor - oriented model treated here , with respect to interpretation as well as statistical procedures . when applying the actor - oriented model to a sequence of two or more repeated observations , these observations are embedded in a continuous - time model . this has clear face validity in applications where changes in the network take place at arbitrary moments between observations . singer ( @xcite ) gives an overview of the use of this principle for continuously distributed data . this approach yields a major advantage over the cited longitudinal erg models : our probability model is defined independently of the observational design . data from irregularly spaced observation intervals can be analyzed without the need to make any adaptations . another advantage is that the dynamic process is defined parsimoniously as a function of its elementary constituents in this case , the conditional probability of a single tie change . the random utility interpretation of ( [ choiceprob ] ) , discussed above , gives an interpretation in terms of myopic optimization of an objective function to which a random disturbance is added , and which can be used to represent a `` social mechanism '' that could have given rise to the observed dynamics . the discrete - time erg models constitute exponential families , which has the advantage that many standard techniques can be directly applied . the distribution of the continuous - time process @xmath63 for the actor - oriented model constitutes an exponential family , so for discrete observation moments our model can be regarded as an incompletely observed exponential family . for both types of model , inference is computer - intensive and time - consuming because of the need to implement elaborate mcmc procedures . the definition of the model given above can be used directly to simulate data from the probability distribution , conditional on an initial state of the network . this contrasts with models in the erg family , which can be simulated only indirectly , by regarding them as the stationary distribution of an auxiliary markov chain , and applying a gibbs or metropolis hastings algorithm . the near degeneracy problem [ snijders et al . ( @xcite ) ] which plagues some specifications of erg models is , in practice , not a problem for the actor - oriented model . this section presents an algorithm for mcmc approximation of the maximum likelihood estimate . estimation is done conditional on the first observation @xmath64 . this has the advantage that no model assumptions need to be invoked concerning the probability distribution that may have led to the first observed network @xmath64 , and the estimated parameters refer exclusively to the dynamics of the network . the algorithm proposed below can be sketched as follows . for each @xmath65(@xmath66 ) the observed data are augmented with random draws from the sequence of intermediate digraphs @xmath67 that could have led from one observation , @xmath68 , to the next , @xmath1 ( section [ s_aug ] ) . these draws can be simulated using a metropolis hastings algorithm ( section [ s_sim ] ) . they are then used in the updating steps of a robbins monro algorithm , following ideas of gu and kong ( @xcite ) , to find the solution of the likelihood equation ( section [ s_appr ] ) . these elements are put together in section [ s_multi ] . the likelihood for the observed data @xmath69 conditional on @xmath64 can not generally be expressed in a computable form . therefore , the observed data will be augmented with data such that an easily computable likelihood is obtained , employing the general data augmentation principle proposed by tanner and wong ( @xcite ) . the data augmentation can be done for each period @xmath61 separately and , therefore , this section considers only the observations @xmath64 and @xmath70 . denote the time points of an opportunity for change by @xmath71 and their total number between @xmath72 and @xmath73 by @xmath74 , the time points being ordered increasingly so that @xmath75 . the model assumptions imply that at each time @xmath71 , there is one actor , denoted @xmath76 , who gets an opportunity for change at this time moment . define @xmath77 as the actor toward whom the tie variable is changed , and define formally @xmath78 if there is no change [ i.e. , if @xmath79 ) ] : @xmath80 given @xmath64 , the outcome of the stochastic process @xmath81 , @xmath82 completely determines @xmath83 . the augmenting data consist of @xmath74 and @xmath84 , @xmath82 , without the time points @xmath71 . it may be noted that this differs from the definition of a sample path in koskinen and snijders ( @xcite ) in that the times in between opportunities for change are integrated out ; the reason is to obtain a simpler mcmc algorithm . the possible outcomes of the augmenting data are determined by the condition @xmath85 for all @xmath86 with @xmath87 . the stochastic process @xmath88 , @xmath89 will be referred to as the sample path ; the elements for which @xmath90 , although redundant to calculate @xmath70 from @xmath64 , are retained because they facilitate the computation of the likelihood . define @xmath91 ; the digraphs @xmath92 and @xmath93 differ in element @xmath94 provided @xmath95 , and in no other elements . the probability function of the sample path , conditional on @xmath64 , is given by @xmath96 where @xmath97 is defined in ( [ wi ] ) and @xmath98 in and just after ( [ choiceprob ] ) . denote the first component of ( [ full_like1 ] ) by @xmath99\\[-8pt ] & & \qquad = p_{\alpha , \beta}\bigl\{t_{r } \leq t_2 < t_{r+1 } \vert x^{(0 ) } , ( i_1 , j_1 ) , \ldots , ( i_r , j_r ) \bigr\ } .\nonumber\end{aligned}\ ] ] conditioning on @xmath100 [ and not on @xmath70 ! ] , the differences @xmath101 are independently exponentially distributed with parameters@xmath102 . hence , under this conditioning the distribution of @xmath103 is the convolution of exponential distributions with parameters @xmath102 for @xmath104 . in the special case that the actor - level rates of change @xmath22 are constant , denoted by @xmath105 , @xmath74 has a poisson distribution with parameter @xmath106 ; ( [ kappa1 ] ) then is given by @xmath107 in the general case where the change rates are nonconstant , the probability ( [ kappa1 ] ) can be approximated as follows . denote by @xmath108 the density function of @xmath109 , conditional on @xmath110 , @xmath111 or , equivalently , on the embedded process @xmath112 the distribution of @xmath103 is a convolution of exponential distributions and , therefore , the central limit theorem implies that the density function @xmath108 is approximately the normal density with expected value @xmath113 and variance @xmath114 hence , @xmath115 probability ( [ kappa1 ] ) now can be expressed as @xmath116 the approximations are valid for large @xmath74 and @xmath117 , under boundedness conditions on the rate functions @xmath118 . the first approximation in ( [ kappa2 ] ) is based on splitting the integration interval into two intervals @xmath119 and @xmath120 for a bounded but large @xmath121 ; the first interval then gives an asymptotically negligible contribution and on the second interval @xmath122 is approximately constant since @xmath123 . the second approximation uses that @xmath117 is large . combining the preceding equations yields @xmath124\\[-8pt ] & & \qquad \approx \frac{1}{\lambda(\alpha , x^{(r ) } ) \sqrt{(2\pi\sigma ^2_\alpha ) } } \exp\biggl ( \frac{-(t_2 - t_1 - \mu_\alpha)^2}{2\sigma^2_\alpha } \biggr ) .\nonumber\end{aligned}\ ] ] this shows that , for observed data @xmath125 augmented by the sample path , the likelihood conditional on @xmath64 can be expressed directly , either exactly or in a good approximation . an mcmc algorithm will be used that finds the ml estimator based on augmented data . several methods for mcmc maximum likelihood estimation have been proposed in the literature . we shall use the markov chain stochastic approximation ( mcsa ) algorithm proposed by gu and kong ( @xcite ) and used ( in a slightly different specification ) also by gu and zhu ( @xcite ) . this algorithm is based on the missing information principle of orchard and woodbury ( @xcite ) and louis ( @xcite)going back to fisher ( @xcite ) ; cf . efron ( @xcite ) . the principle can be summarized as follows . suppose that @xmath5 is observed , with probability distribution parameterized by @xmath126 and having probability density @xmath127 w.r.t . some @xmath128-finite measure . to facilitate estimation , the observed data is augmented by extra data @xmath129 ( regarded as missing data ) such that the joint density is @xmath130 . denote the observed data score function @xmath131 by @xmath132 and the total data score function @xmath133 by @xmath134 . it is not hard to prove ( see the cited literature ) that @xmath135 this is the first part of the missing information principle . this equation implies that the likelihood equation can be expressed as @xmath136 and , therefore , ml estimates can be determined , under regularity conditions , as the solution of ( [ mle1 ] ) . this is applied in situations where the observed data score function @xmath132 is too difficult to calculate , while the total data score function @xmath134 is computable . in our case , we condition on @xmath137 , so this is treated as being fixed ; we observe @xmath138 ; and between each pair of consecutive observations @xmath57 and @xmath10 the data are augmented , as discussed in the previous section , by the sample path that could have brought the network from @xmath57 to @xmath10 . these sample paths combined for @xmath139 to @xmath18 constitute @xmath140 . the following two subsections supply the additional elements for how the augmenting data are used . in the mcsa algorithm of gu and kong ( @xcite ) , the solution of ( [ mle1 ] ) is obtained by stochastic approximation [ robbins and monro ( @xcite ) ; kushner and yin ( @xcite ) ] which is defined by the updating step @xmath141 where @xmath142 is generated according to the conditional distribution of @xmath140 , given , with parameter value @xmath143 . the sequence @xmath144 , called the gain sequence , consists of positive numbers , tending to 0 . the matrix @xmath145 is a suitable matrix . it is efficient [ see kushner and yin ( @xcite ) ] to use a gain sequence tending to zero as @xmath146 for a @xmath147 , and to estimate @xmath126 not by the last element @xmath143 produced by the algorithm , but by a tail average @xmath148 . for fixed @xmath149 and @xmath150 , this average converges to the solution of ( [ mle1 ] ) for a wide range of positive definite matrices @xmath145 . the second part of the missing information principle [ orchard and woodbury ( @xcite ) ] is that the observed fisher information matrix for the observed data can be expressed as @xmath151\\[-8pt ] & = & e_\theta \{d_{xv}(\theta ) \mid x = x \ } - \mbox{cov}_\theta \ { s_{xv}(\theta ; x , v ) \mid x = x \ } , \nonumber\end{aligned}\ ] ] where @xmath152 is the complete data observed fisher information matrix , @xmath153 expression ( [ fi2 ] ) can be interpreted loosely as `` information is total ( but partially unobserved ) information minus missing information . '' this formula allows us to calculate standard errors . the mcsa method relies on monte carlo simulations of the missing data @xmath140 . the markov property allows us to treat all periods @xmath61 separately and , therefore , we simplify the treatment and notation here by treating only one of those periods , temporarily assuming that @xmath154 . the missing data then is the sample path @xmath155 which specifies the sequence of tie changes that brings the network from @xmath156 to @xmath157 . the advantage of having integrated out the time steps @xmath71with the use of the expressions ( [ kappa4 ] ) and ( [ kappa2])is that less random noise is introduced , and the simulated variable @xmath140 is discrete rather than including a euclidean vector with a varying dimension @xmath74 , which would require a more complicated mcmc procedure . the set of all sample paths connecting @xmath64 and @xmath70 is the set of all finite sequences of pairs @xmath86 , @xmath158 , where for @xmath87 the parity of the number of occurrences of @xmath86 is given by ( [ x2 ] ) . such sequences will be denoted by @xmath159 , and the set of all these sequences is denoted @xmath160 . the probability of the sample path conditional on @xmath161 is proportional to ( [ full_like1 ] ) , rewritten here as @xmath162 where @xmath163 is given by ( [ kappa4 ] ) or approximated by ( [ kappa3 ] ) , depending on the specification of @xmath118 . draws from this distribution can be generated by the metropolis hastings algorithm , provided that a proposal distribution is used which connects any two elements of @xmath160 . such a proposal distribution will now be given , denoting the proposal probabilities by @xmath164 and the target probabilities , which are proportional to ( [ full_like2 ] ) , by @xmath165 . then the acceptance probabilities in the metropolis hastings algorithm , for a current state @xmath166 and a proposed state @xmath167 , are @xmath168 a proposal distribution is used that consists of small changes in @xmath166 . the construction of the proposal distribution was based on the considerations that the proposal distribution should mix well in the set of all sample paths , and the metropolis hastings ratios in ( [ mh ] ) should be computable relatively easily . this led to proposal distributions consisting of the following types of small changes in @xmath166 : 1 . _ paired deletions_. of all pairs of indices @xmath169 with @xmath170,@xmath171 , one pair is randomly selected , and @xmath172 and @xmath173 are deleted from @xmath166 . paired insertions_. a pair @xmath174 with @xmath87 is randomly chosen , two indices @xmath169 are randomly chosen , and the element @xmath86 is inserted immediately before @xmath175 and before @xmath176 . single insertions_. at a random place in the path ( allowing beginning and end ) , the element @xmath177 is inserted for a random @xmath178 . single deletions_. of all elements @xmath179 satisfying @xmath180 , a randomly chosen one is deleted . permutations_. for randomly chosen @xmath181 , where @xmath182 is bounded from above by some convenient number to avoid too lengthy calculations , the segment of consecutive elements @xmath183 is randomly permuted . it is evident that these five operations yield new sequences within the permitted set @xmath160 [ cf . ( [ x2 ] ) ] . paired insertions and paired deletions are each others inverse operations , and the same holds for single insertions and single deletions . these four types of operation together are sufficient for all elements of @xmath160 to communicate . permutations are added to achieve better mixing properties . the detailed specification of how these elements are combined in the proposal distribution can be obtained from the authors . two considerations guided this specification . in the first place , transparency of the algorithm . paired insertions and paired deletions are not always unique inverses of each other . nonuniqueness of the inverse operation would lead to complicated counting procedures to determine proposal probabilities . therefore , the restriction is made that pairs of elements @xmath86 can only be inserted at , or deleted from a pair of positions in the sequence , if there are no occurrences of the same @xmath86 between these positions ; and only if at least one other @xmath184 ( with @xmath185 or @xmath186 ) occurs in between these positions . in this restricted set of operations , paired insertions and paired deletions are each other s unique inverses , which simplifies proposal probabilities . due to the presence of permutations , all elements of the space @xmath160 still are reachable from any element . the second consideration is computational efficiency . when proposing that a pair of elements @xmath86 is inserted or deleted at certain positions , then also proposing to permute a segment between those positions entails no increase in computational load for calculating the metropolis hastings ratios , and this permutation will lead to larger changes in @xmath166 , and thereby , hopefully , better mixing for a given number of computations . this subsection combines the elements presented in the preceding subsections to define an algorithm for mcmc approximation of the ml estimate . it is now assumed that an arbitrary number @xmath14 of repeated observations has been made : @xmath187 . as argued before , we condition on the first observation @xmath64 . the further data is denoted by @xmath188 . the parameter @xmath189 is denoted by @xmath126 . the markov property entails that the observed data score function , conditional on @xmath190 , can be decomposed as @xmath191\\[-8pt ] & & \qquad = \sum_{m=2}^m s_{x(t_m ) \vert x(t_{m-1})}(\theta ; x(t_m ) \vert x(t_{m-1 } ) ) , \nonumber\end{aligned}\ ] ] where @xmath192 is the score function based on the conditional distribution of @xmath193 , given @xmath57 . for the period from @xmath194 to @xmath60 , the data set is augmented by @xmath195 which defines the order in which ties are changed between these time points , as described in section [ s_aug ] ; this can be denoted by @xmath196 with outcome @xmath197 . the augmenting variable as used in section [ s_aug ] is @xmath198 . denote the probability ( [ full_like1 ] ) for the period from @xmath194 to @xmath60 by @xmath199 , and the corresponding total data score function by @xmath200 from ( [ mi1 ] ) and ( [ score_dec ] ) , and using the markov property , it can be concluded that the observed data score function now can be written as @xmath201\\[-8pt ] & & \qquad = \sum_{m=2}^m e_\theta\ { s_m(\theta ; x(t_{m-1 } ) , v_m ) \vert x(t_{m-1 } ) = x(t_{m-1 } ) , x(t_m ) = x(t_m ) \ } . \nonumber\hspace{-20pt}\end{aligned}\ ] ] the ml estimate is obtained as the value of @xmath126 for which ( [ mle4 ] ) equals 0 [ cf . ( [ mle1 ] ) ] . the algorithm is iterative , and the @xmath202th updating step now can be represented as follows : 1 . for each @xmath66 , make a large number of the metropolis hastings steps as described in section [ s_sim ] , yielding @xmath203 . 2 . compute @xmath204 using ( [ mle3 ] ) with @xmath205 defined by ( [ full_like1 ] ) for the period from @xmath194 to @xmath60 , and using ( [ kappa1 ] ) and ( [ kappa4 ] ) or ( [ kappa3 ] ) . 3 . update the provisional parameter estimate by @xmath206 as mentioned in section [ s_appr ] , the estimate @xmath207 is calculated as a tail average of the values @xmath143 generated by this algorithm . the covariance matrix of the ml estimator is estimated using ( [ fi2 ] ) , where the expected values in the right - hand side are approximated by monte carlo simulation of the conditional distribution of @xmath140 given @xmath208 , for @xmath209 . this involves the matrix of partial derivatives ( [ fi1 ] ) , which can be estimated by a score - function method as elaborated in schweinberger and snijders ( @xcite ) . the main implementation details are the following : 1 . the method of moments ( mom ) estimator [ snijders ( @xcite ) ] , in practice , yields a good initial value @xmath210 . 2 . to avoid long burn - in times for step ( 1 . ) , the metropolis hastings algorithm for generating @xmath142 can be started from the previous value , @xmath211 , rather than independently . this leads to some autocorrelation in the updates defined in step ( 3 . ) , and the number of metropolis hastings steps must be large enough that this autocorrelation is not too high . it may be noted that the robbins monro algorithm is robust to moderate dependence between successive updates [ kushner and yin ( @xcite ) ] . we have found good results when the number of steps is tuned so that the autocorrelations between the elements of @xmath212 are less than 0.3 . 3 . for @xmath145 in step ( 3 . ) we use a monte carlo estimate of the complete data observed fisher information matrix ( [ fi1 ] ) , estimated for @xmath213 before making the iteration steps . 4 . for the other numerical parameters of the algorithm we use the same values as described , for the stochastic approximation algorithm to compute the mom estimator , in snijders ( @xcite ) . this algorithm is implemented in siena version 3.3 [ snijders et al . ( @xcite ) ] . the executable program as well as the code can be found athttp://www.stats.ox.ac.uk / siena/. as an illustrative example , the data set is used that was also analyzed in van de bunt , van duijn and snijders ( @xcite ) and in snijders ( @xcite ) . this is a friendship network between 32 freshman students in a given discipline at a dutch university . initially most of the students were unknown to each other . there were six waves denoted @xmath214 of data collection , with 3 weeks between waves for the start of the academic year , and 6 weeks in between later . the relation studied is `` being friends or close friends . '' the data set is obtainable from website http://www.stats.ox.ac.uk/siena/[http://www.stats.ox.ac.uk/ ] http://www.stats.ox.ac.uk/siena/[siena/ ] . the transitions between observations @xmath73 to @xmath215 , and @xmath215 to @xmath216 , will here be studied separately . to identify the rate function , we assume ( arbitrarily but conveniently ) that the duration of the time periods is unity , @xmath217 . to keep the model specification relatively simple , the rate function is supposed to be constant across actors , given by @xmath105 from @xmath73 to @xmath215 and by @xmath218 from @xmath215 to @xmath216 ; and the objective function ( [ fi ] ) is chosen independent of the previous state @xmath32 , containing contributions of the effects of outdegree , number of reciprocated ties , number of transitive triplets , number of 3-cycles , gender of the sender of the tie ( `` ego '' ) , gender of the receiver of the tie ( `` alter '' ) , and gender similarity : @xmath219 where variable @xmath220 indicates the gender of actor @xmath8 ( @xmath221 , @xmath222 ) and @xmath223 its average over the 32 individuals . this model illustrates two types of triadic dependence ( transitive triplets and 3-cycles ) and the use of covariates ( gender ) . for this model the parameters are estimated for the two transitions @xmath224 and @xmath225 separately both by the method of moments ( mom ) and by maximum likelihood . for models where the functions @xmath226 used in ( [ fi ] ) depend only on @xmath5 and not on @xmath32 , so that they can be expressed as @xmath227 , the mom estimator as defined by snijders ( @xcite ) is based on the vector with components @xmath228 for @xmath66 and @xmath229 for @xmath230 . the mom estimator is defined as the parameter vector for which the observed and expected values of this statistic are equal , and can be determined by stochastic approximation . both the moment equation and the likelihood equation can be represented as @xmath231 , where @xmath232 is the difference between observed and estimated moments or the complete - data score function , respectively . for both estimators , after running the stochastic approximation algorithm , convergence was checked by simulating the model for the obtained parameters for 2000 runs and calculating , for each component @xmath233 of @xmath232 , the ratio of the average simulated @xmath233 to its standard deviation . in all cases , this ratio was less than 0.1 , indicating adequate convergence . @lld2.2cd2.2cd2.2cd2.2c@ & * effect * & & + & & & + & & & & & + & & & & & + & & & & & & & & & + & + @xmath23 & rate parameter & 3.95 & 0.64 & 2.61 & 0.36 & 3.43 & 0.59 & 5.67 & 0.75 + & + @xmath234 & outdegree & -1.66 & 0.26 & -1.02 & 0.26 & -2.19 & 0.30 & -2.00 & 0.22 + @xmath235 & reciprocated ties & 2.06 & 0.47 & 1.94 & 0.39 & 2.26 & 0.55 & 1.70 & 0.35 + @xmath236 & transitive triplets & 0.30 & 0.06 & 0.18 & 0.06 & 0.36 & 0.07 & 0.26 & 0.05 + @xmath237 & 3-cycles & -0.59 & 0.23 & -0.42 & 0.17 & -0.59 & 0.27 & -0.31 & 0.14 + @xmath238 & gender alter & 0.28 & 0.30 & 0.14 & 0.36 & 0.70 & 0.39 & 0.61 & 0.33 + @xmath239 & gender ego & -0.34 & 0.32 & -0.41 & 0.41 & -0.04 & 0.38 & -0.10 & 0.33 + @xmath240 & gender similarity & 0.33 & 0.30 & 0.34 & 0.36 & 0.48 & 0.37 & 0.44 & 0.32 + parameter estimates and standard errors are reported in table [ tab1 ] . assuming that the estimators are approximately normally distributed ( which is supported by the simulations reported in the next section , although we have no proof ) , the significance can be tested by referring the ratios of estimate to standard error to a standard normal distribution . the method of moments and maximum likelihood estimates are different but lead to the same substantive conclusions . the parameters reflecting network structure , @xmath234 to @xmath237 , give the same picture for both transitions . the negative @xmath241 indicates that an outgoing tie which is not reciprocated and not transitively embedded is not considered attractive ; the positive @xmath242 and @xmath243 indicate that there is evidence for tendencies toward reciprocation and transitivity of friendship choices , when controlling for all other effects in the model . for interpreting the 3-cycles effect , note that closed 3-cycles are structures denying a hierarchically ordered relation . the negative @xmath244 indicates a tendency away from closed 3-cycles , which in conjunction with the positive transitive triplets parameter could be interpreted as a tendency toward a local ( i.e. , triadic ) popularity hierarchy in friendship . for gender , there is only a close to significant effect of gender alter for the @xmath215@xmath216 transition , suggesting that male students tend to be more popular as friends , when controlling for the other effects . from this single data example , of course no conclusions can be drawn concerning the relative value of these two estimation methods . a comparison between the finite sample behavior of the mom and the ml estimators can be based on simulations . here we present a small simulation study as a very limited exploration of the relative efficiency of the two estimators , which is expected to be in favor of the ml estimator . the limited nature of this simulation study does not allow generalization , but the study design is meant to be typical for applications to friendship networks in rather small groups , and replicates approximately the empirical study of the previous section . the model is identical to that of the previous section : there are three repeated observations , 32 actors , one binary covariate called gender , and the first observed network as well as the distribution of the covariate are identical to the van de bunt data set at observation @xmath73 . therefore , the observation moments are again referred to as @xmath245 . the parameter values are rounded figures close to the estimates obtained in the preceding section . networks for times @xmath215 and @xmath216 are generated and parameters estimated under the assumption that parameters @xmath246 are the same in periods @xmath247 and @xmath248 . the simulation model has parameters @xmath249 and @xmath250 , @xmath251 , and @xmath252 . a total of 1000 data sets were generated and the estimates calculated by both methods . table [ tab2 ] reports the average estimates , the root mean squared errors , the rejection rates for testing the data - generating value of the parameter as the null hypothesis ( estimating type - i error rates ) , and the rejection rates for testing that the parameters equal 0 ( estimating power ) , where the tests were two - sided tests based on the @xmath24-ratio for the corresponding parameter estimate , assuming a standard normal reference distribution , at a nominal significance level of 5% . @lld2.3d1.3ccd2.3d1.3cc@ & * effect * & & + & & & + & & & & & & & & & + & + @xmath249 & rate @xmath224 & 2.46 & 0.44 & 0.063 & & 2.37 & 0.44 & 0.114 & + @xmath250 & rate @xmath225 & 3.47 & 0.55 & 0.045 & & 3.39 & 0.54 & 0.099 & + & + @xmath253 & outdegree & -2.01 & 0.15 & 0.053 & 1.00 & -1.96 & 0.14 & 0.047 & 1.00 + @xmath254 & reciprocation & 1.03 & 0.26 & 0.044 & 0.96 & 0.97 & 0.26 & 0.042 & 0.95 + @xmath255 & transitivity & 0.186 & 0.052 & 0.043 & 0.94 & 0.180 & 0.054 & 0.064 & 0.93 + @xmath256 & 3-cycles & 0.02 & 0.14 & 0.042& & 0.03 & 0.14 & 0.060 & + @xmath257 & gender alter & 0.53 & 0.24 & 0.043 & 0.65 & 0.51 & 0.25 & 0.046 & 0.57 + @xmath258 & gender ego & -0.28 & 0.26 & 0.062 & 0.21 & -0.28 & 0.27 & 0.059 & 0.20 + @xmath259 & gender sim . & 0.52 & 0.24 & 0.031 & 0.64 & 0.52 & 0.25 & 0.052 & 0.60 + out of the 1000 generated data sets , 5 were excluded because they did not produce well converging results in the default specification of the algorithm for one or both estimators . table [ tab2 ] shows that the results for the two estimators are very similar , and the type - i error rates are close to the nominal value , except for the inflated type - i rates of the ml estimators for the two rate parameters . the latter is related to the skewed distribution of the rate parameter estimators . the correlations between the estimators are more than 0.93 for all coordinates ; note that correlations are attenuated due to the stochastic nature of the algorithms . it can be concluded that for this type of model , characterized by 32 actors and 3 waves with rates of change 2.5 and 3.5 , with 7 parameters in the objective function , mom and ml estimation yield quite similar results . to explore data sets with less information , a similar simulation study was conducted with 20 actors , where the first network was induced by the @xmath73 network of 20 of the actors in the van de bunt data set , and the rest of the simulation design differed from the previous study by including an interaction effect of reciprocity by gender similarity , represented by the term @xmath260 with parameter @xmath261 . this effect is included to achieve a higher correlation of the parameter estimators , which together with the smaller number of actors should lead to greater difficulties in estimation . @lld2.3d1.3ccd2.3d1.3cc@ & * effect * & & + & & & + & & & & & & & & & + & + @xmath249 & rate @xmath224 & 2.49 & 0.50 & 0.059 & & 2.37 & 0.48 & 0.093 & + @xmath250 & rate @xmath225 & 3.51 & 0.70 & 0.044 & & 3.36 & 0.67 & 0.094 & + & + @xmath253 & outdegree & -2.09 & 0.29 & 0.009 & 0.98 & -2.02 & 0.21 & 0.025 & 1.00 + @xmath254 & reciprocation & 1.08 & 0.35 & 0.043 & 0.86 & 0.97 & 0.33 & 0.052 & 0.82 + @xmath255 & transitivity & 0.170 & 0.099 & 0.022 & 0.51 & 0.158 & 0.096 & 0.059 & 0.52 + @xmath256 & 3-cycles & 0.01 & 0.23 & 0.034& & 0.05 & 0.23 & 0.057 & + @xmath257 & gender alter & 0.59 & 0.44 & 0.020 & 0.15 & 0.60 & 0.38 & 0.048 & 0.43 + @xmath258 & gender ego & -0.35 & 0.50 & 0.024 & 0.00 & -0.34 & 0.36 & 0.034 & 0.11 + @xmath259 & gender sim . & 0.61 & 0.50 & 0.013 & 0.12 & 0.63 & 0.43 & 0.051 & 0.35 + @xmath262 & g. sim . @xmath263 rec . & 0.47 & 0.63 & 0.039 & 0.10 & 0.50 & 0.60 & 0.043 & 0.14 + the results are shown in table [ tab3 ] . of the 1000 generated data sets , 20 were excluded from the results because of questionable convergence of the algorithm . the table shows that the ml estimator here performs clearly better than the mom estimator . the estimated relative efficiency of mom compared to ml expressed as ratio of mean squared errors ranges for these 10 parameters from 0.50 to 0.99 . the correlation between the estimators ranges from 0.71 ( gender ego ) to 0.98 ( rate period 1 ) . for all parameters except @xmath235 , the estimated power of the test based on the ml estimator is higher than that of the mom - based test , which can be traced back to the combination of a smaller mean squared error and a less conservative test . the surprisingly low power of the mom - based test for the gender - ego effect reflects high standard errors that tend to be obtained for estimating @xmath239 . one of the important advantages of a likelihood approach is the possibility of elaborating model selection procedures . here we only explain how to conduct a likelihood ratio test . a convenient way to estimate the likelihood ratio is by a simple implementation of the idea of the path sampling method described by gelman and meng ( @xcite ) , who took this method from the statistical physics literature where it is known as thermodynamical integration . for arbitrary @xmath264 and @xmath265 , defining the function @xmath266 , with @xmath267 , this method is based on the equation @xmath268 this integral can be approximated by replacing the integral by a finite sum and using ( [ mi1 ] ) . applying a simulation procedure which for each @xmath269 , @xmath270 , generates @xmath271 draws @xmath272 from the conditional distribution of @xmath140 given @xmath208 under parameter @xmath269 , the log likelihood ratio ( [ llr ] ) can be approximated by @xmath273 burn - in time can be considerably reduced when starting the mcmc algorithm for generating @xmath274 by the end result of the algorithm used to generate @xmath275 . this was applied to the van de bunt data set used in section [ sec4 ] to test the null hypothesis that all parameters @xmath276 are for period @xmath248 the same as for @xmath247 against the alternative that they are allowed to be different . the rate parameters @xmath23 were allowed to be different under the null as well as the alternative hypothesis . for calculating ( [ path_samp ] ) we used @xmath277 , @xmath278 . the estimated likelihood ratio was 18.7 , which for a @xmath279 distribution yields @xmath280 , leading to rejecting the null hypothesis at conventional levels of significance . this article has presented a model for longitudinal network data collected in a panel design and an algorithm for calculating the ml estimator . the model can represent triadic and other complicated dependencies that are characteristic for social networks . it is designed for applications in social network analysis , but related models may be applied for modeling networks in other sciences , such as biology . earlier , a method of moments ( mom ) estimator for this model was proposed by snijders ( @xcite ) and bayesian inference methods by koskinen and snijders ( @xcite ) . the algorithm was constructed using stochastic approximation according to the approach proposed by gu and kong ( @xcite ) , and employs monte carlo simulations of the unobserved changes between the panel waves , conditional on the observed data . the simulation design used is more efficient than that used in koskinen and snijders ( @xcite ) because here the waiting times between unobserved changes are integrated out . no proof is yet available for the consistency and asymptotic normality of the ml estimator , which are intuitively plausible for the situation where @xmath281 tends to infinity , @xmath72 and @xmath73 are fixed , and the parameters are such that the average degree @xmath282 tends to a positive finite limit . limited simulation results do support the expectation that the estimators are asymptotically normal . the nonstandard assumptions ( lack of independence ) imply , however , that a proof may be expected to be rather complicated . our experience in the reported simulations and in working with empirical data sets is that the algorithm converges well unless data sets are small given the number of estimated parameters , but the algorithm is time - consuming ( e.g. , each ml estimation for one data set in table [ tab2 ] took about 35 minutes on a regular personal computer , while each mom estimation took about 2 minutes ) . further improvements in the algorithm are desirable for the practical use of ml estimators in these models . this is the subject of future work . judging from our very limited simulations , the advantages of ml estimation over mom estimation in terms of root mean squared error and power of the associated tests seem strong for small data sets and small for medium to large data sets . further simulation studies are necessary . however , even if the statistical efficiency is similar , likelihood - based estimation can have additional advantages , for example , the possibility of extensions to more complicated models and of elaborating model selection procedures . software implementing the procedures in this paper is available athttp://www.stats.ox.ac.uk / siena/. part of this work was done while m. schweinberger was working at the university of groningen and the university of washington . part of this work was done while j. koskinen was working at the university of stockholm and the university of melbourne . 99 airoldi , e. , blei , d. m. , fienberg , s. e. , goldenberg , a. , xing , e. p. and zheng , a. x. ( 2007 ) . _ statistical network analysis : models , issues and new directions ( icml 2006)_. _ lecture notes in computer science _ * 4503. springer , berlin . hanneke , s. and xing , e. p. ( 2007 ) . discrete temporal models of social networks . in _ statistical network analysis : models , issues and new directions ( icml 2006 ) _ ( e. airoldi , d. m. blei , s. e. fienberg , a. goldenberg , e. p. xing and a. x. zheng , eds . ) . _ lecture notes in computer science _ 4503 * 115125 . springer , berlin . krackhardt , d. and handcock , m. s. ( 2007 ) . heider vs simmel : emergent features in dynamic structures . in _ statistical network analysis : models , issues and new directions ( icml 2006)_. ( e. airoldi , d. m. blei , s. e. fienberg , a. goldenberg , e. p. xing and a. x. zheng , eds . ) _ lecture notes in computer science _ * 4503 * 1427 . springer , berlin . orchard , t. and woodbury , m. a. ( 1972 ) . a missing information principle : theory and applications . in _ proceedings sixth berkeley sympos . math . probab . _ * 1 * 697715 california press , berkeley . snijders , t. a. b. and van duijn , m. a. j. ( 1997 ) . simulation for statistical inference in dynamic network models . in _ simulating social phenomena _ ( r. conte , r. hegselmann and p. terna , eds . ) 493512 . springer , berlin . van de bunt , g. g. , van duijn , m. a. j. and snijders , t. a. b. ( 1999 ) . friendship networks through time : an actor - oriented statistical network model . _ computational and mathematical organization theory _ * 5 * 167192 . van duijn , m. a. j. , zeggelink , e. p. h. , huisman , m. , stokman , f. m. and wasseur , f. w. ( 2003 ) . evolution of sociology freshmen into a friendship network . _ journal of mathematical sociology _ * 27 * 153191 . wasserman , s. ( 1979 ) . a stochastic model for directed graphs with transition rates determined by reciprocity . _ sociological methodology1980 _ ( k. f. schuessler , ed . ) 392412 . jossey - bass , san francisco .
the schrdinger equation of an interacting many - body system can be solved analytically only for specific interaction potentials and for a very restricted number of particles . a wide range of approximation methods has been developed to determine the ground and excited states of many - body systems found in nuclear , atomic , and condensed matter systems . here , we focus on quasi two - dimensional electronic systems , which are experimentally realized in quantum - dots . quantum dots are often embedded in layered semiconductor structures . the strong confinement along the vertical direction perpendicular to the layer suggests an effectively two - dimensional description along the remaining two lateral dimensions . besides the electron - electron interactions , an additional external confinement potential along the lateral directions is created by etching and gating of the semiconductor device . quantum dots can be regarded as _ artificial atoms _ , albeit with a central potential different from the nuclear coulomb attraction . their excitation spectra are probed by electronic and optical measurements , also under the influence of additional magnetic fields , which gives rise to a landau - level structure . with the discovery of the fractional quantum hall effect ( fqhe ) in systems of interacting electrons in quasi two - dimensional systems , many theoretical approaches for studying interacting electrons have been proposed , including the celebrated laughlin wave - function for electrons in the lowest landau level . in the experimentally realized hall devices translational symmetry is broken by the current source and drain contacts , which lead to the formation of hot - spots with high electric field values @xcite . a comparison of the various theoretical approaches with numerical methods is often performed for few - electron quantum dots . analytic @xcite and numerical @xcite solutions for the two electron case show the interplay of coulomb interaction and confinement potential , leading to alternating spin polarization of the ground state in a quantum - dot as function of magnetic field @xcite . by separation of the two - electron case into the center - of - mass motion and the relative part , semiclassical solutions have been constructed @xcite and compared to the exact solutions @xcite . for three electrons , the relative - coordinate basis set and expressions for the coulomb matrix elements are given in @xcite . in the popular _ configuration interaction _ ( ci ) approach a diagonalization of the hamiltonian including the coulomb - interaction is done with respect to a product basis of single - particle slater determinants . convergence is generally checked by systematically enlarging the size of the basis set , for instance by increasing the value of the cut - off energy up to which all states are included for the non - interacting system @xcite . with increasing number of basis states , the computation of all interacting matrix - elements and the final diagonalization of the hamiltonian quickly exhausts computer resources . thus a reduction of the basis set while maintaining accuracy is required for studying the excitation spectra of few body systems . if the reduction is based on underlying group - theoretical properties , an exact decomposition of the basis set into irreducible blocks is possible . a diagonalization of the smaller blocks is more easily achievable and the decomposition provides physical insight into the system . one example is the separation of the spatial part of the wavefunction from the spinor states for electrons . if we require that the product of spatial and spin - part represents electrons ( antisymmetric states ) , the basis set is decomposed into separate spin states . this results in spin - adapted ( sa ) configuration methods , for instance implemented in the saci program code @xcite . at this point saci methods used in condensed matter theory for more than 3 electrons stop and do not exploit additional properties of the basis set , such as a decomposition into relative and ( interaction - free ) center - of - mass ( _ cm _ ) states . the separation of the basis set into a product state of relative and center - of - mass part is of particular use if the hamiltonian of the @xmath0-electron system is given as a sum of operators acting on center - of - mass and relative coordinates respectively . for electrons confined to a two - dimensional plane perpendicular to a uniform magnetic field this separation of the hamiltonian is possible as long as the confinement potential in the two - dimensional plane does not lead to an admixing of relative and center - of - mass coordinates @xcite . for molecular or atomic systems moving in three dimensions across uniform magnetic fields , additional couplings between relative and center - of - mass coordinates occur , i.e. the motional stark effect @xcite . the importance of a systematic separation of relative coordinates from the center - of - mass excitation for studying correlated two - dimensional electron systems for three electrons was pointed out early by laughlin @xcite and noticed recently again for more than three electrons @xcite . laughlin s trial wave function by construction is expressed only in relative coordinates and can be interpreted and numerically handled similar to a classical plasma of interacting particles @xcite . we note that both in many - electron and nuclear physics , the interactions between the particles are exclusively functions of relative coordinates . then the same shift vector applied to any single particle coordinate does not change the hamiltonian . by the well - known noether theory of symmetry it then follows that the total momentum @xmath1 of the system is conserved . for an unusual but powerful example of a conserved quantity we refer to @xcite . relative coordinates , in contrast to the single particle coordinates , are free of the @xmath2 coordinate . in quantum mechanics , the components of @xmath1 are proportional to the partial derivatives of the @xmath2 coordinate . therefore , if the states depend on relative coordinates only , we have @xmath3 . another possible option which we shall adopt is to demand the @xmath2 state to have zero oscillator excitation . the paradigms of nuclear theory show @xcite that relative coordinates are the natural tools to understand and compute nuclear scattering and reactions . in nuclear structure , such coordinates may be adapted and serve to describe few - particle systems @xcite , clustering , and decay processes . the construction of orbital states can be achieved by use of group and representation theory @xcite . the orbital states are coupled with spin states to antisymmetric states . the adaption to orbital permutational symmetry for composite systems in relative coordinates becomes more involved , but was extensively studied in @xcite . we report here work on 2d electrons in an external perpendicular magnetic field , so that the 2d total momentum is still conserved @xcite . we use harmonic oscillator states and the methods of bargmann hilbert space . the electron - electron coulomb repulsion is computed in matrix form and included into the diagonalization . for @xmath4 electrons we use the relative tetrahedral coordinates from @xcite with simple properties under permutations . the spin states for total spin @xmath5 are explicitly determined by projection operators . + single - particle coordinates are used to construct antisymmetric states as slater determinants of single - particle states . if these states are taken as harmonic oscillator states , slater determinants can not avoid the intrusion of @xmath2 excitations into state space which give raise , in addition to the proper states , to so - called spurious states of spurious angular momentum . the true role of spurious states can be seen only by the parallel use of relative and @xmath2 coordinates . we shall present in section 7 in particular for total spin @xmath6 , a complete analysis of these spurious states . their intrusion of state space is demonstrated in the spectrum of determinantal eigenstates of the standard saci approach @xcite . moreover , if determinantal eigenstates are variationally computed in a subspace defined by a fixed maximal oscillator excitation , we find that the admixture of spurious states distorts this subspace and the variational results . in our proper treatment , any eigenstate in relative coordinates is the basis to a set of equidistant spurious states of increasing @xmath2 excitation . the total spin of @xmath0 electrons ranges from the maximum @xmath7 to the minimum value , @xmath8 for @xmath0 odd and @xmath6 for @xmath0 even . the permutational symmetry of the spin state is characterized by a partition @xmath9,\ : f_1^s+f_2^s = n,\ : f_1^s \geq f_2^s$ ] of two rows @xmath10 $ ] only , because there are for any electron only the two spin ( up , down ) states . the total spin @xmath11 is related one - to - one to the spin partition @xmath12 by @xmath13 since the electrons are fermions , the combined orbital plus spin state must be antisymmetric with partition @xmath14 $ ] . then is follows that the only allowed orbital partition @xmath15 is associate to the spin partition @xmath12 , that is , its young diagram consists of @xmath0 square boxes , arranged in @xmath16 rows of length 2 , and @xmath17 rows of length 1 . to the orbital partition @xmath15 there belong orthogonal partner functions whose number we denote by @xmath18 . they may be labelled by young tableaus @xmath19 . any young tableau is a filling of the boxes of the young diagram @xmath15 by the numbers @xmath20 , such that the numbers increase both to the right and downwards . the application of permutations to an orbital state of @xmath0 particles , denoted as @xmath21 , is completely determined by the tableau @xmath22 . the antisymmetric fermion state , with respect to permutations , takes the form of a sum of products of orbital and spin states , @xmath23\alpha^n s \rangle= \|f\|^{-\frac{1}{2 } } \sum_r \|\alpha^n f r\rangle \|s f^s r^s\rangle,\ ] ] where @xmath24 is the young tableau of the spin partition @xmath12 , determined from the orbital tableau @xmath22 by reflection in the main diagonal of its orbital young diagram . the operators @xmath25 for the external magnetic field , the central field and the interactions are all invariant under permutations and independent of the spin . since the spin basis states @xmath26 are orthogonal , the matrix elements of all spin - independent operators @xmath25 between antisymmetric states reduce to the spin - free sum of orbital matrix elements @xmath27 here the total spin @xmath11 is fixed from eq . [ es1 ] by half the difference of the column length of the orbital partition @xmath15 . from eq . [ es3 ] it follows that we never need the spin states but must compute @xmath18 orbital diagonal matrix elements . _ example _ : for @xmath4 particles , the possible values of the total spin are @xmath5 . the corresponding spin partitions are @xmath28,[31],[22]$ ] , and the orbital partitions are @xmath29 , [ 211],[22 ] $ ] of dimensions @xmath18 = 1 , 3 , 2 respectively . so we need the construction of @xmath18 orbital states with an orbital partition @xmath15 , fixed by the total spin @xmath11 , and by its young tableaus @xmath22 . to project states of orbital symmetry @xmath15 with the young tableau @xmath30 , we use the matrix elements of the representation @xmath31 which are determined by the tableaus . a young diagram is an arrangement of square boxes in rows of length @xmath32 . in a young tableau @xmath22 , we label these boxes by @xmath20 such that the numbers increase to the right and downwards . there are exactly @xmath18 young diagrams for the partition @xmath15 . the row and column of the box @xmath33 are then @xmath34 . for a pair of boxes @xmath35 we define the axial distance @xcite p.28 by @xmath36 we first determine the matrices of the representation for the @xmath37 transpositions @xmath38 of @xmath39 . these transposition by multiplication generate all elements of @xmath39 , and so all representation matrices can be built by matrix multiplication from the representations of these transpositions . for the permutation which interchanges @xmath33 with @xmath40 and the young tableau @xmath22 , there may be a second tableau @xmath41 with boxes @xmath35 interchanged . if there is no such second tableau , the two boxes must appear in succession in the same row or the same column of the young tableau . in these two cases the matrix element of the representation are given by @xmath42 otherwise , the non - trivial part of the representation is given by the @xmath43 matrix @xmath44.\ ] ] the operator of the transposition @xmath45 interchanges the coordinates of the single particle indexed by the number @xmath46 with those of the particle indexed by @xmath47 . following the wigner @xcite pp.102 - 6 prescription , to assure a homomorphic action we must associate to any permutation @xmath48 the inverse action @xmath49 of the permutation on the particle indices . the young operators @xcite pp . 25 - 26 are given by @xmath50^{\dagger}= c^f_{sr}.\ ] ] these operators must be applied to the highest polynomials eq . [ es21 ] of fixed pseudo angular momentum @xmath51 ( see sect . [ sec : pseudo ] ) , provided that from @xcite we know that @xmath51 can yield the orbital partition @xmath15 . if a young operator gives zero when applied to the polynomial , we first apply a lowering operator eq . [ es23 ] and then again try the young operator . exhausting all values of the pseudo angular component we must find all the states of orbital symmetry . for an antisymmetric state eq . [ es2 ] we use orbital transposition operators @xmath52 to write @xmath53s\|\sum_{s < t } v(s , t)\|\alpha^n[1^n]s\rangle \\ \nonumber & & = \frac{1}{\|f\|}\sum_{s < t } \sum_r \langle f r\|t(s,1 ) t(t,2 ) v(1,2)t(t,2)t(s,1)\|f r\rangle \\ \nonumber & & = \frac{1}{\|f\| } \sum_{s < t } \sum_{r}\sum_{r',r '' } \langle fr\| t(s,1 ) t(t,2)\|fr'\rangle \langle f r'\|v(1,2)\|f r''\rangle \langle fr''\|t(t,2)t(s,1)\|fr\rangle \\ \nonumber & & = \frac{n(n-1)}{2 } \frac{1}{\|f\|}\sum_r \langle f r\|v(1,2)\|f r\rangle . \end{aligned}\ ] ] the third line of eq . [ es13e ] follows from the orthogonality of the spin states in eq . [ es2 ] . in the last line we use the involutive property of transpositions and the orthogonality of young tableaus , @xmath54 ^ 2=[t(t,2)]^2=1 , \\ \nonumber & & \langle f r''\|fr'\rangle = \delta_{r'',r'}.\end{aligned}\ ] ] * prop * : the matrix element of orbital twobody interactions , invariant under permutations , between antisymmetric states of fixed total spin @xmath11 , can be reduced to a sum over diagonal orbital matrix elements of the interaction of the first pair between states characterized by young tableaus , multiplied by the number of pairs . to construct states with other tableaus of the same partition , we provide a ladder procedure by employing the properties of the representation . given an initial basis function @xmath55 , we know from eq . [ es12 ] that the permutation @xmath56 connects it at most to a second basis function @xmath57 whose young tableau differs by an interchange of the numbers @xmath56 . we rewrite eq . [ es12 ] in the form @xmath58 therefore once we know how to apply @xmath59 to the state @xmath55 , we can construct @xmath57 as the linear combination eq . [ es14 ] , second line . with the operator eq . [ es13 ] we have constructed the state with the highest young tableau . * prop * : with steps as in eq . [ es14 ] we can generate all other young tableaus from an initial one . examples are given in the following subsections . in an orbital young operator eq . [ es13 ] , the sum runs over all permutations of @xmath39 . this sum may be broken into classes @xmath60 , characterized by @xmath61 independent and commuting cyclic permutations of length @xmath62 where @xmath63 any cyclic permutations in turn can be factorized into transpositions as @xmath64 and any transposition is conjugate to one of the @xmath0 generators @xmath65 of @xmath39 , see eq . [ es28 ] . by the representation @xmath66 , all the factorizations of permutations are converted into matrix factorizations . a systematic enumeration of all @xmath67 permutations can go in terms of classes @xmath60 of conjugate elements . the number of permutations in the class @xmath60 corresponding to eq . [ es14a ] according to weyl @xcite p.329 is @xmath68 for a single electron in 2d , we label its state by the oscillator excitation @xmath69 , by the numbers @xmath70 of right and left circular oscillator quanta , and by the angular momentum @xmath71 . note the following relations for coordinates and derivatives of a single particle in 2d , passed to a circular setting : @xmath72 and the quadratic relations @xmath73 the last relations are easily extended to scalar products of several vectors , labelled and contracted wrt lower indices . these yield generators of the symplectic group @xmath74 . the states of @xmath0 electrons are then labelled by the summed up number @xmath75 of oscillator quanta and by the summed up angular momentum @xmath76 it is desirable to find other ( integer ) orbital quantum numbers which if possible can label at least in part an orthogonal set of orbital states . for this purpose we use the scheme of @xcite whose quantum numbers arise from group representations . the unitary group of all orbital degrees of freedom is the unitary group @xmath77 , its commuting subgroups acting on the 2 orbital or the @xmath37 relative motion degrees of freedom are @xmath78 . their representations are both given by the same two - component integer partitions @xmath79 , h_1+h_2=n,\ : h_1\geq h_2\geq 0 $ ] . the representation of the subgroup @xmath80 is characterized by the angular momentum @xmath81 . for given partition @xmath79 $ ] the angular momentum ranges from its maximum @xmath82 to its minimum value @xmath83 or @xmath84 for @xmath85 odd or even respectively . in the 2d magnetic field , the lowest landau level @xmath86 arises if the summed up angular momentum @xmath81 equals the summed up excitation @xmath75 . in terms of the partition @xmath79 $ ] for the unitary group @xmath87 , we have @xmath88 . the weights , i.e. the degrees @xmath89 of the vectors indexed @xmath90 , are restricted to the range @xmath91 , and the maximum angular momentum is @xmath92 . the lowest landau level therefore enforces @xmath93 . this condition selects among the partitions of @xmath75 the single case @xmath79=[n0 ] , l = n$ ] . only for this partition can the excitation @xmath75 be carried by a single vector . for higher landau levels this condition is relaxed and new unitary partitions are allowed . the group @xmath94 has an orthogonal subgroup @xmath95 whose representation label we term the pseudo angular momentum @xmath51 . the possible values of @xmath51 were given in @xcite . here the authors use the subgroup chain @xmath96 with branching of representation labels @xmath97 > ( \lambda',\mu ' ) > \lambda,\ : ( \lambda',\mu')=(h_1-h_2 , h_2).\ ] ] the range of these labels is determined by specific branchings from the representation in the next higher subgroup which we shall discuss in more detail . we treat the oscillator states by the methods of bargmann hilbert space of analytic functions @xcite . we specifically note that in what follows we must _ interchange the role of the groups and subgroups _ @xmath98 compared to @xcite , and _ convert angular into pseudo angular momentum . _ this allows us to take advantage of the group theory given in @xcite , but in a new and different interpretation . by _ upper indices _ @xmath99 ( or upper @xmath100 ) , see below , we denote the cartesian vector components in 2d , and by _ lower cartesian indices _ @xmath101 or spherical indices @xmath102 three ( or more ) relative vectors of the four ( or more ) particles . the normalized states of angular momentum @xmath103 of a single particle with oscillator excitation @xmath75 in the bargmann representation are @xmath104^{\frac{1}{2}}.\ ] ] with @xmath105 a solid spherical harmonic @xcite p.69 . the coordinates may then be taken as the six linear combinations of cartesian coordinates @xmath106 , \\ \nonumber & & z^+_-= \sqrt{\frac{1}{2}}[(z_1 ^ 1-iz_2 ^ 1)+ i(z_1 ^ 2-iz_2 ^ 2)],\ : z^+_0=z^+_3 , \\ \nonumber & & z^-_+= \sqrt{\frac{1}{2}}[(z_1 ^ 1+iz_2 ^ 1)- i(z_1 ^ 2+iz_2 ^ 2 ) ] , \\ \nonumber & & z^-_-= \sqrt{\frac{1}{2}}[(z_1 ^ 1-iz_2 ^ 1)- i(z_1 ^ 2-iz_2 ^ 2 ) ] , \ : z^-_0=z^-_3 . \\ \nonumber \end{aligned}\ ] ] the lower cartesian indices @xmath107 denote three relative coordinates like the jacobi or tetrahedral ones . the permutations of particles act on these coordinates and introduce linear combinations wrt the lower indices . next we introduce from @xcite p.187 for @xmath4 the elementary polynomials @xmath108_+ \\ \nonumber & s:=z_+^1z_-^1 + z_0 ^ 1z_0 ^ 1 , \\ \nonumber & t:= ( z^1\cdot z^1)(z^2\cdot z^2)-(z^1\cdot z^2)^2 , \\ \nonumber \end{aligned}\ ] ] in @xcite p. 187 we find in the scheme eq . [ es16 ] the linearly independent polynomials @xmath109 here all numbers in exponentials must be non - negative , and @xmath110 is an integer which labels the multiplicity of a repeated pseudo angular momentum @xmath51 for a given partition @xmath79 $ ] of @xmath94 . these multiple states of equal pseudo angular momentum are indexed by @xmath110 and linearly independent , but not yet orthogonal . orthogonality must be achieved by standard matrix methods . the permutation group @xmath111 acting on relative coordinates is a subgroup of the pseudo rotation group @xmath95 . the multiplicity of its representations , see table [ table1 ] , was given for all orbital partitions @xmath15 of @xmath112 particles in @xcite . the polynomials eq . ( [ es21 ] ) have the maximum component @xmath51 of the pseudo angular momentum . lowering of the component is achieved by applying the first order lowering differential operator @xmath113 finally we can also lower the angular momentum @xmath81 from its maximum value @xmath114 in steps of @xmath115 by application of the angular momentum operators @xmath116 @xmath117 & [ 211 ] & [ 22 ] & [ 31 ] & [ 4 ] & \lambda^{\pi}[1 ^ 4]&\lambda^{\pi}[4]\\ \hline \lambda^+ & [ 4 ] & [ 31 ] & [ 22 ] & [ 211]&[1 ^ 4 ] & & \\ \hline 0 & 1 & & & & & 0 ^ -&0^+\\ 1 & & & & 1 & & & \\ 2 & & 1 & 1 & & & & \\ 3 & & 1 & & 1 & 1 & 3^+&3 ^ -\\ 4 & 1 & 1 & 1 & 1 & & 4 ^ -&4^+\\ 5 & & 1 & 1 & 2 & & & \\ 6 & 1 & 2 & 1 & 1 & 1 & 6^+6 ^ -&6 ^ -6^+\\ 7 & & 2 & 1 & 2 & 1 & 7^+&7 ^ -\\ 8 & 1 & 2 & 2 & 2 & & 8 ^ -&8^+ \\ 9 & 1 & 2 & 1 & 3 & 1 & 9^+9 ^ -&9 ^ -9^+\\ 10 & 1 & 3 & 2 & 2 & 1 & 10 ^ -10^+&10^+10 ^ -\\ 11 & & 3 & 2 & 3 & 1 & 11^+&11 ^ - \\ 12 & 2 & 1 & 1 & 1 & 1 & ( 12 ^ -)^212^+&(12^+)^212 ^ -\\ 13 & 1 & 1 & 1 & 2 & 1 & 13 ^ -13^+&13^+13 ^ -\\ 14 & 1 & 2 & 2 & 1 & 1 & 14 ^ -14^+&14^+14 ^ -\\ 15 & 1 & 2 & 1 & 2 & 2 & 15 ^ -(15^+)^2&15^+(15-)^2\\ 16 & 1 & 2 & 2 & 2 & 1 & 16 ^ -16^+&16^+16 ^ -\\ 17 & 1 & 2 & 2 & 3 & 1 & 17 ^ -17^+&17^+17 ^ -\\ 18 & 2 & 3 & 2 & 2 & 2 & ( 18 ^ -)^2(18^+)^2&(18^+)^2(18 ^ -)^2\\ 19 & 1 & 3 & 2 & 3 & 2 & 19 ^ -(19^+)^2&19^+(19 ^ -)^2\\ 20 & 2 & 3 & 3 & 3 & 1 & ( 20 ^ -)^220^+&(20^+)^2 20 ^ -\\ 21 & 2 & 3 & 2 & 4 & 2 & ( 21 ^ -)^2(21^+)^2&(21^+)^2(21 ^ -)^2\\ 22 & 2 & 4 & 3 & 3 & 2 & ( 22^+)^2(22 ^ -)^2&(22 ^ -)^2(22^+)^2\\ 23 & 1 & 4 & 3 & 4 & 2 & 23 ^ -(23^+)^2&23^+(23 ^ -)^2\\ 24 & 3 & 2 & 2 & 2 & 2 & ( 24 ^ -)^3(24^+)^3&(24^+)^2(24 ^ -)^2\\ \hline \end{array } $ ] the orbital partitions @xmath118,[4]$ ] do not apply for electrons with spin @xmath119 . the entries for higher values of @xmath51 than given in table [ table1 ] are found by the recursive methods from @xcite eq . a full orbital 4-particle state is characterized by the quantum numbers @xmath120 , \lambda , l , f r\rangle,\ ] ] where @xmath22 ranges over the orbital young tableaus of the diagram @xmath15 . the total spin @xmath11 is determined by antisymmetry and by the orbital partition @xmath15 , see section 2 . the orbital state has partition @xmath121 $ ] of dimension @xmath122\|=1 $ ] . the operator eq . [ es13 ] applied to the initial state @xmath123 becomes @xmath124 is the sign of the permutation @xmath48 , @xmath125 for even and @xmath126 for odd permutations . the orbital partition is @xmath127 $ ] of dimension @xmath128\|=3 $ ] . the three young diagrams ( written for simplicity in rectangular frames ) read @xmath129 , & r_2=\left [ \begin{array}{ll } 1 & 3\\ 2&\\ 4&\\ \end{array } \right ] , & r_3=\left [ \begin{array}{ll } 1 & 4\\ 2&\\ 3&\\ \end{array } \right]\\ \end{array}.\ ] ] assume that the initial state @xmath130 r_1\rangle$ ] has been generated by application of a young operator eq.[es13 ] . we find the two partner states of @xmath130r_1\rangle$ ] by the operator relations eq [ es14 ] and the axial distances , @xmath131 r_2\rangle = \|\alpha^4[211]r_1\rangle \sqrt{\frac{1}{3 } } + t(2,3)\|\alpha^4[211]r_1\rangle \sqrt{\frac{4}{3}},\:\ : \tau_{23}(r_1)=-2,\\ \nonumber & \|\alpha^4[211 ] r_3\rangle = \|\alpha^4[211]r_2\rangle \sqrt{\frac{1}{8 } } + t(3,4)\|\alpha^4[211]r_2\rangle \sqrt{\frac{9}{8}},\:\ : \tau_{34}(r_2)= -3.\end{aligned}\ ] ] the orbital partition is @xmath132 $ ] of dimension @xmath133\|=2 $ ] . the two young diagrams read @xmath134 , r_2=\left [ \begin{array}{ll } 1&3\\ 2&4\\ \end{array } \right].\ ] ] again we generate the initial orbital state @xmath135r_1\rangle$ ] by application of a young operator eq . we find the second state by the operator relation and the axial distance @xmath136r_1\rangle \sqrt{\frac{4}{3}},\:\ : \tau_{23}(r_1)=-2.\ ] ] since we are dealing with particle - particle interactions , the _ cm _ vector is not affected , the total momentum is conserved , and the system is amenable to a description in relative coordinates . a standard choice are the jacobi relative coordinates @xmath137 in @xcite there was used a particular tetrahedral set of three doubledot relative coordinates . this set has the virtue that under the application of any permutation these relative coordinates are at most permuted , eq . the transformation properties of the doubledot coordinates are summarized in @xcite table 3 . the relative tetrahedral coordinates @xmath138 from @xcite , but now enumerated in the order of @xcite for @xmath4 , are @xmath139 it is found from @xcite , table 3 , that the tetrahedral coordinates eq . [ es30 ] have extremely simple matrix transformations under the generating transpositions @xmath140 : @xmath141 = \left[\begin{array}{rrr } 1&0&0\\ 0&0&-1\\ 0&-1&0\\ \end{array } \right ] \left[\begin{array}{l } \eta_1\\ \eta_2\\ \eta_3\\ \end{array}\right ] , \\ \nonumber & & t^{\eta}(2,3)\left[\begin{array}{l } \eta_1\\ \eta_2\\ \eta_3\\ \end{array } \right ] = \left[\begin{array}{lll } 0&1&0\\ 1&0&0\\ 0&0&1\\ \end{array } \right ] \left[\begin{array}{l } \eta_1\\ \eta_2\\ \eta_3\\ \end{array}\right ] , \\ \nonumber & & t^{\eta}(3,4)\left[\begin{array}{l } \eta_1\\ \eta_2\\ \eta_3\\ \end{array } \right ] = \left[\begin{array}{lll } 1&0&0\\ 0&0&1\\ 0&1&0\\ \end{array } \right ] \left[\begin{array}{l } \eta_1\\ \eta_2\\ \eta_3\\ \end{array}\right].\end{aligned}\ ] ] the relative coordinates in gusev @xcite extend the tetrahedral coordinates eq . [ es30 ] to @xmath142 particles . see appendix b for their transforms under permutations . we give now a first example of the polynomial state construction in the tetrahedral coordinates eq . [ es30 ] for the lowest excitation @xmath143=[30 ] , \lambda=3 , \ : f=[211]$ ] with spin @xmath144 and the orbital young tableaus @xmath145 of eq . [ es16 ] : @xmath146 from matrix elements of the coulomb interaction to the energy level structure . ------------------------------------------------------------------------------ with the polynomial states for a specific set of quantum numbers at hand , we proceed to evaluate the various parts of the four - electron hamiltonian in the presence of a magnetic field @xmath147 perpendicular to the @xmath148 plane , @xmath149 with the larmor frequency @xmath150 , the effective mass @xmath151 and the combined frequency @xmath152 due to an external confinement potential @xmath153 . in all following expressions we work within the symmetric gauge , which allows one to express the uniform magnetic field in terms of two - dimensional oscillator excitations @xcite . the resulting four particle hamiltonian is readily decomposed into _ cm _ and relative parts in the plane perpendicular to the magnetic field direction . without the coulomb part it resembles the addition of 2d oscillator systems with frequencies @xmath154 . the magnetic field induces an angular - momentum dependent energy shift and the non - interacting eigenenergies read @xmath155 where we followed fock s notation . in terms of the excitations @xmath156 these quantum - numbers read @xmath157 the total energy is given by the addition of _ cm _ and relative parts @xmath158 for the evaluation of the coulomb interaction we express the states as polynomials in the jacobi relative vectors . there is , independent of the @xmath2 coordinate , a direct linear matrix transformation from relative tetrahedral to jacobi vectors obtained from eqs . [ es29b ] , [ es30 ] , which for short we denote as @xmath159 the first jacobi vector from eq . [ es29b ] is the relative vector between particles 1,2 and according to eq . [ es13e ] it suffices to compute the full coulomb interaction . in the relative - coordinate polynomials the dimensionless coulomb interaction reads @xmath160 where the factor @xmath161 arises from the number of pairs for the two - body operator in eq . [ es13e ] . the coulomb matrix element between two states expressed in the three jacobi vectors becomes @xmath162 , { \rm min}[m^+_1 , m^-_1 ] , \|n^-_1 - n^+_1\|),\end{aligned}\ ] ] with @xmath163 we construct for a given set of relative - coordinate excitations all relative - coordinate states @xmath164 and express each state @xmath165 as sum of polynomials in jacobi coordinates with exponents @xmath166 @xmath167 then the interaction matrix is given by @xmath168 the coulomb interaction does not connect states with different angular momenta and so the blocks of states @xmath169 fall into subblocks labelled by @xmath170 . however , different excitations of the relative oscillator are coupled by the coulomb interaction . the addition of the relative oscillators , including the coulomb interaction is given by diagonalization of the matrix @xmath171 if we augment the relative eigenvalues with the analytic results for the _ cm _ energies , we are in a position to compare the results with the configuration - interaction methods using slater determinantes of single - particle states . .list of 34 saci coulomb eigenvalues ( @xmath173 ) in units of @xmath174 for @xmath172 obtained by diagonalization of the slater determinants of the 34 single - particle product states @xmath175 yielding @xmath176 within the lowest landau level . the entries in boldface indicate the seven states with the _ cm _ oscillator in the ground state obtained by the polynomial method described in the text for the lowest landau level ( @xmath177 ) , while the remaining 27 eigenvalues correspond to spurious states . [ cols="^,^,^,^ " , ] the construction of the basis set of relative coordinate polynomials for @xmath0 electrons entailed the following steps : + ( 1 ) spin , partition : given the number @xmath0 of electrons , we determine the possible spins @xmath11 , orbital partitions @xmath15 and angular momenta @xmath81 . the state space falls into independent and orthogonal subspaces labeled by @xmath178 . steps ( 1 - 7 ) work for n=4 . for @xmath142 and steps ( 4 - 7 ) an extended analysis needs to be developed , some results are given in the appendices . \(2 ) unitary quantum numbers : we provide the orbital quantum numbers of the system . for a maximal relative oscillator excitation @xmath75 we list the unitary partitions @xmath79 , h_1+h_2=n$ ] for the groups @xmath179 . a restriction to the lowest landau level @xmath86 with @xmath79 \rightarrow [ n0]$ ] greatly simplifies the analysis . \(3 ) angular momenta : the unitary partition @xmath180 $ ] from @xmath181 fixes the ladder of angular momenta @xmath81 from the maximum value @xmath182 in steps of @xmath115 down to @xmath84 or @xmath83 and provides the final blocks @xmath178 . \(4 ) pseudo angular momentum and partition : the pseudo angular momentum @xmath51 is obtained from @xmath180 $ ] by the representation theory of the rotation subgroup in @xmath183 . for @xmath4 , any fixed pseudo angular momentum @xmath51 yields from @xcite p. 260 a precise list of possible orbital partitions @xmath15 . multiple values of pseudo angular momentum yield linearly independent polynomials . \(5 ) state polynomials : given this list of quantum numbers , we construct state polynomials in relative coordinates . the initial polynomial is known from @xcite . the component of pseudo angular momentum and of total angular momentum are lowered independently . \(6 ) young tableau : for the known orbital partitions @xmath15 , by applying permutation operators we project states with definite young tableau . \(7 ) coulomb interaction and energies : for state polynomials in a block @xmath178 , we isolate the first jacobi vector in bra and a ket states , evaluate the matrix elements of the coulomb interaction , determine the hamiltonian matrix , diagonalize it and find the eigenstates of relative motion . the study of the four - electron quantum - dot with coulomb interactions in the relative - coordinate basis elucidates and overcomes some of the problems inherent in the commonly used configuration interaction methods such as saci @xcite : if the @xmath2 excitations , as in determinantal state space , can not be separated , a large cloud of spurious @xmath2 excited states , augmented by spurious angular momenta , intrudes the variational state space and obscures the variational energy spectrum reflecting the coulomb interaction . within saci , higher eigenenergies are distorted by the coulomb interaction and a loss in accuracy is seen , which can be avoided by the separation of the relative and @xmath2 coordinates . this allows one to construct polynomial states with labels from representations of unitary and rotation groups and to diagonalize the coulomb - interaction part in smaller subblocks . our approach opens a clear view on the energy spectrum as shown in figs . 1 and 2 and preserves the exact ladder spacing @xmath184 of _ cm _ copies of the relative coordinates variational subspace . by increasing the basis size with respect to the relative excitations , the eigenenergies and eigenvectors obtained by diagonalization converge toward the exact eigenstates of the system . tk has been supported by the heisenberg program of the dfg ( kr 2889/5 ) . an independent view on orbital basis states for @xmath112 electons under permutations arises by use of the subgroup @xmath186 from @xcite . this method can be extended to higher excitations . it allows in tables [ table8],[table9 ] to identify states of permutational symmetry for any pseudo angular momentum almost without projection . we use @xmath187 to denote a standard state of pseudo angular momentum @xmath188 with component @xmath189 , @xmath190 for parity , and @xmath191 with the states defined in eq . as mentioned in @xcite p. 263 , the group @xmath192 is the largest subgroup of @xmath111 that leaves invariant the double - dot @xmath193 axis up to a reflection . with the frobenius multiplicity expression we assign from the characters of @xmath192 , table [ table6 ] , in table [ table7 ] the reduction of ir f of @xmath111 to those of @xmath192 . the numbers @xmath194 are taken from @xcite and used in table [ table6 ] . they determine the irreducible bases of representation of @xmath192 but from table [ table7 ] also basis states of permutational symmetry . the basis states @xmath118_3 $ ] and @xmath195_3 $ ] are unique , and so their partner states may be derived by the ladder procedure [ es14 ] . the basis functions for @xmath196,[211]$ ] here transform like the doubledot representation . by @xcite eqs . ( 6.1 , 6.9 ) we can always return from the doubledot basis to the jacobi basis . the basis states of @xmath132 $ ] can be separated from the bases of @xmath197 $ ] and @xmath198 $ ] by projection . we fix for the moment the labels @xmath199 and introduce @xmath191 in eq . motivated by @xcite , but independent of it , we redefine two new linear combinations of spherical harmonics @xcite ( 2.5.7 ) by @xmath200,\ : \mu \geq 0,\\ \nonumber & & \rho=-1 : \|-1(\lambda,\mu)\rangle = i^{-1}g_{\mu}[\|(\lambda,\mu)\rangle-\|(\lambda,-\mu)\rangle ( -)^{\lambda+\mu } ] , \ : \mu>0 . \end{aligned}\ ] ] these states by consulting @xcite table 3 may be shown to be orthogonal eigenstates of the permutation operator @xmath201 with eigenvalue @xmath202 , which from @xcite fig 1 reverses the axis @xmath193 . states characterized by @xmath203 we show to belong to irreps of the subgroup @xmath192 . for the proof we employ the class operators @xcite , which are group operators @xmath204 summed over a class of equivalent group elements , and commute with any group operator . the five class operators @xmath205 @xcite of the group @xmath192 , written as sums of permutation operators from @xcite , table 4 are @xmath206 by use of the representation matrices of permutations we express the class operators eq . [ a2 ] in matrix form , table [ table12 ] . the matrix forms depend on the chosen representation , and we make convenient choices : for the representations @xmath207}$ ] and @xmath208 } , d^{[211]}(i , j)= i d^{[31]}((i , j)$ ] we use the basis eq . [ es30 ] of @xmath209 , for the representation @xmath210}$ ] the young representation @xcite p. 226 . by virtue of these choices we find : all class operators eq . [ a2 ] have diagonal representations , see table [ table12 ] . this proves : + * prop * : all basis states of the irreps of the subgroup @xmath192 in the chosen representations are basis states of irreducible representations of the bigger group @xmath111 . + this remarkable result allows in table [ table7 ] , up to certain ambiguities , to almost avoid the use and projection with young operators for the bigger group @xmath111 . in tables [ table8],[table9 ] we use it in relation with the full scheme of groups including @xmath211 and subgroups to assign orbital symmetry to the oscillator states . @xmath212 @xmath213}(ki ) & d^{[31]}(ki ) & d^{[211]}(ki ) & d^{[22]}(ki ) & d^{[1 ^ 4]}(ki)\\ \hline k1 & \begin{array}{\|l\| } \hline 1\\ \hline \end{array } & \begin{array}{\|l\|l\|l\| } \hline 1 & & \\ \hline & 1 & \\ \hline & & 1\\ \hline \end{array } & \begin{array}{\|l\|l\|l\| } \hline 1 & & \\ \hline & 1 & \\ \hline & & 1\\ \hline \end{array } & \begin{array}{\|l\|l\| } \hline 1 & \\ \hline & 1\\ \hline \end{array } & \begin{array}{\|l\| } \hline 1\\ \hline \end{array } \\ \hline k2 & \begin{array}{\|l\| } \hline 1\\ \hline \end{array } & \begin{array}{\|l\|l\|l\| } \hline -1 & & \\ \hline & -1 & \\ \hline & & 1\\ \hline \end{array } & \begin{array}{\|l\|l\|l\| } \hline -1 & & \\ \hline & -1 & \\ \hline & & 1\\ \hline \end{array } & \begin{array}{\|l\|l\| } \hline 1 & \\ \hline & 1 \\ \hline \end{array } & \begin{array}{\|l\| } \hline 1\\ \hline \end{array}\\ \hline k3 & \begin{array}{\|l\| } \hline 2\\ \hline \end{array } & \begin{array}{\|l\|l\|l\| } \hline & & \\ \hline & & \\ \hline & & -2\\ \hline \end{array } & \begin{array}{\|l\|l\|l\| } \hline & & \\ \hline & & \\ \hline & & 2\\ \hline \end{array } & \begin{array}{\|l\|l\| } \hline 1 & \\ \hline & -1\\ \hline \end{array } & \begin{array}{\|l\| } \hline -2\\ \hline \end{array}\\ \hline k4 & \begin{array}{\|l\| } \hline 2\\ \hline \end{array } & \begin{array}{\|l\|l\|l\| } \hline & & \\ \hline & & \\ \hline & & -2\\ \hline \end{array } & \begin{array}{\|l\|l\|l\| } \hline & & \\ \hline & & \\ \hline & & -2\\ \hline \end{array } & \begin{array}{\|l\|l\| } \hline 1 & \\ \hline & 1\\ \hline \end{array } & \begin{array}{\|l\| } \hline 2\\ \hline \end{array}\\ \hline k5 & \begin{array}{\|l\| } \hline 2\\ \hline \end{array } & \begin{array}{\|l\|l\|l\| } \hline & & \\ \hline & & \\ \hline & & 2\\ \hline \end{array } & \begin{array}{\|l\|l\|l\| } \hline & & \\ \hline & & \\ \hline & & -2\\ \hline \end{array } & \begin{array}{\|l\|l\| } \hline 2 & \\ \hline & -2\\ \hline \end{array } & \begin{array}{\|l\| } \hline -2\\ \hline \end{array}\\ \hline \end{array } $ ] @xmath214_2 \| [ 4 ] & [ 211]_3 \\ \hline 0 & -1 & [ 22]_1 \|[1 ^ 4 ] & [ 31]_3 \\ \hline 0 & 0 & [ 31]_1 \| [ 211]_1&[31]_2 \| [ 211]_2 \\ \hline 1 & + 1 & [ 22]_1 \| [ 1 ^ 4 ] & [ 31]_3 \\ \hline 1 & -1 & [ 22]_2 \| [ 4 ] & [ 211]_3 \\ \hline 1 & 0 & [ 211]_1\| [ 31]_1&[211]_2 \| [ 31]_2 \\ \hline \end{array } $ ] of two states separated by @xmath215 , one and only one can belong to the listed tableau . the states @xmath216,[1 ^ 4]$ ] are identified as eigenstates under the transposition @xmath217 with eigenvalue @xmath218 respectively . if a state is not reproduced under @xmath217 , it necessarily belongs to @xmath132 $ ] and spin @xmath6 . we conclude that the states eq . [ a1 ] yield all the bases of the orbital young tableaus . @xmath219 + & - & ( 0,0 ) & 0^+ & [ 4]\\ \hline 1 ^ -&1 ^ - & 1 & 1 & 0 & [ 31]1 + & [ 31]2 + & ( 1,0 ) & 1 ^ - & [ 31]\\ \hline & & 0 & 1 & 1 & - & [ 31]3 + & & & \\ \hline 2^+&2^+ & 2 & 0 & -1 & [ 22]1 + & [ 31]3 + & ( 2,0 ) & 2^+ & [ 31][22]\\ \hline & & 1 & 0 & 0 & [ 31]1 + & [ 31]2 + & & & \\ \hline & & 0 & 0 & 1 & [ 22]2 + & - & & & \\ \hline & 1^+ & 1 & 0 & 0 & [ 211]1 + & [ 211]2 + & ( 0,1 ) & 1^+ & [ 211 ] \\ \hline & & 0 & 0 & 1 & - & [ 211]3 + & & & \\ \hline & 0^+ & 0 & 0 & 1 & [ 4]+ & - & ( 2,0 ) & 0^+ & [ 4 ] \\ \hline 3 ^ -&3 ^ - & 3 & 1 & 0 & [ 211]1\|[31]1&[211]2\|[31]2&(3,0 ) & 3 ^ - & [ 211][31][4]\\ \hline & & 2 & 1 & -1 & [ 4]+ & [ 211]3 + & & & \\ \hline & & 1 & 1 & 0 & [ 211]1\|[31]1&[211]2\|[31]2 & & & \\ \hline & & 0 & 1 & 1 & - & [ 31]3 + & & & \\ \hline & 2 ^ - & 2 & 1 & -1 & [ 22]2 + & [ 211]3 + & ( 1,1 ) & 2 ^ - & [ 211][22 ] \\ \hline & & 1 & 1 & 0 & [ 211]1 + & [ 211]2 + & & & \\ \hline & & 0 & 1 & 1 & [ 22]1 + & - & & & \\ \hline & 1 ^ - & 1 & 1 & 0 & [ 31]1 + & [ 31]2 + & ( 3,0),(1,1 ) & ( 1 ^ -)^2 & [ 31]^2 \\ \hline & & 0 & 1 & 1 & - & [ 31]3 + & & & \\ \hline 4^+&4^+ & 4 & 0 & 1 & [ 22]2\|[4]&[211]3 & ( 4,0 ) & 4^+ & [ 4][31][22][211]\\ \hline & & 3 & 0 & 0 & [ 31]1\|[211]1&[31]\|2[211]2 & & & \\ \hline & & 2 & 0 & -1 & [ 22]1 + & [ 31]3 + & & & \\ \hline & & 1 & 0 & 0 & [ 31]1\|[211]1&[31]2\|[211]2 & & & \\ \hline & & 0 & 0 & 1 & [ 22]2\|[4 ] * & [ 211]3 * & & & \\ \hline & 3^+ & 3 & 0 & 0 & [ 31]1\|[211]1&[31]2\|[211]2&(2,1 ) & 3^+ & [ 31][211][1 ^ 4]\\ \hline & & 2 & 0 & -1 & [ 1 ^ 4 ] & [ 31]3 + & & & \\ \hline & & 1 & 0 & 0 & [ 31]1\|[211]1&[31]2\|[211]2 & & & \\ \hline & & 0 & 0 & 1 & - & [ 211]3 + & & & \\ \hline & 2^+ & 2 & 0 & -1 & [ 22]1 + & [ 31]3 + & ( 4,0),(2,1),(0,2)&(2^+)^3 & [ 31]^3[22]^3 \\ \hline & & 1 & 0 & 0 & [ 31]1 + & [ 31]2 + & & & \\ \hline & & 0 & 0 & 1 & [ 22]2 + & - & & & \\ \hline & 1^+ & 1 & 0 & 0 & [ 211]1+&[211]2 + & ( 2,1 ) & 1^+ & [ 211]\\ \hline & & 0 & 0 & 1 & - & [ 211]3 + & & & \\ \hline & 0^+ & 0 & 0 & 1 & [ 4]+ & - & ( 4,0),(0,2 ) & ( 0^+)^2 & [ 4]^2\\ \hline \end{array } $ ] @xmath2201\|[31]1&[211]2\|[31]2 & ( 5,0 ) & 5 ^ - & [ 211][22][31]^2\\ \hline & & 4 & 1 & 1 & [ 22]1 + & [ 31]3 + & & & \\ \hline & & 3 & 1 & 0 & [ 211]1\|[31]1&[211]2\|[31]2 & & & \\ \hline & & 2 & 1 & -1 & [ 22]2 + & [ 211]3 + & & & \\ \hline & & 1 & 1 & 0 & [ 211]1\|[31]1&[211]2\|[31]2 & & & \\ \hline & & 0 & 1 & 1 & [ 22]1 + & [ 31]3 + & & & \\ \hline & 4 ^ - & 4 & 1 & 1 & [ 1 ^ 4]\|[22]1 * & [ 31]3 + & ( 3,1 ) & 4 ^ - & [ 1 ^ 4][211][22][31]\\ \hline & & 3 & 1 & 0 & [ 211]1\|[31]1&[211]2\|[31]2 & & & \\ \hline & & 2 & 1 & -1 & [ 22]2 + & [ 211]3 + & & & \\ \hline & & 1 & 1 & 0 & [ 211]1\|[31]1&[211]2\|[31]2 & & & \\ \hline & & 0 & 1 & 1 & [ 22]1\|[1 ^ 4 ] * & [ 31]3 + & & & \\ \hline & 3 ^ - & 3 & 1 & 0 & [ 211]1\|[31]1&[211]2\|[31]2&(5,0)(3,1)(1,2)&(3 ^ -)^3 & [ 211]^3[31]^3[4]^3 \\ \hline & & 2 & 1 & -1 & [ 4]+ & [ 211]3 + & & & \\ \hline & & 1 & 1 & 0 & [ 211]1\|[31]1&[211]2\|[31]2 & & & \\ \hline & & 0 & 1 & 1 & - & [ 31]3 + & & & \\ \hline & 2 ^ - & 2 & 1 & -1 & [ 22]2 + & [ 211]3 + & ( 3,1)(1,2 ) & ( 2 ^ -)^2 & [ 211]^2[22]^2 \\ \hline & & 1 & 1 & 0 & [ 211]1 + & [ 211]2 + & & & \\ \hline & & 0 & 1 & 1 & [ 22]1 + & - & & & \\ \hline & 1 ^ - & 1 & 1 & 0 & [ 31]1 + & [ 31]2 + & ( 5,0)(3,1)(1,2 ) & ( 1 ^ -)^3 & [ 31]^3 \\ \hline & & 0 & 1 & 1 & - & [ 31]3 + & & & \\ \hline \end{array } $ ] the efficiency of the tetrahedral coordinates raises the question if similar relative coordinates exist for @xmath142 . as a generalization of the tetrahedral coordinates from @xcite , new symmetrized coordinates for @xmath0 particles were proposed by gusev et al @xcite . the matrix that gives the @xmath0 new coordinates @xmath221 in terms of the old ones @xmath222 reads @xmath223 , \\ \nonumber & & c = c^t,\ : c^{-1}=c,\ : c^2=i , \\ \nonumber & & a=[1-\sqrt{n}]^{-1},\ : b = a+ \sqrt{n}. \end{aligned}\ ] ] the _ cm _ coordinate is included as @xmath224 . we shall explore the properties of these coordinates under the action of permutations . all permutations are generated by products of the @xmath225 transpositions @xmath226 . we need the action of these transpositions on the new relative coordinates . we shall see in eq . [ b4 ] that under permutations @xmath227 acting on particles @xmath228 the symmetrized coordinates transform like single particle coordinates . so we focus on the remaining action @xmath229 of the first transposition @xmath230 with respect to the new coordinates . for @xmath231 we obtain for the linear action of the transposition @xmath230 on the new coordinates @xmath232 \left [ \begin{array}{lllll } 0&1&0&0&0\\ 1&0&0&0&0\\ 0&0&1&0&0\\ 0&0&0&1&0\\ 0&0&0&0&1\\ \end{array } \right ] \left [ \begin{array}{lllll } 1&1&1&1&1\\ 1&b&a&a&a\\ 1&a&b&a&a\\ 1&a&a&b&a\\ 1&a&a&a&b\\ \end{array } \right ] \\ \nonumber & & = \frac{1}{5 } \left [ \begin{array}{l\|l\|l\|l\|l } \hline 5 & 1 + 3a+b & 1 + 3a+b & 1 + 3a+b & 1 + 3a+b\\ \hline 1 + 3a+b&2b+3a^2 & a+b+2a^2+ab&a+b+2a^2+ab&a+b+2a^2+b^2\\ \hline 1 + 3a+b&a+b+2a^2+ab & 2a+2a^2+b^2&2a+a^2 + 2ab & 2a+a^2 + 2ab\\ \hline 1 + 3a+b&a+b+2a^2+ab & 2a+a^2 + 2ab & 2a+2a^2+b^2&2a+a^2 + 2ab\\ \hline 1 + 3a+b&a+b+2a^2+b^2&2a+a^2 + 2ab & 2a+a^2 + 2ab & 2a+2a^2+b^2\\ \hline \end{array } \right ] \end{aligned}\ ] ] this matrix is symmetric and orthogonal by construction , its square is the unit matrix . we compute from eqs.[es21],[es26],[es30 ] the numbers @xmath233,\ : b= -\frac{1}{4}[1 - 3\sqrt{5 } ] , \\ \nonumber & & ( 002300)=2b+3a^2=\frac{1}{8}[5-\sqrt{5 } ] , \\ \nonumber & & ( 011201)=a+b+2a^2+ab=\frac{5}{8}[-1+\sqrt{5 } ] , \\ \nonumber & & ( 020210)=2a+2a^2+b^2=\frac{5}{8}[5-\sqrt{5 } ] , \\ \nonumber & & ( 020102)=2a+a^2 + 2ab=-\frac{5}{8}[3+\sqrt{5}].\end{aligned}\ ] ] for @xmath231 , all the numbers and matrices can be decomposed into a rational part and a second part rational but proportional to @xmath234 . all these numbers form a module that closes under addition , multiplication and division , compare the example eq . we give this decomposition in the following examples . a similar module decomposition applies for any @xmath0 whose square root is not an integer . @xmath235 , b=\frac{1}{4\cdot 5 } \left [ \begin{array}{rrrrr } 4 & 4 & 4 & 4 & 4\\ 4 & -1&-1&-1&-1\\ 4 & -1&-1&-1&-1\\ 4 & -1&-1&-1&-1\\ 4 & -1 & -1&-1 & -1\\ \end{array } \right ] , \\ \nonumber & & d^{\eta}(1,2)= a^{\eta}+ \sqrt{5}b^{\eta } , \\ \nonumber & & a^{\eta}= \frac{1}{5\cdot8 } \left [ \begin{array}{rrrrr } 5\cdot 8&0 & 0 & 0 & 0\\ 0 & 5 & -5&-5 & -5\\ 0 & -5&25&-15&-15\\ 0 & -5&-15&25&-15\\ 0 & -5&-15&-15&25\\ \end{array } \right ] , \ : b^{\eta}= \frac{1}{5\cdot 8 } \left [ \begin{array}{rrrrr } 0 & 0 & 0 & 0 & 0\\ 0 & -1 & 5 & 5 & 5\\ 0 & 5 & -5 & -5 & -5\\ 0 & 5 & -5 & -5 & -5\\ 0 & 5 & -5 & -5 & -5\\ \end{array } \right].\end{aligned}\ ] ] we find with eq.[b3 ] from the values of @xmath236 for @xmath231 , @xmath237 and so the product matrix eq.[b2 ] reduces to its diagonal @xmath238 submatrix that involves only the relative coordinates @xmath239 . this is always necessary , since the transposition @xmath230 can not affect the _ cm _ coordinate @xmath224 . for the other transpositions @xmath240 we find the result @xmath241 so all transpositions @xmath56 in single particle coordinates @xmath242 act on the relative coordinates @xmath243 like transpositions of relative coordinates which we denote as @xmath244 . we summarize eq . [ b4 ] in symbolic form as @xmath245 the transposition @xmath246 , missing in eq . [ b7 ] , and acting on the symmetrized coordinates , is given by the matrix @xmath229 eq . [ b10 ] . [ [ permutations - of - symmetrized - relative - coordinates - for - n - electrons . ] ] permutations of symmetrized relative coordinates for @xmath0 electrons . ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ the results of the last subsection , in particular eq . [ b7 ] generalize to any @xmath0 . since the transpositions eq . [ b4 ] generate the subgroup @xmath247 , all permutations of particles @xmath248 are mapped by eq . [ b7 ] into permutations of the relative coordinates @xmath249 . the correspondence eq . [ b7 ] must be kept in mind when working with the new relative coordinates . we could easily antisymmetrize a polynomial state wrt to the ( n-1 ) particles @xmath250 , but must augment the antisymmetrizer @xmath251 to include particle @xmath83 , see the next section . in terms of the symmetric group , we have the subgroup @xmath247 acting on particles @xmath250 . all elements @xmath252 can be written as @xmath253 with @xmath254 , multiplied by the @xmath0 coset generators which may be chosen as transpositions @xmath255 . these coset generators can be rewritten as @xmath256 and so their matrices @xmath257 acting on relative coordinates reduce to the element @xmath229 whose matrix we compute in eq . [ b7 ] . the representations @xmath258 appearing in eq . [ b5 ] simply mean the interchange of two rows or two columns . so our main task is to give the generalization of eq . [ b2 ] for any @xmath0 . we denote the matrix elements of @xmath229 for short by six integers @xmath259^{-1 } , b = a+ \sqrt{n},\end{aligned}\ ] ] and obtain for the matrix representation in @xmath260-coordinates and general @xmath0 the result @xmath261 \left [ \begin{array}{llllll } 0&1&0&0&0&.\\ 1&0&0&0&0&.\\ 0&0&1&0&0&.\\ 0&0&0&1&0&.\\ 0&0&0&0&1&.\\ .&.&.&.&.&.\\ \end{array } \right ] \left [ \begin{array}{llllll } 1&1&1&1&1&.\\ 1&b&a&a&a&.\\ 1&a&b&a&a&.\\ 1&a&a&b&a&.\\ 1&a&a&a&b&.\\ .&.&.&.&.&.\\ \end{array } \right ] \\ \nonumber & & = \frac{1}{n } \left [ \begin{array}{l\|l\|l\|l\|l\|l } \hline ( n00000 ) & ( 1n\!-\!21000 ) & ( 1n\!-\!21000 ) & ( 1n\!-\!21000 ) & ( 1n\!-\!21000 ) & .\\ \hline ( 1n\!-\!21000)&(002n\!-\!200)&(011n\!-\!301)&(011n\!-\!301)&(011n\!-\!301)&.\\ \hline ( 1n\!-\!21000)&(011n\!-\!301)&(020n\!-\!310)&(020n\!-\!402)&(020n\!-\!402)&.\\ \hline ( 1n\!-\!21000)&(011n\!-\!301)&(020n\!-\!402)&(020n\!-\!310)&(020n\!-\!402)&.\\ \hline ( 1n\!-\!21000)&(011n\!-\!301)&(020n\!-\!402)&(020n\!-\!402)&(020n\!-\!310)&.\\ \hline .. & .. & .. & .. & .. & .\\ \hline \end{array } \right]\end{aligned}\ ] ] the matrix elements are given in the notation eq . [ b9 ] . from eq . [ b10 ] the full matrix @xmath229 is constructed for any @xmath0 : the numbers @xmath236 are inserted from eq . [ b1 ] as functions of @xmath0 . the diagonal elements indexed by @xmath262 are repeated along the diagonal , the elements in the lines on top , and columns below the diagonal , are repeated in rows to the right and in columns downwards respectively by the same functions eq . [ b9 ] of @xmath0 . again @xmath229 is orthogonal and symmetric , its square is the unit matrix . we find in general from eq . [ b1 ] for the entries of the first row and column of @xmath229 @xmath263 and so the @xmath264 matrix @xmath229 reduces to its @xmath265 diagonal submatrix acting exclusively on the relative coordinates @xmath243 . in building orbital states from the symmetrized gusev coordinates we must take care in the application of permutation operators . 99 aguilera - navarro v c , kramer p , moshinsky m + _ harmonic oscillator states and the @xmath266-particle ii . configuration - space states of arbitrary symmetry . _ ann of physics ( n y ) * 54 * ( 1969 ) 379393 avron j e , herbst i w , and simon b + _ separation of center of mass in homogeneous magnetic fields _ annals of physics * 114 * ( 1978 ) 431451 bargmann v and moshinsky m + _ group theory of harmonic oscillators ( i ) . _ + nuclear physics * 18 * ( 1960 ) 697712 bargmann v and moshinsky m + _ group theory of harmonic oscillators ( ii ) . _ + nuclear physics * 23 * ( 1961 ) 177199 barut a o and raczka r + _ theory of group representations and applications . _ + pwn polish scientific publishers , warsaw ( 1977 ) edmonds a r + _ angular momentum in quantum mechanics . _ + princeton university press . princeton ( 1957 ) escobar - ruiz m a and turbiner a v + _ two charges on plane in a magnetic field i. `` quasi - equal '' charges and neutral quantum system at rest cases . _ + ann phys * 340 * ( 2014 ) 3759 grossmann f and kramer t + _ spectra of harmonium in a magnetic field using an initial value representation of the semiclassical propagator . _ + j phys a * 44 * ( 2011 ) 445309 gusev a et al . + _ symbolic - numerical algorithm for generating cluster eigenfunctions : identical particles with pair oscillator interactions . _ in : computer algebra in scientific computing , springer , berlin ( 2013 ) 155168 hamermesh m + _ group theory and its application to physical problems . _ + addison - wesley , reading ( 1962 ) 182278 johnson b r , hirschfelder j o , and yang k - h + _ interaction of atoms , molecules , and ions with constant electric and magnetic fields . _ rev mod phys * 55 * ( 1983 ) 109153 klama s and mishchenko e + _ two electrons in a quantum dot : a semiclassical approach . _ + j. phys . : condens . matter * 10 * ( 1998 ) 34113416 kohn w + _ cyclotron resonance and de haas - van alphen oscillations of an interacting electron gas . _ + phys rev * 123 * ( 1961 ) 12421244 kramer p and moshinsky m + _ group theory of harmonic oscillators ( iii ) : states with permutational symmetry . _ nuclear physics * 82 * ( 1966 ) 241273 kramer p , john g and schenzle d + _ group theory and the interaction of composite nucleon systems . _ + vieweg , braunschweig ( 1971 ) kramer p and moshinsky m + _ group theory of harmonic oscillators ( iv ) : dynamics of the 2,3 , and 4 particle systems . _ nuclear physics * a 107 * ( 1968 ) 481522 kramer p and moshinsky m + _ group theory of harmonic oscillators and nuclear structure . _ in : group theory and its application , + ed e m loebl , academic press , new york ( 1968 ) 340368 kramer p + _ permutation group in light nuclei . in : symmetry properties of nuclei , proc 15th solvay conf . on physics 1970 _ , gordon and breach , new york ( 1974 ) 211232 kramer p , kramer t , and manko v i + _ propagation in time of some classical / quantum systems and nielsen s conserved quantity . _ + phys scr * 79 * ( 2009 ) 055006 kramer t + _ two interacting electrons in a magnetic field : comparison of semiclassical , quantum , and variational solutions . _ + aip conference proceedings * 1323 * ( 2010 ) 178 kramer t , krueckl v , heller e j and parrott r e , + _ self - consistent calculation of electric potentials in hall devices . _ + phys rev b * 81 * ( 2010 ) 205306 kramer t , kreisbeck c , krueckl v , heller e j , parrott r e and liang c - t + _ theory of the quantum hall effect in finite graphene devices . _ + phys rev b * 81 * ( 2010 ) 081410(r ) kramer t + _ interacting electrons in a magnetic field : mapping quantum mechanics to a classical ersatz - system . _ + aip conference proceedings * 1488 * ( 2012 ) 394 laughlin r + _ anomalous quantum hall effect : an incompressible quantum fluid with fractionally charged excitations . _ + phys rev lett * 50 * ( 1983 ) 13951398 moshinsky m and quesne c + _ oscillator systems . in : symmetry properties of nuclei , proc 15th solvay conf . on physics 1970 . _ gordon and breach , new york ( 1974 ) 233300 moshinsky m and smirnov y + _ the harmonic oscillator in modern physics . _ + harwood academic publishers , amsterdam ( 1996 ) muhandiramge r and wang j + _ electronic structure of multi - electron quantum dots . _ + mathematica journal * 10:2 * ( 2006 ) 396414 muhandiramge r d and wang j + _ spinadapted states : a basis for quantum dot structure calculation . _ international journal of quantum chemistry * 106.1 * ( 2006 ) 2753 su b , goldman v j , and cunningham j e + _ evidence for spin singlet - triplet transitions of two electrons in a quantum dot observed via single - electron tunneling _ surf sci * 305 * ( 1994 ) 566 - 570 taut m + _ two electrons in a homogeneous magnetic field : particular analytical solutions . _ + j phys a : math gen * 27 * ( 1994 ) 47234724 tavernier m b et al . + _ four - electron quantum dot in a magnetic field . _ + phys . rev b * 68 * ( 2003 ) 205305 simonovi n s + _ three electrons in a two - dimensional parabolic trap : the relative motion solution . _ + few - body systems * 38 * ( 2006 ) 139145 verin a + _ two anyons in a static , uniform magnetic field . exact solution . _ + physics letters b * 260 * ( 1991 ) , 120124 wagner m , merkt u , and chaplik a v + _ spin - singlet spin - triplet oscillations in quantum dots . _ + phys rev b * 45 * ( 1992 ) 19511954 weyl h + _ the theory of groups and quantum mechanics . _ dover , new york ( 1930 ) wigner e p + _ group theory and its application to the quantum mechanics of atomic spectra . _ academic press , new york ( 1959 ) yannouleas c and landman u + _ quantal molecular description and universal aspects of the spectra of bosons and fermions in the lowest landau level . _ a * 81 * ( 2010 ) 023609 yukich j n , kramer t , and bracher c + _ observed photodetachment in parallel electric and magnetic fields . a * 68 * ( 2003 ) 033412
in theoretical astrophysics the steady increase in computational power has sparked a proportional interest and progress in the study of large - scale structure formation . in particular , as @xmath0-body simulations of cold dark matter halos have become more detailed , several universal properties have emerged . we highlight three important characteristics . firstly , numerous cosmological studies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) revealed similar density profiles over several orders of magnitude in halo mass , with a central cusp and an a @xmath7 falloff at large radii . these generalized nfw models can be described by a general 3-parameter family , referred to as the zhao models ( or @xmath8-models ) @xcite . they are defined by the density @xmath9 or in terms of the logarithmic slope , @xmath10 although in recent years alternative profiles based on the srsic law have produced equally good results @xcite , the generalized nfw models remain very popular and successful to represent dark matter halos . a second relation was found by @xcite . these authors identified that the quantity @xmath11 , which has become known as the pseudo phase - space density , behaves as a power law over 2 - 3 orders of magnitude in radius inside the virial radius , @xmath12 other studies ( e.g. * ? ? ? * ; * ? ? ? * ) have confirmed the scale - free nature of @xmath13 , and their results indicate that its slope lies in the range @xmath14 . this property is remarkable since the density @xmath15 nor the velocity dispersion @xmath16 separately show a power - law behavior . finally , the velocity anisotropy profiles @xmath4 of dark matter systems also evolve toward a similar shape , steepening gradually from isotropic in the center to radially anisotropic in the outer regions . @xcite suggest a nearly linear relation between the logarithmic density slope @xmath17 and the velocity anisotropy profile , based on various types of equilibrated simulations . they proposed the @xmath2 relation @xmath18 several theoretical studies have been made to investigate whether solutions of the jeans equation exist that encompass the observed properties of dark matter halos . in particular , @xcite investigated the anisotropic jeans equation constrained by a slightly different form of the pseudo phase - space density , namely @xmath19 with @xmath20 the radial velocity dispersion . they found a special solution , namely an analytical self - consistent potential - density pair of the form ( [ zhao ] ) with an anisotropy profile @xmath21 that also has an exactly linear @xmath2 relation ( in other words , @xmath22 and @xmath23 ) . as these jeans models satisfy the three universal relations mentioned above , we can consider them as representative models for realistic dark matter halos . the goal of this paper is to take the analytical study of dark matter systems a step further , i.e. to look for full dynamical models that encompass the universal properties found in @xmath0-body simulations . in concreto we will look for phase - space distribution functions ( dfs ) @xmath24 that self - consistently generate the required density , potential , and anisotropy profiles encountered in dark matter halos . such dynamical models would provide a very useful complementary approach to gain insight into the structure of dark matter halos . we shall focus on spherical models , for which a dynamical description simplifies substantially as in this case the df can be expressed as a function @xmath25 of the binding energy and the angular momentum . but even if we limit ourselves to the spherical case , it is not straightforward to obtain such dynamical models . a simple approach is to solve the jeans equation and approximate the velocity dispersion profiles by a multivariate gaussian @xcite . @xcite demonstrated however that these systems are far from equilibrium , thus losing their initial dynamical structure as they evolve to non - gaussian velocity distributions in subsequent @xmath0-body simulations . we therefore need the tools to derive full self - consistent equilibrium models , governed by dfs with a sufficiently general velocity anisotropy profile . simple analytical models are found for only a few special cases , such as the plummer , hernquist , isochrone , and @xmath26-models , none of which are able to describe realistic dark matter halos . for general potential - density pairs , @xcite and @xcite provided an algorithm to construct dynamical models with a specific velocity anisotropy profile of the form ( [ betagen ] ) , with @xmath27 , @xmath28 and @xmath29 , leaving @xmath30 as a free parameter . while the osipkov - merritt method has been widely adopted ( e.g. * ? ? ? * ; * ? ? ? * ) , comparison with simulations shows that such anisotropy profiles are too steep to describe dark matter @xcite . halos are not completely radial at infinity ( i.e. @xmath31 ) and the linear @xmath2 relation can only be realized with a lower transition rate @xmath32 . on a dynamical note , the associated dfs are of the form @xmath33 with @xmath34 , creating an unphysical cut - off boundary for orbits with @xmath35 . hence , the osipkov - merritt framework is too limited to generate realistic dark matter systems , and a more extensive method is needed . recently , @xcite presented an interesting approach to generate dynamical models for potential - density pairs with a more general anisotropy profile . they proposed to express the df as a separable function of the form @xmath36 where @xmath37 is a double power - law function with three parameters @xmath5 , @xmath6 , and @xmath38 . once their values have been determined , the function @xmath39 is derived from the observed density profile by a numerical inversion . this technique yields a three - parameter anisotropy profile that resembles eq . ( [ betagen ] ) , where @xmath38 has a similar role as @xmath30 , and a fixed transition rate @xmath40 . in this manner , the authors were able to construct models with an nfw density with velocity dispersion profiles that agreed with their dark matter simulations . in this paper , we present a technique that enables us to obtain dynamical models with exactly the four - parameter anisotropy profile ( [ betagen ] ) . we demonstrated in a previous paper how this can be achieved , by postulating a separable parameterized form of the so - called augmented density , which provides an equivalent description of a dynamical system . in certain special cases the transformation from the augmented density to the df is analytically tractable . we were thus able to derive a family of dfs for the generalized plummer models ( also called @xmath41-models or veltmann models , @xcite ) with a linear @xmath2 relation . since the generalized plummer potential - density pairs form a special class of the zhao models ( [ zhao ] ) , we will now demonstrate that this result is a first step toward more representative dark matter profiles . in particular , we seek a family of dfs for the dehnen - mclaughlin systems , because of their unique property to satisfy a universal density , a power law @xmath42 , and a linear @xmath2 relation . we use a quadratic programming technique @xcite to build dfs as a linear combination of base functions of the form derived in . in this manner , we demonstrate that it is indeed possible to generate full dynamical models for the dehnen - mclaughlin jeans models , thus providing dfs that encompass the observed properties of dark matter halos . our paper is organized as follows . in section [ preliminaries.sec ] we describe the notion of a dynamical model and we briefly summarize the main aspects of the dehnen - mclaughlin halos . next we outline our modeling technique in section [ modeling.sec ] : we explain the quadratic programming algorithm and we recapitulate the functions that we derived in . with these components , we can build a library to apply the qp - method to the dehnen - mclaughlin halos . in section [ results.sec ] we present our results for a set of systems with different velocity anisotropy profiles , and we discuss the moments , phase - space dfs , and energy and angular momentum distributions for these models . finally , we formulate our conclusions in section [ conclusions.sec ] . the dynamical structure of a gravitational equilibrium system is completely determined by the df @xmath24 , which describes the probability distribution of particles in six - dimensional phase space . a dynamical model is only physical if its df is non - negative everywhere . in the case of spherical symmetry , this df can be written as a function @xmath25 of two isolating integrals , namely the binding energy and the angular momentum @xmath43 with @xmath44 the transverse velocity , and @xmath45 the positive binding potential . for circular orbits the integrals of motion can be written in function of the radius @xmath46 , @xmath47 which can be solved to obtain @xmath48 and @xmath49 . all dynamical properties can be derived from the df , such as the anisotropic velocity moments @xmath50 with @xmath51 the total mass . if the system is self - consistent , then the density @xmath52 is connected to the potential via the poisson equation @xmath53 furthermore , the second - order moments determine the radial and tangential velocity dispersions @xmath54 and @xmath55 , and the velocity anisotropy profile @xmath56 a df is only fully determined when all velocity moments ( [ momdf ] ) are known @xcite . however , it is already far from trivial to obtain _ any _ non - negative df that adequately generates a given density @xmath15 and anisotropy profile @xmath4 . in this paper , we will focus our attention to one specific family of halos , namely the dehnen - mclaughlin jeans models , and we demonstrate that a family of dfs can indeed be constructed for these systems . in the context of theoretical dark matter studies , the model derived by @xcite is of particular interest . we summarize their main results in this section . instead of fitting a parameterized density profile to @xmath0-body simulations , they investigated the solution space of the jeans equation to search for models that explicitly obey the power - law behavior of the pseudo phase - space density . with the extra condition of a linear @xmath2 relation they found a critical solution that satisfies the condition @xmath57 with @xmath58 a scale radius in the remainder of this paper , we adopt the common value @xmath59 and use the notation @xmath19 . dehnen & mclaughlin derived for this case the exponent @xmath60 with @xmath5 the central velocity anisotropy . the corresponding potential - density pair is fully analytical , given by @xmath61 where @xmath62 , @xmath63 is the incomplete beta function and @xmath64 the density can be equivalently written in terms of the slope @xmath17 , which has the same elegant form as the velocity anisotropy profile @xmath4 @xmath65 finally , the authors derived the corresponding velocity dispersions @xmath66 eqs . ( [ pseudodens])-([sigmatdehnen ] ) characterize the dehnen - mclaughlin models ; we refer to their paper for more details . these systems are determined by five parameters : the exponent @xmath67 in the pseudo phase - space density , two scaling constants i.e. the total mass @xmath68 and the scale radius @xmath58 , and the asymptotic anisotropy parameters @xmath5 and @xmath6 . the authors also noted the remarkable property that the shape of the density profile ( and hence the gravitational potential ) only depends on @xmath5 and not on @xmath6 . while the dehnen - mclaughlin jeans models are derived from theoretical considerations , they also fit adequately galaxy - sized and certain cluster - sized halos generated by @xmath0-body simulations @xcite . but as mentioned in the previous section , the density and velocity dispersions alone do not determine the complete dynamical state of dark matter systems . we therefore aim to incorporate these profiles into self - consistent dynamical models , described by non - negative dfs . to find dfs that describe the dehnen - mclaughlin jeans models , we adopt the mathematical framework that we derived in . these tools enabled us to construct a family of components with the anisotropy profile ( [ betagen ] ) , by means of the powerful augmented density concept . we summarize these results , and we demonstrate how to build a linear combination of these components using a quadratic programming ( qp ) technique @xcite . in this manner , we can fit a dynamical model to a given halo . since our dynamical models have to reproduce the moments ( [ rhodehnen ] ) , ( [ sigmardehnen ] ) and ( [ sigmatdehnen ] ) , we first seek dfs that are specifically designed for this task . this can be done by introducing the augmented densities @xmath69 @xcite , which extend the densities to explicit functions of both the radius and the gravitational potential . like the df , the augmented density uniquely determines the dynamical state of a spherical equilibrium system . both functions are connected by the relation @xmath70 if we define that all functions are zero when their arguments lie outside their physical bounds , we can use the laplace - mellin transforms @xmath71 with the inverse transforms @xmath72 combining these expressions with ( [ augdens ] ) , we obtain @xmath73 interchanging the integrals yields the formal relation @xmath74 the main strength of the augmented density formalism is the ability to impose very specific conditions on these functions , to obtain our objective . more precisely , we demonstrated in that separable functions of the form @xmath75 are particularly interesting , since in this case @xmath76 is directly related to the velocity anisotropy profile , @xmath77 with this formalism it is therefore at least formally possible to generate a df for a dynamical system with a given potential @xmath45 , density @xmath15 , and velocity anisotropy profile @xmath4 . if we postulate an anisotropy profile of the form ( [ betagen ] ) , the modeling procedure reduces to the computation of @xmath78 such that @xmath79 with @xmath80 and @xmath81 evidently , the exact solution @xmath78 is generally of numerical form , except for a few special cases . this creates a problem to recover the df from the augmented density , because a direct inversion of the laplace - mellin transformation in ( [ rhodf ] ) is in general numerically unstable @xcite . indeed , the double integration in ( [ augdens ] ) smooths features in @xmath25 and the inversion procedure of determining @xmath25 from @xmath69 has the delicate job of unsmoothing the information contained in @xmath69 . consequently , a direct inversion can only be performed safely for sufficiently simple forms of @xmath78 . we therefore take an alternative approach . instead of solving ( [ rhoisrho ] ) directly , we will approximate the given density profile by a linear combination of simple base functions @xmath82 for which the laplace - mellin inversions are analytical . the corresponding df is then simply the same linear combination of the associated base dfs @xmath83 , resulting in an analytically tractable function . this can be achieved by a least squares fit to a set of density data points , defining a quadratic programming problem in the unknown coefficients . various authors have successfully used a similar modeling technique @xcite . a qp - algorithm , suited for this task , has been developed in our department @xcite , which we outline in the following section . consider a given potential @xmath45 , an anisotropy profile @xmath4 of the form ( [ betagen ] ) and a set of @xmath84 density data points @xmath85 , @xmath86 . to model these data , we first construct a library of @xmath87 base functions @xmath88 with @xmath89 sufficiently simple to compute the associated distributions @xmath83 . from these components , we can extract the corresponding values @xmath90 our aim is now to construct a linear combination of @xmath0 components from this library that provides an adequate fit to the given data . such a fit can be obtained in a statistically meaningful way by minimizing the quantity @xmath91 which is a quadratic function of the coefficients @xmath92 , consequently defining a quadratic programming problem . the data points are given equal weights by setting the constants @xmath93 . to find such a set , we use an iterative algorithm @xcite . with this method , a set of components is successively built in @xmath0 steps . in the first step , @xmath87 @xmath94-minimizations are performed using in turn each component of the library . the base function that yields the lowest @xmath94-value ( hereafter denoted @xmath95 ) is retained as the first element of our best - fitting set , with corresponding coefficient @xmath96 . in each next iteration , the functions in this set are preserved , while the coefficients are allowed to vary . the set is subsequently extended by adding the component from the library that yields the most improvement of the fit , minimizing over all coefficients . in other words , suppose we have obtained the best - fitting set of @xmath97 base functions , with indices @xmath98 . then , we add in turn the remaining @xmath99 components from the library , and calculate the @xmath0 coefficients for each combination by means of the @xmath94-minimizations @xmath100 from these @xmath99 values , we determine the best fit @xmath101 and add the base function with index @xmath102 to the best - fitting set , denoting @xmath103 . by renaming the indices of this set , we thus obtain a linear combination of @xmath0 base functions @xmath104 with coefficients @xmath92 and a goodness of fit @xmath105 . the corresponding df is then simply @xmath106 the qp - algorithm allows additional linear constraints on the coefficients . in particular , we impose upper and lower boundaries @xmath107 these constraints are not necessary , but they greatly reduce the computational cost in the calculation of the df . the reason for this is straightforward : if the components are computed with numerical errors @xmath108 then the total numerical error of the df is @xmath109 the higher the absolute values of the coefficients @xmath110 , the smaller the errors @xmath108 need to be to obtain a given @xmath111 , which increases the computational time . sensible boundary values ( [ qp4 ] ) enable efficient calculations of the df , while maintaining satisfactory fits . the resulting df also needs to be physical , i.e. non - negative everywhere in phase space . we found that all our dfs automatically satisfy this condition without imposing explicit constraints . this procedure has several advantages . the resulting df remains analytically tractable , which also simplifies the computation of all the moments of this dynamical model . furthermore , only a limited number of data points are required , rather than the entire density profile . the same algorithm can also be applied to data extracted from simulations or observations , and a variety of different moments besides the density can be used in the fitting procedure . finally , since all components have a priori the desired @xmath4 , their linear combination will automatically generate the same anisotropy profile . this is a very significant benefit . indeed , practice has shown that it is particularly difficult to construct radially anisotropic systems from base functions that are too simple ( such as fricke components , @xcite ) with different constant anisotropies . in that case an additional fitting is required to the second - order moments . but more importantly , because radial orbits influence the density both at small and large radii , the summation of components with different anisotropies results into a delicate fine - tuning to obtain adequate fits to both the density and the velocity anisotropy . these problems are avoided if the components already have the correct anisotropy profile , and although it is more intricate to design such functions , this approach greatly reduces the complexity of the quadratic programming procedure . we are now left with the construction of a library of adequate base functions @xmath104 . the success of our modeling is largely determined by this library . as stated above , the components have to generate the anisotropy profile ( [ betagen ] ) , while being sufficiently simple to retrieve the corresponding dfs @xmath83 from eq . ( [ rhodf ] ) . in addition , the subsequent densities @xmath112 need to be able to reproduce the specific characteristics of the given data to obtain a satisfactory fit . in the particular case of the dehnen - mclaughlin halos ( [ rhodehnen ] ) , this implies that the components should incorporate the central density cusp and the asymptotic density slope at large radii , using the given potential ( [ psidehnen ] ) . in we derived a family of base functions that meet all these requirements . consider the family of augmented densities @xmath113 where @xmath114 are normalization constants such that @xmath115 and @xmath116 denotes the depth of the ( finite ) potential well , @xmath117 . the normalization constants are chosen such that the sum of the coefficients of a good fit should approximate the total mass of the input data , i.e. @xmath118 . apart from the four fixed parameters that determine the anisotropy profile ( [ betadehnen ] ) , the @xmath78-part of these functions contains three additional free parameters @xmath119 , @xmath120 and @xmath121 , which respectively determine the asymptotic behavior at infinity , the inner slope , and the transition rate between these two regions . they satisfy the conditions @xmath122 , @xmath123 and @xmath124 . all other dynamical properties can be calculated from these augmented densities . the augmented higher - order moments are derived in appendix [ augmom.sec ] , and we demonstrated in the appendices of that for these @xmath125 the inverse laplace - mellin transforms can be performed analytically in eq . ( [ rhodf ] ) . the corresponding distribution functions can be expressed as a series of fox @xmath126-functions @xcite : @xmath127 \left(-\frac{\beta_0}{\delta},\frac{1}{\delta}\right ) & \left(0,1\right ) \end{hmatrix } \right.\!\ ! \right ) , \label{dffoxh}\end{aligned}\ ] ] which can be written as a double series @xmath128 where we used the auxiliary notation @xmath129 the double summation @xmath130 can be computed by changing the indices to @xmath131 , so that the inner summation becomes a finite sum of @xmath132 terms for each value of @xmath133 ; the index @xmath133 is increased until the total sum alters by less than a required numerical error @xmath108 . due to the double summation , the computational time is an inverse quadratic function of @xmath108 . we proved in that these base dfs are continuous and non - negative everywhere in physical phase space . moreover , we showed that they generate exactly the generalized plummer potential - density pairs @xcite @xmath134 with a linear @xmath2 relation . since these systems are closely related to the dehnen - mclaughlin halos , this is a promising result for the success of the quadratic programming routine . the fortran source code with the df base functions and the augmented moments is available on request . every given dehnen - mclaughlin halo requires a specific component library . in particular , the parameters @xmath119 , @xmath120 are constrained by the potential . if we examine the asymptotic behavior of the dehnen - mclaughlin potential ( [ psidehnen ] ) in more detail , we find @xmath135 introducing these asymptotic expansions in the expression ( [ rhofam ] ) we find for the inner and outer slopes of the density @xmath136 evidently , the parameters @xmath119 stipulate the density slope at large radii . because the models fall as @xmath137 , no components can be used in the fitting routine that fall less rapidly . using eq . ( [ gammadehneni ] ) , this puts a boundary on the @xmath119 , @xmath138 conversely , the density slope at small radii depends on the parameters @xmath120 . the density diverges toward the center as @xmath139 , and we can not use components in the fitting routine that have a steeper slope . thus we obtain from eq . ( [ gammadehnen0 ] ) @xmath140 so if a fit to a halo has at least one component with parameter @xmath141 and one with @xmath142 , this fit has the same slope as the given density at small and large radii . finally , the parameters @xmath121 have a similar role as @xmath143 , in the sense that they control the transition rate between the inner and outer density slopes . their value can be chosen freely , but we found that excellent results are obtained with a single fixed value @xmath144 for all components . this is the same choice as the generalized plummer models in . it also simplifies the computation of the dfs ( [ dfbase1])-([dfbase2 ] ) , and it further facilitates the fitting process , leaving only @xmath119 and @xmath120 as free parameters . crrrrrrrrrr @xmath145 & @xmath146 & @xmath147 & @xmath148 & @xmath149 & @xmath150 & @xmath151 & @xmath152 & @xmath153 & @xmath154 & @xmath155 + @xmath114 & @xmath156 & @xmath157 & @xmath158 & @xmath159 & @xmath160 & @xmath161 & @xmath162 & @xmath163 & @xmath164 & @xmath165 + @xmath119 & @xmath166 & @xmath167 & @xmath168 & @xmath167 & @xmath169 & @xmath167 & @xmath170 & @xmath167 & @xmath167 & @xmath168 + @xmath120 & @xmath171 & @xmath171 & @xmath172 & @xmath173 & @xmath171 & @xmath174 & @xmath171 & @xmath175 & @xmath172 & @xmath171 + @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath180 & @xmath181 & @xmath182 & @xmath183 & @xmath184 & @xmath185 & @xmath186 + @xmath145 & @xmath187 & @xmath188 & @xmath189 & @xmath190 & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 & @xmath196 + @xmath114 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 & @xmath205 & @xmath206 + @xmath119 & @xmath166 & @xmath207 & @xmath167 & @xmath167 & @xmath167 & @xmath207 & @xmath207 & @xmath167 & @xmath167 & @xmath169 + @xmath120 & @xmath171 & @xmath172 & @xmath171 & @xmath173 & @xmath172 & @xmath171 & @xmath173 & @xmath208 & @xmath174 & @xmath171 + @xmath176 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 & @xmath216 & @xmath185 & @xmath217 + @xmath145 & @xmath218 & @xmath219 & @xmath220 & @xmath221 & @xmath222 & @xmath223 & @xmath224 & @xmath225 & @xmath226 & @xmath227 + @xmath114 & @xmath228 & @xmath229 & @xmath230 & @xmath231 & @xmath232 & @xmath233 & @xmath234 & @xmath235 & @xmath236 & @xmath237 + @xmath119 & @xmath166 & @xmath238 & @xmath239 & @xmath239 & @xmath238 & @xmath240 & @xmath239 & @xmath166 & @xmath239 & @xmath166 + @xmath120 & @xmath171 & @xmath172 & @xmath171 & @xmath173 & @xmath171 & @xmath171 & @xmath175 & @xmath174 & @xmath174 & @xmath173 + @xmath176 & @xmath209 & @xmath241 & @xmath242 & @xmath212 & @xmath243 & @xmath244 & @xmath184 & @xmath245 & @xmath246 & @xmath247 + @xmath145 & @xmath248 & @xmath249 & @xmath250 & @xmath251 & @xmath252 & @xmath253 & @xmath254 & @xmath255 & @xmath256 & @xmath257 + @xmath114 & @xmath258 & @xmath259 & @xmath260 & @xmath261 & @xmath156 & @xmath262 & @xmath263 & @xmath264 & @xmath265 & @xmath266 + @xmath119 & @xmath267 & @xmath268 & @xmath239 & @xmath239 & @xmath268 & @xmath268 & @xmath240 & @xmath267 & @xmath240 & @xmath240 + @xmath120 & @xmath171 & @xmath172 & @xmath171 & @xmath173 & @xmath171 & @xmath173 & @xmath171 & @xmath173 & @xmath175 & @xmath208 + @xmath176 & @xmath177 & @xmath269 & @xmath270 & @xmath271 & @xmath272 & @xmath273 & @xmath245 & @xmath215 & @xmath274 & @xmath275 + @xmath145 & @xmath276 & @xmath277 & @xmath165 & @xmath278 & @xmath279 & @xmath280 & @xmath281 & @xmath282 & @xmath283 & @xmath284 + @xmath114 & @xmath285 & @xmath286 & @xmath287 & @xmath288 & @xmath289 & @xmath290 & @xmath291 & @xmath292 & @xmath293 & @xmath294 + @xmath119 & @xmath295 & @xmath296 & @xmath239 & @xmath239 & @xmath296 & @xmath296 & @xmath170 & @xmath295 & @xmath296 & @xmath296 + @xmath120 & @xmath171 & @xmath172 & @xmath171 & @xmath173 & @xmath171 & @xmath174 & @xmath175 & @xmath175 & @xmath173 & @xmath208 + @xmath176 & @xmath297 & @xmath298 & @xmath299 & @xmath300 & @xmath301 & @xmath302 & @xmath303 & @xmath304 & @xmath305 & @xmath247 + @xmath145 & @xmath306 & @xmath307 & @xmath308 & @xmath309 & @xmath310 & @xmath311 & @xmath312 & @xmath313 & @xmath314 & @xmath315 + @xmath114 & @xmath316 & @xmath317 & @xmath318 & @xmath319 & @xmath320 & @xmath321 & @xmath322 & @xmath323 & @xmath324 & @xmath325 + @xmath119 & @xmath326 & @xmath170 & @xmath326 & @xmath326 & @xmath239 & @xmath240 & @xmath326 & @xmath170 & @xmath170 & @xmath295 + @xmath120 & @xmath171 & @xmath171 & @xmath173 & @xmath175 & @xmath171 & @xmath172 & @xmath174 & @xmath173 & @xmath172 & @xmath208 + @xmath176 & @xmath327 & @xmath328 & @xmath329 & @xmath330 & @xmath331 & @xmath273 & @xmath245 & @xmath332 & @xmath333 & @xmath334 + now that we have derived the necessary mathematical tools , we can present the results for the dehnen - mclaughlin halos . without loss of generality , we can work in dimensionless units @xmath335 , and we limit ourselves to @xmath59 . consequently , the models are determined by the anisotropy parameters @xmath5 and @xmath6 . although we are able to generate models with arbitrary values for these parameters , realistic dark matter halos are nearly isotopic near the center and radially anisotropic at large radii , so that we concentrate on six representative models with @xmath27 and @xmath336 , 0.2 , 0.4 , 0.6 , 0.8 , 1 . we verified that the modeling procedure works equally well for models with non - zero values of @xmath5 . finally , it is evident from eq . ( [ etadehnen ] ) that @xmath27 sets the parameters @xmath337 . as we demonstrated above , the very specific form of the base functions ( [ rhofam ] ) simplifies our qp - algorithm considerably for these models . only the parameters @xmath119 and @xmath120 remain to construct a library of components , and we have found that only 30 components are sufficient to yield excellent fits . the parameters @xmath119 take five values , ranging from @xmath338 to 10 or 12 , depending on the model . the parameters @xmath120 take six values from @xmath339 to @xmath340 . we list these values for each model in the headers of table [ models.tbl ] . we extract @xmath341 values of the density ( [ rhodehnen ] ) at radii @xmath342 , distributed logarithmically between and , that serve as input data @xmath85 in each qp - procedure . evidently , this range is much larger than the virialized region in @xmath0-body simulations . this larger range is therefore not intended to be realistic , but rather to demonstrate that our models are accurate to arbitrary distances . furthermore , this makes it possible to create discrete equilibrium systems from the dfs , by means of monte carlo simulators , that trace very closely the dehnen - mclaughlin halos . after calculating the densities ( [ rhofam ] ) of every library component at these radii @xmath343 , we can perform the qp - algorithm for the six values of @xmath6 , constructing iteratively the best - fitting linear combination ( [ qpbest ] ) of @xmath0 components with additional constraints of the form ( [ qp4 ] ) , @xmath344 we show the results for the six models in table [ models.tbl ] . the columns list the components of the fits . every @xmath345 denotes the goodness of fit of the best linear combination of the components in columns @xmath346 to @xmath0 . each component is determined by the parameters @xmath119 , and @xmath120 , which in turn define the normalization constants @xmath114 from eq . ( [ normalize ] ) . the coefficients are only given for the final fit with 10 components , i.e. @xmath145 . it can be checked that for each model @xmath347 indicating excellent fits . combining this result with eqs . ( [ qpacc ] ) and ( [ qp4b ] ) , it can be seen that if @xmath348 , the numerical errors of the base functions @xmath108 need at most be a factor @xmath349 smaller than a given error @xmath350 , allowing efficient computations of the df with sufficient accuracy . the resulting @xmath345 for each model are also displayed in fig . [ f1.eps ] . evidently , @xmath348 components are more than sufficient to obtain very accurate dynamical models . as an example , fig . [ f2.eps ] shows the 10 individual components of the qp - model with @xmath351 . although this fit has the highest @xmath352 of our set , its total density is a very close approximation to the given data over the entire range in radius . [ f3.eps ] displays several moments for our six models , with 10 components . the top row shows the density @xmath15 , the pseudo phase - space density @xmath42 , and the @xmath2 relation . below each graph , we calculated the residual errors between the qp - fits and the theoretical curves , i.e. for each profile @xmath353 we have @xmath354 as can be seen , the relative errors on the densities are less than @xmath355 along 7 orders of magnitude in radius , and the exact asymptotic slopes of the models ensure excellent fits even beyond this range . the power - law trend of @xmath42 and the @xmath2 relations are also reproduced very accurately with errors @xmath356 . note that the small offset between the pseudo phase - space density profiles for the different models is due to the dependence of @xmath20 on @xmath6 . in the central row , we display the velocity dispersion profiles @xmath20 , @xmath357 , and the anisotropies @xmath4 . it is striking that , while the dispersion data were not used in the fit , the deviations of these moments from the theoretical values are even smaller , less than @xmath358 . evidently , since the models have the anisotropies ( [ betadehnen ] ) by construction , the @xmath4 profiles are exact , without errors . note also that all tangential velocity dispersion profiles @xmath357 intersect at a common radius @xmath359 . while the density and dispersions are defined by the dehnen - mclaughlin halos , the higher - order moments are determined by the qp - models . the fourth - order moments ( see appendix [ augmom.sec ] ) allow us to derive the radial and tangential kurtosis , @xmath360 and the fourth - order anisotropy @xmath361 interestingly , as a result of the separable form of the augmented densities , we find that the @xmath362 profiles are only a function of the @xmath4 , @xmath363 these profiles are shown in the bottom row of fig . [ f3.eps ] . the kurtosis values describe the non - gaussianity of the velocity distributions at a certain radius . our radial kurtosis values are very large in the center , which indicates that the @xmath364-distributions are very peaked ( leptokurtic ) at small radii . the @xmath365 curves decrease rapidly as a function of radius : they reach zero at radii between 0.26 - 0.36 and become negative at larger radii , leading to flat - topped ( platykurtic ) radial velocity distributions . this behavior is in accordance with @xmath0-body simulations @xcite . clearly , the value of @xmath6 has little influence on the radial kurtosis , as in the case of @xmath42 . in contrast , the tangential kurtosis @xmath366 curves do depend significantly on @xmath6 . all @xmath367-distributions are highly peaked at small radii . for @xmath368 the tangential kurtosis decreases to slightly negative values , i.e. at larger radii the tangential velocity distributions become slightly flat - topped . models with @xmath369 have nearly gaussian @xmath367-distributions for @xmath370 . if @xmath371 , the @xmath366 profiles reach a minimum value and increase again for larger radii . the six top panels of fig . [ f4.eps ] show the distribution functions @xmath25 of our radially anisotropic systems with @xmath27 and @xmath372 , expressed as logarithmic isoprobability contours and a logarithmic color gradient in the integral space , with @xmath373 scaled to @xmath49 , denoting the angular momentum of a circular orbit with energy @xmath374 . all models are clearly physical , i.e. the dfs are non - negative everywhere . this means that the dehnen - mclaughlin jeans models can actually be realized by full dynamical models . moreover , contrary to the osipkov - merritt models , these functions fill the entire integral space . in the isotropic case , the contours are horizontal ( no dependence on angular momentum ) , and their orientation alters gradually with increasing @xmath6 in an intuitive way , as orbits with high eccentricities ( i.e. low angular momentum ) become more abundant . the dfs describe the probability distributions of particles in phase space , but not in the integral space . it is more physically meaningful to consider the true orbital distributions @xmath375 with the so - called density of states function @xmath376 where @xmath377 and @xmath378 are the pericenter and apocenter of an orbit with energy @xmath374 and angular momentum @xmath373 . in other words , the functions @xmath379 express the likelihood of an orbit with energy @xmath374 and scaled angular momentum @xmath380 . the results are displayed in the bottom panels of fig . [ f4.eps ] as logarithmic isoprobability contours and a logarithmic color gradient in the integral space . for high binding energies , all orbital distributions contain increasing probabilities toward circular orbits ( high angular momentum ) . for low binding energies , i.e. at large radii , the abundance of orbits with low angular momentum gradually increases from the isotropic case to models with high values of @xmath6 . we conclude the discussion of our dehnen - mclaughlin dfs with an analysis of the marginal distributions . the differential energy and angular momentum distributions are the integrals of the orbital distributions , @xmath381 with @xmath382 these curves are displayed in fig . [ f5.eps ] . the differential energy distributions are all monotonously decreasing functions of @xmath374 . it is striking that these profiles are almost identical , regardless of the anisotropy @xmath6 . this result reinforces previous dynamical studies @xcite , and suggests that @xmath15 , @xmath42 and @xmath365 are linked to a universal differential energy distribution , independent of @xmath6 , caused by the same physical processes . in contrast , the angular momentum distributions depend on @xmath6 , most notably for high radial anisotropies . for increasing values of @xmath6 , the fraction of orbits with low angular momentum increases . @xcite proposed a universal form for the integrated angular momentum distribution in dark matter halos @xmath383 . alternatively , @xcite found a differential distribution @xmath384 our models indicate a similar profile , with @xmath385 , although the functions ( [ angularss ] ) fall steeper than ours as @xmath373 increases . in this paper we presented a set of anisotropic distribution functions @xmath25 for the dehnen - mclaughlin jeans models . we constructed these dfs as a linear combination of base functions , which we fitted to the halo density by means of a quadratic programming algorithm . the base functions were specifically designed for this task , derived from a separable augmented density that generates exactly the general four - parameter velocity anisotropy profiles ( [ betagen ] ) . we demonstrated that the resulting fits are very accurate , from the center to large radii , far beyond the range of @xmath0-body simulations . this method has several advantages . the dfs can be written as a sum of a double series , without numerical integrations or inversions . consequently , the dfs and all subsequent moments can be computed with high accuracy . moreover , the advanced form of the base functions makes the qp - fitting very fast , requiring only a small number of library components . in this manner , we have constructed a family of dynamical models that incorporate the observed features in @xmath0-body simulations . we summarize their properties : 1 . the models a priori have the universal properties encountered in @xmath0-body studies of dark matter halos , which formed the building bricks of the jeans models by @xcite : they generate a universal density profile , a power - law pseudo phase - space density @xmath42 , four - parameter velocity anisotropy profiles with arbitrary values of @xmath5 and @xmath6 , and a linear @xmath2 relation . in particular , we analyzed six models that are isotropic in the center and radially anisotropic at large radii . the dfs are physical , i.e. they are non - negative over the entire phase space . this means that the dehnen - mclaughlin jeans models can actually be realized by self - consistent dynamical systems . in addition , the dfs are continuous and smooth functions and , contrary to the popular osipkov - merritt models , they fill the entire accessible phase space . the energy distributions are monotonously decreasing functions , and like the radial kurtosis profiles , these functions are nearly independent of @xmath6 . this suggests that @xmath15 , @xmath42 , @xmath365 , and @xmath386 have a universal form , caused by the same physical processes . in addition , we have also been able to generate dynamical models with a non - linear @xmath2 relation , i.e. @xmath387 and @xmath388 , albeit with higher @xmath345 values . from these successful results for the dehnen - mclaughlin halos , we expect equally adequate fits for other zhao models , such as the nfw halos . other profiles , such as the srsic - type densities , might require a modified family of base functions . such a systematic study could unravel more hitherto hidden properties of dark matter halos , and the various connections between these characteristics . our set of dynamical models is not only useful for a purely theoretical analysis . the dfs can also serve to generate initial conditions with monte carlo simulators ( e.g. * ? ? ? * ; * ? ? ? * ) to investigate the physical processes within these equilibrium models by means of controlled numerical @xmath0-body simulations . finally , our algorithm and base functions can be used in other dynamical studies , using different data moments than the density . this could lead to a significant improvement in the dynamical modeling of observed stellar systems , such as clusters of galaxies . in a spherical dynamical model , the anisotropic velocity moments of the df are @xmath389 using the augmented density formalism , these moments can be calculated as @xmath390 , with @xmath391 { { \mathrm{d}}}\psi',\ ] ] where @xmath392 denotes the @xmath393th differentiation with respect to @xmath394 . applying this to our base functions ( [ rhofam ] ) , we obtain for the augmented second - order moments @xmath395 with @xmath396 , and @xmath397 the augmented fourth - order moments have the form @xmath398 , \label{mu40fam}\nonumber\\ \\ \tilde\mu_{22,i}(\psi , r ) & = & 2\left(1-\beta(r){\vphantom{1 ^ 1_1}}\right)\,\tilde\mu_{40,i}(\psi , r ) , \label{mu22fam } \\ \tilde\mu_{04,i}(\psi , r ) & = & \frac{2}{3}\left[{\vphantom{\frac{1}{1}}}\left(1-\beta(r){\vphantom{1 ^ 1_1}}\right)\left(2-\beta(r){\vphantom{1 ^ 1_1}}\right ) - \frac{1}{\beta_\delta } \left(\beta_0-\beta(r){\vphantom{1 ^ 1_1}}\right ) \left(\beta_\infty-\beta(r){\vphantom{1 ^ 1_1}}\right ) \right]\,\tilde\mu_{40,i}(\psi , r ) . \label{mu04fam}\end{aligned}\ ] ] the true velocity moments are connected with the anisotropic velocity moments through the relation @xmath399 so that @xmath400 finally , the fourth - order anisotropy profile can be derived by combining eqs . ( [ mu04fam ] ) , ( [ mu400 ] ) and ( [ mu040 ] ) , @xmath401
in statistical mechanics , thermodynamic laws are recovered at the level of ensemble averages . the past decade has brought new insights into nonequilibrium statistical mechanics due to the discovery of various types of fluctuation relations valid arbitrarily far from equilibrium @xcite . these relations identify , at the level of the single realization of a statistical ensemble , the `` trajectory entropy '' which upon ensemble averaging reproduce the thermodynamic entropy . they therefore quantify the statistical significance of nonthermodynamic behaviors which can become significant in small systems @xcite . various experimental verifications of these fts have been reported @xcite . + in this paper , we consider an open system , described by a master equation ( me ) , exchanging matter and energy with multiple reservoirs . the system can be externally driven by varying its energies or the different temperature or chemical potentials of the reservoirs . there are three mechanisms for bringing such a system out of equilibrium : preparing it in a nonequilibrium state , externally driving it , or putting it in contact with multiple reservoirs at different temperatures or chemical potentials thus breaking the detailed balance condition ( dbc ) . we show that each of these mechanism makes a distinct contribution to the total entropy production ( ep ) generated by the nonequilibrium dynamics of the system . the two first contributions are nonzero only if the system is not in its steady state and are therefore called nonadiabatic . the third contribution is equal to the ep for slow transformation during which the system remains in a steady - state and is therefore called adiabatic . we derive three fts , for the total ep and its nonadiabatic and adiabatic contribution and show that they lead to exact inequalities valid arbitrary far from equilibrium . previously derived fts are recovered by considering specific types of nonequilibrium transformations . steady state fts @xcite are obtained for systems maintained in a nonequilibrium steady state ( ness ) between reservoirs with different thermodynamic properties . the jarzynski or crooks type fts @xcite are derived for systems initially at equilibrium with a single reservoir which are externally driven out of equilibrium by an external force . the hatano - sasa ft @xcite is recovered for externally driven systems initially in a ness with multiple reservoirs . + to calculate the statistical properties of the various contributions to the total ep and to demonstrate the fts , we extend the generating function ( gf ) method @xcite to driven open systems . apart from providing clear proofs of the various fts , this method is useful for simulations because it does not require to explicitly generate the stochastic trajectories . some additional insight is provided by using an alternative derivation of the fts similar to the crooks derivation @xcite , where the total ep and its nonadiabatic part can be identified in terms of forward - backward trajectory probabilities . by doing so , we connect the trajectory approach previously used for driven closed systems @xcite with the gf approach used for steady state systems @xcite . + we propose to experimentally test these new fts in a driven single orbital quantum dot where the various entropy probability distributions can be measured by the full electron counting statistics which keeps track of the four possible types of electron transfer ( in and out of the dot through either lead ) . such measurements of the bidirectional counting statistics have become feasible recently @xcite . we calculate the entropy probability distributions , analyze their behavior as the driving is varied between the sudden and the adiabatic limits , and verify the validity of the fts . + in section [ mastereq ] we present our stochastic model and in section [ entropy ] we describe the various contributions to the total ep generated during a nonequilibrium transformation and the inequalities that these contributions satisfy . in section [ trajectory ] , we define the various trajectory entropies which upon ensemble averaging give the various contributions to the ep . we then present the gf formalism used to calculate the statistical properties of these trajectory entropies . in section [ ft ] , we derive the various fts and the implied inequalities . alternative derivations of fts in terms of forward - backward trajectories are given in appendix [ fttrajectory ] . by considering specific nonequilibrium transformations , we recover most of the previously derived fts . finally in section [ qd ] , we apply our results to the full counting statistics of electrons in a driven quantum dot . conclusions are drawn in section [ conclusions ] . we consider an externally driven open system exchanging particles and energy with multiple reservoirs . each state @xmath0 of the system has a given energy @xmath1 and @xmath2 particles . the total number of states @xmath0 is finite and equal to @xmath3 . the probability to find the system in a state @xmath0 at time @xmath4 is denoted by @xmath5 . the evolution of this probability is described by the me @xmath6 where the rate matrix satisfies @xmath7 we assume that if a transition from @xmath0 to @xmath8 can occur , the reversed transition from @xmath8 to @xmath0 can also occur . various parameters , such as the energies @xmath1 of the system or the chemical potential @xmath9 and the temperature @xmath10 of the @xmath11 reservoir can be varied in time externally according to a known protocol . this is described by the dependence of the rate matrix on several time - dependent parameters @xmath12 . if the transition rates are kept constant , the system will eventually reach the unique steady state solution @xmath13 which satisfies @xmath14 @xcite . + the transition rates will be expressed as sums of contributions from different reservoirs @xmath11 @xmath15 each satisfying @xmath16 \bigg\ } \;. \label{aaaad}\end{aligned}\ ] ] if all reservoirs have the same thermodynamic properties ( temperature @xmath17 and chemical potential @xmath18 ) , the steady state distribution coincide with the equilibrium distribution @xmath19 which satisfies the detailed balanced condition ( dbc ) @xmath20 as a consequence of ( [ aaaad ] ) and ( [ aaaae ] ) , the equilibrium distribution then assumes the grand canonical form @xmath21}}{\xi(\lambda ) } \label{aaaac}\;,\end{aligned}\ ] ] where @xmath22 is the grand canonical partition function . however , in the general case where the reservoirs have different @xmath23 and @xmath24 , the dbc does not hold and @xmath13 is a ness . we introduce the probability flux due to matter or energy exchange between the system and the @xmath11 reservoir during transitions between state @xmath0 and @xmath8 @xmath25 we can rewrite eq . ( [ aaaaa ] ) as @xmath26 at steady state as well as at equilibrium , we have that @xmath27 . however , for a ness state @xmath28 , whereas at equilibrium @xmath29 . the _ gibbs entropy _ of the system is a state function defined as @xmath30 using ( [ aaaaa ] ) and ( [ aaaaarate ] ) , the _ system ep _ reads @xmath31 this can be partitioned as @xcite @xmath32 with the _ total ep _ @xmath33 and the _ reservoir ep _ ( also called medium entropy or entropy flow ) @xmath34 we note that @xmath35 follows from @xmath36 if @xmath37 and @xmath38 for @xmath39 ( if @xmath40 the log in zero ) , by using the fact that @xmath41 and @xmath42 . @xmath43 is the contribution to @xmath44 coming from the changes in the system probability distribution and @xmath45 is the contribution coming from matter and energy exchange processes between the system and its reservoirs . + we further separate the reservoir ep into two components @xcite @xmath46 with the _ excess ep _ @xmath47 and the _ adiabatic ep _ ( also called housekeeping entropy @xcite ) @xmath48 the positivity of ( [ baaaf ] ) follows from the same reason as ( [ baaac ] ) . if a transformation is done very slowly , the system remains at all times in the steady state distribution @xmath49 . such a transformation is called adiabatic . we then have @xmath50 and @xmath51 . we also notice that @xmath52 when the dbc is satisfied . + we next define the state function quantity @xmath53 which is obviously zero when the system is at steady state . when considering a transformation between steady states , @xmath54 . we call @xmath55 the _ boundary ep _ ( the terminology will be explain shortly ) and separate it in two parts @xmath56 where the _ nonadiabatic ep _ is @xmath57 and the _ driving ep _ is @xmath58 with @xmath59 the positivity of ( [ baaah ] ) is again shown in the same way as for ( [ baaac ] ) and ( [ baaaf ] ) . if no external driving acts on the system , @xmath60 is time independent and from ( [ baaajb ] ) , @xmath61 . for an adiabatic transformations , since @xmath49 , from ( [ baaaha ] ) and ( [ baaah ] ) , we see that @xmath62 as well as @xmath63 . from ( [ baaahb ] ) , this also means that @xmath61 . therefore , @xmath64 only for nonadiabatic driving . using ( [ baaah ] ) with ( [ baaaab ] ) and ( [ baaag ] ) , we find @xmath65 it is clear from the last equality why we call @xmath66 the nonadiabatic ep . the inequality @xmath67 which follows from the first line is a generalization of the `` second law of steady state thermodynamics '' @xcite derived for transitions between steady states . we next summarize our results @xmath68 the total ep is always positive and can be separated into two positive contributions , adiabatic ( which are nonzero only when the dbc is violated ) and nonadiabatic effects . the latter can be due to a nonadiabatic external driving acting on the system or to the fact that one considers transformation during which the system is initially or finally not in a steady state . we therefore have a minimum ep principle stating that the total ep for arbitrary nonequilibrium transformations takes its minimal value if the transformation is done adiabatically ( very slowly ) . the equality sign in ( [ baaai ] ) is satisfied for adiabatic transformations which occur at equilibrium . the equality sign in ( [ baaak ] ) holds for adiabatic transformations . the equality sign in ( [ baaaj ] ) only occurs when the dbc is satisfied . the evolution described by the me can be represented by an ensemble of stochastic trajectories involving instantaneous jumps between states . this will allow us to define fluctuating ( trajectory ) entropies . + we denote a trajectory taken by the system between @xmath69 and @xmath70 by @xmath71 . at @xmath69 the system is in @xmath72 , and stays there until it jumps at @xmath73 to @xmath74 , _ etc . _ , jumps at @xmath75 from @xmath76 to @xmath77 and stays in @xmath77 until @xmath70 [ see fig . [ figtraj ] ] . @xmath78 is the total number of jumps during this trajectory . + we next introduce various type of trajectory entropy production ( tep ) . we will see at the end of this section that when ensemble averaged , these correspond to the various ep defined in section [ entropy ] . + the _ trajectory gibbs entropy _ is defined as @xmath79 \equiv - \ln p_{m_{(\tau)}}(t ) \ ; , \label{trajentr_aaaaa}\end{aligned}\ ] ] where @xmath80 represents the value of @xmath81 along the trajectory @xmath82 . the _ system tep _ is given by @xmath83 = -\frac{\dot{p}_{m}(t)}{p_{m}(t ) } \bigg\vert_{m_{(\tau ) } } + \sum_{j=1}^{n } \delta(t-\tau_j ) \ln \frac{p_{m_{j-1}}(\tau_{j})}{p_{m_{j}}(\tau_{j } ) } \;. \label{trajentr_aaaab}\end{aligned}\ ] ] the first term represents the smooth changes of @xmath84 $ ] along the horizontal segments of the trajectory on fig . [ figtraj ] during which the system is in a well defined state . these changes are only due to the time dependence of the probability to be on a given state . the notation @xmath85 means that the @xmath0 in the expression changes depending on which horizontal segment along the trajectory one considers . the second term represents the discrete changes of @xmath84 $ ] along the vertical segments of the trajectory . these changes are singular and only due to the change in the system state . + separating the trajectory system tep similarly as the system ep in section [ entropy ] , we get @xmath86 = \dot{s}[m_{(\tau)},t ] + \dot{s}_r[m_{(\tau)},t ] \label{trajentr_aaaac}\;,\end{aligned}\ ] ] where the _ total tep _ is @xmath86 & \equiv & -\frac{\dot{p}_{m}(t)}{p_{m}(t ) } \bigg\vert_{m_{(\tau ) } } \label{trajentr_aaaaf}\\ & & \hspace{-0.5cm}+\sum_{j=1}^{n } \delta(t-\tau_j ) \ln \frac{w_{m_{j},m_{j-1}}^{(\nu_j)}(\lambda_{\tau_j } ) p_{m_{j-1}}(\tau_{j } ) } { w_{m_{j-1},m_{j}}^{(\nu_j)}(\lambda_{\tau_j } ) p_{m_{j}}(\tau_{j } ) } \nonumber\;,\end{aligned}\ ] ] and the _ reservoir tep _ @xmath87 \equiv \sum_{j=1}^{n } \delta(t-\tau_j ) \ln \frac{w_{m_{j},m_{j-1}}^{(\nu_j)}(\lambda_{\tau_j } ) } { w_{m_{j-1},m_{j}}^{(\nu_j)}(\lambda_{\tau_j } ) } \;. \label{trajentr_aaaag}\end{aligned}\ ] ] we further separate the reservoir tep into @xmath88 = \dot{s}_{a}[m_{(\tau)},t ] + \dot{s}_{ex}[m_{(\tau)},t ] \label{trajentr_aaaad}\;,\end{aligned}\ ] ] with the _ adiabatic tep _ @xmath89 \equiv \sum_{j=1}^{n } \delta(t-\tau_j ) \ln \frac{w_{m_{j},m_{j-1}}^{(\nu_j)}(\lambda_{\tau_j } ) p_{m_{j-1}}^{\rm st}(\lambda_{\tau_{j } } ) } { w_{m_{j-1},m_{j}}^{(\nu_j)}(\lambda_{\tau_j } ) p_{m_{j}}^{\rm st}(\lambda_{\tau_{j } } ) } \ ; , \label{trajentr_aaaah}\end{aligned}\ ] ] and the _ excess tep _ @xmath90 \equiv \sum_{j=1}^{n } \delta(t-\tau_j ) \ln \frac{p_{m_{j}}^{\rm st}(\lambda_{\tau_{j } } ) } { p_{m_{j-1}}^{\rm st}(\lambda_{\tau_{j } } ) } \;. \label{trajentr_aaaai}\end{aligned}\ ] ] the _ nonadiabatic tep _ @xmath91 & \equiv & -\frac{\dot{p}_{m}(t)}{p_{m}(t ) } \bigg\vert_{m_{(\tau ) } } \label{trajentr_aaaaj}\\ & & + \sum_{j=1}^{n } \delta(t-\tau_j ) \ln \frac{p_{m_{j-1}}(\tau_{j } ) p_{m_{j}}^{\rm st}(\lambda_{\tau_{j } } ) } { p_{m_{j-1}}^{\rm st}(\lambda_{\tau_{j } } ) p_{m_{j}}(\tau_{j } ) } \nonumber\end{aligned}\ ] ] is made of the sum of two terms @xmath91 \equiv \dot{s}_d [ m_{(\tau)},t ] + \dot{s}_{b}[m_{(\tau)},t ] \label{trajentr_aaaaja } \end{aligned}\ ] ] the _ boundary tep _ @xmath92 & \equiv & -\frac{\dot{p}_{m}(t)}{p_{m}(t ) } \bigg\vert_{m_{(\tau ) } } - \dot{\lambda}_{t } \frac{\partial \phi_{m_{(\tau)}}(\lambda_{t})}{\partial \lambda_{t } } \label{trajentr_aaaajb}\\ & & + \sum_{j=1}^{n } \delta(t-\tau_j ) \ln \frac{p_{m_{j-1}}(\tau_{j } ) p_{m_{j}}^{\rm st}(\lambda_{\tau_{j } } ) } { p_{m_{j-1}}^{\rm st}(\lambda_{\tau_{j } } ) p_{m_{j}}(\tau_{j } ) } \ ; , \nonumber \end{aligned}\ ] ] and the _ driving tep _ @xmath93 & \equiv & \dot{\lambda}_{t } \frac{\partial \phi_{m_{(\tau)}}(\lambda_{t})}{\partial \lambda_{t } } \nonumber \;.\end{aligned}\ ] ] as in section [ entropy ] , since @xmath91 = \dot{s } [ m_{(\tau)},t ] + \dot{s}_{ex}[m_{(\tau)},t ] \label{trajentr_aaaae}\;,\end{aligned}\ ] ] we get @xmath86 = \dot{s}_{na}[m_{(\tau)},t ] + \dot{s}_{a}[m_{(\tau)},t ] \label{trajentr_aaaacb } \;.\end{aligned}\ ] ] we generically denote these tep by @xmath94 $ ] . the change of @xmath94 $ ] along a trajectory @xmath82 of length @xmath95 is given by @xmath96 & = & \int_{0}^{t } dt \dot{a}[m_{(\tau)},t ] \label{trajaveaaaaag } \;.\end{aligned}\ ] ] notice that @xmath97 $ ] and @xmath98 $ ] are state function tep @xmath99 = \ln \frac{p_{m_0}(0)}{p_{m_n}(t ) } = s_{m_n}(t ) - s_{m_0}(0 ) \ ; , \label{caaaj}\end{aligned}\ ] ] where @xmath100 , and @xmath101 & = & \big(s_{m_n}(t)-\phi_{m_n}(\lambda_{t})\big ) \label{caaajb } \\ & & \hspace{1.5cm}- \big(s_{m_0}(0)-\phi_{m_0}(\lambda_{0})\big ) \;.\nonumber\end{aligned}\ ] ] the other tep are path functions . + using ( [ aaaad ] ) , the reservoir tep can be expressed as @xmath102= \label{trajentr_aaaagb}\\ & & \hspace{0.5cm}-\sum_{\nu } \beta_{\nu}(\lambda_{t } ) \bigg ( i_{heat}^{(\nu)}[m_{(\tau)},t ] - \mu_{\nu}(\lambda_{t } ) i_{mat}^{(\nu)}[m_{(\tau)},t ] \bigg)\;,\nonumber\end{aligned}\ ] ] where the heat current between the @xmath11 reservoir and the system is @xmath103 = \sum_{j=1}^{n } \delta_{\nu,\nu_j}(t-\tau_j ) \big ( \epsilon_{m_{j}}(\lambda_{\tau_j})-\epsilon_{m_{j-1}}(\lambda_{\tau_j } ) \big ) \label{heatcurr}\end{aligned}\ ] ] and the matter current between the @xmath11 reservoir and the system @xmath104 = \sum_{j=1}^{n } \delta_{\nu,\nu_j}(t-\tau_j ) \big(n_{m_{j}}-n_{m_{j-1}}\big ) \;. \label{mattercurr}\end{aligned}\ ] ] the currents are positive if the system energy ( matter ) increases . @xmath105 is a dirac distribution centered at time @xmath106 only if the transition is due to the reservoir @xmath107 . otherwise it is zero . we thus confirm that the reservoir ep is the entropy associated to system - reservoir exchange processes . + we assume that the parametric time dependence of the energies , temperatures and chemical potentials is known . except in degenerate cases for which two different transitions between states have the same energy difference and number of particle difference , the trajectory of the system can be uniquely determined by measuring the heat and matter currents between the system and the reservoirs . the system steady state probability distribution can be calculated by recording the steady state currents for sufficiently long times for different values of the energies , temperatures or chemical potentials . the driven system probability distribution can in principle be calculated by reproducing the measurement of the currents multiple times . all trajectory entropies containing the logarithm of the transition rates can be expressed in term of a combination of the reservoir ep ( directly measurable via current ) and other trajectory entropies which can be expressed in term of the system probability distribution ( actual or steady state ) . therefore , provided the current measurements can be repeated often enough to get a good statistics , all the trajectory entropies are in principle measurable . the gf formalism allows to compute the probability distributions and all statistical properties of the tep without having to generate the trajectories themselves . it further provides a direct means for proving the fts . + the gf associated with the changes of @xmath94 $ ] along a trajectory is given by @xmath108\ } } \rangle } \label{trajaveaaaaah } \;,\end{aligned}\ ] ] where @xmath109 denotes an average over all possible trajectories . the probability that the system follows a trajectory with the constraint @xmath110 $ ] at time @xmath4 , can be obtained from the gf using @xmath111 \rangle } \label{trajaveaaaaai}\\ & = & \frac{1}{2 \pi } \int_{-\infty}^{\infty } d\gamma { { \rm e}}^{-{{\rm i}}\gamma \delta a } g({{\rm i}}\gamma , t ) \nonumber \;.\end{aligned}\ ] ] by inverting ( [ trajaveaaaaai ] ) , we get @xmath112 the moments of the distribution are given by derivatives of the gf @xmath113 \rangle } = \left . \frac{\partial^k g(\gamma , t ) } { \partial \gamma^k } \right\|_{\gamma=0 } \label{trajaveaaaaak } \ ; , \ ; \ \ k=1,2,\cdots \;.\end{aligned}\ ] ] in order to compute the gf , we recast it in the form @xmath114 where @xmath115\ } } \rangle}_m \label{trajaveaaaaalb } \end{aligned}\ ] ] is the product of the probability to find the system in state @xmath0 at time @xmath4 multiplied by the expectation value of @xmath116\}}$ ] conditional on the system being in state @xmath0 at time @xmath4 . since @xmath117=0 $ ] for a trajectory of length @xmath69 , we have @xmath118 . we also have @xmath119 and @xmath120 . + the time derivative of ( [ trajaveaaaaal ] ) gives @xmath121 where @xmath122 depends on the tep of interest . below , we will derive equations of motion for the @xmath123 s associated to the various tep . the generating function associated with a state function tep @xmath117=a_{m}(t)-a_{n}(0)$ ] like @xmath124 $ ] or @xmath125 $ ] may be straightforwardly obtained using ( [ trajaveaaaaal ] ) with ( [ trajaveaaaaalb ] ) . we get @xmath126 @xmath127 $ ] acquires an amount @xmath128 each time a transition from a state @xmath8 to @xmath0 occurs and it remains constant along a given state @xmath0 of the system . this means that @xmath129\ } } \rangle}_{m ' } \nonumber \;,\end{aligned}\ ] ] which using ( [ trajaveaaaaalb ] ) can be rewritten @xmath130 each time a transition along the system trajectory occurs , @xmath131 $ ] acquires an amount @xmath132 . similarly to the excess entropy , we get @xmath133 this equation has been used in the study of steady state fts @xcite . + in ( [ trajentr_aaaagb ] ) , we expressed the reservoir tep in terms of currents . the time integrated individual currents give the heat and matter transfer between the @xmath11 reservoir and the system @xmath134=\int_{0}^{t}dt ' i_{heat}^{(\nu)}[m_{(\tau)},t']$ ] and @xmath135=\int_{0}^{t}dt ' i_{mat}^{(\nu)}[m_{(\tau)},t']$ ] . their statistics can be calculated using @xmath136 where @xmath137 is a vector who s elements are the different @xmath138 s and @xmath139 s . the gf calculated from ( [ genaaaagcurr ] ) is therefore associated with the join probability distribution for having a certain heat and matter transfer with each reservoir . each time a transition along the system trajectory occurs , @xmath140 $ ] acquires an amount @xmath141 . we therefore get @xmath142 each time a transition along the system trajectory occurs , @xmath143 $ ] acquires an amount @xmath144 . in addition it also changes by an amount @xmath145 during an infinitesimally small time on a given state @xmath0 of the system . combining the two , we have @xmath146 like the total tep , @xmath147 $ ] acquires an amount @xmath148 each time a transition from a states @xmath8 to @xmath0 occurs , and also changes by an amount @xmath145 during an infinitesimally small time on a given state @xmath0 . this gives @xmath149 since @xmath150 $ ] exclusively accumulates along the segments of the system trajectory , we get @xmath151 it follows from ( [ trajaveaaaaak ] ) that the average change of a tep is obtained from its gf by differentiation with respect to @xmath152 at @xmath153 . by differentiating the gf evolution equations of this section one recover the evolution equation for the eps of section [ entropy ] . the eps are therefore the ensemble average of the teps introduced in this section @xmath154 \rangle}$ ] and @xmath155 \rangle}$ ] . we can easily verify that @xmath156 is the solution of the evolution equations ( [ genaaaak ] ) and ( [ genaaaan ] ) . it immediately follows from probability conservation and ( [ trajaveaaaaao ] ) that @xmath157 . summing both side of ( [ genaaaai ] ) over @xmath0 , we also verify that @xmath158 . because @xmath159 , @xmath160 where @xmath161 . therefore , we find that @xmath162 . using ( [ trajaveaaaaah ] ) , this results in the three fts @xmath163 \ } } \rangle}=1 \ ; , \label{fti}\\ & & { \langle \exp{\{- \delta s_{na}[m_{(\tau)},t ] \ } } \rangle}=1 \ ; , \label{ftx}\\ & & { \langle \exp{\{- \delta s_{a}[m_{(\tau)},t ] \ } } \rangle}=1 \;. \label{fthk}\end{aligned}\ ] ] these fts hold irrespective of the initial condition and the type of driving . using jensen s inequality @xmath164 , they imply the inequalities ( [ baaai])-([baaaj ] ) ( [ fti ] ) is the generalization to open systems of the integral ft for the tep obtained earlier for closed systems @xcite . the tep ( [ trajentr_aaaaf ] ) needs to specify which reservoir is responsible for the transitions occurring along the trajectory ( by labeling the rates with reservoir index ) . this point , made earlier for open system at steady state @xcite , is generalized here for driven systems with an arbitrary initial condition . ( [ ftx ] ) will be shown in next section to reduce to the integral hatano - sasa ft @xcite for systems initially in a steady state . ( [ fthk ] ) generalizes the integral ft for the adiabatic entropy @xcite previously derived for closed system initially in a steady state . + some additional insights can be gained by an alternative proof of the fts ( [ fti ] ) and ( [ ftx ] ) which use a forward - backward trajectory picture of the dynamics . this is given for completeness in appendix [ fttrajectory ] . we consider a system initially ( t=0 ) at steady state and subjected to an external driving force between @xmath165 and @xmath166 . for @xmath167 , the system remains in the steady state corresponding to @xmath168 . the time protocol of @xmath12 during the driving @xmath169 is arbitrary . if @xmath170 is the characteristic transient time needed for the system to reach a steady state from an arbitrary distribution , for @xmath171 , the system is in the new steady state corresponding to @xmath172 . the system is measured between @xmath69 and @xmath70 . we restrict our analysis to cases where the system is at steady state at @xmath69 and the driving starts at least a time @xmath170 after the measurement started : @xmath173 . + we use the braket notation where @xmath174 is the probability vector with components @xmath5 and @xmath175 denotes the rate matrix . @xmath176 denotes a vector with all components equal to one . the me ( [ aaaaa ] ) now reads @xmath177 the generating function for the reservoir ep ( [ genaaaag ] ) for @xmath178 evolves according to the adjoint equation of ( [ vecme ] ) @xmath179 the initial condition of ( [ vecme ] ) and ( [ vecmead ] ) is @xmath180 . the formal solution of ( [ vecmead ] ) for @xmath181 before the driving starts reads @xmath182 we now insert a closure relation in term of right and left eigenvectors of the adjoint rate matrix between the evolution operator and the initial condition . because the rate matrix and its adjoint have the same eigenvalues ( all negative and one zero ) , for @xmath183 , only the right and left eigenvector associated with the zero eigenvalue survive . since the right [ left ] eigenvector of @xmath184 is @xmath176 [ @xmath185 , we get for @xmath183 @xmath186 for longer times , even when the system starts to be driven , @xmath176 remains invariant under the time evolution operator as can be seen using ( [ aaaaarate ] ) in ( [ vecmead ] ) . we get @xmath187 where @xmath188 is the total number of states . this implies the following integral ft for the reservoir tep @xmath189 \ } } \rangle } = m \sum_{m=1}^{m } \big ( p^{\rm st}_m ( \lambda_{t_{di } } ) \big)^2 \geq 1 \;. \label{ssaaaai}\end{aligned}\ ] ] the equality on the l.h.s ( r.h.s ) is satisfied if @xmath190 ( @xmath191 ) . jensen s inequality implies @xmath192 . note that since @xmath97 $ ] is a state function , it is easily verified that @xmath193 \ } } \rangle } \label{ssaaaaj } \;.\end{aligned}\ ] ] we assume that the driving starts at the same time or later as the measurement ( @xmath194 ) . we define @xmath195 . + we have pointed out at the end of section [ entropy ] that for transitions between steady states @xmath196 , so that @xmath197 we used the fact that @xmath198 starts ( stops ) evolving at @xmath199 ( @xmath200 ) . the same is true at the trajectory level , since from ( [ caaajb ] ) we have @xmath201=0 $ ] and therefore @xmath202 = \delta s_d[m_{(\tau)},t ' ] \ ; , \label{caaana}\end{aligned}\ ] ] where @xmath203 & \equiv & \sum_{j=0}^{n } \ln \frac{p_{m_{j}}^{\rm st}(\lambda_{\tau_j } ) } { p_{m_{j}}^{\rm st}(\lambda_{\tau_{j+1}})}\label{caaav}\\ & = & \sum_{j=0}^{n } \big ( \phi_{m_{j}}(\lambda_{\tau_{j+1 } } ) - \phi_{m_{j}}(\lambda_{\tau_{j } } ) \big ) \nonumber \\ & = & \int_{0}^{t ' } dt \dot{\lambda}_{t } \frac{\partial \phi_{m_{(\tau)}}(\lambda_{t})}{\partial \lambda_{t } } \nonumber \\ & = & \int_{t_{di}}^{t_{df } } dt \dot{\lambda}_{t } \frac{\partial \phi_{m_{(\tau)}}(\lambda_{t})}{\partial \lambda_{t } } \nonumber \;.\end{aligned}\ ] ] the integrand in the third line contributes only during the time intervals between jumps provided the driving is changing . therefore if without loss of generality we choose the measurement time such that @xmath204 ( if @xmath205 one can redefine @xmath200 as equal to @xmath95 ) , @xmath206=\delta s_d[m_{(\tau)},t']$ ] . this means that for a transition between steady states , the ft ( [ ftx ] ) reduces to the hatano - sasa ft @xcite @xmath207 \ } } \rangle}=1 \label{fty } \;. \end{aligned}\ ] ] alternatively , ( [ fty ] ) can be proved from ( [ genaaaad ] ) by showing that when @xmath178 , @xmath208 is solution of ( [ genaaaad ] ) . the ft ( [ fty ] ) holds for an arbitrary driving protocol . let us consider the two extremes . ( infinitely slow ) driving , the inequality in ( [ caaanaa ] ) becomes an equality . in the other extreme of a _ sudden _ driving , where @xmath209 and @xmath210 , @xmath211 = \phi_{m_0}(\lambda_{t_{df } } ) - \phi_{m_0}(\lambda_{t_{di } } ) \label{caabj } \end{aligned}\ ] ] becomes a state function and its average takes the simple form @xmath212 using ( [ trajentr_aaaae ] ) with ( [ caaana ] ) , and since @xmath213 = \phi_{m_t}(\lambda_{t_{df } } ) - \phi_{m_0}(\lambda_{t_{di } } ) \ ; , \label{caabi}\end{aligned}\ ] ] we find that @xmath214 = \phi_{m_0}(\lambda_{t_{df}})- \phi_{m_t}(\lambda_{t_{df } } ) \label{caabk}\end{aligned}\ ] ] also becomes a state function and its average becomes @xmath215 for a system coupled to a single reservoir ( or multiple reservoirs with identical thermodynamical properties ) , the dbc ( [ aaaae ] ) is satisfied . a non - driven system in an arbitrary state will reach after some transient time @xmath216 the equilibrium grand canonical distribution ( [ aaaac ] ) . we again choose @xmath195 and @xmath204 . from the tep of section [ trajectory ] , we find in this case @xmath217 & = & 0 \label{daaaab } \\ \delta s_{r}[m_{(\tau)},t ' ] & = & \delta s_{ex}[m_{(\tau)},t ' ] \nonumber\\ \delta s_{tot}[m_{(\tau)},t ' ] & = & \delta s_{na}[m_{(\tau)},t ' ] \;.\nonumber\end{aligned}\ ] ] the two ft ( [ fti ] ) and ( [ ftx ] ) become identical and the ft ( [ fthk ] ) becomes trivial . using ( [ aaaac ] ) , we also find that eq . ( [ caaanb ] ) becomes @xmath218 where @xmath219 is the thermodynamic grand canonical potential . + we next consider transitions between equilibrium states , so that the procedure is the same as in [ hssub ] but with the dbc ( [ aaaae ] ) now satisfied . we therefore have @xmath220=\delta s_d[m_{(\tau)},t']$ ] . the driving implies externally modulating the system energies , the chemical potential or the temperature of the reservoir . + when driving the system energy , using ( [ caaav ] ) and ( [ daaaa ] ) , we find @xmath211 = \delta s_d[m_{(\tau)},t ' ] = \beta w[m_{(\tau ) } ] - \beta \delta \omega \label{daaab}\end{aligned}\ ] ] where the work is given by @xmath221 = \int_{t_{di}}^{t_{df } } dt \dot{\epsilon}_{m_{(\tau)}}(\lambda_t)$ ] and @xmath222-\omega[\epsilon(t_{di})]$ ] . both ft , ( [ fti ] ) and ( [ ftx ] ) , lead to the same jarzynski relation @xcite @xmath223}\ } \rangle } = \exp{\{-\beta \delta \omega \ } } \;. \label{daaad}\end{aligned}\ ] ] when driving the reservoir chemical potential , eq . ( [ caaav ] ) and ( [ daaaa ] ) give @xmath211 = \delta s_d[m_{(\tau)},t ' ] = \beta \tilde{w}[m_{(\tau ) } ] - \beta \delta \omega \label{daaaf } \end{aligned}\ ] ] where @xmath224 = -\int_{t_{di}}^{t_{df } } dt \dot{\mu } n_{m_{(\tau)}}$ ] and @xmath225-\omega[\mu(t_{di})]$ ] . both ft , ( [ fti ] ) and ( [ ftx ] ) , now lead to @xmath226 \ } } \rangle } = \exp{\{-\beta \delta \omega \ } } \;. \label{daaah}\end{aligned}\ ] ] the case where reservoir temperature is driven can be calculated similarly . driving the reservoir temperature using ( [ aaaac ] ) , eq . ( [ caaanb ] ) becomes @xmath227 eq . ( [ caaav ] ) gives @xmath228 & = & \int_{0}^{t } d\tau \dot{\lambda}_{\tau } \frac{\partial \beta(\lambda_{\tau})}{\partial \lambda_{\tau } } \big ( \epsilon_{m_{(\tau ) } } - \mu n_{m_{(\tau ) } } \big ) \label{daaaj}\\ & & - \big ( \beta(\lambda_{t } ) \omega ( \lambda_{t } ) - \beta(\lambda_{0 } ) \omega ( \lambda_{0 } ) \big ) \;. \nonumber\end{aligned}\ ] ] using ( [ caaba ] ) , we get @xmath229 in a ness , the relations of section [ trajectory ] give @xmath230&=&\delta s_d[m_{(\tau)},t]=0 \label{ssaaaaa}\\ \delta s [ m_{(\tau)},t]&= & -\delta s_{ex}[m_{(\tau)},t ] \nonumber\\ \delta s_{tot } [ m_{(\tau)},t]&= & \delta s_{a}[m_{(\tau)},t ] \nonumber \;.\end{aligned}\ ] ] furthermore , since @xmath231 , we get @xmath232 we shall rewrite the gf evolution equation for the reservoir tep ( [ genaaaag ] ) in the bracket notation @xmath233 so that @xmath234 where @xmath176 is a vector with all elements equal to one . since from ( [ genaaaag ] ) the generator has the property @xmath235 , its eigenvalues have the symmetry @xmath236 . furthermore , since @xmath237 is a positive matrix , the frobenious - perron theorem @xcite ensures that all eigenvalues are negative or zero and that the left and the right eigenvectors , @xmath238 and @xmath239 , associated with the largest eigenvalue @xmath240 exist . adopting the normalization @xmath241 , we find for long times @xmath242 and that @xmath243 this means that the cumulant generating function @xmath244 satisfies the symmetry @xmath245 using the theory of large fluctuations this symmetry implies the detailed steady state ft @xcite @xmath246 where @xmath247 is the probability for a trajectory of the system to produce a reservoir tep equal to @xmath248 . + @xmath249 $ ] grows in average with time because it depends on the number of jumps along the trajectory . @xmath124 $ ] is bounded . the ft ( [ ssaaaah ] ) can therefore be viewed as a consequence of the detailed ft for @xmath250 derived in ( [ caaap ] ) . the long time limit is needed in order to neglect the contribution from @xmath124 $ ] to @xmath251 $ ] . + the ft ( [ ssaaaai ] ) remains valid at steady state . fts for currents can also be derived @xcite . we have seen in section [ measure ] that the various entropies can be calculated by measuring the different currents between the system and the reservoirs . the counting statistics of electrons through quantum dots has recently raised considerable theoretical @xcite as well as experimental @xcite interest . the single electrons entering and exiting a quantum dot connected to two leads can be measured . one can therefore calculate all currents , deduce the system trajectories and calculate the various trajectory entropies presented earlier . + we will analyze the probability distribution for the various trajectory entropies in a single level quantum dot of energy @xmath252 connected to two leads with different chemical potentials @xmath253 , where @xmath254 . we neglect spin so that the dot can either be empty @xmath255 or filled @xmath256 . the me is of the form ( [ aaaaa ] ) @xcite @xmath257 where @xmath258 the coefficients @xmath259 characterize the coupling between the dot and the lead @xmath11 with fermi distribution @xmath260 . if @xmath261 , @xmath262=[{\rm energy}]=[{\rm time}^{-1}]$ ] . by renormalizing energies by @xmath252 , all parameters of our model become dimensionless . the steady state distribution of the system is @xmath263 and the steady state currents are given by @xmath264 we switch the chemical potential of the left lead @xmath265 using the protocol @xmath266 while holding the right lead chemical potential fixed @xmath267 . we can therefore calculate all the trajectory entropies probability distributions using the gf method described in section [ statprop ] . we solve numerically the evolution equations for the @xmath268 s associated with the different entropies for different values of @xmath152 with the initial condition @xmath269 . after calculating the @xmath270 s using ( [ trajaveaaaaal ] ) , the probability distribution is obtained by a numerical inverse fourier transform ( [ trajaveaaaaai ] ) . in all calculations we used @xmath271 , @xmath272 , @xmath273 and @xmath274 . + we start by analyzing the different contributions to the ep as defined in section [ entropy ] for the @xmath275 protocol shown in fig . ( [ aver])a . + the system is initially in a nonequilibrium distribution different from the steady state . the solution of the me ( [ yaaaa ] ) as well as its steady state solution are displayed in fig . [ aver]b . between @xmath69 and @xmath276 , @xmath277 is essentially constant and the system undergoes an exponential relaxation to the steady state . between @xmath276 and @xmath278 , @xmath277 changes from @xmath279 to @xmath280 fast enough for the system distribution to start differing again from the instantaneous steady state distribution ( adiabatic solution ) . after @xmath278 , @xmath277 remains constant and the system again undergoes a transient relaxation to the new steady state corresponding to @xmath280 . [ aver]c shows the time dependent ep @xmath281 and its adiabatic @xmath282 and nonadiabatic contribution @xmath283 . as predicted , these three quantities are always positive [ see ( [ baaai])-([baaaj ] ) ] . we also demonstrate that @xmath283 only contributes when nonadiabatic effects are significant i.e. when the actual probability distribution is different from the steady state one [ @xmath284 . @xmath285 only once at @xmath286 , when @xmath287 and the dbc is satisfied . otherwise the dbc is broken and @xmath288 . in fig . [ aver]d , we present the two contributions to the nonadiabatic ep @xmath283 , the driving ep @xmath289 and the boundary ep @xmath290 [ see ( [ baaak ] ) ] . the driving ep @xmath289 only contributes when @xmath277 changes in time . one can also guess that @xmath291 due to the fact that the change of boundary ep during an interval between two steady state is zero . @xmath292 because the system is initially not in a steady state . [ aver]e shows the alternative partitioning of the nonadiabatic ep into the system ep @xmath293 and the excess ep @xmath294 [ see ( [ baaakk ] ) ] . finally the splitting of the total ep in the reservoir ep and the system ep [ see ( [ baaab ] ) ] is shown in fig . we see that at steady state @xmath295 so that @xmath296 . + we next study the statistical properties of the different tep for transitions between steady states . the probability distributions are obtained using the gf method presented in section [ statprop ] . the five driving protocols used to change @xmath275 from @xmath279 to @xmath280 are presented in fig . [ profilefig ] . they range from sudden switch in ( i ) to slow ( adiabatic ) switch in ( v ) . the system is always initially in the steady state corresponding to @xmath279 . we will consider measurements which end when the system reaches its new steady state at @xmath280 . different measurement times are represented by a , b , c , d . [ ps ] , we display @xmath297 . since @xmath298 is a state function , @xmath297 is the same for the various protocol . because we consider a two level system , @xmath298 can only take four possible values which correspond to the four possible change in the system state between its initial and final condition . the transitions @xmath299 and @xmath300 are much more probable because the probability to initially find the system in the empty state @xmath255 is much higher ( @xmath301 ) than finding it in the filled state @xmath256 ( @xmath302 ) . the transition @xmath299 is more probable than @xmath300 because the system has a final probability @xmath303 to be in its empty state and @xmath304 to be in its filled state . + in the left column of fig . [ psexpsd ] , we depict @xmath305 . here @xmath306 so that @xmath307 [ see ( [ baaahb ] ) ] . all curves ( i)-(v ) satisfy the ft ( [ ftx ] ) . for the sudden switch ( i ) , @xmath308 becomes a state function which only depends on the initial state of the system [ see ( [ caabj ] ) ] . @xmath308 can therefore take two possible values corresponding to the empty @xmath255 or filled @xmath256 orbital with a respective probability @xmath301 or @xmath302 . when the driving speed slows down in ( ii ) the peaks are broadened . in the adiabatic limit ( v ) , @xmath305 becomes a broad distribution with zero average . in the right column of fig . [ psexpsd ] , we depict @xmath309 . for sudden switch ( i ) , @xmath310 turn to a state function which only depends on the final steady state distribution [ see ( [ caabk ] ) ] . it is clear from ( [ caabk ] ) that the transitions @xmath299 and @xmath311 leads to @xmath312 and @xmath313 and @xmath300 to the same @xmath310 with opposite sign . the probabilities to observe these transitions follow from the fact that the system is initially more likely to be in @xmath255 ( prob @xmath301 ) than in @xmath256 . the probability for the final state @xmath255 ( @xmath256 ) is @xmath303 ( @xmath304 ) . therefore the most likely transition is @xmath299 followed from @xmath300 . as the driving speed slows down like in ( ii ) , the peaks get broadened . since @xmath314 [ see ( [ baaakk ] ) ] and since in the adiabatic switch limit ( v ) @xmath305 is centered around zero , @xmath309 ( v ) has the same peak structure as @xmath315 [ see fig . [ ps ] ] , but broadened by @xmath308 . + in fig . [ pse ] , we display @xmath316 for different measurement times and protocols . plots with same driving but different measurement times [ ( i)a and ( ii)d or ( ii)a and ( ii)d or ( iii)b and ( iii)d ] show the evolution of @xmath316 in the final steady state . the plots ( ii)-(v ) satisfy the ft ( [ ssaaaai ] ) which for our parameters imply @xmath317 . the ft is not satisfied for ( i ) because the driving starts at the same time as the measurement [ see section [ fte ] ] . to understand the structure of @xmath316 , we time integrate ( [ trajentr_aaaagb ] ) and use the fact that in our model the heat current is proportional to the matter current between the @xmath11 reservoir and the system @xmath318 where @xmath319 we get @xmath320 in the sudden switch limit ( i ) , we get @xmath321 where @xmath322 is the net number of electron transferred from the reservoir @xmath11 to the system between @xmath255 and @xmath4 . this explains why @xmath323 in ( i ) only take discrete value which are multiples of each other . the distance between the peaks of @xmath324 observed in ( i ) is due to the right lead only @xmath325 because our parameters are such that @xmath326 . the new peaks which appear in ( ii ) with a spacing @xmath327 are due to the fact that the driving starts some time after the measurement so that @xmath328 also contributes . as the driving speed slows down ( iii)-(v ) , the discrete structure broadens and @xmath323 can take continuous values . [ psa ] , we display @xmath329 . all curves satisfy the ft ( [ fthk ] ) . the verification ( not shown ) is best done on the gf ( @xmath330 ) because the numerical accuracy of the tail of the distribution is not sufficient . the peak structure of @xmath331 can be understood from @xmath332 and @xmath309 because @xmath333 . this is particularly clear for the sudden switch ( i ) where the possible values of the entropies are strongly restricted . indeed , in ( i ) each peak of @xmath332 is split in three smaller peaks which have the same structure as @xmath309 . as the speed of the driving decreases ( ii)-(v ) , the peak structure disappears . [ psi ] , we show @xmath334 . all curves satisfy the ft ( [ fti ] ) [ verification was done on the gf ( not shown ) ] . the structure of @xmath334 can be understood using @xmath332 and @xmath297 because @xmath335 . this is clear for the sudden switch ( i ) where the peaks of @xmath332 are split in smaller peaks which have the structure of @xmath297 . for a driven open system in contact with multiple reservoirs and described by a master equation , we have proposed a partitioning of the trajectory entropy production into two parts . one contributes when the system is not in its steady state and contains two contributions due to the external driving and the deviation from steady state in the initial and final probability distribution of the system . the second part comes from breaking of the detailed balance condition by the multiple reservoirs and becomes equal to the total entropy production when the system remains in its steady state all throughout the nonequilibrium process . both parts as well as the total entropy production satisfy an general integral fluctuation theorem which imposes positivity on their ensemble average . this partitioning also provides a simple way to identify which part of the entropy production contributes during a specific type of nonequilibrium process [ see fig . [ shema ] ] . previously derived integral fluctuation theorems can be recovered from our three general fluctuation theorems and in addition we derived a new integral fluctuation theorem for the part of the entropy production due to exchange processes between the system and its reservoirs ( reservoir entropy production ) . our results strictly apply to systems described by a master equation ( [ aaaaa ] ) . however , as has often be the case for previous fluctuation relations , one could expect similar results to hold for other types of dynamics . for electron transport trough a single level quantum dot between two reservoir with time dependent chemical potentials , we have simulated and analyzed in detail the probability distributions of the various trajectory entropies and showed how they can be measured in electron counting statistics experiments . the support of the national science foundation ( grant no . che-0446555 ) and nirt ( grant no . eec 0303389 ) is gratefully acknowledged . m. e. is partially supported by the fnrs belgium ( collaborateur scientifique ) . we show that the ft ( [ fti ] ) and ( [ ftx ] ) have an interesting interpretation in term of the ratio of the probability of a forward dynamics generating a given trajectory and the probability of the time - reversed trajectory during some backward dynamics . this is an alternative to the gf approach which connects the detailed form to the integral form of the fts . + the forward dynamics is described by the me ( [ aaaaa ] ) . we introduce the probability ( in trajectory space ) @xmath336 $ ] that the system follows a trajectory @xmath82 @xmath337 & = & p_{m_0}(0 ) \bigg [ \prod_{j=1}^{n } \label{caaaa}\\ & & \hspace{-2 cm } \exp{\bigg(\int_{\tau_{j-1}}^{\tau_{j } } d\tau ' w_{m_{j-1},m_{j-1}}(\lambda_{\tau ' } ) \bigg ) } w_{m_{j},m_{j-1}}^{(\nu_{j})}(\lambda_{\tau_j } ) \bigg ] \nonumber \\&&\hspace{0 cm } \exp{\big(\int_{\tau_{n}}^{t } d\tau ' w_{m_{n},m_{n}}(\lambda_{\tau ' } ) \big ) } \;.\nonumber\end{aligned}\ ] ] the @xmath338 factors in this expression represent the probability that the system undergoes a given transition whereas the exponentials describe the probability for the system to remain in a given state between two successive jumps . summation over all possible trajectories will be denoted by @xmath339 . it consists of time - ordered integrations over the @xmath78 time variables @xmath340 from @xmath255 to @xmath95 ( this gives the probability of having a path with @xmath78 transitions ) and then summing over all possible @xmath78 from @xmath255 to @xmath341 . normalization in trajectory - space implies that @xmath342 = 1 $ ] . + the backward dynamics is described on the time interval @xmath343 $ ] by the me @xmath344 where the new rate matrix @xmath345 satisfies @xmath346 . we require that the parametric time dependence ( via the driving protocol @xmath347 ) of the rate matrix in eq . ( [ caaab ] ) is time - reversed compared that of eq . ( [ aaaaa ] ) and that the diagonal part of the rate matrix in eq . ( [ aaaaa ] ) and ( [ caaab ] ) is the same @xmath348 this still leaves room for different choices of @xmath345 . we will later specify two choice of @xmath349 [ ( [ caaaoa ] ) and ( [ caaax ] ) ] that will result in two fts . + we define the time - reversed trajectory of @xmath82 by @xmath350 @xmath351 . the probability @xmath352 $ ] that the system described by ( [ caaab ] ) follows the time - reversed trajectory @xmath353 is given by @xmath354 & = & \tilde{p}_{m_n}(0 ) \bigg [ \prod_{j=1}^{n } \label{caaac } \\&&\hspace{-1.5 cm } \exp{\bigg(\int_{t-\tau_{n - j+2}}^{t-\tau_{n - j+1 } } d\tau ' \tilde{w}_{m_{n - j+1},m_{n - j+1}}(\lambda_{t-\tau ' } ) \bigg ) } \nonumber \\&&\hspace{2 cm } \tilde{w}_{m_{n - j},m_{n - j+1}}^{(\nu_{n - j+1 } ) } ( \lambda_{t-\tau_{n - j+1 } } ) \bigg ] \nonumber \\&&\hspace{1 cm } \exp{\big(\int_{t-\tau_{1}}^{t } d\tau ' \tilde{w}_{m_{0},m_{0}}(\lambda_{t-\tau ' } ) \big ) } \;,\nonumber\end{aligned}\ ] ] where @xmath355 . normalization in the reverse path ensemble implies @xmath356 = 1 $ ] . + we consider the ratio of the two probabilities ( [ caaaa ] ) and ( [ caaac ] ) , @xmath357 & \equiv & \ln \frac{{\cal p}[m_{(\tau)}]}{\tilde{{\cal p}}[\bar{m}_{(\tau ) } ] } \label{caaad } \;.\end{aligned}\ ] ] due to ( [ caaaba ] ) , the contributions from the exponentials ( which represent the probabilities to remain on a given state ) in ( [ caaad ] ) cancel , so that @xmath357= \ln \frac{p_{m_0}(0)}{\tilde{p}_{m_n}(0 ) } + \sum_{j=1}^{n } \ln \frac{w_{m_{j},m_{j-1}}^{(\nu_{j})}(\lambda_{\tau_j } ) } { \tilde{w}_{m_{j-1},m_{j}}^{(\nu_{j})}(\lambda_{\tau_{j } } ) } \;.\label{caaae}\end{aligned}\ ] ] we can partition ( [ caaae ] ) in the form @xmath357 & = & \ln \frac{p_{m_0}(0)}{\tilde{p}_{m_n}(0 ) } + \sum_{j=1}^{n } \ln \frac{p_{m_{j}}^{\rm st}(\lambda_{\tau_j } ) } { p_{m_{j-1}}^{\rm st}(\lambda_{\tau_j } ) } \label{caaaf}\\ & & + \bigg(\sum_{j=1}^{n } \ln \frac{p_{m_{j-1}}^{\rm st}(\lambda_{\tau_j } ) w_{m_{j},m_{j-1}}^{(\nu_{j})}(\lambda_{\tau_j } ) } { p_{m_{j}}^{\rm st}(\lambda_{\tau_j } ) \tilde{w}_{m_{j-1},m_{j}}^{(\nu_{j})}(\lambda_{\tau_{j}})}\bigg ) \;. \nonumber\end{aligned}\ ] ] we assume for the moment that @xmath358 $ ] can be expressed exclusively in terms of quantities of the dynamics ( [ aaaaa ] ) i.e. a recipe has to be provided to express the tilde quantities in ( [ caaae ] ) [ @xmath359 and @xmath360 in terms of non - tilde quantities . + in analogy with ( [ caaad ] ) , we define @xmath361 \equiv \ln \frac{\tilde{{\cal p}}[\bar{m}_{(\tau ) } ] } { { \cal p}[m_{(\tau ) } ] } \label{caaag}\end{aligned}\ ] ] for the tilde dynamics . the previous recipe also implies that @xmath362 $ ] can be exclusively expressed in terms of quantities of the tilde dynamics ( [ caaab ] ) . ( [ caaad ] ) together with ( [ caaag ] ) implies that @xmath358=-\tilde{r}[\bar{m}_{(\tau)}]$ ] . + the probability @xmath363 to observe a trajectory such that @xmath358=r$ ] during the forward dynamics is related to the probability @xmath364 to observe a trajectory such that @xmath362=-r$ ] during the backward dynamics @xmath365 \delta(r - r[m_{(\tau ) } ] ) \label{caaah}\\ & = & \sum_{m_{(\tau ) } } \tilde{{\cal p}}[\bar{m}_{(\tau ) } ] { { \rm e}}^{r[m_{(\tau ) } ] } \delta(r - r[m_{(\tau ) } ] ) \nonumber \\ & = & { { \rm e}}^{r } \sum_{m_{(\tau ) } } \tilde{{\cal p}}[\bar{m}_{(\tau ) } ] \delta(r - r[m_{(\tau)}])\nonumber \\ & = & { { \rm e}}^{r } \sum_{\bar{m}_{(\tau ) } } \tilde{{\cal p}}[\bar{m}_{(\tau ) } ] \delta(r+\tilde{r}[\bar{m}_{(\tau ) } ] ) \nonumber \\ & = & { { \rm e}}^{r } \tilde{{\rm p}}(-r ) \nonumber \;.\end{aligned}\ ] ] by integrating @xmath366 over @xmath367 , we get @xmath368 } \rangle}=1 \label{caaai } \;.\end{aligned}\ ] ] it follows from jensen s inequality @xmath164 , that @xmath369 \rangle } \geq 0 $ ] . + we now make a first choice of @xmath370 in the backward dynamics ( [ caaab ] ) @xmath371 in this case the backward dynamics is identical to the original one , except that the driving protocol is time reversed . if we also choose the initial conditions of the backward dynamics to be the final conditions of the forward dynamics @xmath372 , using ( [ caaae ] ) and ( [ trajentr_aaaaf ] ) , we find @xmath357 = \delta s_{tot } [ m_{(\tau)},t ] \;. \label{caaao}\end{aligned}\ ] ] the ft ( [ fti ] ) previously derived using gfs now follows from eq . ( [ caaai ] ) . using ( [ caaah ] ) , we also get the detailed form of the ft @xmath373 we now make a second choices of @xmath370 in ( [ caaab ] ) @xmath374 in the theory of mes , ( [ caaab ] ) with ( [ caaax ] ) is called the time reversal me of ( [ aaaaa ] ) @xcite . we again choose the initial condition of the tilde dynamics to be the final conditions of the original dynamics @xmath372 . using ( [ caaaf ] ) with ( [ caaax ] ) and ( [ trajentr_aaaaj ] ) , we get @xmath357 = \delta s_{na}[m_{(\tau)},t ] \;. \label{caaay}\end{aligned}\ ] ] the previously derived ft ( [ ftx ] ) follows now from ( [ caaai ] ) . from eq . ( [ caaah ] ) we find the detailed form of the ft @xmath375 we can interpret the change in the total tep during the @xmath255 to @xmath95 time interval as the logarithm of the ( forward ) probability that the driven system follows a given trajectory divided by the backward probability that the system , initially in the final probability distribution of the forward evolution , and driven in a time reversed way compared to the forward evolution , follows the time - reversed trajectory . + the nonadiabatic tep is interpreted as the logarithm of the ( forward ) probability that the driven system follows a given trajectory divided by the backward probability that the system , initially in the final probability distribution of the forward evolution , and described by the time - reversed me , follows the time - reversed trajectory . + it should be noted that the backward me ( [ caaab ] ) with ( [ caaax ] ) is different from the backward me ( [ caaab ] ) with ( [ caaaoa ] ) only for systems interacting with multiple reservoirs which break the dbc . only in this case ( [ fti ] ) is different from ( [ ftx ] ) .
it is well known that relativistic quantum field theory models with four - fermion ( 4f ) interactions serve as effective theories for low energy considerations of different real phenomena in a variety of physical branches . for example , the meson spectroscopy , neutron star and heavy - ion collision physics are often investigated in the framework of ( 3 + 1)-dimensional 4f theories @xcite , known as nambu jona - lasinio ( njl ) models @xcite . in particular , low dimensional 4f field theories provide a powerful tool for investigations in condensed - matter physics . indeed , physics of ( quasi)one - dimensional organic peierls insulators ( the best known material of this kind is polyacetylene ) is well described in terms of the ( 1 + 1)-dimensional 4f gross - neveu ( gn ) model @xcite . the quasirelativistic treatment of electrons in planar systems like high - temperature superconductors or in graphene ( a planar monoatomic layer of carbon atoms ) is also possible in terms of ( 2 + 1)-dimensional gn models @xcite . notice that there are physical effects , which were observed for the first time just in the framework of njl- and gn - type models . in particular , the phenomenon of a dynamical generation of a fermion mass is well - known for strong interaction physics since the time , when nambu and jona - lasinio @xcite proved a dynamical breaking of the continuous @xmath2-symmetry on the basis of a generic 4f interaction theory . later on , this effect served as the basis of a qualitatively successful description of the low - energy meson spectrum of quantum chromodynamics ( qcd ) @xcite . the effect of dynamical symmetry breaking and generation of a fermion mass is also known in low dimensional , d=1 + 1 and d=2 + 1 , gn theories @xcite , where the four - fermion theory is renormalizable and asymptotic free @xcite in the case of d=1 + 1 , whereas for d=2 + 1 the 4f gn models are perturbatively nonrenormalizable but become renormalizable in the framework of the @xmath3 expansion technique @xcite ( @xmath4 being the number of fermion fields ) . another example of this kind is the effect of spontaneous chiral symmetry breaking induced by external magnetic or chromomagnetic fields . this effect was for the first time studied also in terms of the ( 2 + 1)-dimensional gn model @xcite . note in addition that due to a rather simple structure , different low dimensional gn models provide a good laboratory for a deeper study of dense baryon matter and , in particular , for the consideration of such phenomena as ( color)superconductivity @xcite , charged pion condensation @xcite , etc . moreover , these theories are very useful in developing new quantum field theoretical techniques like the @xmath3 expansion @xcite , the optimized perturbation theory @xcite , and so on . it is necessary to point out that there is one more , not yet mentioned above , phenomenon that is still a well - known feature of only some ( 1 + 1)-dimensional 4f theories . this is the duality correspondence between chiral symmetry breaking and superconductivity @xcite . ( in order to avoid the prohibition of cooper pairing as well as spontaneous breaking of a continuous symmetry in ( 1 + 1)-dimensional models @xcite , the consideration should be performed in the leading order of the @xmath3-technique , i.e. in the large-@xmath4 limit assumption . in this case quantum fluctuations , which would otherwise destroy a long - range order corresponding to spontaneous symmetry breaking , are suppressed by @xmath3 factors . ) to formulate this phenomenon , let us imagine that there is a microscopic 4f theory ( for d=1 + 1 see , e.g. , @xcite ) which describes a competition between fermion - antifermion ( or chiral ) and difermion ( or superconducting ) channels of interaction . moreover , we suppose that the ground state of the model is characterized by nonzero fermion number- and chiral charge densities ( the last is an imbalance between densities of left- and right - handed fermions ) , i.e. there are two external parameters , fermion number- @xmath0 and chiral charge- @xmath1 chemical potentials , respectively . then the duality phenomenon means that the phase structure of this 4f model is symmetric with respect to the following simultaneous transformations @xmath5 , @xmath6 and chiral symmetry breaking @xmath7 superconductivity ( here @xmath8 and @xmath9 are the coupling constants in the chiral and superconducting channels , respectively ) . thus , there is a correspondence between properties ( phase structure ) of the model , e.g. , at @xmath10 and @xmath11 , etc . moreover , knowing condensates and other dynamical and thermodynamical quantities of the system , e.g. in the chirally broken phase , one can obtain the corresponding quantities in the superconducting phase of the model , by simply performing there the duality transformation . it is worth to note that in recent years properties of media with nonzero chiral chemical potential @xmath1 , i.e. chiral media , attracted considerable interest ( see , e.g. , @xcite and references therein ) . in nature , chiral media might be realized in heavy - ion collisions , compact stars , condensed matter systems , etc . @xcite ( see also the review @xcite ) . in particular , one can expect that in the quark - gluon - plasma phase of qcd a chirality imbalance is produced , i.e. there appears a nonzero chiral charge density @xmath12 . the combined effect of the chiral imbalance and of the very strong magnetic field , which can be produced at noncentral heavy ion collision , results in the so - called chiral magnetic effect ( see , e.g. , the paper @xcite ) . it means that there might be induced an electric current along the direction of the magnetic field . it is important to note that this phenomenon is described effectively in the framework of 4f models with a chiral chemical potential @xmath1 , i.e. in terms of a quantity conjugated to the chiral charge density @xmath12 . recently , it was established that in planar condensed matter systems there may exists a phenomenon , which can be regarded as an analogue of the qcd chiral magnetic effect @xcite . this is the pseudo chiral magnetic effect , which can be observed , e.g. , in a distorted graphene sheet under the influence of external in - plane magnetic field . indeed , due to a mechanical distortion of the lattice structure of graphene , there might appears a nonzero chiral charge density @xmath12 ( chiral imbalance ) as well as a conserved electric current along the in - plane external magnetic field ( for details , see the paper @xcite ) . notice that , similar to the qcd chiral magnetic effect , one might expect that the pseudo chiral magnetic effect can be effectively described in terms of ( 2 + 1)-dimensional 4f models with nonzero chiral chemical potential @xmath1 ( in addiion to the usual chemical potential @xmath0 ) . finally , it is important to remark that a mechanical distortion of the lattice structure in graphene - like materials ( as well as other external perturbations ) can also lead to opening of different superconducting channels in the system @xcite . thus , it makes sense to study a competition between chiral symmetry breaking and superconductivity in ( 2 + 1)-dimensional quantum field theories with chiral chemical potential @xmath1 . in particular , it is important to predict / observe the possible duality relations between these qualitatively different phenomena accompanied by @xmath1 in field theories in spacetime dimensions d@xmath131 + 1 . in this paper we demonstrate that there exists a dual correspondence ( or dual symmetry ) between the phenomena of the chiral symmetry breaking and superconductivity in the framework of some ( 2 + 1)-dimensional 4f models . the consideration is performed at zero temperature @xmath14 , however it can be easily generalized to the case @xmath15 . we hope that our investigations shed some new light on the physical effects in planar condensed matter systems . the paper is organized as follows . in sec . ii a ( 2 + 1)-dimensional 4f gn - type model , containing both fermion - antifermion ( or chiral ) and difermion ( or superconducting ) interaction channels and including two kinds of chemical potentials , @xmath16 , is presented . here we show that under the pauli - grsey transformations of fermi fields there is a dual relationship between the relevant 4f structures . next , the unrenormalized thermodynamic potential ( tdp ) of the gn - type model is given in the leading order of the large - n expansion . in sec iii the dual symmetry of the model tdp is established . it means that it is invariant under the interchange of coupling constants , @xmath16 chemical potentials and chiral and superconducting condensates . moreover , the renormalization of the tdp is performed . iv contains a detailed numerical investigation of various phase portraits with particular emphasis on the role of the duality symmetry of the tdp . some technical details are relegated to appendices . moreover , in appendix [ apd ] we present an alternative ( 2 + 1)-dimensional gn model with dual relationship between other chiral and superconducting channels of 4f interaction . our investigation is based on a ( 2 + 1)-dimensional gn type model with massless fermions belonging to a fundamental multiplet of the auxiliary @xmath17 flavor group . its lagrangian describes the interaction both in the fermion antifermion and difermion ( or superconducting ) channels : @xmath18\psi_k&+ & \frac { g_1}n\left ( 4f\right ) _ { ch}+\frac { g_2}n\left ( 4f\right ) _ { sc},\label{1}\end{aligned}\ ] ] where the four - fermion structures @xmath19 and @xmath20 are used , @xmath21 in addition , @xmath0 and @xmath1 in ( 1 ) denote a fermion number chemical potential and a chiral ( axial ) chemical potential , respectively . @xmath0 is conjugated to a fermion number density @xmath22 , whereas @xmath1 is conjugated to a nonzero density of chiral charge @xmath23 , which represents an imbalance in densities of the right- and left - handed fermions . as it is noted above , all fermion fields @xmath24 ( @xmath25 ) form a fundamental multiplet of the @xmath17 group . moreover , each field @xmath24 is a four - component ( reducible ) dirac spinor ( the symbol @xmath14 denotes the transposition operation ) . therefore , in eqs . ( 1)-(2 ) the quantities @xmath26 ( @xmath27 ) and @xmath28 are matrices in the four - dimensional spinor space . moreover , @xmath29 is the charge conjugation matrix . the algebra of these matrices as well as their particular representations are given in appendix [ apa ] . clearly , the lagrangian @xmath30 is invariant under transformations from the internal auxiliary @xmath17 group , which is introduced here in order to make it possible to perform all the calculations in the framework of the nonperturbative large-@xmath4 expansion method . physically more interesting is that the model ( 1 ) is invariant under transformations from the @xmath31 group , where @xmath32 is the fermion number conservation group , @xmath33 ( @xmath34 ) , and @xmath35 is the group of continuous chiral transformations , @xmath36 ( @xmath25 ) . it means that in the framework of the model ( 1 ) both the particle number density @xmath37 and the density of chiral charge @xmath38 are conserved quantities . , which anticommutes with @xmath26 ( @xmath39 ) and @xmath28 , one can consider another continuous chiral @xmath40 transformation group of the spinor fields , @xmath41 ( @xmath25 ) . alternatively , there exists a 4f model with another fermion antifermion and difermion channels of interaction , which , in addition to @xmath32 , is invariant under the continuous chiral @xmath40 group ( see appendix [ apd ] ) . it is worth mentioning that in the case of graphene reducible four - component spinors just describe the two sublattice ( or pseudospin ) and two valley ( or dirac point ) degrees of freedom of the hexagonal honeycomb lattice of carbon atoms ( see e.g. @xcite and references therein ) . ] before studying the thermodynamics of the model , we would like to point out that there is a so - called duality correspondence between chiral and superconducting channels of the model ( 1)-(2 ) . to see this , it is very useful to form an infinite set @xmath42 composed of all lagrangians @xmath43 when the free model parameters @xmath44 and @xmath1 take arbitrary admissible values , i.e. @xmath45 at arbitrary fixed values of coupling constants @xmath46 , @xmath47 and chemical potentials @xmath48 . then , let us perform in ( 1)-(2 ) the so - called pauli - grsey ( pg ) transformation of spinor fields @xcite , @xmath49 taking into account that all spinor fields anticommute with each other , it is easy to see that under the action of the transformations ( [ 0004 ] ) the 4f structures of the lagrangian ( 1 ) are converted into themselves , i.e. @xmath50 and , moreover , each element ( lagrangian ) @xmath43 of the set @xmath51 is transformed into another element of the set @xmath51 according to the following rule @xmath52 i.e. the set @xmath51 is invariant under the field transformations ( [ 0004 ] ) . owing to the relations ( [ 040 ] ) and ( [ 004 ] ) there is a connection between properties of the model at some fixed free model parameters @xmath53 and properties of the model in the case , when @xmath5 and @xmath6 . in particular , if at some fixed @xmath44 and @xmath1 we have the chiral symmetry breaking ( csb ) phase , then at @xmath5 and @xmath6 one can definitely predict the superconducting ( sc ) phase , and vice versa . due to this reason , we will call the relations ( [ 040 ] ) and ( [ 004 ] ) the duality property of the model ( or duality correspondence between csb and sc ) . note also that in appendix [ apd ] we present another ( 2 + 1)-dimensional 4f model , in which the duality correspondence between alternative chiral and superconducting channels is realized . further on , we will study the role and the influence of the duality property of the model ( 1 ) on its phase structure . to this end , we introduce the semi - bosonized version of lagrangian ( [ 1 ] ) that contains only quadratic powers of fermionic fields as well as auxiliary bosonic fields @xmath54 , @xmath55 , @xmath56 and @xmath57 , @xmath58\psi_k -\frac{n(\sigma^2+\pi^2)}{4g_1 } -\frac n{4g_2}\delta^{}\delta- \frac{\delta^{}}{2}[\psi_k^tc\psi_k ] -\frac{\delta}{2}[\bar\psi_k c\bar\psi_k^t ] . \label{2}\end{aligned}\ ] ] ( here and in what follows the summations over repeated indices @xmath25 and @xmath59 are implied . ) clearly , the lagrangians ( 1 ) and ( [ 2 ] ) are equivalent , as can be seen by using the euler - lagrange equations of motion for bosonic fields which take the form @xmath60 one can easily see from ( [ 3 ] ) that the neutral fields @xmath61 and @xmath55 are real quantities , i.e. @xmath62 and @xmath63 ( the superscript symbol @xmath64 denotes the hermitian conjugation ) , but the ( charged ) difermion fields @xmath56 and @xmath65 are hermitian conjugated complex quantities , so that @xmath66 and vice versa . moreover , under the chiral @xmath35 group the fields @xmath67 are singlets , but the fields @xmath68 are transformed in the following way @xmath69 clearly , all the fields ( [ 3 ] ) are also singlets with respect to the auxiliary @xmath17 group , since the representations of this group are real . moreover , with respect to the parity transformation @xmath70 , @xmath71 the fields @xmath61 , @xmath72 and @xmath57 are even quantities , i.e. scalars , but @xmath55 is a pseudoscalar . if the difermion field @xmath56 has a nonzero ground state expectation value , i.e. @xmath73 , then the abelian fermion number conservation @xmath32 symmetry of the model is spontaneously broken down and the superconducting phase is realized in the model . however , if @xmath74 then the continuous @xmath35 chiral symmetry of the model is spontaneously broken . let us now study the phase structure of the four - fermion model ( 1 ) starting from the equivalent semi - bosonized lagrangian ( [ 2 ] ) . in the leading order of the large-@xmath4 approximation , the effective action @xmath75 of the considered model is expressed by means of the path integral over fermion fields : @xmath76[d\psi_l]\exp\bigl(i\int { \cal l}\,d^3 x\bigr),\ ] ] leading to @xmath77 + \widetilde { \cal s}_{\rm { eff}}. \label{5}\end{aligned}\ ] ] the term @xmath78 in ( [ 5 ] ) is the fermion contribution to the effective action and is given by : @xmath79[d\psi_l]\exp\bigl\{i\int\big [ \bar\psi_k(\gamma^\nu i\partial_\nu + \mu\gamma^0+\mu_5\gamma^0\gamma^5 -\sigma - i\gamma^5\pi ) \psi_k - \frac{\delta^{}}{2}(\psi_k^t c\psi_k ) -\frac{\delta}{2}(\bar\psi_kc \bar\psi_k^t)\big ] d^3 x\bigr\}. \label{6}\ ] ] the ground state expectation values @xmath80 , @xmath81 , etc . of the composite bosonic fields are determined by the saddle point equations , @xmath82 for simplicity , throughout the paper we suppose that the above mentioned ground state expectation values do not depend on spacetime coordinates , i.e. @xmath83 where @xmath84 are constant quantities . in fact , they are coordinates of the global minimum point of the thermodynamic potential ( tdp ) @xmath85 . in the leading order of the large-@xmath4 expansion and using ( 13 ) it is defined by the following expression : @xmath86 the tdp ( [ 08 ] ) is invariant with respect to chiral @xmath35 symmetry group . so , as it is clear from ( [ 003 ] ) , it depends on the quantities @xmath87 and @xmath88 through the combination @xmath89 . moreover , without loss of generality , one can suppose that @xmath90 . thus , to find the other ground state expectation values @xmath80 etc . , it is enough to study the global minimum point of the tdp @xmath91 , @xmath92 taking into account the relations ( [ 5 ] ) , ( [ 6 ] ) and ( [ 08 ] ) , we have from ( [ 09 ] ) @xmath93[d\psi_l]\exp\big ( i\int d^3 x\big [ \bar\psi_k d\psi_k\right.\nonumber\\ & & \left.- \frac{\delta^}{2}(\psi_k^tc\psi_k ) -\frac{\delta}{2}(\bar\psi_k c\bar\psi_k^t)\big ] \big ) \right ) , \label{9}\end{aligned}\ ] ] where @xmath94 . to proceed further , let us point out again that without loss of generality the quantities @xmath95 might be considered as real ones . might be eliminated by an appropriate transformation of fermion fields in the path integral ( [ 9 ] ) . ] so , in what follows we will suppose that @xmath96 , where @xmath97 now is already a real quantity . the path integration in the expression ( [ 9 ] ) is evaluated in appendix [ apb ] , we consider for simplicity the case @xmath98 , however the procedure is easily generalized to the case with @xmath99 . ] so we have for the tdp ( [ 9 ] ) the following expression @xmath100 , \label{11}\end{aligned}\ ] ] where @xmath101 are presented in ( [ b12 ] ) and superscription un denotes the unrenormalized quantity . note that the tdp ( [ 11 ] ) describes thermodynamics of the model at zero temperature . taking into account the expressions ( [ b12 ] ) , the tdp ( [ 11 ] ) can be presented in the form @xmath102 where @xmath103 and @xmath104 it is clear from ( [ 013 ] ) that the tdp ( [ 012 ] ) is an even function of each of the quantities @xmath0 , @xmath1 , @xmath87 , and @xmath97 , i.e. without loss of generality we can consider in the following only @xmath105 , @xmath106 , @xmath107 , and @xmath108 values of these quantities . moreover , the tdp ( [ 012 ] ) is invariant with respect to the so - called duality transformation @xcite , @xmath109 notice that this invariance of the tdp is a consequence of the rule ( [ 004 ] ) , according to which the 4f lagrangian ( 1 ) is modified under the pauli - gursey transformation ( [ 0004 ] ) of the spinor field . in powers of @xmath97 the fourth - degree polynomial @xmath110 has the following form @xmath111 expanding the right - hand side of ( [ 17 ] ) in powers of @xmath87 , one can obtain an equivalent alternative expression for this polynomial . namely , @xmath112 note also that according to the general theorem of algebra , the polynomial @xmath110 can be presented in the form @xmath113 where @xmath114 , @xmath115 , @xmath116 and @xmath117 are the roots of this polynomial . the fourth - order polynomial with similar coefficients @xmath118 ( [ 013 ] ) was studied in our paper @xcite , where it was shown that all its roots @xmath119 ( @xmath120 ) are real quantities ( see appendix b in @xcite ) . the roots @xmath119 are the energies of quasiparticle or quasiantiparticle excitations of the system . in particular , it follows from ( [ 17 ] ) that at @xmath121 @xmath122 whereas it is clear from ( [ 18 ] ) that at @xmath123 @xmath124 taking into account in ( [ 012 ] ) the relation ( [ 170 ] ) as well as the formula @xmath125 ( obtained rigorously , e.g. , in appendix b of @xcite and being true up to an infinite term independent of the real quantity @xmath126 ) , it is possible to integrate there over @xmath127 . so the unrenormalized tdp ( [ 012 ] ) can be presented in the following form , @xmath128 first of all , let us obtain a finite , i.e. renormalized , expression for the tdp ( [ 28 ] ) at @xmath130 and @xmath131 , i.e. in vacuum . since in this case a thermodynamic potential is usually called effective potential , we use for it the notation @xmath132 . it follows from ( [ 012 ] ) and ( [ 013 ] ) that at @xmath130 and @xmath131 @xmath132 looks like @xmath133\nonumber\\ & \equiv&\frac{m^2}{4g_1}+\frac{\delta^2}{4g_2}-\int\frac{d^2p}{(2\pi)^2}\big ( \sqrt{\|\vec p\|^2+(m+\delta)^2}+\sqrt{\|\vec p\|^2+(m-\delta)^2}\big ) , \label{25}\end{aligned}\ ] ] where @xmath134 . ( to obtain the second line of this formula , we have integrated there over @xmath127 according to the relation ( [ int ] ) . ) it is evident that the effective potential ( [ 25 ] ) is an ultraviolet divergent quantity . so , we need to renormalize it . this procedure consists of two steps : ( i ) first of all we need to regularize the divergent integral in ( [ 25 ] ) , i.e. we suppose there that @xmath135 , @xmath136 . ( ii ) second , we must suppose also that the bare coupling constants @xmath8 and @xmath9 depend on the cutoff parameter @xmath137 in such a way that in the limit @xmath138 one obtains a finite expression for the effective potential . before performing the steps ( i ) and ( ii ) of the renormalization procedure , it is useful to take into account the following asymptotic expansion at @xmath139 @xmath140 then , after construction of the regularized effective potential @xmath141 ( see below ) , we use there the asymptotic expansion ( [ 19 ] ) and integrate it over @xmath142 and @xmath143 term - by - term . the result reads @xmath144+\delta^2\left [ \frac 1{4g_2}-\frac{2\lambda\ln(1+\sqrt{2})}{\pi^2}\right ] -\frac{2\lambda^3(\sqrt{2}+\ln(1+\sqrt{2}))}{3\pi^2}+{\cal o}(\lambda^0 ) , \label{12}\end{aligned}\ ] ] where @xmath145 denotes an expression which is finite in the limit @xmath146 . now , we should suppose that the bare coupling constants @xmath8 and @xmath9 depend on the cutoff parameter @xmath137 in such a way that in the limit @xmath138 one obtains a finite expressions in the square brackets of ( [ 12 ] ) . clearly , to fulfil this requirement it is sufficient to require that @xmath147 where @xmath148 are finite and @xmath137-independent model parameters with dimensionality of inverse mass . since bare couplings @xmath8 and @xmath9 do not depend on a normalization point , the same property is also valid for @xmath148 . hence , taking into account in ( [ 12 ] ) the relations ( [ 13 ] ) and ignoring there an infinite @xmath87- and @xmath97-independent constant , one obtains the following _ renormalized _ , i.e. finite , expression @xmath149 for the effective potential , @xmath150 it should also be mentioned that the effective potential ( [ 14 ] ) is a renormalization group invariant quantity . the fact that it is possible to renormalize the effective potential of the initial model ( 1 ) in the leading order of the large @xmath4-expansion is the reflection of a more general property of ( 2 + 1)-dimensional theories with four - fermion interactions . indeed , it is well known that in the framework of the naive perturbation theory ( over coupling constants ) these models are not renormalizable . however , as it was proved in @xcite , in the framework of nonperturbative large @xmath4-technique these models are renormalizable in each order of the @xmath3-expansion . the @xmath145 term in ( [ 12 ] ) can be calculated explicitly , so we have for the renormalized effective potential @xmath149 ( [ 14 ] ) the following expression @xmath151 the global minimum point @xmath152 , where @xmath153 and @xmath154 , of this function was already investigated in @xcite , although in the framework of another ( 2 + 1)-dimensional gn model . so , we present at once the phase structure of the initial model ( 1 ) at @xmath130 and @xmath131 ( see fig . 1 ) . in fig . 1 the phase portrait of the model is depicted depending on the values of the free model parameters @xmath155 and @xmath156 . there the plane @xmath157 is divided into several areas . in each area one of the phases i , ii or iii is implemented . in the phase i , i.e. at @xmath158 and @xmath159 , the global minimum of the effective potential @xmath160 is arranged at the origin . so in this case we have @xmath161 and @xmath162 . as a result , in the phase i both continuous symmetries , chiral @xmath35 and electromagnetic @xmath32 , remain intact and fermions are massless . due to this reason the phase i is called symmetric . in the phase ii , which is allowed only for @xmath163 , at the global minimum point the relations @xmath164 and @xmath162 are valid . so in this phase chiral @xmath35 symmetry is spontaneously broken down and fermions acquire dynamically the mass @xmath165 . finally , in the superconducting phase iii , where @xmath166 , we have spontaneous breaking of the @xmath32 symmetry . this phase corresponds to the following values of the gaps : @xmath161 , @xmath167 . note also that if @xmath168 and , in addition , @xmath169 ( it is just the line _ l _ in fig . 1 ) , then the effective potential ( [ 15 ] ) has two equivalent global minima . the first one , the point @xmath170 , corresponds to a phase with chiral symmetry breaking . the second one , i.e. the point @xmath171 , corresponds to superconductivity . clearly , if the cutoff parameter @xmath137 is fixed , then the phase structure of the model can be described in terms of bare coupling constants @xmath172 instead of finite quantities @xmath173 . indeed , let us first introduce a critical value of the couplings , @xmath174 . then , as it follows from fig . 1 and ( [ 13 ] ) , at @xmath175 and @xmath176 the symmetric phase i of the model is located . if @xmath177 , @xmath176 ( @xmath175 , @xmath178 ) , then the chiral symmetry broken phase ii ( the superconducting phase iii ) is realized . finally , let us suppose that both @xmath177 and @xmath178 . in this case at @xmath11 ( @xmath10 ) we have again the chiral symmetry broken phase ii ( the superconducting phase iii ) . now , few comments about the nature of the superconductivity ( sc ) and chiral symmetry breaking ( csb ) are in order . in the framework of the model ( 1 ) there are two well - known mechanisms for appearing of these phenomena , ( i ) dynamical symmetry breaking , which occurs at strong couplings , and ( ii ) cooper instability of fermi surface , which takes place , in contrast to the case ( i ) , at weak couplings . if @xmath130 and @xmath131 , then the initial model ( 1 ) describes effectively the undoped regime of high temperature cuprate superconductors , etc ( see , e.g. , the papers @xcite , where the phase structure , similar to the phase diagram of our fig . 1 , was obtained ) . in this case fermi surface is absent , so both csb and sc phases appear dynamically in the system at a _ rather strong attraction _ in the fermion - antifermion or difermion channels , i.e. at @xmath177 and @xmath11 or at @xmath178 and @xmath179 , respectively . hence , at zero values of chemical potentials only the mechanism ( i ) for generating both sc and csb is realized in our model . is well known since the papers by nambu and jona - lasinio [ 3 ] . it is also valid in the framework of other ( 3 + 1)-dimensional 4f models with chiral and color superconducting channels of interaction at @xmath130 @xcite . in @xcite the bare coupling constants @xmath8 and @xmath9 are free model parameters . moreover , strong constraints were obtained there on @xmath8 and @xmath9 , at which color sc or csb is generated dynamically in the models . ] the second way , i.e. the mechanism ( ii ) , to break spontaneously a symmetry can be realized only at nonzero chemical potentials . for example , if the doping or impurities are present in the system , then one must introduce a finite chemical potential @xmath0 . in this case the fermi surface can be created . as a result , if there is an _ arbitrary small attraction _ in the sc channel , i.e. even at @xmath176 , there appears the so - called cooper instability of the fermi surface , which destroys the normal ground state in favor of superconductivity . ( below , when discussing the phase portrait of fig . 5 ( see sect [ ivb ] ) , we will see that at @xmath131 and @xmath180 both ( i ) and ( ii ) mechanisms take part in forming the phase structure . ) if @xmath130 but @xmath181 , then csb can also be generated in the model due to the mechanism ( ii ) at arbitrary small values of @xmath8 ( see in @xcite ) . this fact supports the duality correspondence between sc and csb in the model ( 1 ) . + to renormalize the tdp ( [ 28 ] ) in the most general case , it is necessary to imagine how the integrand of ( [ 28 ] ) , i.e. the quantity @xmath182 , behaves at asymptotically high values of @xmath183 . the properties of the roots @xmath119 ( @xmath120 ) of the polynomial @xmath110 appearing in ( [ 012 ] ) can be obtained both numerically and analytically using the methods of appendix b of @xcite . so it is easy to show that at @xmath139 @xmath184 since the asymptotic expansions ( [ 016 ] ) and ( [ 19 ] ) coincide , one can subtract from the integrand of ( [ 28 ] ) the expression @xmath185 , thereby obtaining a convergent integral . taking into account this circumstance , we have identically @xmath186 where @xmath187 is the effective potential ( [ 25 ] ) of the model in vacuum , i.e. at @xmath130 and @xmath131 , and @xmath188.\label{018}\end{aligned}\ ] ] it is clear that the term @xmath189 in ( [ 017 ] ) is a convergent integral and all ultraviolet divergences of the tdp ( [ 28])-([017 ] ) are located in the term @xmath187 . hence , we can renormalize only the first term of ( [ 017 ] ) in the way of the previous subsection and obtain finally : @xmath190 where @xmath149 and @xmath189 are presented in ( [ 15 ] ) and ( [ 018 ] ) , correspondingly . it was already mentioned above that using the method presented in appendix b of the paper @xcite , the roots @xmath119 can be calculated numerically ( at fixed values of @xmath191 , etc ) . as a result , it is also possible to study numerically the whole tdp ( [ 20 ] ) . carrying out this procedure , we have seen that the global minimum of the tdp @xmath192 is always at the point of the form @xmath193 or @xmath194 . so , to get a more detailed information about the phase structure of the model , it is sufficient to investigate the reductions of the tdp ( [ 20 ] ) on the @xmath87- and @xmath97 axes on the @xmath87- and @xmath97 axes is known ( see ( [ 26 ] ) and ( [ 27 ] ) ) . ] , i.e. to consider the functions @xmath195\nonumber\\ & = & v_1(m)-\int\frac{d^2p}{(2\pi)^2 } \sum_{\eta = \pm}\big [ { \cal e}_{\eta}-\sqrt{\|\vec p\|^2+m^2}+(\mu-{\cal e}_{\eta})\theta(\mu-{\cal e}_{\eta})\big ] , \label{31}\\ f_2(\delta)&\equiv&\omega^{ren}(m=0,\delta)= v_2(\delta)-\int\frac{d^2p}{(2\pi)^2 } \sum_{\eta = \pm}\big [ \sqrt{\delta^2+(\|\vec p\|+\eta\mu)^2}-\sqrt{\|\vec p\|^2+m^2}\nonumber\\&+&\left ( \mu_5-\sqrt{\delta^2+(\|\vec p\|+\eta\mu)^2}\right ) \theta\left ( \mu_5-\sqrt{\delta^2+(\|\vec p\|+\eta\mu)^2}\right ) \big ] , \label{32}\end{aligned}\ ] ] respectively , where @xmath196 and @xmath197 is the reduction of the vacuum effective potential ( [ 15 ] ) on the @xmath87 ( in this case @xmath198 , @xmath199 ) or @xmath97 ( in this case @xmath200 , @xmath201 ) axes . moreover , to obtain the second line in ( [ 31 ] ) we use the evident relations , @xmath202 and @xmath203 . after tedious but straightforward calculations , the tdp ( [ 31 ] ) can be presented in the following form ( see appendix [ apc ] ) : @xmath204\nonumber\\ & & -\frac{\theta\left ( \sqrt{m^2+\mu_5 ^ 2}-\mu\right ) } { 2\pi}\left [ \mu_5 ^ 2\sqrt{\mu_5 ^ 2+m^2}+\mu_5 m^2\ln\left ( \frac{\mu_5+\sqrt{\mu_5 ^ 2+m^2}}{m}\right ) \right ] \nonumber\\ & & -\frac{\theta\left ( \mu - m\right ) \theta\left ( \sqrt{m^2+\mu_5 ^ 2}-\mu\right ) } { 2\pi}\left [ \mu_5\mu\sqrt{\mu^2-m^2}-\mu_5 m^2\ln\left ( \frac{\mu+\sqrt{\mu^2-m^2}}{m}\right ) \right ] \label{c18}\end{aligned}\ ] ] ( at @xmath131 this expression coincides with the corresponding tdp ( 22 ) from the paper @xcite ) . moreover , comparing the expressions ( [ 31 ] ) and ( [ 32 ] ) , we see that the quantity @xmath205 can be easily obtained from the tdp @xmath206 by the substitutions @xmath207 , @xmath208 and @xmath6 , i.e. @xmath209 clearly , the connection ( [ 35 ] ) between the reductions of the tdp @xmath192 on the @xmath87- and @xmath97 axes is the consequence of the duality property ( [ 16 ] ) of the model . next , in order to find the phase structure of the model , we will determine the global minimum points of the tdps @xmath210 ( [ c18 ] ) and @xmath211 ( [ 35 ] ) and then compare the minimum values of these functions vs external parameters @xmath212 . suppose now that at some fixed particular values of the model parameters , i.e. at @xmath213 , @xmath214 and @xmath215 , @xmath216 , the global minimum of the tdp ( [ 20 ] ) lies at the point , e.g. , @xmath217 . it means that for such fixed values of the parameters the chiral symmetry breaking ( csb ) phase is realized . then it follows from the duality invariance of the unrenormalized tdp with respect to the transformation @xmath218 ( [ 16 ] ) that the permutation of the coupling constant and chemical potential values for the renormalized tdp ( [ 20 ] ) means @xmath219 , @xmath220 , @xmath221 , under which this tdp is invariant . ] ( i.e. at @xmath222 , @xmath223 and @xmath224 , @xmath225 ) moves the global minimum of the tdp @xmath226 to the point @xmath227 , and the superconducting ( sc ) phase is originated ( and vice versa ) . this is the so - called duality correspondence between csb and sc phases in the framework of the model under consideration . hence , the knowledge of a phase of the model ( 1 ) at some fixed values of external free model parameters @xmath228 is sufficient to understand what phase is realized at rearranged values of external parameters , @xmath229 . moreover , we would like to emphasize once again that there exists an equality of the order parameters ( condensates ) , which characterizes both the initial phase and the phase corresponding to rearranged external parameters . in other words the chiral condensate of the csb phase at fixed @xmath228 is equal to the sc condensate of the phase at @xmath229 ( and vice versa ) . the duality transformation ( [ 16 ] ) of the tdp can also be applied to the arbitrary phase portrait of the model ( see below ) . in particular , it is clear that if we have a most general phase portrait , i.e. the correspondence between any point @xmath230 of the four - dimensional space of external parameters and possible model phases ( csb , sc and symmetric phase ) , then under the duality transformation ( @xmath231 , @xmath6 , csb@xmath7sc ) this phase portrait is mapped to itself , i.e. the most general phase portrait is selfdual . in practice , usually , there are constraints on the model parameters . as a result , if the constraint is dually ( non)invariant , then the phase portrait is also a dually ( non)invariant . below , we will use the dual symmetry of the tdp in order to explain and construct the phase structure of the model in different particular cases . + first of all , taking into account the duality correspondence between csb and sc , let us determine its characteristic features and then find ( analytically and numerically ) the phase structure of the model ( 1 ) in two particular cases , ( i ) @xmath232 and ( ii ) @xmath233 . it is evident that both of these constraints , ( i ) and ( ii ) , are invariant with respect to the duality transformation @xmath231 , @xmath6 . hence , the corresponding phase portraits should be selfdual ( a more detailed explanation of this fact is presented in the next subsection ) . in the case ( i ) we want to study the @xmath157 phase portrait , whereas in the case ( ii ) the @xmath234 phase portrait . it follows from the duality correspondence that in the case ( i ) the region of the csb phase is a mirror image of the sc phase with respect to the line @xmath233 of the @xmath157 plane . however , in the case ( ii ) the regions of these phases are arranged mirror symmetrically with respect to the line @xmath232 of the plane @xmath234 . of course , the specific form of the csb and sc areas depends on the values of the external parameters . for example , at @xmath235 the phase structure is presented in fig ( note that in order to obtain this phase portrait it is sufficient to investigate the behavior of the global minimum point of the tdp @xmath149 ( [ 15 ] ) vs the coupling constants @xmath155 and @xmath156 . ) however , at fixed @xmath236 the phase structure is presented in fig . 2 , where one can see the mirror symmetrical arrangement of the phases ii and iii with respect to the line @xmath233 . in figs 3 , 4 the @xmath234-phase portraits of the model ( 1 ) are presented at fixed @xmath237 and fixed @xmath238 , respectively . the csb and sc phases in these figures are arranged , as defined above , symmetrically with respect to the line @xmath232 of the @xmath234 plane . if @xmath239 , then at the points of the boundary of the phase i in fig . 3 , the second order phase transition takes place . in this case it is possible to give exact analytical expressions for the boundaries of the region i in fig . indeed , the global minimum point @xmath165 of the tdp ( [ c18 ] ) is defined by the equation @xmath240 . inside the region ii of fig . 3 we have @xmath241 and @xmath242 . however , on the boundary between phases i and ii one can see that @xmath161 and , moreover , that @xmath243 . it is the last relation that defines the boundary equation between i and ii phases . this equation looks like @xmath244 in a similar way it is possible to find the equation for the boundary between i and iii phases of fig . 3 : @xmath245 in contrast , on the boundary of the phase i of fig . 4 , a phase transition of the first order takes place . since on the boundary between ii and iii phases of this figure we also have a first order phase transitions , it is clear that the point t of the phase portrait fig . 4 is a so - called triple point , i.e. the point where three different phases , i , ii and iii , coexist . the triple point t of fig . 4 corresponds to the chemical potential values @xmath246 . in these two particular cases the restrictions on the external model parameters are already not dually invariant , so the corresponding @xmath157-phase portraits are not selfdual . however , as we shall see , each of these two phase portraits is a dual mapping of another one . the @xmath157-phase structure of the model at arbitrary fixed nonzero value of @xmath0 ( of @xmath1 ) and at @xmath131 ( at @xmath130 ) is presented in fig . 5 ( in fig . 6 ) let us first compare fig . 1 and fig it is easy to see that if @xmath158 , @xmath159 and @xmath235 , then the system is in the symmetric phase i ( see fig . 1 ) . however , an arbitrary small nonzero value of the chemical potential @xmath0 induces ( at @xmath131 ) in this case the superconducting phase iii ( see fig . 5 as well as fig . 3 for the particular case @xmath248 ) . note that there is a very simple explanation of this fact , which is based on the symmetry breaking mechanism ( ii ) ( see the end of the section iii a ) , i.e. on the cooper instability of a fermi surface . indeed , at @xmath158 , @xmath159 and @xmath180 we have a nonzero particle density ( see , e.g. , in @xcite ) , so there is a fermi sea of particles with energies less or equal to @xmath0 ( fermi surface ) . evidently , in this case there is no energy cost for creating a pair of particles with opposite momenta just over the fermi surface . then , due to an arbitrary weak attraction between these particles ( @xmath249 ) , the cooper pair is formed and @xmath32 symmetry is spontaneously broken , as a result of bose einstein condensation of cooper pairs . since in the energy spectrum of fermions the gap @xmath250 appears , rather small external forces are not able to destroy the superconducting condensate and it is a stable one . if @xmath158 , @xmath166 , then we have a rather strong interaction in the sc channel , i.e. @xmath178 and @xmath175 . as a result , in this case both at @xmath130 and @xmath180 the dynamical symmetry breaking mechnism ( i ) ( see the end of the section iii a ) plays a decisive role in the forming of sc phenomenon . now suppose that @xmath251 and @xmath252 , i.e. @xmath177 . ( in particular , the point @xmath157 might belong to the phase ii of fig . 1 , where chiral symmetry is spontaneously broken at @xmath130 , @xmath131 . ) then it follows from fig . 5 that at @xmath131 there is a sufficiently high critical value @xmath253 of the chemical potential @xmath0 , such that at @xmath254 the point @xmath255 will certainly be in one of the regions iii of fig . hence , the growth ( at @xmath131 ) of the chemical potential @xmath0 leads to the appearance of the sc phase in the model ( 1 ) at arbitrary @xmath256 and @xmath252 . , @xmath131 without duality correspondence . ] analysing in the same manner fig . 6 , we see that in the opposite case when @xmath130 but @xmath257 , the growth ( at @xmath130 ) of the chiral chemical potential @xmath1 leads to the appearance of the csb phase in the model ( 1 ) at arbitrary fixed @xmath256 and @xmath252 . indeed , these two properties of the model ( 1 ) , i.e. ( a ) the appearance of the sc at @xmath131 and at growing values of @xmath0 as well as ( b ) the appearance of the csb at @xmath130 and at growing values of @xmath1 ( in these two cases it is supposed that @xmath256 and @xmath252 are fixed ) , are dually connected ( or dually conjugated ) . to understand this fact , it is necessary to emphasize once again that ( a ) follows from the phase portrait of fig . 5 , whereas ( b ) is the consequence of the phase portrait presented in fig . 6 . now we will demonstrate that under the dual transformation @xmath218 ( @xmath231 , @xmath6 , csb@xmath7sc ) the phase portrait of fig . 5 is transformed into the phase portrait of fig . 6 ( and vice versa ) . note that the application of @xmath218 to fig . 5 can be divided into three more simple steps . ( i ) first , under permutation @xmath258 in ( [ 16 ] ) we have renaming of the phases , i.e. ii@xmath7iii . ( for example , in this case the global minimum point of the csb phase , i.e. the point @xmath259 , is transformed into the point @xmath260 and , as a result , the csb phase is transformed into the sc phase . ) ( ii ) second , performing the @xmath6 and @xmath231 transformations in fig . 5 , we rename the coordinate axes of the figure ( after the duality transformation we have along the vertical axis the quantity @xmath261 and along the horizontal one the quantity @xmath262 ) and change its caption or external conditions , at which the phase portrait is obtained . ( in our case the caption * @xmath131 with @xmath0 being an arbitrary fixed quantity * for fig . 5 is replaced by * @xmath130 with @xmath1 being an arbitrary fixed quantity * , i.e. we obtain the caption for fig . ( iii ) finally , it is necessary to direct the axis corresponding to the quantity @xmath261 ( to the quantity @xmath262 ) horizontally ( vertically ) . as a result of this dual transformation of the fig . 5 we obtain just fig . 6 . + in a similar way it is possible to apply the duality transformation @xmath218 to each of the phase portraits presented in figs . 14 . as a result , we see that these phase structures are selfdual . in the next subsubsections we obtain the phase structure of the model in other more general nonselfdual cases , i.e. the @xmath157-phase portraits at arbitrary fixed values of chemical potentials and the @xmath234-phase portraits at arbitrary fixed values of the coupling constants . -phase portrait of the model at arbitrary fixed values of @xmath0 and for different values of the chiral chemical potential @xmath1 . ( a ) the case @xmath263 . ( b ) the case @xmath264 . ( c ) the case @xmath265 . ( d ) the case @xmath266 . we use the same designations of the phases as in fig . 1 . , title="fig:",scaledwidth=45.0% ] -phase portrait of the model at arbitrary fixed values of @xmath0 and for different values of the chiral chemical potential @xmath1 . ( a ) the case @xmath263 . ( b ) the case @xmath264 . ( c ) the case @xmath265 . ( d ) the case @xmath266 . we use the same designations of the phases as in fig . 1 . , title="fig:",scaledwidth=45.0% ] + -phase portrait of the model at arbitrary fixed values of @xmath0 and for different values of the chiral chemical potential @xmath1 . ( a ) the case @xmath263 . ( b ) the case @xmath264 . ( c ) the case @xmath265 . ( d ) the case @xmath266 . we use the same designations of the phases as in fig . 1 . , title="fig:",scaledwidth=45.0% ] -phase portrait of the model at arbitrary fixed values of @xmath0 and for different values of the chiral chemical potential @xmath1 . ( a ) the case @xmath263 . ( b ) the case @xmath264 . ( c ) the case @xmath265 . ( d ) the case @xmath266 . we use the same designations of the phases as in fig . 1 . , title="fig:",scaledwidth=45.0% ] + let us now fix the fermion number chemical potential @xmath247 and consider how the @xmath157-phase portrait of the model is evolved vs the chiral chemical potential @xmath1 . this phase portrait at @xmath131 is presented in fig . 5 . in figs . 7 we have drawn several @xmath157-phase portraits at different relations between @xmath1 and @xmath0 , ( a ) @xmath263 , ( b ) @xmath264 , ( c ) @xmath265 , ( d ) @xmath266 . analyzing figs . 5 and 7 , we see the following phase evolution of the system vs @xmath1 at fixed @xmath247 . at @xmath131 the sc phase iii fills almost the whole @xmath255 plane ( see fig . it means that at arbitrary fixed @xmath256 and @xmath252 we have superconductivity in the system at sufficiently high values of @xmath0 ( see section [ ivb ] ) . if @xmath1 begins to grow , then for a lot of @xmath157 points there might occur a restoration of the initial symmetry ( the symmetrical phase i arises ) in the model however , the further growth of @xmath1 leads to the final appearance of the csb phase . it is the general property of the chiral chemical potential @xmath1 , which promotes the chiral symmetry breaking in the system . since the above mentioned constraints ( a), .. ,(d ) at which figs . 7 were drawn are not invariant under the duality transformation @xmath267 ( [ 16 ] ) , we would like to note that the phase portraits of fig . 7 are not selfdual . in this case the possible duality transformation ( see the previous subsubsection [ ivb ] ) of figs . 7 might result in the set of four phase portraits , which are a good illustration of the evolution of the model phase structure vs the particle number chemical potential @xmath0 at arbitrary fixed @xmath181 . it is straightforwardly clear from the dual mapping of figs . 7 that at arbitrary fixed values of @xmath257 and @xmath256 , @xmath252 we have in the system the appearance of superconductivity at sufficiently high @xmath0 . this property is dually conjugated to the appearing of csb phase at sufficiently high values of @xmath1 , which is a consequence of the phase portraits fig . 7 ( see the above consideration ) . we have already considered the @xmath234-phase structure of the model in the case of dually symmetrical constraints on the coupling constants , i.e. when @xmath233 ( see figs . 3 , 4 ) . in the present section , special attention is paid to a more general case when @xmath268 . here we construct @xmath234-phase portraits of the model for qualitatively different representative relations between @xmath155 and @xmath156 . namely , we fix in each of the regions of fig . 1 some representative @xmath157 point and draw a corresponding @xmath234-phase diagram . ( note that due to a duality symmetry of the model , it is sufficient to reduce the number of the representative @xmath157 points to three . ) in fig . 8 the @xmath234-phase portrait corresponds to arbitrary fixed @xmath158 and @xmath269 ( it is clear that at @xmath235 the point @xmath270 lies in the symmetrical region i of fig . 1 ) . on the phase boundaries of this figure there are phase transitions of the second order . in fig . 9 the @xmath234-phase portrait corresponds to arbitrary fixed @xmath163 and @xmath271 ( in this case at @xmath235 the point @xmath272 lies in the csb region ii of fig . 1 ) . finally , in fig . 10 the @xmath234-phase portrait corresponds to arbitrary fixed @xmath163 and @xmath273 ( in this case at @xmath235 the point @xmath274 lies in the sc region iii of fig . 1 ) . to obtain the @xmath234-phase portraits of the model when the point @xmath157 is fixed in the two remaining areas of fig . 1 , it is enough to perform the dual mapping of the figs 9 and 10 . in figs . 9 , 10 all phase boundaries are the curves of the first order phase transitions , so the point t on these figures is a triple point , where three phases coexist . the main conclusion from figs 8 - 10 is that the growth of @xmath1 ( the growth of @xmath0 ) induces the chiral symmetry breaking ( induces the superconductivity ) . + -phase portrait of the model at arbitrary fixed @xmath163 and @xmath273 . the notations i , ii and iii are the same as in fig . the letter t denotes a triple point , @xmath275 , @xmath276.,title="fig:",scaledwidth=45.0% ] + in this paper , the duality correspondence phenomenon between csb and sc , which was found earlier in ( 1 + 1)-dimensional 4f theories @xcite , is demonstrated to take place also in the framework of the ( 2 + 1)-dimensional 4f model ( 1 ) . ( an alternative 4f model with other dually conjugated chiral and superconducting channels of interaction is considered in appendix [ apd ] ) . it contains both fermion - antifermion ( with @xmath8 a bare coupling constant ) and difermion ( with @xmath9 a bare coupling constant ) interaction channels and two kinds of chemical potentials , @xmath16 . the duality property of the model means that its thermodynamic potential @xmath277 ( [ 11 ] ) is invariant under the duality transformation ( [ 16 ] ) . as a result , we have established the duality correspondence between csb and sc phases of the model ( 1 ) ( as well as of the alternative model of appendix [ apd ] ) . let us suppose , e.g. , that for some fixed set @xmath230 of external parameters ( the coupling constants @xmath155 and @xmath156 are connected with bare couplings @xmath278 by the relation ( [ 13 ] ) ) the chiral symmetry breaking phase is realized in the model . then for rearranged values of external parameters , @xmath231 , @xmath6 , ( or for dually conjugated set @xmath279 ) we have the so - called dually conjugated superconducting phase ( and vice versa ) . moreover , it must be emphasized that the chiral condensate of the csb phase , realized for the set @xmath230 , is equal to the superconducting condensate of the dually conjugated sc phase , corresponding to the set of external parameters with @xmath231 , @xmath6 ( and vice versa ) . in this way , it is sufficient to have the information about the ground state of the initial phase , which is realized for the set @xmath230 , in order to determine the properties of the ground state of the dually conjugated phase , corresponding to the rearranged external parameter set @xmath279 . in order to study the role and influence of the duality property on the phase structure of the model , for comparison and illustrations , we have demonstrated a variety of phase portraits in the ( @xmath16)- and @xmath157 planes . in particular , if the constraint , under which we obtain a phase portrait , is dually invariant , i.e. at @xmath232 or at @xmath233 , then we have a selfdual phase diagram ( see , e.g. , figs 14 ) . the selfdual phase portrait is mapped into itself under the duality transformation ( it is introduced in sec . [ ivb ] ) , and its most characteristic feature is the mirror symmetrical arrangement of the csb and sc phases with respect to the line @xmath232 ( or @xmath233 ) of the phase diagram . we have also presented a series of nonselfdual phase portraits , which do not transform into themselves under the duality mapping ( see figs . the results obtained lead to the conclusion that the growth of the chiral chemical potential @xmath1 promotes the chiral symmetry breaking , were studied , correspondingly , in the recent papers @xcite and @xcite . it was shown there that the chiral chemical potential @xmath1 plays a role of the catalyst of dynamical chiral symmetry breaking . ] whereas particle number chemical potential @xmath0 induces superconductivity in the system ( see also in @xcite ) . we hope that our investigations may shed some new light on physical effects in planar systems like high - temperature superconductors or graphene . the two - dimensional irreducible representation of the ( 2 + 1)-dimensional lorentz group so(2,1 ) is realized by the following @xmath281 @xmath282-matrices : @xmath283 acting on two - component dirac spinors . they have the properties : @xmath284=-2i\varepsilon^{\mu\nu\alpha } \tilde\gamma_{\alpha};~ ~\tilde\gamma^{\mu}\tilde\gamma^{\nu}=-i\varepsilon^{\mu\nu\alpha } \tilde\gamma_{\alpha}+g^{\mu\nu } , \label{a2}\end{aligned}\ ] ] where @xmath285 . there is also the relation : @xmath286 note that the definition of chiral symmetry is slightly unusual in ( 2 + 1)-dimensions ( spin is here a pseudoscalar rather than a ( axial ) vector ) . the formal reason is simply that there exists no other @xmath281 matrix anticommuting with the dirac matrices @xmath287 which would allow the introduction of a @xmath28-matrix in the irreducible representation . the important concept of chiral symmetries and their breakdown by mass terms can nevertheless be realized also in the framework of ( 2 + 1)-dimensional quantum field theories by considering a four - component reducible representation for dirac fields . in this case the dirac spinors @xmath288 have the following form : @xmath289 with @xmath290 being two - component spinors . in the reducible four - dimensional spinor representation one deals with 4@xmath2914 @xmath280-matrices : @xmath292 , where @xmath293 are given in ( [ a1 ] ) . one can easily show , that ( @xmath294 ) : @xmath295 = diag(-i\varepsilon^{\mu\nu\alpha}\tilde\gamma_\alpha , -i\varepsilon^{\mu\nu\alpha}\tilde\gamma_\alpha ) . \label{a5}\end{aligned}\ ] ] in addition to the dirac matrices @xmath296 there exist two other matrices , @xmath297 and @xmath28 , which anticommute with all @xmath296 and with themselves @xmath298 with @xmath299 being the unit @xmath281 matrix . it is obvious that @xmath300 and @xmath301 . each of the matrices @xmath297 and @xmath28 ( see appendix [ apa ] ) can be selected as a generator for the corresponding @xmath40 and @xmath35 chiral group of spinor field transformations . for example , the lagrangian ( 1 ) is invariant with respect to @xmath35 such that @xmath302 . alternatively , it is possible to construct a 4f model with fermion - antifermion and superconducting channels , symmetric under @xmath40 continuous chiral transformations , @xmath303 ( @xmath34 ) . its lagrangian , e.g. , reads @xmath304\psi_k&+ & \frac { g_1}n\left ( \widetilde { 4f}\right ) _ { ch}+\frac { g_2}n\left ( \widetilde { 4f}\right ) _ { sc},\label{a7}\end{aligned}\ ] ] where the four - fermion structures @xmath305 and @xmath306 are used , @xmath307 here @xmath308 and @xmath0 is the usual particle number chemical potential ( as in ( 1 ) ) . since this lagrangian is invariant under @xmath40 , there exist a corresponding conserved density of chiral charge @xmath309 as well as its thermodynamically conjugate quantity , the chiral ( or axial ) chemical potential @xmath310 . note that in the framework of the model ( [ a7 ] ) there is also the duality between chiral and superconducting channels of interaction , i.e. between @xmath311 and @xmath312 four - fermion structures ( [ a71 ] ) . indeed , performing in ( [ a7])-([a71 ] ) the modified pauli - gursey transformation of spinor fields , @xmath313 we see that @xmath314 the relations ( [ d2 ] ) are the basis for the duality invariance of the thermodynamic potential of the model ( [ a7 ] ) . then , the thermodynamic potential @xmath322 , which describes the ground state of the model ( [ a7])-([a8 ] ) , is analogously defined by the following relation ( for convenience , we omit here and below the tilde in fields ) @xmath323[d\psi_l]\exp\big ( i\int d^3 x\big [ \bar\psi_k d_{\gamma^3}\psi_k\right.\nonumber\\ & & \left.- \frac{\delta}{2}(\psi_k^t\widetilde c\psi_k ) -\frac{\delta}{2}(\bar\psi_k\widetilde c\bar\psi_k^t)\big ] \big ) \right ) , \label{90}\end{aligned}\ ] ] where @xmath324 . in writing down this expression , we used the same arguments that were used in the derivation of the thermodynamic potential in the framework of the model ( 1 ) ( see the derivation of ( [ 9 ] ) in section ii ) . path integration in the expression ( [ 90 ] ) is evaluated in appendix [ apb ] so we have for the tdp ( [ 90 ] ) the following expression @xmath325 , \label{91}\end{aligned}\ ] ] where @xmath326 are presented in ( [ b14 ] ) . comparing this thermodynamic potential with the tdp @xmath327 ( [ 11 ] ) of the model ( 1 ) , we see that @xmath328 as a result , it is clear that the tdp @xmath329 of the 4f model ( [ a7 ] ) is invariant under the following dual transformation @xmath330 furthermore , to find phase portraits of the model ( [ a7 ] ) , it is sufficient to perform in all figs 1, .. ,10 the replacement @xmath331 . let us calculate the following path integral over anticommuting four - component dirac spinor fields @xmath332 , @xmath333 : @xmath334[d\psi ] \exp\big ( i\int d^3 x\big [ \bar\psi d_\gamma\psi-\frac{\delta}{2}(\psi^tc\gamma\psi ) -\frac{\delta}{2}(\bar\psi c\gamma\bar\psi^t)\big ] \big ) , \label{b1}\end{aligned}\ ] ] where @xmath335 is the charge conjugation matrix , @xmath336 . since in the 4@xmath2914 spinor space the product matrix @xmath337 should be an antisymmetric one , it is clear that the @xmath338 matrix in ( [ b1 ] ) might be , e.g. , the unit matrix @xmath339 in the spinor space or @xmath340 ( for notations see appendix [ apa ] ) , etc . note in addition that at @xmath341 and @xmath342 ( see in ( [ 9 ] ) ) the integral @xmath343 is equal to the argument of the @xmath344-function in the formula ( [ 9 ] ) in the particular case @xmath98 , whereas at @xmath340 and @xmath345 ( see ( [ 90 ] ) ) the integral @xmath343 is equal to the argument of the @xmath344-function in the formula ( [ 90 ] ) also in the particular case @xmath98 . recall , there are general gaussian path integrals @xcite : @xmath346\exp\big ( i\int d^3 x\big [ -\frac{1}{2}\psi^t a \psi + \eta^t\psi \big ] \big ) & = & \left(\det(a)\right ) ^{1/2}\exp\big ( -\frac{i}{2}\int d^3 x\big [ \eta^t a^{-1}\eta\big ] \big ) , \label{b2}\\ \int[d\bar\psi]\exp\big ( i\int d^3 x\big [ -\frac{1}{2}\bar\psi a \bar\psi^t + \bar\eta\bar\psi^t\big ] \big ) & = & \left ( \det(a)\right ) ^{1/2}\exp\big ( -\frac{i}{2}\int d^3 x\big [ \bar\eta a^{-1}\bar\eta^t\big ] \big ) , \label{b3}\end{aligned}\ ] ] where @xmath347 is an antisymmetric operator in coordinate and spinor spaces , and @xmath348 , @xmath349 are anticommuting spinor sources which also anticommute with @xmath288 and @xmath350 . first , let us integrate in ( [ b1 ] ) over @xmath288-fields with the help of the relation ( [ b2 ] ) supposing there that @xmath351 , @xmath352 , i.e. @xmath353 . then @xmath354\exp\big ( -\frac{i}{2}\int d^3 x \bar\psi \big [ \delta c\gamma+d_\gamma(\delta c\gamma)^{-1}d_\gamma^t\big ] \bar\psi^t\big ) . \label{b4}\end{aligned}\ ] ] second , the integration over @xmath350-fields in ( [ b4 ] ) can be easily performed with the help of the formula ( [ b3 ] ) , where one should put @xmath355 and @xmath356 . as a result , we have @xmath357\right ) ^{1/2}=\left ( \det [ \delta^2(c\gamma)^2+d_\gamma ( c\gamma)^{-1 } d_\gamma^t c\gamma]\right ) ^{1/2}. \label{b5}\end{aligned}\ ] ] we evaluate the expression ( [ b5 ] ) in two cases , ( i ) when @xmath341 and @xmath342 , where @xmath358 is presented in ( [ 9 ] ) , and ( ii ) when @xmath340 and @xmath345 , where @xmath359 is given in ( [ 90 ] ) . * ( i ) the case @xmath360 and @xmath361. taking into account the relations @xmath362 and @xmath363 ( @xmath59 ) , we obtain from ( [ b5 ] ) @xmath364\right ) ^{1/2}\equiv\left ( \det b\right ) ^{1/2 } , \label{b7}\end{aligned}\ ] ] where @xmath365 . using the general relation @xmath366 , we get from ( [ b7 ] ) : @xmath367 ( a more detailed consideration of operator traces is presented in appendix a of the paper @xcite . ) in this formula the symbol tr means the trace of an operator both in the coordinate and internal spaces . moreover , @xmath368 ( @xmath120 ) in ( [ b8 ] ) are four eigenvalues of the 4@xmath2914 fourier transformation matrix @xmath369 of the operator @xmath370 from ( [ b7 ] ) . namely , @xmath371 where @xmath372 . it is clear from these relations that @xmath101 are even functions vs @xmath0 and/or @xmath1 . so , it is enough to take into account only nonnegative values of the chemical potentials @xmath0 and @xmath1 . ( ii ) the case @xmath373 and @xmath374. in this case @xmath375 . then it follows from ( [ b5 ] ) that @xmath376\right ) ^{1/2}\equiv\left ( \det { \cal b}\right ) ^{1/2 } , \label{b10}\end{aligned}\ ] ] where @xmath377 . then , similar to ( [ b8 ] ) , it is easy to find that in the case under consideration @xmath378 where @xmath379 ( @xmath120 ) are four eigenvalues of the 4@xmath2914 fourier transformation matrix @xmath380 of the operator @xmath381 from ( [ b10 ] ) . it turns out that the eigenvalues @xmath379 are connected with the eigenvalues @xmath368 ( [ b12 ] ) by the simple relations @xmath382 to calculate the improper convergent integral in ( [ 31 ] ) we first use there a polar coordinate system , i.e. @xmath383 , and then restrict the @xmath384-integration region by @xmath137 ( suppose in addition that @xmath385 ) . as a result , we come to the regularized expression @xmath386 of the tdp ( [ 31 ] ) , @xmath387 , where @xmath388,\label{c9}\\ i_\pm&=&-\int_0^\lambda\frac{pdp}{2\pi}\left ( \mu-\sqrt{m^2+(p\pm\mu_5)^2}\right ) \theta\left ( \mu-\sqrt{m^2+(p\pm\mu_5)^2}\right ) . \label{c10}\end{aligned}\ ] ] due to the presence of the @xmath389-functions in ( [ c10 ] ) and sufficiently high values of the cutoff parameter @xmath390 , the quantities @xmath391 indeed do not depend on @xmath137 . moreover , it is evident that @xmath392 is a nonzero quantity only in the case @xmath393 . hence , substituting @xmath394 for the @xmath392-integration in ( [ c10 ] ) , we have @xmath395 ( the upper limit in the integral ( [ c12 ] ) corresponds to a value of the momentum @xmath384 , where the @xmath396-function of the integrand for @xmath392 from ( [ c10 ] ) is equal to zero , i.e. to a value of @xmath384 determined by the condition @xmath397 . ) it is convenient to present the quantity @xmath398 from ( [ c10 ] ) as a sum of two integrals , @xmath399 , where @xmath400 it is clear that @xmath401 is a nonzero quantity only at @xmath402 . then , performing in the integral ( [ c14 ] ) the substitution @xmath403 , we obtain @xmath404 obviously , we have @xmath405 at @xmath406 . to evaluate @xmath407 at other values of @xmath0 , we should , first , substitute @xmath408 in the integral ( [ c13 ] ) and then consider two different regions of the parameter @xmath0 , ( i ) @xmath409 , ( ii ) @xmath410 . as a result , we have @xmath411 summing the expressions ( [ c12 ] ) , ( [ c15 ] ) and ( [ c16 ] ) , we have @xmath412 to calculate the @xmath413-term of the regularized tdp @xmath386 , it is useful to present @xmath413 as a sum of three integrals , each one corresponds to some square root expression of the integrand in ( [ c9 ] ) . then , substituting @xmath394 and @xmath403 in two of these integrals , it is possible to present @xmath413 in the form @xmath414 since @xmath415 , we an use @xmath416 in the last two terms of ( [ c4 ] ) . then , after integrations one can see that the sum of these two last integrals in ( [ c4 ] ) is equal to zero in the limit @xmath146 . hence , @xmath417.\label{c11}\end{aligned}\ ] ] finally , performing in ( [ c17 ] ) trivial table integrations and using the relation @xmath418 we obtain from ( [ c0 ] ) , ( [ c17 ] ) and ( [ c11 ] ) the expression ( [ c18 ] ) ( the function @xmath419 is given in the text below ( [ 32 ] ) ) . d. ebert and m.k . volkov , z. phys . c 16 * , 205 ( 1983 ) ; d. ebert , h. reinhardt and m.k . volkov , prog . phys . * 33 * , 1 ( 1994 ) . m. buballa , phys . * 407 * , 205 ( 2005 ) . y. nambu and g. jona - lasinio , phys . rev.122 , 345 ( 1961 ) ; phys . rev . * 124 * , 246 ( 1961 ) . a. chodos and h. minakata , phys . a * 191 * , 39 ( 1994 ) ; lect . notes phys . * 508 * , 231 ( 1998 ) [ hep - th/9709197 ] ; h. caldas , j.l . kneur , m.b . pinto and r.o . ramos , phys . b * 77 * , 205109 ( 2008 ) . v. schn and m. thies , at the frontier of particle physics : handbook of qcd : boris ioffe festschrift , vol . * 3 * , 1945 , world scientific ( 2001 ) [ hep - th/0008175 ] ; m. thies , j. phys . a * 39 * , 12707 ( 2006 ) . semenoff , i.a . shovkovy and l.c.r . wijewardhana , mod . a * 13 * , 1143 ( 1998 ) . zhukovsky , k.g . klimenko , v.v . khudyakov and d. ebert , jetp lett . * 73 * , 121 ( 2001 ) ; v.c . zhukovsky and k.g . klimenko , theor . math . phys . * 134 * , 254 ( 2003 ) ; e.j . ferrer , v.p . gusynin and v. de la incera , mod . b * 16 * , 107 ( 2002 ) ; eur . j. b * 33 * , 397 ( 2003 ) . marino and l.h.c.m . nunes , nucl . b * 741 * , 404 ( 2006 ) ; l.h.c.m . nunes , r.l.s . farias and e.c . marino , phys . a * 376 * , 779 ( 2012 ) . klimenko , r.n . zhokhov and v.c . zhukovsky , phys . d * 86 * , 105010 ( 2012 ) . klimenko , r.n . zhokhov and v.c . zhukovsky , mod . lett . a * 28 * , 1350096 ( 2013 ) . d. mesterhazy , j. berges , and l. von smekal , phys . b * 86 * , 245431 ( 2012 ) . d. ebert , k.g . klimenko , p.b . kolmakov and v.c . zhukovsky , arxiv:1509.08093 [ cond-mat.mes-hall ] . klimenko , z. phys . c * 37 * , 457 ( 1988 ) ; b. rosenstein , b.j . warr , and s.h . park , phys . d * 39 * , 3088 ( 1989 ) ; phys . lett . * 62 * , 1433 ( 1989 ) . klimenko , z. phys . c * 54 * , 323 ( 1992 ) ; theor . * 89 * , 1161 ( 1992 ) ; theor . math . * 90 * , 1 ( 1992 ) ; v.p . gusynin , v.a . miransky and i.a . shovkovy , phys . lett . * 73 * , 3499 ( 1994 ) ; a.s . vshivtsev , b.v . magnitsky and k.g . klimenko , phys . atom . nucl . * 57 * , 2171 ( 1994 ) ; theor . phys . * 106 * , 319 ( 1996 ) ; v.p . gusynin , d.k . hong and i.a . shovkovy , phys . d * 57 * , 5230 ( 1998 ) . a. chodos , h. minakata , f. cooper , a. singh , and w. mao , phys . d * 61 * , 045011 ( 2000 ) . d. ebert , k.g . klimenko and h. toki , phys . d * 64 * , 014038 ( 2001 ) ; h. kohyama , phys . d * 77 * , 045016 ( 2008 ) ; phys . d * 78 * , 014021 ( 2008 ) . d. ebert , k.g . klimenko , a.v . tyukov and v.c . zhukovsky , phys . d * 78 * , 045008 ( 2008 ) ; d. ebert , k.g . klimenko , phys . * d80 * , 125013 ( 2009 ) ; d. ebert , n.v . gubina , k.g . klimenko , s.g . kurbanov and v.c . zhukovsky , phys . d * 84 * , 025004 ( 2011 ) . gubina , k.g . klimenko , s.g . kurbanov and v.c . zhukovsky , phys . d * 86 * , 085011 ( 2012 ) ; moscow univ . phys . bull . * 67 * , 131 ( 2012 ) . kneur , m.b . pinto , r.o . ramos and e. staudt , d * 76 * , 045020 ( 2007 ) ; phys . lett . b * 657 * , 136 ( 2007 ) . klimenko , z. phys . c * 50 * , 477 ( 1991 ) ; mod . lett . a * 9 * , 1767 ( 1994 ) . i. ojima and r. fukuda , prog . phys . * 57 * , 1720 ( 1977 ) . a.n . vasiliev and g.y . panasyuk , theor . * 103 * , 570 ( 1995 ) [ teor . mat . * 103 * , 295 ( 1995 ) ] . m. thies , phys . d * 68 * , 047703 ( 2003 ) ; phys . d * 90 * , no . 10 , 105017 ( 2014 ) . d. ebert , t.g . khunjua , k.g . klimenko and v.c . zhukovsky , phys . d * 90 * , 045021 ( 2014 ) . n.d . mermin and h. wagner , phys . * 17 * , 1133 ( 1966 ) ; s. coleman , commun . 31 * , 259 ( 1973 ) . andrianov , d. espriu and x. planells , eur . j. c * 73 * , 2294 ( 2013 ) ; eur . j. c * 74 * , 2776 ( 2014 ) ; r. gatto and m. ruggieri , phys . d * 85 * , 054013 ( 2012 ) ; m. ruggieri , arxiv:1110.4907 ; l. yu , h. liu and m. huang , phys . rev . d * 90 * , 074009 ( 2014 ) ; l. yu , h. liu and m. huang , arxiv:1511.03073 [ hep - ph ] ; g. cao and p. zhuang , phys . d * 92 * , 105030 ( 2015 ) . braguta and a.y . kotov , arxiv:1601.04957 [ hep - th ] . zhukovsky , k.g . klimenko and v.v . khudyakov , theor . phys . * 124 * , 1132 ( 2000 ) [ teor . mat . * 124 * , 323 ( 2000 ) ] . b. vanderheyden and a.d . jackson , phys . d * 61 * , 076004 ( 2000 ) ; v.c . zhukovsky , v.v . khudyakov , k.g . klimenko and d. ebert , jetp lett . * 74 * , 523 ( 2001 ) [ pisma zh . teor . fiz . * 74 * , 595 ( 2001 ) ] . g. cao , l. he and p. zhuang , phys . d * 90 * , 056005 ( 2014 ) . a.n . vasiliev,``functional methods in quantum field theory and statistical physics '' , leningrad univ . press , leningrad , 1976 .
the study of nuclei far from stability is an increasingly important part of nuclear physics @xcite . as radioactive beams allow more experiments on these nuclei , theoretical modeling is changing in significant ways . new ideas and progress in computer technology have allowed nuclear theorists to understand bits and pieces of nuclear structure quantitatively @xcite . short - lived exotic nuclei offer unique tests of those aspects of our developing many - body theories that depend on neutron excess @xcite . the major challenge is to predict or describe in detail exotic new properties of nuclei far from the stability valley , and to understand the origins of these properties . for medium - mass and heavy nuclei , an important goal is obtaining a universal energy - density functional , which will be able to describe static and dynamic properties of finite nuclei and extended nucleonic matter with arbitrary neutron - to - proton ratio . self - consistent methods based on density - functional theory are already sophisticated enough to allow precise analysis of ground - state properties ( e.g. binding energies ) in heavy nuclei @xcite . they can also help describe nuclear decays and excited states . their predictions for collective excitations as we approach the neutron drip line are especially interesting . what happens to low- and high - frequency multipole modes when the neutron excess is unusually large ? to address these questions we use the quasiparticle random - phase approximation ( qrpa ) , a powerful tool for understanding both low - lying vibrational states and giant resonances @xcite . the qrpa is a microscopic approach that is nevertheless simple enough to allow `` no - core '' calculations . the approximation , which should be good for collective vibrations as long as their amplitudes are small , is especially effective in conjunction with skyrme energy functionals . our work is a part of a broad program to test and improve these functionals , which thus far have been fitted mainly to ground - state observables , by applying them to collective excitations , particularly near the drip line . this paper lays out our approach and evaluates its accuracy . for these purposes we restrict ourselves to a single skyrme functional , skm@xmath0 . a forthcoming study will examine the performance of skyrme functionals more generally . the qrpa is a standard method for describing collective excitations in open - shell superconducting nuclei with stable mean - field solutions , either spherical or deformed . what is not standard , and at the same time is extremely important for weakly bound nuclei , is the treatment of the particle continuum . continuum extensions of the random phase approximation ( rpa ) or qrpa are usually carried out in coordinate space , facilitating treatment of decay channels and guaranteeing correct asymptotics . surprisingly , as we discuss below , the rich literature on the rpa and qrpa , which includes many coordinate - space calculations , contains few treatments of the continuum that exploit the entire skyrme functional in a fully self - consistent way . to avoid confusion , we state what we mean by a fully self - consistent rpa or qrpa calculation . first , the underlying mean - field calculation must be self - consistent in the usual sense . next , the residual interaction used in the rpa or qrpa must be derived from the same force or energy functional that determines the mean field . an important consequence of this condition , and of other more detailed technical conditions discussed below , is that spurious excitations arising from symmetry breaking by the mean field have zero or nearly zero energy , leaving the physical intrinsic excitations completely uncontaminated by spurious motion . finally , energy - weighted sum rules must be satisfied to high accuracy . we elaborate on these requirements below ; refs . @xcite discuss ways in which rpa calculations commonly violate them . the literature applying rpa or qrpa to nuclear structure is huge , and a complete review is beyond the scope of our paper . we do , however , present an overview of the studies that are related in one way or another to nuclear density functionals , self consistency , pairing , and the key issue of the particle continuum . the standard version of qrpa , the so - called matrix formulation , is carried out in the configuration space @xcite of single - quasiparticle states . a number of papers treat collective states in spherical nuclei in the skyrme - rpa and qrpa matrix formulation ( see refs . @xcite and references cited therein ) , in which the positive - energy continuum is discretized , e.g. by solving the hartree - fock - bogoliubov ( hfb ) and qrpa equations in a harmonic - oscillator single - particle basis . within this group , the first fully self - consistent calculations that properly account for continuum effects are those of refs . @xcite , in which the localized canonical basis of coordinate - space hfb is used to calculate beta - decay rates of neutron - rich r - process nuclei and gamow - teller strength distributions . recently , fully self - consistent hfb+qrpa calculations have also been carried out with the finite - range gogny force @xcite . unlike many previous gogny+hfb studies that employed a harmonic oscillator basis , ref . @xcite solves the hfb equations in the eigenbasis of a woods - saxon potential , the particle continuum of which is discretized by enclosing the system in a box . coordinate - space green s functions as a method of implementing the rpa through linear response were first used in ref . @xcite and subsequently applied to the description of low- and high - energy nuclear modes ( see , e.g. , refs . many of those calculations are not realistic enough , however , because they ignore the spin - orbit and coulomb residual interactions in the rpa @xcite . coordinate - space greens - function qrpa was studied in ref . @xcite , in the bcs approximation , with a phenomenological woods - saxon average potential . coordinate - space hfb+qrpa for spherical nuclei was formulated in refs . @xcite and applied to excitations of neutron - rich nuclei . as in @xcite , the hartree - fock ( hf ) field in refs . @xcite was approximated by a woods - saxon potential . while the calculations of refs . @xcite are based on skyrme - hfb fields , they violate full self consistency by replacing the residual velocity - dependent terms of the skyrme force by the landau - migdal force in the qrpa , and neglecting spin - spin , spin - orbit , and coulomb residual interactions entirely . within this approach , extensive skyrme - hf+bcs qrpa calculations of e1-strength in neutron - rich nuclei were carried out in refs . @xcite . an alternative coordinate - representation approach , also based on green s functions , was formulated in refs . @xcite within migdal s finite - fermi - systems theory . most of practical applications of this method , however , involve approximations that break self consistency in one way or another , including the use of highly truncated pairing spaces , different interactions in hfb and qrpa , and the so - called diagonal pairing approximation @xcite . properties of excited states and strength functions have also been investigated within the relativistic rpa @xcite or qrpa @xcite . the qrpa work employs the matrix formulation and is fully self - consistent , since it uses the same lagrangian in the relativistic hartree - bogoliubov calculation of the ground state and in the qrpa matrix equations , which are solved in the canonical basis . at the present , no fully self - consistent continuum hfb+qrpa calculations exist in deformed nuclei . @xcite studied giant resonances in deformed nuclei within time - dependent hf theory , formulated in coordinate space with a complex absorbing boundary condition imposed . symmetry - unrestricted rpa calculations , with no pairing , were carried out in ref . @xcite in a `` mixed representation '' @xcite on a cartesian mesh in a box , while ref . @xcite contains examples of bcs+qrpa calculations in the single - particle basis of a deformed woods - saxon potential . the work described in this paper is fully self - consistent : among other things we use precisely the same interaction in the hfb and qrpa calculations so as to preserve the small - amplitude limit of time - dependent hfb . we formulate the qrpa in the canonical eigenbasis of the one - body particle - density matrix @xcite which is calculated in the coordinate representation in a large spherical box . as mentioned above , the canonical basis has been used previously to study @xmath1 decay and gamow - teller strength @xcite ; its use in charge - conserving modes near the drip line is more challenging , however , because of the existence of spurious states in the monopole and dipole channels . these zero - energy modes can mix with physical states unless the qrpa equations are solved with high accuracy . a less precise implementation of our approach was used to calculate neutrino - nucleus cross sections in @xmath2pb in ref . @xcite . this paper is organized as follows . section [ sec : method ] below presents our approach . in sec . [ sec : accuracy ] we check the qrpa solutions carefully , focusing on spurious modes . section [ sec : conclusion ] contains the main conclusions of our work . mathematical details are in two appendices , the first of which is on the qrpa equations and the second on calculating the derivatives of the skyrme functionals that enter the formalism . our first step in the self - consistent treatment of excitations is to solve the spherical hfb equations in coordinate space ( without mixing neutron and proton quasiparticle wave functions @xcite ) , with the method developed in ref . @xcite ( see also refs . we can use arbitrary skyrme functionals in the particle - hole and pairing ( particle - particle ) channels . we modify the code used in refs . @xcite so that it solves the hfb equations with higher accuracy , which we need because the qrpa uses all the single - quasiparticle states produced by the hfb equations , even those that are essentially unoccupied . our modifications are : ( i ) the use of quadruple precision ( though in solving the qrpa equations we use double precision ) ; ( ii ) a smaller discretization length ( 0.05 fm ) ; and ( iii ) a high quasiparticle - energy cutoff ( 200 mev ) and a maximum angular momentum @xmath3=15/2 ( @xmath4 ) or 21/2 ( @xmath5 ) . in a 20 fm box , this cutoff corresponds to 200300 quasiparticle states for each kind of nucleon . we include all these quasiparticle states in the hfb calculation because a very large energy cutoff is essential for the accuracy of self - consistent qrpa calculations @xcite . hence , the effective pairing window in our hfb calculations is also very large , with the pairing functional fitted to experimental pairing gaps extracted as in ref . @xcite from the measured odd - even mass differences in several sn , ni , and ca isotopes . next , we construct the canonical basis , the eigenstates of single - particle density matrix @xmath6 . to avoid poor accuracy ( see ref . @xcite ) in the wave functions of the nearly empty canonical particle states , we do not diagonalize @xmath6 directly in coordinate space . instead we construct an intermediate basis by orthonormalizing a set of functions @xmath7 , where @xmath8 and @xmath9 are the upper and lower components of the quasiparticle wave function with energy @xmath10 @xcite . we use the density matrix in coordinate space to calculate the matrix in this basis , which we then diagonalize to obtain the canonical states . the reason for using the sum of @xmath8 and @xmath11 is that solutions of the hfb equations expressed in the canonical basis ( eqs . ( 4.14 ) of ref . @xcite ) are , in the new basis , _ guaranteed _ to be numerically consistent with those of the original hfb problem . this is because the configuration space is the same in both cases , independent of the pairing cutoff ( see ref . @xcite for a discussion relevant to this point ) . without pairing , when either @xmath12 or @xmath9 is equal to zero , our method is equivalent to taking a certain number of hf states , including many unoccupied states . in the canonical basis , the hfb+qrpa equations have a form almost identical to that of the bcs+qrpa approximation , the only difference being the presence of off - diagonal terms in the single - quasiparticle energies . the qrpa+hfb formalism employs more pairing matrix elements than the qrpa+bcs , however . as noted already , full self consistency requires the use of the same interaction in the qrpa as in the hfb approximation . more specifically , this means that the matrix elements that enter the qrpa equation are related to second derivatives of a mean - field energy functional . we describe the densities and the form of the functional carefully in the appendices . but we must meet other conditions as well for qrpa calculation to be self - consistent . essentially all the single - particle or quasiparticle states produced by the hfb calculation must be used in the space of two - quasiparticle qrpa excitations . this requirement is rather stringent , so we truncate the two - quasiparticle space at several levels and check for convergence of the qrpa solution . first we omit canonical - basis wave functions that have occupation probabilities @xmath13 less than some small @xmath14 , ( or hf energies greater than some @xmath15 if there is no pairing ) . then we exclude from the qrpa pairs of canonical states for which the occupation probabilities are both larger than @xmath16 . this second cut is based on the assumption that two - particle transfer modes are not strongly coupled to particle - hole excitations . in addition , if the factors containing @xmath17 and @xmath18 in the qrpa equation see eqs . ( [ a_spherical ] ) and ( [ b_spherical ] ) are very small , in practice smaller than 0.01 , then we set the corresponding matrix elements equal to zero . this does not affect the size of the qrpa space , but significantly speeds up the calculations . for good performance we diagonalize qrpa - hamiltonian matrices of order @xmath19 in neutron - rich sn isotopes . having solved the qrpa equations , we can then calculate the strength function @xmath20 for the multipole operator @xmath21 . the smoothing width @xmath22 is supposed to be large enough to remove spurious oscillations in @xmath23 associated with a finite box radius @xmath24 @xcite . a reasonable form , based on a single - particle estimate , for the smoothing width ( app . b of ref . @xcite ) , is @xmath25 where @xmath26 is the neutron fermi level and @xmath27 is the nucleon mass . in deriving eq . ( [ smwidth ] ) we assumed that the single - proton continuum is effectively shifted up several mev by the coulomb barrier . in other words , we associate the threshold energy with the neutron fermi level . in all the tests below , we use the skyrme functional skm@xmath28 @xcite and a volume pairing functional @xcite ( @xmath29 a constant in eq . ( [ pairing_energy_functional ] ) ) . the pairing parameter in eq . ( [ volpair ] ) is @xmath30 mev @xmath31 . usually we work in a box of radius 20 fm , though we vary this radius below to see its effects . in several tests we examine the weakly bound nucleus @xmath32sn , which is very close to the two - neutron drip line . in this system , the protons are unpaired and the neutrons paired ( with @xmath33=1.016mev ) in the hfb ground state . neutron quasiparticle energies @xmath34 for @xmath35 states in @xmath32sn , calculated by diagonalizing the hfb hamiltonian in the canonical basis , versus the quasiparticle energies @xmath36 obtained by directly solving the hfb equations in coordinate space . standard ( solid line , diamonds ) and improved ( dotted line , dots ) methods are used to obtain the canonical states . see text for details . , width=453 ] benchmark tests of the hfb part of our calculations are reported in ref . since the accuracy of the canonical wave functions , in which the qrpa calculations are carried out , strongly affects the quality of results ( in particular qrpa self consistency ) , we take special care to compute them precisely . as discussed in sec . [ sec : method ] , we obtain canonical states by diagonalizing the single - particle density matrix @xmath6 represented in the orthonormalized set of functions @xmath37 . the accuracy of this method is illustrated in fig . [ fig : ehfb_eigh11 ] , which plots the quasiparticle energies @xmath34 , obtained by diagonalizing the hfb hamiltonian in the canonical basis ( eq . ( 4.20 ) of ref . @xcite ) , versus the quasiparticle energies @xmath36 obtained by solving the hfb differential equations directly in coordinate space ( eq . ( 4.10 ) of ref . two sets of canonical states are used : ( i ) those obtained through the procedure outlined above ( dotted line ) and ( ii ) those obtained in the standard way by diagonalizing the density matrix @xmath38 in discretized coordinate space ( eq . ( 3.24a ) of ref . @xcite ; solid line ) . if the canonical basis is precisely determined , @xmath34=@xmath36 and the two sets of @xmath34 coincide . within the standard approach , however , the high canonical energies deviate visibly from their hfb counterparts , i.e. , the accuracy of the underlying canonical wave functions is poor . on the other hand , the quasiparticle energies and canonical wave functions calculated within the modified approach introduced above are as accurate as the original solutions to the hfb equations , even for high - lying nearly - empty states . ( see also sec . vi.d of ref . @xcite for a discussion relevant to this point . ) having examined the canonical basis , we turn to the accuracy of the qrpa part of the calculation . to test it , we first consider solutions related to symmetries . if a hamiltonian is invariant under a symmetry operator @xmath39 and the hfb state @xmath40 spontaneously breaks the symmetry , then @xmath41 , with @xmath42 an arbitrary c - number , is degenerate with the state @xmath40 . the qrpa equations have a spurious solution at zero energy associated with the symmetry breaking @xcite , while all other solutions are free of the spurious motion . this property is important for strength functions , and gives us a way of testing the calculations . since our qrpa equations , which assume spherical symmetry , are based on mean fields that include pairing and are localized in space , there appear spurious states associated with particle - number nonconservation ( proton and/or neutron ; 0@xmath43 channel ) and center - of - mass motion ( 1@xmath44 channel ) . these two cases are discussed below in sec . [ zeroplus ] and sec . [ oneminus ] . in addition to the spurious state associated with nonconservation of particle - number by the hfb , the @xmath45 channel contains the important `` breathing mode '' . in table [ tab : strength_number ] we display results from a run with @xmath46=@xmath47 for neutrons and @xmath48 mev for protons , resulting in the inclusion of 310 proton quasiparticle states and the same number of neutron states , with angular momentum up to @xmath49 . the table shows the qrpa energies and transition matrix elements of the particle - number operator . the spurious state is below 200 kev , well separated from the other states , all of which have negligible `` number - strength '' . the nonzero number strength in the spurious state , like the nonzero energy of that state , is a measure of numerical error . if the space of two - quasiparticle states is smaller , with @xmath50 mev and @xmath51 , the energy of the spurious state and the number strength barely change . isoscalar @xmath45 strength function in @xmath32sn for ( i ) the single - proton energy cutoff @xmath15=100 mev and the neutron - quasiparticle occupation cutoff @xmath52 ( thin solid line ) ; ( ii ) @xmath15=150 mev and @xmath53 ( dotted line ) ; and ( iii ) @xmath15=200 mev and @xmath54 ( thick solid line ) . results corresponding to ( ii ) and ( iii ) practically coincide.,width=453 ] figure [ fig:0+_sn174_ecrit ] shows the strength function @xmath23 for the isoscalar @xmath45 transition operator , cf . @xcite , @xmath55 we have plotted three curves with successively more quasiparticle levels ( from 246 proton levels and 203 neutron levels to 341 proton levels and 374 neutron levels ) , with cutoff parameters given in the figure caption . the major structures in the strength function are stable . the error remaining after to the gentlest truncation is extremely small . the dependence of the strength function on the box size and quasiparticle cutoff is shown in fig . [ fig:0+_sn174_box ] . the upper part of the figure ( panels a - c ) corresponds to a constant smoothing width of @xmath22=0.5 mev . this relatively small value is not sufficient to eliminate the finite - box effects but it allows us to assess the stability of the qrpa solutions as a function of @xmath24 . the large structure corresponding to the giant monopole resonance ( gmr ) is independent of box size no matter what the cutoff , but increasing the number of configurations magnifies the dependence on box size of local fluctuations in @xmath23 . the lower part of the figure ( panels d - f ) are smoothed more realistically , as in eq . ( [ smwidth ] ) . it is gratifying to see that the resulting strength functions are practically identical , i.e. , the remaining dependence on @xmath24 and the cutoff is very weak . isoscalar @xmath45 strength function in @xmath32sn for the box radii : @xmath24=20fm ( solid line ) and @xmath24=25fm ( dotted line ) . in ( a ) , ( b ) , and ( c ) the smoothing - width parameter @xmath22 is 0.5 mev for all energies , while in ( d ) , ( e ) , and ( f ) @xmath56 is given by eq . ( [ smwidth ] ) . we use the same three sets of cutoff conditions as in fig . [ fig:0+_sn174_ecrit ] , namely ( i ) in parts ( a ) and ( d ) , ( ii ) in parts ( b ) and ( e ) , and ( iii ) in parts ( c ) and ( f ) . , width=566 ] isoscalar @xmath57 strength function in @xmath58sn for the corrected dipole operator in eq . ( [ dipolecorr ] ) ( solid line ) and the uncorrected operator in eq . ( [ dipole ] ) ( dotted line ) . the cutoff @xmath15 is 140 mev and @xmath59 is @xmath60 . the self consistency of our calculations makes the solid and dotted curves coincide nearly exactly over the whole energy range.,width=491 ] isoscalar @xmath57 strength function in @xmath61sn ( left ) and @xmath32sn ( right ) for two box radii : @xmath24=20fm ( solid line ) and @xmath24=25fm ( dotted line ) . in ( a ) and ( b ) the smoothing - width parameter is constant ( @xmath22=0.5 mev ) , while in ( c ) and ( d ) @xmath56 is given by eq . ( [ smwidth ] ) . , width=566 ] isoscalar and isovector strength functions for ( a ) the 0@xmath43 channel of @xmath61sn , ( b ) the 0@xmath43 channel of @xmath32sn , ( c ) the 2@xmath43 channel of @xmath61sn , and ( d ) the 2@xmath43 channel of @xmath32sn . the cutoff @xmath62 is 150 mev and @xmath59 is @xmath47 . , width=529 ] the energy - weighted sum rule ( ewsr ) for the isoscalar @xmath45 mode @xcite is given by @xmath63 where the expectation value is evaluated in the hfb ground state . this sum rule provides a stringent test of self consistency in the qrpa . in @xmath32sn , the right - hand side of eq.([isoscalar_0+_sum_rule ] ) is 35215 @xmath64 mev @xmath65 and the left - hand side 34985@xmath66 @xmath64 mev @xmath65 for all of the calculations of fig . [ fig:0+_sn174_box ] ; the qrpa strength essentially exhausts the sum rule . ( the qrpa values of the ewsr in this paper are obtained by summing up to @xmath67 mev . ) the @xmath57 channel , home of the giant dipole resonance , the isoscalar squeezing resonance , and as yet incompletely understood low - energy peaks in neutron - rich nuclei ( sometimes associated with skin excitations ) , has a spurious isoscalar mode associated with center - of - mass motion that can seriously compromise the low - energy spectrum if not handled with extreme care . we test the ability of our qrpa to do so in @xmath68sn , @xmath61sn , @xmath32sn , and @xmath69sn . ( the nuclei @xmath68sn and @xmath69sn are the two - proton and two - neutron drip - line systems predicted by the hfb calculation with skm@xmath70 . neither nucleus has any static pairing , i.e. , @xmath33=@xmath71=0 . ) in the following calculations , we take @xmath72 mev for the protons and @xmath73 for the neutrons . as discussed above , smoothed strength functions are practically independent of small changes in the cutoff . they are also independent of the cutoff in quasiparticle angular momentum provided we include all states with @xmath74@xmath7515/2 . figure [ fig:1-_sn100_sn176 ] shows the predicted isoscalar dipole strength function for @xmath58sn . for the transition operator , we use @xmath76 and the corrected operator @xmath77 to remove as completely as possible residual pieces of the spurious state from the physical states @xcite . the fact that the strength functions produced by these two operators displayed in fig . [ fig:1-_sn100_sn176 ] coincide so closely shows the extreme accuracy of our qrpa solutions ; they are uncontaminated by spurious motion even without the operator correction . the spurious - state energies @xmath78 are 0.964 mev for @xmath68sn and 0.713 mev for @xmath61sn , and the energies of the first physical excited states are 7.958 mev for @xmath68sn and 7.729 mev for @xmath61sn . in @xmath32sn ( @xmath69sn ) , @xmath78 is 0.319 mev ( 0.349 mev ) and the first physical state is at 3.485 mev ( 2.710 mev ) , lower than in the more stable isotopes . pairing correlations do not affect accuracy ; the neutrons in @xmath61sn and @xmath32sn are paired , while those in @xmath68sn and @xmath69sn are not . we display the fine structure of the isoscalar @xmath57 strength functions in @xmath61sn and @xmath32sn in fig . [ fig:0-_sn174_box ] , which also illustrates the dependence of the results on @xmath24 . the dependence is consistent with that of fig . [ fig:0+_sn174_box ] for the isoscalar 0@xmath43 strength ; the low - amplitude fluctuations in @xmath23 that are unstable as a function of @xmath24 disappear , and the smoothed strength function depends only weakly on @xmath24 . in @xmath61sn , the two sharp peaks below 10 mev correspond to discrete states while the broad maxima centered around 15 mev and 27 mev are in the continuum , well above neutron - emission threshold . a similar three - peaked structure emerges in @xmath32sn , though most of the strength there is concentrated in the low - energy peak at @xmath79 mev . [ fig:1-_sn100_sn176 ] shows ( as we will discuss in our forthcoming paper @xcite ) that the appearance of the low - energy isoscalar dipole strength is a real and dramatic feature of neutron - rich dripline nuclei @xcite . the ewsr for the isoscalar @xmath57 mode @xcite is @xmath80 in @xmath32sn , the right - hand side is 403310 @xmath64 mev @xmath81 , while the left - hand side is 400200 @xmath64 mev @xmath81 . for @xmath69sn , the corresponding numbers are 406576 @xmath64 mev @xmath81 and 407100 @xmath64 mev @xmath81 . this level of agreement is very good . figure [ fig : is_iv_0+_2 + ] displays strength functions for the 0@xmath43 and 2@xmath43 channels in @xmath61sn and @xmath32sn . ( we discussed the isoscalar @xmath45 mode above to illustrate the accuracy of our solutions , but include it here as well for completeness . ) the calculations show the appearance of low - energy @xmath45 strength both isovector and isoscalar and low - energy isovector 2@xmath43 strength in @xmath32sn , though in none of these instances is the phenomenon quite as dramatic as in the isoscalar @xmath57 channel . the ewsr for the isoscalar 2@xmath43 transition operator , @xmath82 can be written as @xcite @xmath83 the sum rule is obeyed as well in the 2@xmath43 isoscalar channel as in the @xmath45 and @xmath57 channels , the only difference being that one needs to include quasiparticle states with @xmath84 for @xmath32sn . for @xmath61sn ( @xmath32sn ) from fig . [ fig : is_iv_0+_2 + ] , the ewsr is 37222 ( 34971 ) @xmath64mev@xmath65 while the qrpa value is 37030 ( 35010 ) @xmath64mev@xmath65 . while on the topic of the sum rule , we display in table [ tab : jmax_dependence_of_ewsr ] the @xmath85-dependence of the ewsr for several channels in @xmath86sn , with @xmath87 fm . by taking @xmath88 we appear to obtain essentially the entire strength in all three cases . in this work we have reported on the development and detailed testing of a fully self - consistent skyrme - qrpa framework that employs the canonical hfb basis . the method can be used to calculate strength distributions in any spin - isospin channel and in any spherical even - even nucleus . a good calculation requires a large single - quasiparticle space . our results show that our space is large enough in nuclei as heavy as the sn isotopes . we are currently investigating the predictions of a range of skyrme functionals across the ca , ni , and sn isotope chains . the initial results presented here point to increases in low - lying strength at the neutron drip line , particularly in the isoscalar - dipole channel . in a forthcoming paper @xcite we will report on the robustness of these effects , on the physics underlying them , on their variation with atomic mass and number , and on their implications for the future of skyrme functionals . we gratefully acknowledge useful discussions with shalom shlomo , nils paar , dario vretenar , and the japanese members of the us - japan cooperative project on mean - field approach to collective excitations in unstable medium - mass and heavy nuclei " . mb is grateful for the warm hospitality at the physics division at ornl and the theory division at gsi darmstadt , where most of his contribution to this work was made . this work was supported in part by the u.s . department of energy under contracts nos . de - fg02 - 97er41019 ( university of north carolina at chapel hill ) , de - fg02 - 96er40963 ( university of tennessee ) , de - ac05 - 00or22725 with ut - battelle , llc ( oak ridge national laboratory ) , de - fg05 - 87er40361 ( joint institute for heavy ion research ) , and w-31 - 109-eng-38 ( argonne national laboratory ) ; by the national science foundation contract no . 0124053 ( u.s .- japan cooperative science award ) ; by the polish committee for scientific research ( kbn ) ; and by the foundation for polish science ( fnp ) . we used the parallel computers of the center for computational sciences at oak ridge national laboratory and information technology services at the university of north carolina at chapel hill . the qrpa equations are the small - oscillations limit of the time - dependent hartree - fock - bogoliubov approximation , see , e.g. , @xcite . in the canonical basis the most general equations take the form @xmath89 @xmath90 @xmath91 @xmath92 } { \delta\rho_{k^\prime k}\delta\rho_{l^\prime l } } , \label{ph_matrix_element_uncoupled}\ ] ] @xmath93 } { \delta\kappa^\ast_{k^\prime k } \delta\kappa_{l^\prime l } } , \label{pp_matrix_element_uncoupled}\ ] ] @xmath94 } { \delta\kappa^\ast_{k^\prime k } \delta\rho_{l l^\prime } } = { \bar{v}^{\rm 1p3h\ \ast}}_{l l^\prime k^\prime k } , \label{3p1h_matrix_element_uncoupled}\ ] ] where @xmath95 and @xmath96 are single - particle indices for the canonical basis , and the states are assumed to be ordered . the symbol @xmath97 refers to the conjugate partner of @xmath95 , @xmath98 and @xmath99 come from the bcs transformation associated with the canonical basis , and the @xmath100 are the one - quasiparticle matrix elements of the hfb hamiltonian ( cf . ( 4.14b ) of ref . @xcite ) . @xmath101 and @xmath102 are the forward and backward amplitudes of the qrpa solution @xmath103 , and @xmath104 is the corresponding excitation energy . @xmath105 $ ] is the energy functional ( see app.[app : matrix_elements ] for an explicit definition ) and @xmath6 and @xmath106 are the density matrix and pairing tensor . after taking the functional derivatives , we replace @xmath6 and @xmath106 by their hfb solutions , in complete analogy with an ordinary taylor - series expansion . to write the equations in coupled form , we introduce the notation @xmath107 where @xmath108 denote spherical quantum numbers . using ( i ) rotational , time - reversal , and parity symmetries of the hfb state , ( ii ) the conjugate single - particle state where @xmath109 and @xmath110 are real radial and spin wave functions . thus , we multiply all hfb @xmath111 by @xmath112 in the qrpa calculations . ] @xmath113 and ( iii ) the relations @xmath114j_k } \times\left\{\begin{array}{cc}\sqrt{2 } , & \mu=\mu^\prime,\\ 1 , & \mbox{otherwise},\end{array}\right . \\ & & y^k_{kk^\prime } = ( -)^{j_\mu - m_\mu}(-)^{j_{\mu^\prime}-m_{\mu^\prime } } \langle j_\mu -m_\mu j_{\mu^\prime } -m_{\mu^\prime}\|j_k m_k\rangle \bar{y}^k_{[\mu\mu^\prime]j_k } \times\left\{\begin{array}{cc}\sqrt{2 } , & \mu=\mu^\prime,\\ 1 , & \mbox{otherwise},\end{array}\right.\end{aligned}\ ] ] with @xmath115 the angular momentum of the state @xmath103 and the factor @xmath116 for convenience @xcite , one can rewrite the qrpa equation as @xmath117j_k,[\nu\nu^\prime]j_k } & b_{[\mu\mu^\prime]j_k,[\bar{\nu}\bar{\nu}^\prime]j_k } \\ -b^\ast_{[\mu\mu^\prime]j_k,[\bar{\nu}\bar{\nu}^\prime]j_k } & -a^\ast_{[\mu\mu^\prime]j_k,[\nu\nu^\prime]j_k } \end{array } \right ) \left ( \begin{array}{c } \bar{x}^k_{[\nu\nu^\prime]j_k}\\ \bar{y}^k_{[\nu\nu^\prime]j_k } \end{array } \right ) = e_k \left ( \begin{array}{c } \bar{x}^k_{[\mu\mu^\prime]j_k}\\ \bar{y}^k_{[\mu\mu^\prime]j_k } \end{array } \right),\ \ \mu \leq \mu^\prime , \label{qrpa_eq}\ ] ] @xmath118j_k,[\nu\nu^\prime]j_k } & = & \frac{1}{\sqrt{1+\delta_{\mu\mu^\prime } } } \frac{1}{\sqrt{1+\delta_{\nu\nu^\prime } } } \left\ { e_{\mu\nu}\:\delta_{\mu^\prime\nu^\prime } -e_{\mu^\prime\nu}\:\delta_{\mu\nu^\prime}(-)^{j_\mu+j_{\mu^\prime}-j_k } \right . \nonumber\\ & & -e_{\mu\nu^\prime}\:\delta_{\mu^\prime\nu}(-)^{j_\mu+j_{\mu^\prime}-j_k } + e_{\mu^\prime\nu^\prime}\:\delta_{\mu\nu}\nonumber\\ & & + g(\mu\mu^\prime\nu\nu^\prime;j_k ) ( u_{\mu^\prime } u_\mu u_\nu u_{\nu^\prime } + v_{\nu } v_{\nu^\prime } v_{\mu^\prime } v_{\mu } ) \nonumber\\ & & + f(\mu\mu^\prime\nu\nu^\prime;j_k ) ( u_{\mu } v_{\nu^\prime } u_\nu v_{\mu^\prime } + u_{\mu^\prime } v_{\nu } u_{\nu^\prime } v_{\mu } ) \nonumber\\ & & -(-)^ { j_{\nu^\prime}+j_\nu - j_k } f(\mu\mu^\prime\nu^\prime\nu;j_k ) ( u_{\mu } v_{\nu } u_{\nu^\prime } v_{\mu^\prime } + u_{\mu^\prime } v_{\nu^\prime } u_{\nu } v_{\mu } ) \nonumber\\ & & -h(\mu\mu'\nu\nu';j_k ) ( v_\nu v_{\nu ' } u_\mu v_{\mu ' } + u_{\mu ' } v_\mu u_\nu u_{\nu ' } ) \nonumber \\ & & + ( -)^{j_\mu+j_{\mu'}-j_k } h(\mu'\mu\nu\nu';j_k ) ( v_\nu v_{\nu ' } u_{\mu ' } v_\mu + u_\mu v_{\mu ' } u_\nu u_{\nu ' } ) \nonumber \\ & & -h^\ast(\nu\nu'\mu\mu';j_k ) ( u_\mu u_{\mu ' } u_{\nu ' } v_\nu + u_\nu v_{\nu ' } v_{\mu ' } v_\mu ) \nonumber \\ & & + ( -)^{j_\nu+j_{\nu'}-j_k}h^\ast(\nu'\nu\mu\mu';j_k ) ( u_\mu u_{\mu ' } u_\nu v_{\nu ' } + u_{\nu ' } v_\nu v_{\mu ' } v_\mu ) \left.\right\ } , \label{a_spherical}\end{aligned}\ ] ] @xmath119j_k,[\bar{\nu}\bar{\nu}^\prime]j_k } & = & \frac{1}{\sqrt{1+\delta_{\mu\mu^\prime } } } \frac{1}{\sqrt{1+\delta_{\nu\nu^\prime } } } \left\ { -g(\mu\mu^\prime\nu\nu^\prime;j_k ) ( u_{\mu^\prime } u_\mu v_\nu v_{\nu^\prime } + u_{\nu } u_{\nu^\prime } v_{\mu^\prime } v_{\mu})\right . \nonumber\\ & & -(-)^ { j_\nu+j_{\nu^\prime}-j_k } f(\mu\mu^\prime\nu^\prime\nu;j_k ) ( u_{\mu } u_{\nu } v_{\mu^\prime } v_{\nu^\prime } + u_{\mu^\prime } u_{\nu^\prime } v_{\nu } v_{\mu } ) \nonumber\\ & & + ( -)^ { j_\nu+j_{\nu^\prime}+j_\mu+j_{\mu^\prime } } f(\mu^\prime\mu\nu^\prime\nu;j_k ) ( u_{\mu^\prime } u_{\nu } v_{\nu^\prime } v_{\mu } + u_{\mu } u_{\nu^\prime } v_{\mu^\prime } v_{\nu } ) \nonumber\\ & & + h(\mu\mu'\nu\nu';j_k)(v_\nu v_{\nu ' } u_{\mu ' } v_\mu + u_\mu v_{\mu ' } u_\nu u_{\nu ' } ) \nonumber \\ & & -(-)^{j_\mu+j_{\mu'}-j_k } h(\mu'\mu\nu\nu';j_k ) ( v_\nu v_{\nu ' } u_\mu v_{\mu ' } + u_{\mu ' } v_\mu u_\nu u_{\nu ' } ) \nonumber\\ & & -h^\ast(\nu\nu'\mu\mu';j_k ) ( v_\mu v_{\mu ' } u_{\nu ' } v_\nu + u_\nu v_{\nu ' } u_\mu u_{\mu ' } ) \nonumber\\ & & + ( -)^{j_\nu + j_{\nu'}-j_k}h^\ast(\nu'\nu\mu\mu';j_k ) ( v_\mu v_{\mu ' } u_\nu v_{\nu ' } + u_{\nu ' } v_\nu u_\mu u_{\mu ' } ) \left.\right\ } , \label{b_spherical}\end{aligned}\ ] ] @xmath120j_k \| \bar{v}^{\rm pp } f(\mu\mu^\prime\nu\nu^\prime;j_k)&=&\sum _ { m_\mu m_{\mu^\prime}m_\nu m_{\nu^\prime } } \langle j_\mu m_\mu j_{\mu^\prime } m_{\mu^\prime } \| j_k m_k\rangle \langle j_\nu m_\nu j_{\nu^\prime } m_{\nu^\prime } \| j_k m_k\rangle \bar{v}^{\rm ph}_{k\bar{l^\prime}\bar{k^\prime}l } \nonumber\\ & = & \sum_{j^\prime } ( -)^{j _ { \mu^\prime}+j_\nu + j^\prime } \left\ { \begin{array}{ccc } j_\mu & j_{\mu^\prime } & j_k \\ j_\nu & j_{\nu^\prime } & j^\prime \end{array } \right\ } ( 2j^\prime + 1 ) \langle [ \mu\nu^\prime]j_k \| \bar{v}^{\rm ph } \| [ \mu^\prime \nu]j_k \rangle , \label{f_matrix_element } \\ h(\mu\mu'\nu\nu';j_k ) & = & \sum_{m_\mu m_{\mu ' } m_\nu m_{\nu ' } } \langle j_\mu m_\mu j_{\mu ' } m_{\mu ' } \| jm \rangle \langle j_\nu m_\nu j_{\nu ' } m_{\nu ' } \| jm \rangle \bar{v}^{\rm 3p1h}_{\bar{l}\bar{l } ' k \bar{k } ' } \nonumber\\ & = & \sum_{j^\prime } ( -)^{j _ { \mu}+j_\nu + 1 + l_{\nu^\prime } -j_k - j^\prime } \left\ { \begin{array}{ccc } j_\mu & j_{\mu^\prime } & j_k \\ j_{\nu^\prime } & j_{\nu } & j^\prime \end{array } \right\ } ( 2j^\prime + 1 ) \langle [ \mu\nu]j_k \| \bar{v}^{\rm 3p1h } \| [ \mu^\prime \nu^\prime]j_k \rangle . \nonumber\\ \label{h_matrix_element}\end{aligned}\ ] ] we have represented the second derivatives of the energy functional @xmath105 $ ] as unsymmetrized matrix elements of effective interactions @xmath121 , @xmath122 , and @xmath123 . these effective interactions are given in app . [ app : matrix_elements ] . although the `` matrix elements '' are unsymmetrized , the underlying two - quasiparticle states are of course antisymmetric . as a consequence , @xmath124j_k,[\nu\nu^\prime]j_k}= b_{[\mu\mu^\prime]j_k , [ \bar{\nu } \bar{\nu}^\prime ] j_k}=0 $ ] if @xmath115 is odd and either @xmath125 or @xmath126 . the nuclear energy functional @xmath105 $ ] is usually separated into particle - hole ( ph ) and pairing pieces ( again , see app . b for explicit expressions ) . if the pairing functional , which we will call @xmath127 $ ] , depends on @xmath6 then the derivatives of @xmath127 $ ] with respect to @xmath128 are called pairing - rearrangement terms @xcite . in the qrpa , two kinds of pairing - rearrangement terms can arise in general . one has particle - hole character and is included in @xmath129 ; the other affects 3-particle-1-hole ( 3p1h ) and 1-particle-3-hole ( 1p3h ) configurations and is represented by @xmath130 and @xmath131 . if the @xmath6-dependence of @xmath127 $ ] is linear , then the ph - type pairing - rearrangement term does not appear . furthermore , the 3p1h and 1p3h pairing - rearrangement terms arise only for @xmath132 modes if the hfb state has @xmath133 . most existing work uses a pairing functional that is linear in @xmath6 , and so needs no pairing - rearrangement terms in @xmath134 channels . in this appendix , we discuss interaction matrix elements coming from @xmath105 $ ] , which we take to contain separate skyrme ( i.e. strong - force , @xmath106-independent ) , coulomb , and pairing energy functionals : @xmath135 = e_{\rm skyrme}[\rho ] + e_{\rm coul}[\rho_{\rm p } ] + e_{\rm pair}[\rho,\kappa,\kappa^\ast],\ ] ] where @xmath136 is the proton density matrix . the most general skyrme energy functional in common use is given by @xmath137 & = & \sum_{t=0,1 } \int d^3r \bigg\ { c_t^{\rho}[\rho_{00}]\ , \rho^2_{t0 } ( { \bm{r } } ) + c_t^{\delta\rho } \rho_{t0}({\bm{r } } ) \delta \rho_{t0}({\bm{r } } ) + c_t^{\tau } \left [ \rho_{t0}({\bm{r } } ) \ , \tau_{t0}({\bm{r } } ) -{\bm{j}}^2_{t0}({\bm{r } } ) \right ] \nonumber \\ & & + c_t^{s}[\rho_{00}]\ , { \bm{s}}^2_{t0}({\bm{r } } ) + c_t^{\delta s}{\bm{s}}_{t0}({\bm{r}})\cdot\delta{\bm{s}}_{t0}({\bm{r } } ) + c_t^{t } \left [ { \bm{s}}_{00}({\bm{r } } ) \cdot { \bm{t}}_{t0}({\bm{r } } ) -\tensor{j}{}^2_{t0}({\bm{r } } ) \right ] \nonumber \\ & & + c_t^{\nabla j } \left [ \rho_{t0}({\bm{r}}){\bm{\nabla}}\cdot{\bm{j}}_{t0}({\bm{r } } ) + { \bm{s}}_{t0 } ( { \bm{r } } ) \cdot { \bm{\nabla } } \times { \bm{j}}_{t0}({\bm{r } } ) \right ] \bigg\ } . \ ] ] ( see , e.g. , @xcite and references therein for a general discussion . ) all densities are labeled by isospin indices @xmath138 , where @xmath139 takes values zero and one and @xmath140 is always equal to 0 . a more general theory could violate isospin at the single - quasiparticle level , leading to additional densities @xmath141 @xcite . we do not consider such densities here . the @xmath142 are the coupling constants for the effective interaction . as usual , two of them are chosen to be density dependent : @xmath143 & = & a_t^\rho + b_t^\rho \ , \rho_{00}^\alpha({\bm{r } } ) , \nonumber \\ c_t^s [ \rho_{00 } ] & = & a_t^s + b_t^s \ , \rho_{00}^\alpha({\bm{r } } ) .\end{aligned}\ ] ] here @xmath144 , @xmath145 , @xmath146 , @xmath147 , @xmath148 , @xmath149 , and @xmath150 are local densities and currents , which are derived from the general density matrices for protons and neutrons @xmath151 where @xmath152 and @xmath153 labels the spin components so that , e.g. , @xmath154 is a spin component of the single - particle wave function associated with the state @xmath95 . defining @xmath155 where @xmath156 is a matrix element of the vector of pauli spin matrices , we write the local densities and currents as @xmath157 the coulomb energy functional is given by @xmath158 = \frac{e^2}{2 } \iint \ ! d^3r \ , d^3 r ' \frac{\rho_{\rm p}({\bm{r } } ) \rho_{\rm p}({\bm{r}}^\prime ) } { \|{\bm{r}}-{\bm{r}}'\| } -\frac{3}{4 } e^2 \left(\frac{3}{\pi}\right)^{\frac{1}{3 } } \int \ ! d^3 r \ ; \rho_{\text{p}}^{4/3 } ( { \bm{r } } ) , \ ] ] where we make the usual slater approximation @xcite for the exchange term . for the pairing functional we take the quite general form @xmath159 = e_{\text{pair}}[\rho,\tilde{\rho},\tilde{\rho}^\ast ] = \int \ ! d^3 r \ ; c^{\tilde\rho}[\rho_{00}({\bm{r } } ) ] \sum_{\tau={\rm p , n } } \|\tilde{\rho}_\tau({\bm{r}})\|^2 , \label{pairing_energy_functional}\ ] ] where the density - dependent pairing coupling constant @xmath160 $ ] is an arbitrary function of @xmath161 . the quantity @xmath162 is defined as @xcite @xmath163 with @xmath164 being the standard pairing tensor in the coordinate representation . the second derivatives of the energy functional in eq.([ph_matrix_element_uncoupled ] ) , as the equation indicates and we ve already noted , can be written as _ unsymmetrized _ matrix elements @xmath165 of an effective interaction between uncoupled pairs of single - particle states . the particle - hole matrix elements take the form @xmath166 the last term contains the pairing rearrangement discussed at the end of the previous appendix . the effective skyrme interaction in eq . ( [ decomp ] ) is given by @xmath167 \ , \rho_{00}^{\alpha-1}({\bm{r } } ) \ , \delta({\bm{r}}-{\bm{r } } ' ) \nonumber\\ & & + \big [ f_3 \ , \rho^2_{10}({\bm{r } } ) + h_3 \ , { \bm{s}}^2_{00}({\bm{r } } ) + n_3 \ , { \bm{s}}^2_{10}({\bm{r } } ) \big ] \ , \rho_{00}^{\alpha-2}({\bm{r } } ) \ , \delta({\bm{r}}-{\bm{r } } ' ) \nonumber\\ & & + ( a_4 + c_4 \ , { \vec{\tau } } \cdot { \vec{\tau } } ' ) \ , ( { \bm{\sigma } } + { \bm{\sigma } } ' ) \cdot { { \bm{k}}^\dagger}\times \delta({\bm{r}}-{\bm{r } } ' ) \ , { \bm{k } } . \label{operator_v_skyrme}\end{aligned}\ ] ] @xmath168 is the vector of pauli matrices in isospin space and @xmath169 the coefficients in eq . ( [ skyrme_operator ] ) are defined in tables [ tab : coefficients_abc ] and [ tab : coefficients_efg ] . equation ( [ operator_v_skyrme ] ) contains the usual skyrme - interaction operators , but , the energy functional ( [ e : skyrme ] ) does not necessarily correspond to a real ( density - dependent ) two - body skyrme interaction because the matrix elements are not antisymmetrized . compared to the case usually discussed in the literature , the more general functional relaxes relations that would otherwise restrict the spin - isospin structure of the effective interaction in eq . ( [ skyrme_operator ] ) ; see , e.g. , @xcite for a discussion of the increased freedom . the densities and currents that appear in eq . ( [ skyrme_operator ] ) come mostly from rearrangement terms and take the values given by the hfb ground state . the isoscalar and isovector spin densities @xmath170 vanish when the hfb ground state is time - reversal invariant or spherical as assumed here . the terms containing them will therefore not appear in the expressions for the matrix elements of the effective interaction for such states given below . the effective coulomb interaction in eq . ( [ decomp ] ) , acting between protons , is given by @xmath175 finally , the ph - type pairing - rearrangement terms in eq . ( [ decomp ] ) come from an effective interaction @xmath176}{d \rho_{00}^2({\bm{r } } ) } \sum_{\tau={\rm p , n}}\|\tilde{\rho}_\tau({\bm{r}})\|^2\ , \delta({\bm{r}}'-{\bm{r}}).\ ] ] the second derivatives with respect to @xmath177 also can be written as unsymmetrized matrix elements of effective interactions , this time in the particle - particle channel . the particle - particle effective interaction entering the matrix elements @xmath178 is obtained from eq . ( [ pairing_energy_functional ] ) through eq . ( [ pp_matrix_element_uncoupled ] ) as @xmath179 \left ( 3 - { \bm{\sigma } } \cdot { \bm{\sigma } } ' - { \vec{\tau } } \cdot { \vec{\tau } } ' - { \bm{\sigma } } \cdot { \bm{\sigma } } ' \ , { \vec{\tau } } \cdot { \vec{\tau } } ' \right ) \ , \delta({\bm{r}}-{\bm{r } } ' ) .\ ] ] in the numerical calculations of this paper , we use a volume pairing - energy functional , i.e. , @xmath180 last of all are the mixed functional derivatives involving both @xmath6 and @xmath106 ( or @xmath181 ) in eq . ( [ 3p1h_matrix_element_uncoupled ] ) . they also can be written as the unsymmetrized matrix elements of an effective interaction : @xmath182 where @xmath183 denotes the time - reversed state of @xmath95 , and the 3p1h effective interaction itself is @xmath184}{d\rho_{00}({\bm{r } } ) } \left [ \tilde{\rho}_{\rm p } ( { \bm{r}})\left(1-\tau_z^\prime\right ) + \tilde{\rho}_{\rm n } ( { \bm{r}})\left(1+\tau_z^\prime\right)\right ] \delta({\bm{r}}'-{\bm{r } } ) , \ ] ] where @xmath185 acts on the single - particle states @xmath186 and @xmath183 in eq . ( [ v_3p1h_matrix_element ] ) , and the eigenvalues 1 and @xmath187 are assigned to the neutron and proton , respectively . to calculate the coupled matrix elements in eqs . ( [ g_matrix_element])([h_matrix_element ] ) , we use an intermediate ls scheme : @xmath188j_k\|\bar{v}\|[\nu\nu^\prime]j_k\rangle } \nonumber\\ & = & \sum_{ll^\prime s } \hat{j}_\mu \hat{j}_{\mu^\prime } \hat{j}_\nu \hat{j}_{\nu^\prime } \hat{l } \hat{l^\prime } \hat{s}^2 \left\ { \begin{array}{ccc } l_\mu & l_{\mu^\prime } & l \\ 1/2 & 1/2 & s \\ j_\mu & j_{\mu^\prime } & j_k \end{array } \right\ } \left\ { \begin{array}{ccc } l_\nu & l_{\nu^\prime } & l^\prime \\ 1/2 & 1/2 & s \\ j_\nu & j_{\nu^\prime } & j_k \end{array } \right\}\nonumber \\ & & \times \langle ( l_\mu l_{\mu^\prime})ls;j_k\|\bar{v}\|(l_\nu l_{\nu^\prime})l^\prime s;j_k\rangle,\end{aligned}\ ] ] @xmath189 eq . ( [ skyrme_operator ] ) gives \i ) proton - proton or neutron - neutron matrix elements : @xmath190 ii ) proton - neutron matrix elements : @xmath191 we use the canonical ( and real ) radial wave functions @xmath192 , the angular wave functions @xmath193 , and the spin wave functions to write the nontrivial matrix elements included in eqs.([v_skyrme_proton_proton ] ) and ( [ v_skyrme_proton_neutron ] ) as @xmath194^\ast_{l0 } \left [ y_{l_\nu}(\omega ) y_{l_{\nu^\prime}}(\omega)\right]_{l0 } , \nonumber\\ \label{matrix_element_delta}\end{aligned}\ ] ] @xmath195^\ast_{l0 } \left [ y_{l_\nu}(\omega ) y_{l_{\nu^\prime}}(\omega)\right]_{l0 } = \frac{1}{4\pi}\hat{l_\mu } \hat{l_{\mu^\prime } } \hat{l_\nu } \hat{l_{\nu^\prime } } \left(\begin{array}{ccc}l_\mu & l_{\mu^\prime } & l \\ 0 & 0 & 0\end{array}\right ) \left(\begin{array}{ccc}l_\nu & l_{\nu^\prime } & l \\ 0 & 0 & 0\end{array}\right ) , \ ] ] @xmath196^\ast_{l0 } \left [ y_{l_\nu}(\omega ) y_{l_{\nu^\prime}}(\omega)\right]_{l0 } \nonumber\\ & & \times \int dr\:r^2 r_\mu(r ) r_{\mu^\prime}(r ) \left\ { \left [ \left ( \frac{d^2}{dr^2}+\frac{2}{r}\frac{d}{dr}-\frac{l_\nu(l_\nu+1)}{r^2 } \right)r_\nu(r)\right]r_{\nu^\prime}(r)\right.\nonumber\\ & & \left.+r_\nu(r)\left [ \left ( \frac{d^2}{dr^2}+\frac{2}{r}\frac{d}{dr } -\frac{l_{\nu^\prime}(l_{\nu^\prime}+1)}{r^2}\right)r_{\nu^\prime}(r ) \right]\right\}\nonumber\\ & & -\sum_{\delta l_\nu=0,1 } \sum_{\delta l_{\nu^\prime}=0,1 } \frac{1}{2}\sqrt{(l_\nu+\delta l_\nu)(l_{\nu^\prime}+\delta l_{\nu^\prime } ) } \left\{\begin{array}{ccc}l_{\nu^\prime}-1 + 2\delta l_{\nu^\prime } & l_{\nu}-1 + 2\delta l_{\nu } & l \\ l_\nu & l_{\nu^\prime } & 1 \end{array}\right\ } \nonumber\\ & & \times\int dr\:r^2 r_\mu(r ) r_{\mu^\prime}(r ) \left\ { ( l_\nu+1-\delta l_\nu)\frac{r_\nu(r)}{r } + ( -)^{\delta l_\nu}\frac{dr_\nu(r)}{dr}\right\}\nonumber\\ & & \times \left\ { ( l_{\nu^\prime}+1-\delta l_{\nu^\prime})\frac{r_{\nu^\prime}(r)}{r } + ( -)^{\delta l_{\nu^\prime}}\frac{dr_{\nu^\prime}(r)}{dr}\right\}\nonumber\\ & & \times \delta_{ll^\prime } \int d\omega \left [ y_{l_\mu}(\omega ) y_{l_{\mu^\prime}}(\omega)\right]^\ast_{l0 } \left [ y _ { l_{\nu^\prime}-1 + 2\delta l_{\nu^\prime } } ( \omega ) y_{l_\nu-1 + 2\delta l_\nu } ( \omega)\right]_{l0 } , \end{aligned}\ ] ] @xmath197 @xmath198 the square brackets around products of spherical harmonics and the parentheses surrounding products of operators indicate angular - momentum coupling . to evaluate eqs . ( [ t2 ] ) and ( [ spin_orbit ] ) , one can use @xmath199 -(-)^{l_\nu+l_{\nu^\prime}+l^\prime}[\nu \leftrightarrow \nu^\prime ] -(-)^{l_\mu+l_{\mu^\prime}+l}[\mu \leftrightarrow \mu^\prime ] \nonumber\\ & & + [ \mu \leftrightarrow \mu^\prime { \rm \ and\ } \nu \leftrightarrow \nu^\prime ] , \end{aligned}\ ] ] where for reduced matrix elements we have used the convention @xmath200 and made the abbreviation @xmath201 -(-)^{l_\nu+l_{\nu^\prime}+l^\prime}[\nu \leftrightarrow \nu^\prime ] -(-)^{l_\mu+l_{\mu^\prime}+l}[\mu \leftrightarrow \mu^\prime ] + [ \mu \leftrightarrow \mu^\prime { \rm \ and\ } \nu \leftrightarrow \nu^\prime ] \nonumber\\ & \equiv & a_{\mu\mu^\prime\nu\nu^\prime } - ( -)^{l_\nu+l_{\nu^\prime}+l^\prime } a_{\mu\mu^\prime\nu^\prime\nu } - ( -)^{l_\mu+l_{\mu^\prime}+l}a_{\mu^\prime\mu\nu\nu^\prime } + a_{\mu^\prime\mu\nu^\prime\nu } .\end{aligned}\ ] ] eq . ( [ matrix_element_delta ] ) , modified to include additional factors in the radial integral , can also be used ( together with the subsequent equations ) to evaluate the matrix elements of the terms involving @xmath202 in @xmath203 , the coulomb - exchange interaction , and the contributions of the pairing functional to the effective ph , pp , and 3p1h interactions . the coulomb - direct term can be evaluated in a similar but slightly more complicated way , via a multipole expansion . in the main part of this paper we used the skyrme functional skm@xmath28 , which is usually parameterized as in interaction in terms of coefficients @xmath204 and @xmath205 . the relations between these coefficients and those used here , if no terms are neglected , are @xcite @xmath206 in the hf fits that originally determined the skm@xmath28 parameters , the effects of @xmath207 ( the `` @xmath208 terms '' ) were neglected because of technical difficulties . these terms have often been included in subsequent rpa calculations . to maintain self consistency here , we have set them to zero both in the hfb calculation and in the qrpa .
let @xmath9 be a group with a generating set @xmath10 , and acting on a set @xmath1 . the ( simplicial ) schreier graph @xmath11 of the action @xmath12 is the graph with the set of vertices @xmath1 , where two vertices @xmath13 and @xmath14 are adjacent if and only if there exists @xmath15 such that @xmath16 . the schreier graphs are generalizations of the cayley graph of a group , which corresponds to the action of a group on itself by the left multiplication . the growth of cayley graphs ( growth of groups ) is one of the central objects of study in geometric group theory . the celebrated theorem of m. gromov characterizes groups of polynomial growth . many classical groups ( like linear groups , solvable groups , hyperbolic groups , etc . ) have either polynomial or exponential growth . the first group of intermediate growth between polynomial and exponential was constructed by r. grigorchuk @xcite . nowadays , the grigorchuk group is the best known example of automaton groups ( also known as self - similar groups , or groups generated by automata ) . in this paper , we consider the growth of schreier graphs of automaton groups . every automaton group @xmath9 generated by an automaton @xmath17 over some alphabet @xmath1 naturally acts on the set @xmath18 for every @xmath19 , and on the space @xmath0 of right - infinite sequences over @xmath1 . we get the associated sequence of finite schreier graphs @xmath20 of the action @xmath21 , and the family of orbital schreier graphs @xmath22 for @xmath2 of the action of @xmath9 on the orbit of @xmath4 . the schreier graphs @xmath23 and @xmath5 provide a large source of self - similar graphs , and were studied in relation to such topics as spectrum , growth , amenability , topology of julia sets , etc . ( see @xcite ) . it was noticed in @xcite that for a few automaton groups the sequence of schreier graphs @xmath23 converges to a nice metric space with fractal structure . further , this observation lead to the notion of a limit space of an automaton group , and , more generally , to a limit dynamical system @xcite . the expanding property of this dynamical system corresponds to the contracting property of a group , what brings us to the important class of contracting automaton groups . namely the strong contracting properties of the grigorchuk group was used to prove that it has intermediate growth , while in general a contracting group may have exponential growth . at the same time , all orbital schreier graphs @xmath5 of contracting groups have polynomial growth @xcite . the degree of the growth is related to the asymptotic characteristics of the group such as the complexity of the word problem , the hausdorff dimension of the limit space , the exponent of divergence of geodesics in the associated gromov - hyperbolic self - similarity complex , etc . ( see @xcite ) . most of the studied automaton groups are generated by polynomial automata . these automata were introduced by s. sidki in @xcite , who tried to classify automaton groups by the cyclic structure of the generating automaton . a finite automaton is called polynomial if the simple directed cycles away from the trivial state are disjoint . the term polynomial " comes from the equivalent definition , where a finite automaton is polynomial if the number of paths of length @xmath24 avoiding the trivial state in the automaton grows polynomially in @xmath24 . it is an open problem whether contracting groups and groups generated by polynomial automata are amenable . however , it is known that these groups do not contain nonabelian free subgroups @xcite ( but may contain free semigroups and be of exponential growth ) , and that groups generated by polynomial automata of degree @xmath25 ( bounded automata ) are amenable @xcite . in @xcite v. nekrashevych introduced a general approach to the existence of free subgroups in automaton groups , and applied it to contracting groups and to groups generated by polynomial automata . in order to eliminate one of the cases , he proved that the orbital schreier graphs @xmath5 are amenable , and conjectured @xcite that these graphs may have subexponential growth . this conjecture was based on the results from @xcite , where it is shown that for the group generated by one of the simplest polynomial automata all schreier graphs @xmath5 have intermediate growth . the main goal of this paper is to prove this conjecture . [ theorem_main ] let @xmath9 be a group generated by a polynomial automaton of degree @xmath7 . there exists a constant @xmath26 such that all orbital schreier graphs @xmath27 for @xmath2 have subexponential growth not greater than @xmath28 . in order to prove theorem 1 we establish certain weak contracting properties of groups generated by polynomial automata . in particular , we prove that the word problem in these groups is solvable in subexponential time bounded above by @xmath29 for some constant @xmath30 . we also prove theorem [ theorem_main ] for a more general class of automata ( see theorem [ theorem_growth_schr_general ] in section [ generalized version ] ) , which generalizes both polynomial automata and contracting groups . in section [ section_examples ] we apply theorem [ theorem_main ] and its generalized version to construct automaton groups with schreier graphs @xmath5 of intermediate growth . in the first example we consider a class of automaton groups , whose schreier graphs @xmath5 are a generalized version of the @xmath8-periodic graph of intermediate growth studied in @xcite . basically , this example shows that for many polynomial automata the schreier graphs @xmath5 have intermediate growth . another example comes from the well - known hanoi tower game on @xmath31 pegs . this game was modeled by automaton groups @xmath32 in @xcite , and , using known estimates on the complexity of the hanoi tower game , it was noticed that the orbital schreier graph @xmath33 for @xmath34 has intermediate growth . the automata generating groups @xmath32 for @xmath34 are not polynomial , but we apply similar arguments to prove that all orbital schreier graphs @xmath35 have intermediate growth . in this section we recall all needed definition , see @xcite for a more detailed introduction . * spaces of words . * let @xmath1 be a finite alphabet with at least two letters . denote by @xmath36 the set of all finite words over @xmath1 ( including the empty word denoted @xmath37 ) . the length of a word @xmath38 is denoted by @xmath39 . let @xmath40 be the space of all infinite sequences ( words ) @xmath41 , @xmath42 , with the product topology of discrete sets @xmath1 . the space @xmath0 is homeomorphic to the cantor set , i.e. , it is compact totally disconnected metrizable topological space without isolated points . * automata . * an invertible _ automaton _ @xmath17 over the alphabet @xmath1 is the triple @xmath43 , where @xmath10 is the set of states of the automaton , @xmath44 is the transition function , and @xmath45 is the output function such that for every @xmath46 the map @xmath47 is a permutation on @xmath1 . all automata in the paper are invertible , and further we omit the term invertible . an automaton is _ finite _ if the set of its states is finite . an automaton @xmath17 is represented by a directed labeled graph ( moore diagram ) , whose vertices are the states of @xmath17 , and for every state @xmath46 and every letter @xmath48 there is an arrow @xmath49 labeled by the pair @xmath50 . this diagram contains complete information about the automaton , and we identify the automaton with its moore diagram . the notation @xmath17 is also used for the state set of the automaton @xmath17 , so that one can talk about a state @xmath51 . a subset @xmath52 ( with induced edges ) is a _ subautomaton _ of @xmath17 if @xmath53 for every @xmath54 and @xmath48 . * automaton groups . * for every state @xmath51 and every finite word @xmath55 there exists a unique path in the automaton @xmath17 starting at the state @xmath56 and labeled by @xmath57 , @xmath58 , , @xmath59 for some @xmath60 . then the word @xmath61 is called the image of @xmath62 under @xmath56 , and the end state of this path is denoted by @xmath63 . in other words , for every finite word @xmath64 we define the _ image @xmath65 of @xmath66 under @xmath56 _ and the _ state _ @xmath63 recursively by the rules @xmath67 for every @xmath48 and @xmath64 . the action of states of @xmath17 on the set @xmath18 can be given by the automaton @xmath68 obtained from @xmath17 by passing to the power @xmath18 of the alphabet @xmath1 . the automaton @xmath68 is defined over the alphabet @xmath18 , and has the same states as @xmath17 , but the arrows are @xmath69 labeled by @xmath70 for every @xmath51 and @xmath71 . in the same way we get an action of every state @xmath51 on the space @xmath0 by looking at the infinite paths in the automaton . since the automata are invertible , all transformations @xmath51 are also invertible . the group @xmath9 generated by all states @xmath51 is called the _ automaton group _ generated by @xmath17 . since every state preserves the length of words in its action on @xmath72 , the states act by isometries on the space @xmath0 , and every automaton group is a subgroup of @xmath73 . the automaton groups generated by @xmath17 and @xmath68 coincide viewed as subgroups of @xmath73 with the natural identification @xmath74 . all automata in the paper are supposed to be minimized , i.e. , different states of a given automaton act differently on @xmath0 . hence we identify the states with the respective transformations of @xmath0 . an alternative approach is through self - similar actions . a faithful action of a group @xmath9 on the set @xmath75 is called _ self - similar _ if for every @xmath76 and @xmath64 there exist @xmath77 and @xmath78 such that @xmath79 for all @xmath80 . the element @xmath81 is called the _ restriction ( state ) of @xmath82 at @xmath66 _ and is denoted by @xmath83 . we are using left actions , i.e. , @xmath84 , and hence the restrictions have the property @xmath85 every self - similar action of a group @xmath9 is given by the _ complete automaton _ @xmath86 of the group , whose states are the elements of @xmath9 , and the arrows are @xmath87 labeled by @xmath88 for every @xmath76 and @xmath48 . an automaton group @xmath9 is called _ contracting _ if there exists a finite subset @xmath89 with the property that for every @xmath76 there exists @xmath90 such that @xmath91 for all words @xmath64 of length @xmath92 . the smallest set @xmath93 with this property is called the _ nucleus _ of the group . notice that a finitely generated contracting group can be generated by a finite automaton . a group @xmath9 generated by a finite automaton is contracting if and only if the complete automaton @xmath86 of the group contains only finitely many simple directed cycles . throughout the paper by a _ cycle _ in an automaton we mean a simple directed cycle . states that lie on cycles are called _ circuit_. an element @xmath82 of an automaton group @xmath9 is circuit if there exists a nonempty word @xmath64 such that @xmath94 . * polynomial automata . * a cycle in an automaton is called trivial if it is a loop at the state acting trivially on @xmath0 ( trivial state , denoted @xmath95 ) . a finite automaton @xmath17 is called _ polynomial _ if different nontrivial cycles in @xmath17 are disjoint . a polynomial automaton @xmath17 is _ of degree @xmath7 _ if the largest number of nontrivial cycles in @xmath17 connected by a directed path is equal to @xmath96 . a finite automaton @xmath17 is polynomial ( of degree @xmath97 ) if and only if the number of directed paths in @xmath17 of length @xmath24 that do not pass through the trivial state is bounded by a polynomial in @xmath24 ( of degree @xmath97 ) . the set of all states ( transformations of @xmath0 ) of all polynomial automata of degree @xmath97 forms a group @xmath98 called the _ group of polynomial automata of degree _ every finitely generated subgroup of @xmath98 is a subgroup of some automaton group generated by a polynomial automaton . we classify the states of a polynomial automaton @xmath17 as follows . a state @xmath51 is _ finitary _ if there exists @xmath90 such that @xmath99 for all @xmath71 . the finitary states are precisely the elements of @xmath100 . the states from @xmath101 are called _ polynomial of degree @xmath7_. for every polynomial automaton @xmath17 there exists @xmath24 such that for every @xmath51 and @xmath71 the state @xmath63 is either circuit or has degree less than the degree of @xmath56 . notice , that if @xmath51 is a nontrivial circuit state , then for every @xmath24 there exists precisely one word @xmath71 such that @xmath56 and @xmath63 have the same degree as polynomial states , and @xmath102 for all @xmath103 , @xmath104 , has degree less than the degree of @xmath56 . * schreier graphs . * let @xmath9 be an automaton group generated by a finite subset @xmath10 . the _ schreier graph @xmath105 _ is the graph with the set of vertices @xmath18 , where two vertices @xmath66 and @xmath106 are adjacent if there exists @xmath15 such that @xmath107 . orbital schreier graph @xmath108 _ for @xmath2 is the graph whose vertex set is the orbit @xmath109 and two vertices @xmath66 and @xmath106 are adjacent if there exists @xmath15 such that @xmath107 . the orbital schreier graphs @xmath110 are precisely the connected components of the schreier graph @xmath111 of the action @xmath112 . every orbital schreier graph @xmath5 is a limit of the finite schreier graphs @xmath23 in the local gromov - hausdorff topology on pointed graphs . * growth of graphs . * let @xmath113 be a locally finite connected graph . the _ growth function _ @xmath114 of @xmath113 with respect to its vertex @xmath66 is equal to the number of vertices in the closed ball @xmath115 of radius @xmath24 centered at @xmath66 . there is a partial order on the growth functions . given two functions @xmath116 we say that @xmath117 has _ growth _ not greater than @xmath82 ( denoted @xmath118 ) if there exists a constant @xmath119 such that @xmath120 for all @xmath90 . if @xmath118 and @xmath121 then @xmath117 and @xmath82 are called equivalent @xmath122 and have the same growth . formally , by the growth we can understand the equivalence class of a function . then , for any two vertices of the graph @xmath113 , the respective growth functions are equivalent , and one can talk about the growth of @xmath113 . a graph @xmath113 has _ subexponential growth _ if its growth function has growth not greater than the exponential growth @xmath123 with @xmath124 , and is not equivalent to it . the growth is _ superpolynomial _ if it is greater than every polynomial function , and the growth is _ intermediate _ if it is superpolynomial and subexponential . let @xmath9 be a finitely generated group with finite generating set @xmath10 . the _ length _ @xmath125 of @xmath76 with respect to @xmath10 is equal to the distance between the trivial element @xmath95 and @xmath82 in the cayley graph @xmath126 , i.e. , @xmath127 and @xmath128 the growth function @xmath129 of the group @xmath9 is the growth function of the cayley graph @xmath126 with respect to the vertex @xmath95 , i.e. , @xmath130 is equal to the number of elements @xmath76 of length @xmath131 . the growth functions of the cayley graphs @xmath132 and @xmath133 , for any two finite generating sets @xmath134 and @xmath135 , are equivalent . the growth of schreier graphs @xmath136 of an automaton group @xmath9 also does not depend on the choice of a finite generating set @xmath10 of the group . working with automaton groups it is useful to assume that a generating set @xmath10 is _ self - similar _ ( automaton ) , i.e. , @xmath137 for every @xmath46 and @xmath64 . for example , we will frequently use the following observation . if there is a presentation @xmath138 for @xmath139 then for every @xmath64 we get the _ induced presentation _ @xmath140 where @xmath141 , and @xmath142 for @xmath143 . in particular , if @xmath10 is self - similar then @xmath144 and @xmath145 . also , it is usually assumed that @xmath10 is _ symmetric _ , i.e. , @xmath146 . all logarithms in the paper are with base @xmath147 , except if directly indicated . usually , a logarithm appears as @xmath148 , and the base can be hidden in the constant @xmath149 . also ( to avoid some multiple brackets ) we use the convention that @xmath150 and @xmath151 so that @xmath152 for all @xmath153 , otherwise one can just replace @xmath154 by @xmath155 . let us recall how to prove that for contracting groups the schreier graphs @xmath5 have polynomial growth ( see ( * ? ? ? * section 2.13.4 ) , @xcite ) . let @xmath9 be a finitely generated contracting group with nucleus @xmath93 , and let @xmath10 be a finite symmetric self - similar generating set of @xmath9 that contains @xmath93 . we can choose a constant @xmath149 such that @xmath156 for all @xmath157 and every word @xmath64 of length @xmath158 . consider an element @xmath76 and let @xmath138 for @xmath139 with @xmath159 . then @xmath160 and hence the element @xmath83 has length @xmath161 . it follows that @xmath91 for all words @xmath64 of length @xmath162 . this justifies the term contracting group : the length of restrictions exponentially decreases ( contracts " ) until they become elements of the nucleus . now consider the schreier graph @xmath163 for a sequence @xmath164 . let @xmath165 be the ball of radius @xmath24 in the graph @xmath5 centered at the vertex @xmath4 . notice that if @xmath166 for @xmath167 and @xmath168 then @xmath169 . let @xmath31 be the least integer greater than @xmath148 . then each sequence in the ball @xmath165 is of the form @xmath170 for some @xmath171 and @xmath172 for some @xmath173 . hence @xmath174 which is a polynomial in @xmath24 . more precise estimate can be given using the contracting coefficient of the group ( see ( * ? ? ? * proposition 2.13.8 ) ) . we want to apply similar arguments to groups generated by polynomial automata , and first we will establish certain weak contracting properties of these groups . [ lemma_weak_contr_poly ] let @xmath9 be a finitely generated subgroup of @xmath98 . there exists a constant @xmath149 such that for every @xmath76 and every word @xmath64 of length @xmath175 the state @xmath83 is either circuit or belongs to @xmath176 . to prove lemma [ lemma_weak_contr_poly ] we use induction on @xmath7 . however there is one slight difficulty , that in order to apply the induction hypothesis we may need to consider the length of the element @xmath83 for a different generating set than the length of @xmath82 . to overcome this problem we will make a few assumptions so that the groups involved in the induction have consistent generating sets . without loss of generality , we can assume that the group @xmath9 is an automaton subgroup of @xmath98 . let @xmath10 be a finite symmetric self - similar generating set of @xmath9 . let @xmath177 be an automaton subgroup of @xmath9 generated by @xmath178 for @xmath179 . the generating set @xmath180 is again symmetric and self - similar . notice that , in general , the group @xmath177 may not coincide with @xmath181 . let us prove that the length of cycles in the complete automaton @xmath86 of the group @xmath9 is bounded . without loss of generality , by passing to a power of the alphabet , we can assume that the alphabet @xmath1 and the generating set @xmath10 satisfy the following conditions : + 1 ) for every @xmath46 and @xmath48 the state @xmath182 is either circuit or has degree less than the + degree of @xmath56 ; + 2 ) for every nontrivial circuit state @xmath46 there exists ( unique ) @xmath48 such that @xmath183 + ( every cycle in @xmath10 is a loop ) . + ( for example , one can pass to the alphabet @xmath184 , where @xmath31 is a multiple of the length of every cycle in @xmath10 , and is greater than the diameter of @xmath10 ) . let us prove that then every cycle in the automaton @xmath86 is a loop , i.e. , for every circuit element @xmath76 there exists a letter @xmath48 such that @xmath185 . assume @xmath94 for a nonempty word @xmath186 and let @xmath138 for @xmath139 with @xmath159 . notice that @xmath187 for every @xmath188 , because otherwise @xmath94 can be expressed as a product @xmath189 with @xmath190 . over all presentations of @xmath82 as a product @xmath191 we consequently for @xmath192 choose the one with the largest number of circuit generators from @xmath193 and then with the largest number of the rest generators from @xmath180 . consider the induced presentation : @xmath194 where @xmath142 with @xmath141 . by our choice of the presentation @xmath138 we should get @xmath195 then @xmath196 where @xmath197 is the first letter of @xmath198 , and we get @xmath199 . it also follows that @xmath200 for @xmath201 , @xmath202 , and in its induced presentation @xmath203 every generator @xmath204 belongs to the set @xmath205 . in particular , the length of @xmath206 with respect to the generating set @xmath205 of the group @xmath207 is not greater than @xmath24 . further we will use this property without specifying the generating set . to the rest of this section we always assume that the group @xmath9 and the generating set @xmath10 satisfy the above assumptions . we will prove a slightly stronger formulation of lemma 1 . * lemma 1. _ there exists a constant @xmath149 such that for every @xmath208 for @xmath139 and every word @xmath64 of length @xmath209 either @xmath210 and in the induced presentation @xmath211 every generator @xmath212 belongs to @xmath205 , or the element @xmath83 is circuit , @xmath213 for some @xmath48 , and the induced presentations of @xmath83 and @xmath214 coincide as words in generators . _ the proof goes by induction on @xmath7 . for @xmath215 the group @xmath9 is a subgroup of the group @xmath100 of finitary elements . there exists a constant @xmath149 such that for all words @xmath66 of length @xmath158 we have @xmath99 for every @xmath46 ( actually by our assumption 1 ) on the generating set @xmath10 we can take @xmath216 ) . hence for every product @xmath208 for @xmath139 we get @xmath217 and in the induced presentation of @xmath83 every element is trivial . we assume the lemma holds for the groups @xmath218 ( with @xmath7 replaced by @xmath31 ) for @xmath219 with some common constant @xmath220 . let @xmath76 be a circuit element and @xmath185 for a letter @xmath48 . let @xmath221 , @xmath139 , be a presentation such that the induced presentation of @xmath222 coincides with @xmath223 as a word in generators . then for every word @xmath64 either @xmath224 and every generator in the induced presentation of @xmath83 belong to @xmath205 , or @xmath94 ( in the case @xmath225 ) and the induced presentation of @xmath83 coincides with @xmath223 . consider the product @xmath226 for @xmath227 , @xmath228 , and its restrictions @xmath229 for words @xmath64 of length @xmath230 . we have the following cases . if @xmath231 then @xmath232 , so further assume @xmath233 and let us look at @xmath234 . by the assumption on the length of the word @xmath66 the element @xmath234 either belongs to @xmath235 or it is circuit ( with the specific induced presentation ) . + if @xmath236 then @xmath237 for @xmath238 . + if @xmath234 is circuit and @xmath239 then @xmath240 is circuit , here @xmath241 . + if @xmath234 is circuit and @xmath242 then @xmath243 for @xmath238 . + hence in all cases the element @xmath244 for @xmath245 is either circuit , or belongs to @xmath207 , or it is of the form @xmath246 for @xmath238 . in the last case we can apply the same arguments to the product @xmath246 . after @xmath7 steps we get either a circuit state or an element of @xmath207 . in the worst case we need to take the words of length @xmath247 choose a constant @xmath248 such that @xmath249 is greater than the value of the equation above . then the element @xmath240 for words @xmath64 of length @xmath250 is either circuit or belongs to @xmath207 , and the conditions on the induced presentation of @xmath240 are satisfied . let @xmath251 be circuit elements and consider the product @xmath252 for @xmath253 . assume @xmath254 is expressed as a product in generators of length @xmath255 with the same properties as the one of @xmath82 above , and @xmath81 is expressed as a product of @xmath24 generators from @xmath205 . then the element @xmath256 for @xmath64 of length @xmath257 with @xmath258 is either circuit or belongs to @xmath207 with the specific induced presentation . now consider an arbitrary element @xmath259 for @xmath139 . let us partition the product @xmath260 on blocks @xmath261 such that every block @xmath262 contains only generators from @xmath205 ( may be trivial ) , and @xmath254 is either a generator from @xmath263 , or @xmath254 is circuit and the induced presentation of @xmath264 coincides with the presentation @xmath254 for some @xmath48 . notice that the sum of lengths of all @xmath262 and @xmath254 is equal to @xmath24 . by passing to the induced presentation of the state @xmath265 on any letter @xmath48 we can assume that every element @xmath254 is circuit . consider every product @xmath266 , @xmath267 , and so on , in the presentation ( [ eqn_lemma_g = ghgh ... ] ) . every such block restricted to a word of length @xmath268 is either circuit or belongs to @xmath207 . hence the state @xmath83 for words @xmath64 of length @xmath269 can be expressed as a product ( [ eqn_lemma_g = ghgh ... ] ) with @xmath270 positions with some circuit elements @xmath254 . applying the same procedure @xmath271 times we get either a circuit state or an element of @xmath207 . choose a constant @xmath149 such that @xmath272 ( here we use @xmath273 ) . then @xmath83 for @xmath64 of length @xmath274 is either circuit or belongs to @xmath207 , and in both cases it has the required induced presentation . the word problem in every finitely generated subgroup of @xmath98 is solvable in subexponential time . we use the same notations and assumptions as above . consider an element @xmath259 for @xmath139 . let @xmath31 be the least integer greater than @xmath275 . the element @xmath82 is trivial if and only if it acts trivially on @xmath184 and every element @xmath83 for @xmath276 is trivial . notice that the size of @xmath184 is subexponential in @xmath24 . consider the induced presentation @xmath277 with @xmath278 . if @xmath279 for all @xmath188 , then the problem reduces to the word problem in the group @xmath207 . otherwise , @xmath83 is circuit and there exists @xmath48 such that the induced presentation @xmath280 coincides with the presentation @xmath277 letter by letter , i.e. , @xmath281 for all @xmath188 . then @xmath83 is trivial if and only if it acts trivially on @xmath1 and every element @xmath282 for @xmath201 , @xmath283 , is trivial . again the problem reduces to the word problem in @xmath207 . by induction we conclude that the word problem is solvable in subexponential time with an upper bound @xmath284 for some constant @xmath220 . we are ready to prove the main result . [ theorem_growth_schreier_poly ] let @xmath9 be a finitely generated subgroup of @xmath98 . there exists a constant @xmath149 such that every orbital schreier graph @xmath27 for @xmath2 has subexponential growth not greater than @xmath285 . the proof goes by induction on @xmath7 . for @xmath215 the group @xmath286 is finite , and every schreier graph @xmath5 has at most @xmath287 vertices . we can also start the induction from @xmath288 using the fact that in this case the group @xmath289 is contracting ( see @xcite ) . we suppose by induction that the statement holds for the group @xmath207 and every schreier graph @xmath290 has subexponential growth not greater than @xmath291 with some constant @xmath220 . fix a sequence @xmath164 and consider the ball @xmath165 in the graph @xmath163 of radius @xmath24 centered at the vertex @xmath4 . if @xmath166 for @xmath167 and @xmath168 then @xmath169 . hence for every fixed @xmath31 each sequence in the ball @xmath165 is of the form @xmath170 for some @xmath171 and @xmath172 for some @xmath292 , where @xmath293 it follows that @xmath294 . let @xmath295 be the ball of radius @xmath24 in the group @xmath207 with respect to its generating set @xmath205 . let @xmath31 be the least integer greater than @xmath296 given in lemma [ lemma_weak_contr_poly ] . then for every @xmath297 there exists @xmath48 such that @xmath185 . hence @xmath298 where @xmath299 for @xmath48 . by induction hypothesis the size of the orbit @xmath300 is not greater than @xmath301 . let us estimate the size of the orbits @xmath302 . let @xmath48 be the first letter of the word @xmath66 and consider the following cases . if @xmath303 then for @xmath304 we have @xmath305 for some @xmath306 ; and for @xmath307 with @xmath201 , @xmath308 , we have @xmath309 for some @xmath306 and @xmath310 . hence we get estimates @xmath311 for @xmath201 , @xmath283 . if @xmath312 for @xmath313 then for @xmath304 we have @xmath314 for some @xmath315 and @xmath310 ; and for @xmath307 with @xmath201 , @xmath283 , we have @xmath316 for some @xmath306 and @xmath310 . hence we get estimates @xmath317 for some constant @xmath248 . as a corollary we recover the following result from ( * ? ? ? * corollary 4.6 ) . the orbital schreier graphs @xmath5 for @xmath2 of groups generated by polynomial automata are amenable . theorem [ theorem_main ] can be generalized to automaton groups with a certain combination of contracting and polynomial properties . [ theorem_growth_schr_general ] let @xmath17 be a finite automaton with subautomaton @xmath318 such that different cycles in @xmath319 are disjoint , and the group generated by @xmath318 is contracting . let @xmath9 be the automaton group generated by @xmath17 . there exists a constant @xmath149 such that every orbital schreier graph @xmath27 for @xmath2 has subexponential growth bounded above by @xmath285 , where @xmath7 is the maximal number of different cycles in @xmath319 connected by a directed path . the proof goes by induction on @xmath7 following the same strategy as in the proof of theorem [ theorem_growth_schreier_poly ] . first , we make a few assumptions about the generating sets to make our life easier . we can assume that the generating sets @xmath17 and @xmath318 are symmetric , and @xmath318 contains the nucleus @xmath93 of the contracting group @xmath320 . let @xmath180 for @xmath321 be the largest subautomaton of @xmath17 such that @xmath322 contains at most @xmath31 cycles connected by a directed path . then every @xmath180 is also symmetric and self - similar . we pass to a power of the alphabet so that every cycle in @xmath319 , and hence in @xmath323 , is actually a loop ; and for every @xmath324 and @xmath48 either @xmath325 or @xmath182 is circuit . let @xmath177 be the group generated by @xmath180 . in particular , @xmath326 and we use notation @xmath327 . the group @xmath328 is contracting with nucleus @xmath93 . indeed , by construction , there are no cycles in @xmath329 , and hence @xmath330 for all @xmath331 and all words @xmath66 of length greater than the diameter of @xmath329 . [ lemma_cycles_are_loops_generalized ] the length of cycles in the complete automaton @xmath86 of the group @xmath9 is bounded . take a circuit element @xmath76 . let @xmath94 for a nonempty word @xmath66 , and @xmath332 for every beginning @xmath106 of @xmath66 . over all presentation of @xmath82 as a product @xmath191 for @xmath139 with @xmath159 , we consequently choose the one with the largest number of circuit elements @xmath333 and then with the largest number of any elements @xmath333 , for @xmath321 . consider the induced presentation : @xmath334 where @xmath142 with @xmath141 . notice , that this presentation also satisfies the above assumptions , as is the induced presentation @xmath335 for every iteration @xmath336 of the word @xmath66 . in particular , every @xmath337 is actually circuit , because otherwise the presentation given by @xmath335 would contain more circuit generators than the chosen one . two consecutive elements @xmath337 and @xmath338 in the presentation can not both lie in @xmath93 . indeed , in this case @xmath339 for all sufficiently large @xmath106 . hence @xmath335 can be expressed as a product of less than @xmath24 generators for large enough iteration @xmath336 , contradicting @xmath159 . our choice of the presentation @xmath138 and the assumptions on the generating set @xmath10 force @xmath340 for every @xmath341 . hence @xmath342 ( here @xmath343 ) , where @xmath42 is the unique letter such that @xmath344 . if @xmath345 for @xmath346 then @xmath347 and we still get @xmath342 with @xmath348 . assume @xmath349 , and hence @xmath350 with @xmath351 . consider the induced presentations of @xmath352 for every @xmath353 . if @xmath341 then the @xmath188-th generator in the presentation of every @xmath352 remains the same and is equal to @xmath337 . let us trace the positions with elements from @xmath93 . if @xmath345 then the @xmath188-th element in the presentation of @xmath352 is equal to @xmath354 . hence these @xmath188-th elements change according to the vertices of the path in the nucleus @xmath93 that starts at the state @xmath337 and goes along arrows with left label @xmath197 . it follows that every position is pre - periodic . we can eliminate the pre - periods by passing from the presentation @xmath355 to the induced presentation @xmath356 for large enough @xmath357 , and hence we assume that every position repeats periodically . if @xmath357 is a multiple of the lengths of every cycle in @xmath93 , then the induced presentation of @xmath352 coincides with @xmath191 letter by letter , and hence @xmath356 . it follows that the length of the word @xmath66 is bounded by the least common multiple of the length of cycles in the nucleus @xmath93 . if @xmath358 then @xmath359 , @xmath360 , and we can apply the same arguments . let @xmath361 be the upper bound on the length of cycles in @xmath86 . just to notice , in general , we can not pass to a power of the alphabet to transform every cycle into a loop , i.e. , to make @xmath362 , as was possible in the case of polynomial automata . further , we will use the fact proved in lemma [ lemma_cycles_are_loops_generalized ] , that if @xmath363 for @xmath364 is circuit and @xmath94 then @xmath365 for every @xmath77 , @xmath366 , and the length of @xmath367 with respect to @xmath368 is not greater than the length of @xmath82 with respect to @xmath180 . the formulation of lemma 1 remains the same ( with @xmath369 ) . [ lemma_weak_contr_poly_generalized ] there exists a constant @xmath149 such that for every @xmath259 for @xmath139 and every word @xmath64 of length @xmath209 the element @xmath83 is either circuit , or @xmath224 . the proof is basically the same as the one of lemma 1. let us indicate only the main argument . we assume the lemma holds for the groups @xmath218 for @xmath370 with some common constant @xmath220 . let @xmath76 be a circuit element . consider the product @xmath226 for @xmath253 and its restrictions @xmath229 for words @xmath64 of length @xmath371 . we have the following cases . if @xmath224 then @xmath232 . so further assume @xmath372 , and hence @xmath83 and @xmath82 lie on the same cycle . there exists a nonempty word @xmath106 of length @xmath373 such that @xmath374 . let us look at @xmath234 , which either belongs to @xmath235 or is circuit . if @xmath236 then @xmath375 for circuit @xmath376 and @xmath377 . otherwise @xmath234 is circuit , and let @xmath378 be the length of the cycle at @xmath234 . then the length of the word @xmath379 is a multiple of @xmath380 and of @xmath357 . if @xmath381 then @xmath382 and the state @xmath240 is circuit ; otherwise @xmath383 . hence in all cases the state @xmath384 for a word @xmath4 of length @xmath385 is either circuit , or it belongs to @xmath207 , or it is of the form @xmath386 for circuit @xmath387 and @xmath238 with @xmath388 . eventually , in the worst case , we will need to consider the product @xmath226 for circuit @xmath82 and @xmath389 . in this case we can apply the same arguments as in the proof of lemma [ lemma_cycles_are_loops_generalized ] . the rest of the proof is the same . now we can return to the growth of schreier graphs . if @xmath288 then the group @xmath9 is contracting and the statement holds . we suppose by induction that the statement holds for the group @xmath207 with some constant @xmath220 . take a sequence @xmath164 and consider the schreier graph @xmath163 . we will use the estimate . let @xmath295 be the ball of radius @xmath24 in the group @xmath207 with respect to its generating set @xmath205 . let @xmath31 be the least integer greater than @xmath296 from lemma [ lemma_weak_contr_poly_generalized ] . then for every @xmath297 there exists a nonempty word @xmath106 of length @xmath373 such that @xmath390 . hence @xmath391 where @xmath392 . by induction hypothesis the size of the orbit @xmath300 is not greater than @xmath301 . let us estimate the size of the orbits @xmath393 . if @xmath394 then for every @xmath395 we get @xmath305 for some @xmath396 . if @xmath397 for @xmath398 , @xmath399 , and @xmath400 , then for every @xmath395 we get @xmath314 for some @xmath401 and @xmath310 . summarizing all estimates we get @xmath402 for some constant @xmath248 . omega - periodic graphs. let @xmath403 be the binary alphabet . for every finite word @xmath404 over @xmath1 consider the automaton @xmath405 shown in figure [ fig_automatonomega ] , where we use notation @xmath406 and @xmath407 . every automaton @xmath405 is polynomial of degree @xmath408 . the automaton @xmath405 is a subautomaton of @xmath409 , and we get an increasing chain of polynomial automata @xmath410 for every @xmath42 . hence every orbital schreier graph @xmath411 for @xmath2 is a subgraph of @xmath412 . this allows us to construct the schreier graph @xmath411 consequently as @xmath413 , by looking at the action of the states @xmath414 . we start from the orbital schreier graphs @xmath415 of the automaton @xmath416 . the transformation @xmath417 is called the adding machine , because its action on a sequence @xmath418 corresponds to the addition of @xmath419 to the binary number @xmath420 . in particular , the infinite cyclic group generated by @xmath416 acts faithfully on every its orbit on @xmath0 . hence every schreier graph @xmath415 for @xmath2 is a line , i.e. , its vertices can be identified with @xmath421 via the map @xmath422 , and the edges are @xmath423 for all @xmath424 . notice that for every state @xmath425 of the automaton @xmath405 if @xmath426 , for @xmath427 , and we take the first position @xmath24 with @xmath428 , then @xmath429 and @xmath430 . it follows that each @xmath425 preserves the orbits of the action of @xmath417 on @xmath0 . hence the orbits of @xmath431 for every word @xmath66 coincide with the orbits of @xmath432 , and we can identify the vertex set of every schreier graph @xmath411 for @xmath2 with the set @xmath421 . every state @xmath425 of the automaton @xmath433 acts on the infinite sequences as follows @xmath434 for every @xmath2 and @xmath435 . notice that every sequence from @xmath0 appears in for suitable numbers @xmath436 . further , for every fixed @xmath437 , the edges defined by , when @xmath4 runs through @xmath0 , are periodic , namely , these edges are invariant under the shift @xmath438 on @xmath421 . it follows that all schreier graphs @xmath411 for @xmath2 are unions of periodic subgraphs on @xmath421 , and hence they are @xmath8-periodic in the terminology of @xcite . the schreier graph @xmath439 ( see figure [ fig_omegagraph ] ) was defined in @xcite as an example of a graph of intermediate growth ( its growth is equivalent to @xmath440 ) connected to the long range percolation on @xmath421 . all orbital schreier graphs @xmath441 for @xmath2 of the automaton @xmath442 were studied in @xcite , where it was proved that the family of these graphs contains uncountably many pairwise nonisomorphic graphs , all of them except @xmath443 are locally isomorphic , and all have intermediate growth . the groups generated by @xmath442 and @xmath444 are the same , and the automata @xmath445 are just different generating sets of this group . hence the schreier graphs @xmath446 also have intermediate growth . the schreier graphs @xmath411 for every nonempty word @xmath66 have growth not less than the growth of @xmath441 or of @xmath446 . hence their growth is superpolynomial , and then theorem [ theorem_main ] implies that all orbital schreier graphs @xmath411 for @xmath2 and @xmath447 have intermediate growth . actually , if we take the automaton shown in figure [ fig_automatonomega ] and change the labels of the edges passing from the states @xmath448 in any way we want , we still get the automaton whose all schreier graphs @xmath5 have intermediate growth . more generally , if we take any polynomial automaton @xmath17 over @xmath1 that contains @xmath449 ( with any labels at @xmath450 ) as a subautomaton , then all orbital schreier graphs @xmath451 have intermediate growth . nonpolynomial example . * consider the automaton @xmath318 over the alphabet @xmath403 shown in figure [ fig_automnonpoly ] . the automaton @xmath318 is not polynomial , but satisfies the conditions of theorem [ theorem_growth_schr_general ] . the group @xmath452 is the infinite cyclic group , which coincides with the automaton group generated by the subautomaton of @xmath318 with states @xmath453 . the group @xmath454 is contracting with nucleus @xmath455 , while the group @xmath9 generated by @xmath318 is not contracting . theorem [ theorem_growth_schr_general ] implies that all orbital schreier graphs @xmath456 for @xmath2 have subexponential growth . to obtain a lower bound we can use the previous example . we notice that the group @xmath454 acts faithfully on every its orbit on @xmath0 . the states @xmath457 act similarly to the states @xmath458 , namely , @xmath459 and @xmath460 for @xmath2 and @xmath461 . it follows that the schreier graph @xmath441 can be seen as a subgraph of the schreier graph @xmath456 via the map @xmath462 for every @xmath2 . hence all orbital schreier graphs @xmath456 have intermediate growth . * hanoi graphs. the hanoi tower game on @xmath31 pegs is played with @xmath24 disks of distinct size placed on @xmath31 pegs , @xmath463 . initially , all disks are placed on the first peg according to their size so that the smallest disk is at the top , and the largest disk is at the bottom . a player can move only one top disk at a time from one peg to another peg , and can never place a bigger disk over a smaller disk . the goal of the game is to transfer the disks from the first peg to another peg . more generally , one can take any two configurations of disks and ask a player to transform one configuration to another one , where a configuration is any arrangement of @xmath24 disks on @xmath31 pegs such that a bigger disk does not lie on a smaller disk . for more information about this game , its history , solutions , and open problems , we refer the reader to @xcite and the references therein . the hanoi tower game is modeled by the hanoi graphs @xmath464 . the vertices of the graph @xmath464 are the configurations of @xmath24 disks on @xmath31 pegs , and two configurations are adjacent if one can pass from one to another by a single disk move . then the hanoi tower game is equivalent to finding a path in the graph @xmath464 between two given vertices . in particular , the diameter of the hanoi graph gives an upper bound on the optimal number of steps in the hanoi tower game . it was noticed in @xcite that the game can be also modeled by a finite automaton @xmath465 over the alphabet @xmath466 , called the hanoi towers automaton on @xmath31 pegs . in this model , the pegs are identified with the letters of the alphabet @xmath1 , and the disks are labeled by @xmath467 according to their size . then the configurations of @xmath24 disks on @xmath31 pegs can be encoded by words of length @xmath24 over @xmath1 , where the word @xmath62 corresponds to the unique configuration in which the disk @xmath188 is placed on the peg @xmath197 . the automaton @xmath465 has the trivial state @xmath95 and the state @xmath468 for every transposition @xmath469 on @xmath1 with two arrows @xmath470 labeled by @xmath471 and @xmath472 , and the loop at @xmath468 labeled by @xmath473 for every @xmath474 . for example , the automaton @xmath475 is shown in figure [ fig_automhanoi ] ( the loops at the trivial state @xmath95 are not drawn ) . the action of the state @xmath468 on a word of length @xmath24 corresponds to a single disk move between the pegs @xmath188 and @xmath476 . hence the schreier graph @xmath477 is precisely the hanoi graph @xmath464 . the complexity of the hanoi tower game highly depends on whether @xmath478 or @xmath34 . in the case @xmath478 , the hanoi graph @xmath479 has diameter @xmath480 , and this is the smallest number of steps to win the game . the automaton @xmath481 is bounded and generates a contracting group . all orbital schreier graphs @xmath482 for @xmath2 have polynomial growth of degree @xmath483 . for @xmath34 , the hanoi tower game can be solved in subexponential number of moves , asymptotically in @xmath484 moves ( see @xcite ) . in particular , the diameter of the hanoi graph @xmath464 and the schreier graph @xmath477 is asymptotically @xmath484 . one can apply this asymptotic estimates to the orbital schreier graphs @xmath485 for @xmath2 . it is easy to see that the growth function @xmath129 of each graph @xmath485 satisfies @xmath486 , where @xmath487 is the diameter of the graph @xmath477 . hence the upper bound on the diameter of @xmath477 implies a superpolynomial lower bound on the growth of @xmath485 , namely @xmath488 with some constant @xmath119 . this was used in ( ? ? ? * theorem 2.1 ) to conclude that the schreier graph @xmath489 for @xmath34 has intermediate growth between @xmath490 and @xmath491 for some constants @xmath492 . the automaton @xmath465 for @xmath34 is not polynomial , and does not satisfy the assumptions of theorem [ theorem_growth_schr_general ] . however , we can apply similar ideas to give the subexponential upper bound @xmath493 on the growth of every schreier graph @xmath485 for @xmath2 . to provide an upper bound we follow similar arguments as in the proofs on theorems [ theorem_growth_schreier_poly ] and [ theorem_growth_schr_general ] . we will show only the case @xmath494 , the general case is analogous . let @xmath32 be the automaton group generated by @xmath465 . notice that for every state @xmath56 of the automaton @xmath465 and every word @xmath64 if @xmath495 then @xmath496 and @xmath497 . it immediately follows that every cycle in the complete automaton @xmath498 of the group @xmath32 is a loop labeled by @xmath473 for some letter @xmath48 . the group @xmath499 is contracting with nucleus @xmath500 . then for every @xmath501 , @xmath502 , the restriction @xmath83 belongs to @xmath93 for all words @xmath66 of length @xmath503 . hence every schreier graph @xmath482 has subexponential ( polynomial ) growth not greater than @xmath504 and the automaton @xmath475 . for @xmath48 denote by @xmath505 the subautomaton of @xmath475 consisting of states that fix the letter @xmath13 . the group generated by @xmath505 acts on the words over @xmath506 in the same way as the group @xmath499 acts on @xmath507 , and that is where we can apply the inductive arguments . in particular , if @xmath138 for @xmath508 then for every word @xmath64 that contains at least @xmath154 letters different from @xmath13 the restriction @xmath83 belongs to @xmath475 , in other words , @xmath83 fixes one more letter except @xmath13 . take two different letters @xmath509 and consider elements @xmath221 for @xmath510 and @xmath511 for @xmath512 . take an arbitrary word @xmath64 of length @xmath513 with @xmath514 . if the word @xmath66 contains at least @xmath154 letters not equal to @xmath13 then @xmath515 . otherwise , @xmath66 contains at least @xmath154 letters equal to @xmath13 , and then @xmath516 contains at least @xmath154 letters equal to @xmath13 . since @xmath517 , we get @xmath518 . it follows that for every word of length @xmath519 there exists a letter @xmath306 such that every generator in the induced presentation of @xmath520 belongs to @xmath521 ( i.e. , @xmath240 is circuit ) . let us prove that there exists a constant @xmath248 such that for every product @xmath522 , @xmath523 , and every word @xmath64 of length @xmath524 there exists @xmath48 such that every generator @xmath212 in the induced presentation @xmath525 belongs to @xmath505 ( hence @xmath83 is circuit , here @xmath213 ) . we partition the presentation @xmath138 on blocks @xmath526 such that every @xmath527 contains only generators from @xmath528 for some letter @xmath48 ( depending on @xmath188 ) . consider the products @xmath529 , @xmath530 , , and their restrictions on words @xmath66 of length @xmath531 . using the above property we get that the element @xmath83 can be expressed as a product with @xmath532 elements @xmath254 . applying the same procedure @xmath533 times , we get an element with needed properties . the existence of the constant @xmath248 follows . consider the schreier graph @xmath534 for @xmath164 . we will use estimate with @xmath31 being the least integer greater than @xmath535 . then @xmath536 where @xmath537 consists of those elements @xmath538 that can be expressed as a product of no more than @xmath24 elements from @xmath505 . every element @xmath304 fixes every occurrence of the letter @xmath13 in the sequence @xmath539 , and changes every other letter in the same way as the group @xmath499 acts on @xmath506 . hence the orbit @xmath540 has subexponential growth not greater than @xmath541 . we get the upper bound @xmath542 for some constant @xmath149 . * weakly contracting groups . * one of the main properties used in the proof of theorems [ theorem_main ] and [ theorem_growth_schr_general ] leads us to the following definition . we say that a group @xmath9 generated by a finite automaton is _ weakly contracting _ if the length of cycles in the complete automaton @xmath86 of the group is bounded . every finitely generated contracting group is weakly contracting . the groups generated by polynomial automata , the groups from theorem [ theorem_growth_schr_general ] , and the hanoi towers groups are also weakly contracting . it is natural to ask which properties can be generalized to this class of groups . is the word problem in weakly contracting groups solvable in subexponential time ? do weakly contracting groups not contain nonabelian free subgroups ? do the orbital schreier graphs @xmath5 of weakly contracting groups have subexponential growth ?
the ability to simulate hydrodynamic flow is key to studying most astrophysical objects . supernova explosions , gamma - ray bursts , x - ray bursts , classical novae , the outbursts of luminous blue variables ( lbvs ) , and stellar winds are just a few phenomena for which understanding and numerical tools evolve in tandem . this is due in part to the physical complexity , multi - dimensional character , and instabilities of such dynamical fluids . moreover , rotation is frequently a factor in the dynamical development of astrophysical transients . one - dimensional hydrodynamics is by and large a solved problem , but multi - dimensional hydrodynamics is still a challenge . in the context of astrophysical theory , this is due primarily to the need to address time - dependent gravitational potentials , complicated equations of state ( eoss ) , flexible grids , multi - d shock structures , and chaotic and turbulent flows . as a result , theorists who aim to understand the universe devote much of their time to code development and testing . one of the outstanding and complex problems in theoretical astrophysics is the mechanism for core - collapse supernovae . for more than two decades , the preferred mechanism of explosion has been the delayed neutrino mechanism @xcite . one - dimensional ( 1d ) simulations generally fail to produce explosions @xcite . however , 2-dimensional ( 2d ) simulations , and the accompanying aspherical instabilities , suggest that the neutrino - mechanism may indeed be viable @xcite , though this has yet to be proven . in fact , @xcite have recently reported an acoustic mechanism , which seems to succeed on long timescales when and if other mechanisms fail . these authors identified two primary reasons why this mechanism might have been missed before : 1 ) 2d radiation - hydrodynamic simulations with reasonable approximations are still expensive to run , and with limited resources simulations are rarely carried to late enough times ; and 2 ) , a noteworthy feature of the code they used , vulcan/2d , is its arbitrary lagrangian - eulerian ( ale ) structure . vulcan/2d incorporates non - standard grids that liberate the inner core from the courant and resolution limitations of standard spherical grids . in this context , @xcite claim that simulating all degrees of freedom leads to a new mechanism in which the gravitational energy in aspherical accretion is converted to explosion energy by first exciting protoneutron star core g - modes . these modes then radiate acoustic power and revive the stalled shock into explosion . remarkably , the acoustic mechanism , given enough time , leads to successful core - collapse supernovae in all progenitors more massive than @xmath19 m@xmath2 simulated to date @xcite . however , this is a remarkable claim , and given the implications , must be thoroughly investigated . for instance , one question to ask is : could the results seen by @xcite be numerical artifacts of vulcan ? therefore , to address this and other issues that surround the acoustic supernova mechanism , as well as a host of other outstanding astrophysical puzzles , we have designed the new hydrodynamic code , bethe - hydro , from the bottom up . bethe - hydro will be coupled with the 2d mixed - frame radiation transport scheme of @xcite to create the 2d radiation / hydrodynamics code bethe , and the merger of these two codes will be the subject of a future paper . in this paper , we present and test the hydrodynamic and gravity algorithms of bethe - hydro . since bethe - hydro uses arbitrary grids and a general gravity solver , we expect it to be a powerful and flexible numerical tool , able to configure the grid to suit the computational challenge . the core of bethe - hydro is a 1d and 2d ale hydrodynamics solver . first , solutions to the lagrangian hydrodynamic equations are obtained on an arbitrary polygonal grid . then , to avoid tangled grids , the hydrodynamic variables can be remapped to another grid . a unique and powerful feature of ale schemes is their flexibility to tailor an arbitrary grid to the computational challenge and to alter the grid dynamically during runtime . hence , purely lagrangian , purely eulerian , or arbitrarily moving grids ( chosen to optimize numerical performance and resolution ) are possible . these grids can be non - orthogonal , non - spherical , and non - cartesian . some of the earliest 2d hydrodynamic simulations in astrophysics were , in fact , performed to address the core - collapse problem . while some employed standard fixed - grid schemes @xcite , other 2d simulations were calculated with adaptive grids . although many utilized moving grids , for the most part , the differencing formulations were eulerian @xcite . on the other hand , @xcite did employ an `` euler - lagrange '' method involving a lagrangian hydrodynamic solve , followed by a remapping stage . even though these simulations did exploit radially dynamical grids , they were restricted to be orthogonal and spherical . many of the current hydrodynamic algorithms used in astrophysics are based upon either zeus @xcite or higher - order godunov methods , in particular the piecewise - parabolic method ( ppm ) @xcite . both approaches have been limited to orthogonal grids . one concern with ppm - based codes is that they solve the hydrodynamic equations in dimensionally - split fashion . as a result , there have been concerns that these algorithms do not adequately resolve flows that are oblique to the grid orientation . rectifying this concern , recent unsplit higher - order godunov schemes , or approximations thereof , have been developed @xcite . despite this and the employment of adaptive mesh refinement ( amr ) , these codes must use orthogonal grids that are often strictly cartesian . in bethe - hydro , we use finite - difference schemes based upon the support - operator method @xcite of @xcite and @xcite . differencing by this technique enables conservation of energy to roundoff error in the absence of rotation and time - varying gravitational potentials . similarly , momentum is conserved accurately , but due to the artificial viscosity scheme that we employ , it is conserved to roundoff error for hydrodynamic simulations using cartesian coordinates only . moreover , using the support - operator method and borrowing from an adaptation of the support - operator technique for elliptic equations @xcite , we have developed a gravity solver for arbitrary grids . unfortunately , by including a general gravity capability , we sacrifice strict energy and momentum conservation . however , we have performed tests and for most cases the results are reasonably accurate . furthermore , we have included rotational terms and we use a modified version of the subcell pressure scheme to mitigate hourglass instabilities @xcite . to resolve shocks , we employ an artificial viscosity method , which is designed to maintain grid stability as well @xcite , and there are no restrictive assumptions made about the equation of state . higher - order godunov techniques employ riemann solvers to resolve shocks , but frequently need an artificial viscosity scheme to eliminate unwanted post - shock ringing . moreover , the inner workings of riemann solvers often make local approximations that the eos has a gamma - law form , which stipulates that as the internal energy goes to zero so does the pressure . for equations of state appropriate for core - collapse supernovae , this artificially imposed zero - point energy can pose problems for the simulations @xcite . two codes in astrophysics which have already capitalized on the arbitrary grid formulations of ale are djehuty @xcite and vulcan/2d @xcite . in both cases , the grids employed were designed to be spherical in the outer regions , but to transition smoothly to a cylindrical grid near the center . these grid geometries reflected the basic structure of stars , while avoiding the cumbersome singularity of a spherical grid . with this philosophy , @xcite , using djehuty , have studied the helium core flash phase of stellar evolution in 3d . in the core - collapse context , @xcite , using vulcan/2d , performed 2d radiation / hydrodynamic and radiation / mhd simulations with rotation . several gravity solvers have been employed in astrophysics . the most trivial are static or monopole approaches . for arbitrary potentials , the most extensively used are n - body schemes , which fit most naturally in smooth particle hydrodynamics ( sph ) codes @xcite . as a virtue , sph solves the equations in a grid - free context , and while sph has opened the way to three - dimensional ( 3d ) hydrodynamic simulations in astrophysics , including core - collapse simulations @xcite , the smoothing kernels currently employed pose serious problems for simulating fundamental hydrodynamic instabilities such as the kelvin - helmholtz instability @xcite . another approach to the solution of poisson s equation for gravity is the use of multipole expansions of the potential @xcite . this technique achieves its relative speed by calculating simple integrals on a spherical grid once . then , the stored integrals are used in subsequent timesteps . in bethe - hydro , we construct finite - difference equations for poisson s equation using the support - operator method and find solutions to the resulting linear system of equations for the potentials via an iterative multigrid - preconditioned conjugate - gradient method @xcite . in [ section : bethe_hydro ] , we give a summary of bethe - hydro and sketch the flowchart of the algorithm . the coordinates and mesh details are discussed in [ section : coords_mesh ] . we then describe in [ section : lag_hydro ] the discrete lagrangian equations , including specifics of the 2@xmath0-order time - integration and rotational terms . section [ section : gravity ] gives a complete description of the 1d and 2d gravity solvers . hydrodynamic boundary conditions are discussed in [ section : hydro_boundary ] . the artificial viscosity algorithm that provides shock resolution and grid stability is described in [ section : artificial_viscosity ] . in [ section : hourglass ] , we present the subcell pressure scheme that suppresses hourglass modes . remapping is described in depth in [ section : remap ] . in [ section : tests ] , we demonstrate the code s strengths and limitations with some test problems . finally , in [ section : discussion ] we summarize the central characteristics and advantages of bethe - hydro . in ale algorithms , the equations of hydrodynamics are solved in lagrangian form . within this framework , equations for the conservation of mass , momentum , and energy are : @xmath3 @xmath4 and @xmath5 @xmath6 is the mass density ( which we refer to simply as `` density '' and is distinct from the energy or momentum densities ) , @xmath7 is the fluid velocity , @xmath8 is the gravitational potential , @xmath9 is the isotropic pressure , @xmath10 is the specific internal energy , and @xmath11 is the lagrangian time derivative . the equation of state may have the following general form : @xmath12 where @xmath13 denotes the mass fraction of species @xmath14 . therefore , we also solve the conservation equations : @xmath15 completing the set of equations for self - gravitating astrophysical flows is poisson s equation for gravity : @xmath16 where @xmath17 is newton s gravitational constant . all ale methods have the potential to solve eqs . ( [ eq : mass_lag]-[eq : ene_lag ] ) using arbitrary , unstructured grids . in this lies the power and functionality of ale methods . the solutions involve two steps and are conceptually quite simple : 1 ) a lagrangian solver moves the nodes of the mesh in response to the hydrodynamic forces ; and 2 ) to avoid grid tangling , the nodes are repositioned , and a remapping algorithm interpolates hydrodynamic quantities from the old grid to this new grid . of course , the challenge is to find accurate solutions , while conserving energy and momentum . in constructing bethe - hydro to satisfy these requirements , we use the ale hydrodynamic techniques of @xcite , @xcite , @xcite , and @xcite . to summarize the overall structure of bethe - hydro , we present a schematic flowchart , fig . [ flowchart ] , and briefly describe the key steps . establishing the structure for all subsequent routines , the first step is to construct the arbitrary , unstructured grid . this leads to the next step , problem initialization . then , the main loop for timestep integration is entered . after a call to the eos to obtain the pressure , solutions to poisson s equation for gravity are calculated using either the 2d or 1d algorithms in [ section : gravity ] . both use the support - operator method to discretize eq . ( [ eq : poisson ] ) , and the resulting system of linear equations is solved by a tridiagonal solver in 1d and a multigrid pre - conditioned conjugate - gradient iterative method in 2d ( [ section : gravity ] ) . after the timestep is calculated ( [ section : predictor_corrector ] ) , the lagrangian equations of hydrodynamics are solved on an arbitrary grid using the compatible hydrodynamics algorithm of @xcite ( see [ section : lag_hydro ] for further discussion ) . to ensure 2nd - order accuracy in both space and time , we employ a predictor - corrector iteration ( [ section : predictor_corrector ] ) , in which a second call to the eos and the gravity solver are required . other than the gravity solver block , the lagrangian hydrodynamics solver ( [ section : lag_hydro ] ) is represented by the set of steps beginning with predictor and ending with corrector ( fig . [ flowchart ] ) . after finding lagrangian hydrodynamic solutions at each timestep , one could proceed directly to the next timestep , maintaining the lagrangian character throughout the simulation . however , in multiple spatial dimensions , grid tangling will corrupt flows with significant vorticity . in these circumstances , we conservatively remap the lagrangian solution to a new arbitrary grid that mitigates tangled grids . the new grid may be the original grid , in which case we are solving the hydrodynamic equations in the eulerian frame , or it may be a new , arbitrarily - defined , grid . in bethe - hydro , eqs . ( [ eq : mass_lag]-[eq : poisson ] ) may be solved with any of several coordinate systems and assumed symmetries . included are the standard 1d & 2d cartesian coordinate systems . for astrophysical applications , we use 1d spherical and 2d cylindrical coordinate systems . irrespective of the coordinate system , we indicate a position by @xmath18 . the components of the cartesian coordinate system are @xmath19 and @xmath20 ; the spherical components are @xmath21 , @xmath22 , and @xmath23 ; and the cylindrical coordinates are @xmath21 , @xmath24 , and @xmath23 . distinct from the coordinate system is the grid , or mesh , which defines an arbitrary discretization of space . it is this unique feature and foundation of ale techniques that provides their flexibility . figure [ grid_ex ] shows two example grids : the butterfly mesh on the left and a spiderweb mesh on the right . either may be used in a 2d cartesian or 2d cylindrical simulations , although the placement of the nodes is neither cartesian nor cylindrical . instead , these meshes have been designed to simulate spherically or cylindrically convergent phenomena without the limitation of small zones near the center . when using cylindrical coordinates in ale algorithms , one may use control - volume differencing ( cvd ) or area - weighted differencing ( awd ) ; we use cvd . thorough comparisons of these two differencing schemes may be found in @xcite . here , we give a basic justification for choosing cvd . for cvd , the volumes of the subcell , cell , and node are calculated by straightforward partitioning of these regions by edges . hence , volumes and masses are exact representations of their respective regions . while this discretization is natural and easy to comprehend , it does not preserve strict spherical symmetry when using a spherical grid and cylindrical coordinates . on the other hand , awd is designed to preserve spherically symmetric flows , but only with the use of spherical grids that have equal spacing in angle . since this prevents the use of arbitrary grids , and the asymmetries of cvd are small , we use cvd for discretization . construction of the grid begins with the arbitrary placement of nodes . at these nodes , coordinate positions , velocities , and accelerations are defined . simply specifying node locations does not completely define a mesh , since there is not a unique way to assign cells and masses to these nodes . therefore , the user must specify the connectivity among nodes , which define the arbitrarily - shaped polygonal cells . it is within these cells that cell - centered averages of @xmath6 , @xmath10 , @xmath9 , @xmath13 , and @xmath8 are defined . with the node positions , their connectivity , and the cells defined , a mesh is completely specified and all other useful descriptions follow . cells are denoted by @xmath24 , and nodes are indexed by @xmath25 . the set of nodes that defines cell @xmath24 are @xmath26 , where the nodes are ordered counterclockwise . conversely , the set of cells that shares node @xmath25 is denoted by @xmath27 . each cell has @xmath28 nodes that define it , and each node has @xmath29 cells that share it . the sample sub - grid depicted in fig . [ fig : grid ] helps to illustrate the nodal and cell structure . the filled circles indicate node positions , @xmath30 , and the crosses indicate the cell - center positions , @xmath31 . the solid lines are direction - oriented edges that separate cells from one another , and the open circles indicate their mid - edge locations . partitioning the cell into subcells , the dashed lines connect the mid - edges with the cell centers . no matter how many sides a cell has , with this particular division the subcells are always quadrilaterals . naturally , the cell volumes ( see appendix [ section : integrals ] for formulae calculating discrete volumes ) , @xmath32 , are related to the subcell volumes , @xmath33 , by @xmath34 furthermore , each node has a volume , @xmath35 , defined by the adjoining subcells that share the node @xmath25 : @xmath36 for calculating pressure forces and fluxes , vectors are assigned to each half edge on either side of node @xmath25 : @xmath37 and @xmath38 , where @xmath39 indicates the half edge in the counterclockwise direction around the cell and @xmath40 indicates its opposite counterpart . their magnitudes are the areas represented by the half edges , and their directions point outward and normal to the surface of cell @xmath24 . from these half - edge area vectors , an area vector , @xmath41 , that is associated with zone @xmath24 and node @xmath25 is then defined : @xmath42 similarly , vectors are associated with the lines connecting the mid - edges and the cell centers : @xmath43 . again , their magnitudes are the corresponding area , but while the vector is normal to this line , the direction is oriented counterclockwise around the cell . the fundamental assumption of lagrangian algorithms is that the mass , @xmath44 , of a discrete volume @xmath32 is constant with time . for staggered - grid methods , in which scalars are defined as cell - centered averages and vectors are defined at the nodes , it is necessary to define a lagrangian mass , @xmath45 , for the nodes as well . this nodal mass is associated with the node s volume @xmath35 ( eq . [ eq : volume_node ] ) . conservation of mass ( eq . [ eq : mass_lag ] ) implies zero mass flux across the boundaries , @xmath46 , of either the cell volume or nodal volume . therefore , the region of overlap for @xmath32 and @xmath35 , which is the subcell volume @xmath33 , is bounded by surfaces with zero mass flux . consequently , the most elemental lagrangian mass is the subcell mass , @xmath47 , and the cell mass ( @xmath44 ) and node mass ( @xmath45 ) are constructed as appropriate sums of subcell masses : @xmath48 and @xmath49 hence , we arrive at three discrete forms of mass conservation ( eq . [ eq : mass_lag ] ) : @xmath50 in defining the discrete momentum and energy equations , we use the compatible hydrodynamics algorithms developed by @xcite . specifically , the discrete divergence and gradient operators are _ compatible _ in that they faithfully represent their analog in continuous space _ and _ their definitions are expressly related to one another using the hydrodynamic expressions for conservation of momentum and energy . as a result , this approach leads to discretizations that satisfy momentum and energy conservation to machine accuracy . this is accomplished with the support - operator method @xcite . given the integral identity , @xmath51 where @xmath8 is any scalar and @xmath52 is some vector , there is an incontrovertible connection between the divergence and gradient operators . for many choices of discretization , the discrete counterparts of these operators could violate this integral identity . simply put , the goal of the support - operator method is to define the discrete operators so that they satisfy eq . ( [ eq : intidentity ] ) . the first step is to define one of the discrete operators . it is often , but not necessary , that the discrete divergence operator is defined via gauss s law : @xmath53 and then eq . ( [ eq : intidentity ] ) is used to compatibly define the other discrete operator . discretizing the hydrodynamic equations , @xcite begin by defining @xmath54 at the nodes and use the integral form of energy conservation and an equation equivalent to eq . ( [ eq : intidentity ] ) to define for each cell the discrete divergence of the velocity , @xmath55 . to begin , @xcite integrate eq . ( [ eq : mom_lag ] ) ( excluding gravity and rotational terms ) over the volume of node @xmath25 , producing the discrete form for @xmath56 and the momentum equation : @xmath57 where @xmath58 is the change in velocity from timestep @xmath59 to the next timestep @xmath60 , the timestep is @xmath61 , and @xmath62 is the pressure in cell @xmath24 . in other words , the net force on node @xmath25 is a sum of the pressure times the directed zone areas that share node @xmath25 . hence , the subcell force exerted by zone @xmath24 on point @xmath25 is @xmath63 . a more complete description of the subcell force , however , must account for artificial viscosity ( [ section : artificial_viscosity ] ) : @xmath64 where @xmath65 is the subcell force due to artificial viscosity . furthermore , in this work , we include gravity and rotation for 2d axisymmetric simulations , and the full discrete momentum equation becomes @xmath66 where @xmath67 and @xmath68 are the gravitational and rotational accelerations , respectively . for simplicity , and to parallel the discussion in @xcite , we ignore these terms in the momentum equation and proceed with the compatible construction of the energy equation ( see [ section : consenerot ] and [ section : consenegravity ] for discussions of total energy conservation including rotation and gravity , respectively ) . to construct the discrete energy equation , @xcite integrate eq . ( [ eq : ene_lag ] ) over the discrete volume of cell @xmath24 : @xmath69 where @xmath70 . then , the objective is to determine a discrete form for the right hand side of this equation that conserves energy and makes the discrete gradient and divergence operators compatible . @xcite accomplish this with the integral for conservation of energy ( neglecting gravity ) : @xmath71 where the second expression is the integral identity , eq . ( [ eq : intidentity ] ) , that defines the physical relationship between the gradient and divergence operators . neglecting boundary terms , the discrete form of this integral is @xmath72 where we used @xmath73 , @xmath74 , and substituted in eq . ( [ eq : mom_discrete1 ] ) . if we set the expression for each zone in eq . ( [ eq : enecons_discrete ] ) to zero , we arrive at the compatible energy equation : @xmath75 the rhs of eq . ( [ eq : ene_discrete ] ) is merely the @xmath76 work term of the first law of thermodynamics . its unconventional form is a consequence of the support - operator method . thus , we derive two significant results . first , inspection of eqs . ( [ eq : ene_discrete ] ) and ( [ eq : mom_discrete1 ] ) leads to compatible definitions for the discrete gradient and divergence operators . specifically , the discrete analogs of the gradient and divergence operators are @xmath77 and @xmath78 second , eqs . ( [ eq : mass_discrete ] ) , ( [ eq : mom_discrete1 ] ) , and ( [ eq : ene_discrete ] ) form the compatible discrete equations of lagrangian hydrodynamics . in its current form , bethe - hydro strictly conserves momentum for simulations that use cartesian coordinates and not cylindrical coordinates . @xcite show that the requirement for strict conservation is the use of control - volume differencing , or that exact representations of cell surfaces are employed . as a simple example , consider momentum conservation with pressure forces only : @xmath79 for cartesian coordinates , the sum @xmath80 is exactly zero for each cell as long as the surface - area vectors are exact representations of the cell s surface . this is not the case for area - weighted differencing , and hence momentum is not conserved in that case . for cylindrical coordinates , the sum of surface - area vectors gives zero only for the @xmath24-component . however , the assumption of symmetry about the cylindrical axis ensures momentum conservation for the @xmath21-component . therefore , control - volume differencing ensures momentum conservation , even with the use of cylindrical coordinates . based upon similar arguments , the subcell - pressure forces that eliminate hourglass motions ( [ section : hourglass ] ) are formulated to conserve momentum for both cartesian and cylindrical coordinates . on the other hand , because we multiply the cartesian artificial viscosity force by @xmath81 to obtain the force appropriate for cylindrical coordinates , the artificial viscosity scheme ( [ section : artificial_viscosity ] ) that we employ is conservative for cartesian coordinates only . we test momentum conservation with the sedov blast wave problem ( [ section : sedov ] ) using cartesian and cylindrical coordinates . symmetry dictates conservation in the @xmath21-component of the momentum . consequently , we compare the @xmath24-component of the total momentum at @xmath82 s with subsequent times . since the initial momentum of this test is zero , we obtain a relative error in momentum by @xmath83 , where @xmath84 is the @xmath24-component of the total momentum at time @xmath85 and @xmath86 is a similar quantity for @xmath87 . as expected , the relative error using cartesian coordinates reflects roundoff error . for cylindrical coordinates and 35,000 cells , the relative error in momentum is @xmath88 after 0.8 s and 71,631 timesteps . although momentum is strictly conserved for simulations using cartesian coordinates only , the sedov blast wave problem demonstrates that momentum is conserved very accurately , even when cylindrical coordinates are used . our method is explicit in time . therefore , for numerical convergence , we limit the timestep , @xmath61 , to be smaller than three timescales . by the courant - friedrichs - levy ( cfl ) condition we limit @xmath61 to be smaller than the shortest sound - crossing time and the time it takes flow to traverse a cell . the latter condition is useful to avoid tangled meshes and is necessary when remapping is used . in addition , it is limited to be smaller than the fractional change in the volume represented by @xmath55 . in calculating these timescales a characteristic length for each cell is needed . given arbitrary polygonal grids , we simply use the shortest edge for each cell . in practice , the timestep is the shortest of these timescales multiplied by a scaling factor , cfl . in our calculations , we set @xmath89 . to ensure second - order accuracy and consistency with the time levels specified in eqs . ( [ eq : mom_discrete2 ] ) and ( [ eq : ene_discrete ] ) , we employ predictor - corrector time integration using the finite difference stencil from @xcite . for each timestep , the current time level of each quantity is identified with @xmath59 in superscript . the time - centered quantities are indicated with a @xmath90 superscript , while the predicted values are labeled @xmath91 , and the fully updated values for the next timestep are labeled with a @xmath60 superscript . table [ table : predictor - corrector ] lists the steps for the predictor - corrector integration . the predictor step begins by defining the forces on nodes , @xmath92 , using current pressures @xmath93 , areas @xmath94 , and artificial viscosity forces @xmath95 . ( [ eq : mom_discrete2 ] ) is then used to calculate the predicted velocity , @xmath96 . the predicted node coordinate is then calculated by @xmath97 , where @xmath98 is the time - centered velocity computed as @xmath99 . then , using @xmath92 and @xmath98 in eq . ( [ eq : ene_discrete ] ) , the predicted specific internal energy , @xmath100 , is obtained . using the predicted node positions , the predicted rotational acceleration ( @xmath101 ) , volume ( @xmath102 ) , and density ( @xmath103 ) , are computed , which in turn are used to calculate the predicted gravitational acceleration , @xmath104 . calling the eos with @xmath100 and @xmath103 gives the predicted pressure @xmath105 . we finish the predictor step and prepare for the corrector step by calculating the time - centered positions @xmath106 , pressures @xmath107 , rotational accelerations @xmath108 , and gravitational accelerations @xmath109 as simple averages of their values at the @xmath110 and @xmath59 time levels . the corrector step then proceeds similarly to the predictor step . @xmath107 and @xmath111 are used to compute @xmath112 , which in turn is used to compute the updated velocity , @xmath113 . after computing the new time - centered velocity , the node coordinates are finally updated using the time - centered velocity . then , the specific internal energy , @xmath114 , is updated using @xmath112 and @xmath98 . finally , the volume , @xmath115 , and density @xmath116 are computed for the next timestep . in 2d simulations using cylindrical coordinates , we have extended bethe - hydro to include rotation about the axis of symmetry . for azimuthally symmetric configurations , all partial derivatives involving @xmath23 are zero . consequently , the conservation of mass and energy equations are @xmath117 and @xmath118 where the @xmath119 subscript on the del operator reminds us that the derivatives with respect to @xmath23 vanish . for concise notation , we define a pseudo lagrangian derivative : @xmath120 while eqs . ( [ eq : mass_lag_rot ] ) and ( [ eq : ene_lag_rot ] ) are simple extensions of the true mass ( eq . [ eq : mass_lag ] ) and energy ( eq . [ eq : ene_lag ] ) equations , the presence of `` christoffel '' terms in the momentum equation give it extra terms : @xmath121 where the extra rotational terms are represented by @xmath122 : @xmath123 to see the origin of these extra terms , one must begin with the momentum equation in cylindrical coordinates and eulerian form : @xmath124 again , dropping terms involving partial derivatives of @xmath23 , and moving all terms not involving partial derivatives to the right - hand side ( rhs ) , eq . ( [ eq : mom_cyl ] ) reduces to @xmath125 hence , using our definition for the pseudo lagrangian derivative , @xmath126 , and @xmath127 in eq . ( [ eq : mom_azsymmetric ] ) we obtain eq . ( [ eq : mom_lag_rot ] ) . it is now apparent why @xmath126 is a pseudo lagrangian derivative . strictly , the `` extra '' rotational terms are a result of `` christoffel '' terms in the true lagrangian derivative . therefore , in defining @xmath126 without them , we ve defined a pseudo lagrangian derivative . conveniently , the @xmath23 component of eq . ( [ eq : mom_lag_rot ] ) , @xmath128 is simply a statement of the conservation of angular momentum about the axis of symmetry : @xmath129 where @xmath130 is the angular momentum about the @xmath24-axis of a lagrangian mass . other than the appearance of an extra term on the rhs of the @xmath21-component and of an equation for the @xmath23 component in eq . ( [ eq : mom_lag_rot ] ) , the 2d hydrodynamic equations including rotation about the @xmath24-axis are similar to the equations without rotation . therefore , the algorithms for dynamics in the 2d plane remain the same , except for the inclusion of the conservation of angular momentum equation and the fictitious force term . it seems natural to define @xmath131 and @xmath132 , where @xmath133 is the angular velocity of node @xmath25 . in fact , our first implementation of rotation did just this . however , results indicated that this definition presents problems near the axis of rotation . whether @xmath134 or @xmath133 is defined , the angular momentum at node @xmath25 is associated with the region of subcells that comprise the node s volume . for the nodes on the axis , this definition states that @xmath135 since @xmath136 , which implies that the angular momentum for the region of subcells near the axis is zero . this awkward feature causes no problems for lagrangian calculations with no remapping . however , during the subcell remapping step , the subcells near the axis contain no angular momentum , yet physically there should be some angular momentum to be remapped . therefore , we have devised an alternative method for including rotational terms in 2d simulations . just as for mass , the angular momentum of a subcell , @xmath137 , is a primary indivisible unit , which by eq . ( [ eq : cons_ang ] ) is constant during a lagrangian hydrodynamic timestep . the angular momentum for a cell is @xmath138 and for each node is @xmath139 to relate angular velocity and @xmath140 to the angular momentum , we treat each subcell as a unit with constant angular velocity . for such regions of space the angular momentum is related to the angular velocity by @xmath141 where @xmath142 is the moment of inertia for subcell @xmath143 . since we are interested in accelerations and velocities at the nodes , we choose that @xmath144 , which naturally leads to @xmath145 substituting eq . ( [ eq : define_jzp ] ) into eq . ( [ eq : define_jp ] ) , we define a relationship between the angular momentum at the nodes and the angular velocity at the nodes : @xmath146 where @xmath147 by conservation of angular momentum , @xmath148 , the angular velocity at the @xmath60 timestep , @xmath149 , is @xmath150 finally , a new form for the @xmath21-component of @xmath68 is constructed . the previously defined form is @xmath151 , and having @xmath21 in the denominator poses numerical problems near and at @xmath152 . since @xmath153 , the new half timestep acceleration is @xmath154 without rotation and time - varying gravity , the equations for 2d simulations using cylindrical coordinates conserve energy to machine precision . however , with rotation , strict energy conservation is lost . using the equations for energy and momentum including rotation , eqs . ( [ eq : ene_lag_rot ] ) and ( [ eq : mom_lag_rot ] ) the energy integral becomes @xmath155 since the terms @xmath156 and @xmath157 are equal in magnitude and opposite in sign , the rate of change for the total energy reduces to the usual surface integral : @xmath158 if the form of the energy equation is to remain the same , discrete versions of @xmath156 and @xmath157 must be derived that cancel one another . in the previous section , we made arguments not to use @xmath159 . however , it is simpler and more intuitive to illustrate why the 2d equations , including rotation , can not satisfy strict energy conservation , while satisfying angular momentum conservation . with conservation of angular momentum we have @xmath160 or @xmath161 the discrete time derivative of @xmath140 is then @xmath162 where we have used @xmath163 . this specifics the required time levels for the components of @xmath164 to conserve angular momentum . considering time integration , the discrete analog of @xmath165 is @xmath166 . to ensure that this sums to zero , the discrete analog of @xmath167 is @xmath168 with these time levels , the 2d equations , which include rotation , would satisfy energy and angular momentum conservation to round - off error however , @xmath169 and @xmath170 are not known at the beginning of the timestep . it is only after the predictor step of the predictor - corrector method that @xmath170 and @xmath171 are obtained . as a result , @xmath167 , including the correct time levels and predictor values , is @xmath172 but to conserve angular momentum and energy , @xmath169 appears in the denominator . therefore , the terms @xmath173 and @xmath174 will not be exactly opposite and equal in magnitude , and angular momentum and energy conservation can not be enforced to machine precision at the same time . in the previous section , we motivated a different discretization of angular momentum , angular velocity , and @xmath140 . however , the basic results of this section do not change . there is still the fundamental problem that to strictly conserve angular momentum , positions at the @xmath60 time level are required , while one must use the predicted , @xmath91 , values in updating the radial velocity . sacrificing strict energy conservation , we have chosen to strictly conserve angular momentum for simulations including rotation . we have designed a test that isolates and tests rotation . using cylindrical coordinates , the calculational domain lies in the @xmath175 plane and extends from the symmetry axis , @xmath176 , to @xmath177 cm and from @xmath178 cm to @xmath179 cm . the initial density is homogeneous with @xmath180 g @xmath181 . isolating the rotational forces , we eliminate pressure gradients by setting the internal energy everywhere to zero . consequently , the trajectory of a lagrangian mass is a straight line in 3d and has a very simple analytic evolution for @xmath182 , determined only by its initial position and velocity . we set @xmath183 and give @xmath140 and @xmath184 homologous profiles : @xmath185 cm s@xmath186 and @xmath187 cm s@xmath186 . this produces a homologous self - similar solution for @xmath184 and @xmath140 that our algorithm for rotation should reproduce . in figure [ rottestplot ] , plots of @xmath140 vs. @xmath21 ( top panel ) and @xmath188 vs. @xmath21 panel ( bottom ) show that bethe - hydro performs well with this test . simultaneously testing the rotational remapping ( [ section : ang_vel_remap ] ) and hydrodynamic algorithms , we employ three remapping regions . the inner 0.1 cm is eulerian , zones exterior to 0.2 cm follow lagrangian dynamics , and the region in between provides a smooth transition between the two domains . both @xmath140 and @xmath188 profiles are presented at @xmath189 , and @xmath190 s. as designed the angular momentum is conserved to machine accuracy . in the top panel ( @xmath140 vs. @xmath21 ) , it is hard to discern any deviation of the simulation ( crosses ) from the analytic solution ( solid line ) . in fact , the maximum error as measured by @xmath191 ranges from @xmath192 near the beginning of the simulation to @xmath193 at the end . @xmath188 vs. @xmath21 ( bottom panel ) shows similar accuracy , except for the region near the axis . at the center @xmath194 reaches a maximum value of 0.2 rads s@xmath186 at @xmath195 s. on the whole , fig . [ rottestplot ] demonstrates that this code reproduces the analytic result with and without remapping . there are two general strategies for solving poisson s equation for gravity on arbitrary grids . the first is to define another regular grid , interpolate the density from the hydro grid to this new gravity grid , and use many of the standard poisson solvers for regular grids . however , in our experience this approach can lead to numerical errors , which may lead to unsatisfactory results . the second approach is to solve eq . ( [ eq : poisson ] ) explicitly on the arbitrary hydro grid , eliminating an interpolation step . we prefer the latter approach , giving potentials and accelerations that are more consistent with the hydrodynamics on an unstructured mesh . in determining the discrete analogs of eq . ( [ eq : poisson ] ) , we use the support - operator method . after a bit of algebra these discrete equations may be expressed as a set of linear equations whose unknowns are the discretized gravitational potentials . solving for these potentials is akin to solving a matrix equation of the form , @xmath196 , and since the solutions change little from timestep to timestep , we use an iterative matrix inversion approach in which the initial guess is the solution from the previous timestep . in solving poisson s equation for gravity , we seek an algorithm satisfying the following conditions : 1 ) the potential should be defined at zone centers and the gravitational acceleration should be defined at the nodes ; 2 ) we should conserve momentum and energy as accurately as possible ; and 3 ) the solver must be fast . to this end , we employ the method described in @xcite , @xcite , and @xcite for elliptic solvers . specifically , we follow the particular implementation of @xcite . applying the support - operator method , we first define the divergence operator and , in particular , define @xmath197 . poisson s equation for gravity may be written as an equation for the divergence of the acceleration , @xmath198 where @xmath199 , the gravitational acceleration , is the negative of the gradient of the potential . @xmath200 integrating eq . ( [ eq : poisson_div ] ) over a finite volume gives @xmath201 and if the volume of integration , @xmath202 , is chosen to be that of a cell , the rhs of eq . ( [ eq : divtheorem ] ) becomes @xmath203 . in discretizing the middle term in eq . ( [ eq : divtheorem ] ) , @xmath204 an expression for the divergence of @xmath199 follows naturally : @xmath205 where @xmath206 is the area of edge @xmath207 , and @xmath208 is the component of the acceleration parallel to the direction of the edge s area vector . therefore , substituting eq . ( [ eq : discretediv ] ) into eq . ( [ eq : divtheorem ] ) suggests a discrete expression for eq . ( [ eq : poisson_div ] ) : @xmath209 having defined the discrete divergence of a vector , we then use the support - operator method to write the discrete equivalent of the gradient . once again , the integral identity relating the divergence and gradient operators is given by eq . ( [ eq : intidentity ] ) . applying the definition for the discrete divergence , the first term on the lhs of eq . ( [ eq : intidentity ] ) is approximated by @xmath210 where @xmath211 is the component of @xmath52 perpendicular to edge @xmath207 , and the rhs of eq . ( [ eq : intidentity ] ) is approximated by @xmath212 it is already apparent that the discrete potential is defined for both cell centers ( @xmath213 ) and edges ( @xmath214 ) . this fact seems cumbersome . however , we will demonstrate below that in the final equations the number of unknowns may be reduced by eliminating the cell - centered potentials , @xmath213 , in favor of the edge - centered potentials , @xmath214 . completing the discretization of eq . ( [ eq : intidentity ] ) , the second term on the lhs is @xmath215 where @xmath216 is a volumetric weight and @xmath217 and @xmath67 are defined at the nodes . the volumetric weight associated with each corner is defined as one quarter of the area of the parallelogram created by the sides that define the corner . additionally , the weights are normalized so that @xmath218 . the remaining task is to define the dot product @xmath219 at each corner when the vector is expressed in components of a nonorthogonal basis set : @xmath220 where @xmath221 and @xmath222 are basis vectors located at the center of the edges on either side of the corner associated with node @xmath25 with orientations perpendicular to the edges , and @xmath223 and @xmath211 are the corresponding magnitudes in this basis set . with this basis set , the inner product is @xmath224 where @xmath225 \ , , \ ] ] and @xmath226 is the angle formed by edges @xmath207 and @xmath227 . using eqs . ( [ eq : h_basis])-([eq : smatrix ] ) in eq . ( [ eq : intidentity_lhs2 ] ) , the second term on the lhs of eq . ( [ eq : intidentity ] ) is @xmath228 where @xmath229 and @xmath230 . combining eqs . ( [ eq : intidentity_lhs1 ] ) , ( [ eq : intidentity_rhs ] ) , and ( [ eq : intidentity_lhs2 ] ) , we have the discrete form of the identity eq . ( [ eq : intidentity ] ) : @xmath231 = 0 \ , .\end{gathered}\ ] ] to find the equation for @xmath208 on each edge we set the corresponding @xmath232 and set all others to zero . this gives an expression for @xmath208 in terms of @xmath214 , @xmath213 , and the other edge - centered accelerations of cell @xmath24 : @xmath233 = 0\end{gathered}\ ] ] or @xmath234 \\ = ( \phi_e - \phi_z)a_e \ , .\end{gathered}\ ] ] together , the set of equations for the edges of cell @xmath24 forms a matrix equation of the form on the @xmath24-axis are zero because the areas on the axis of symmetry are zero . therefore , for the zones along the axis , one less equation and variable appear in the set of equations for @xmath208 . ] @xmath235 where @xmath236 is a vector based upon the edge values , @xmath237 . inverting this equation gives an expression for the edge accelerations as a function of the cell- and edge - centered potentials : @xmath238 after finding the edge acceleration for each zone , these values are inserted into eq . ( [ eq : discretepoisson_div ] ) . the stencil of potentials for this equation for each cell includes the cell center and all edges . while eq . ( [ eq : gacc_phiz_phie ] ) gives the edge accelerations associated with cell @xmath24 , there is no guarantee that the equivalent equation for a neighboring cell will give the same acceleration for the same edge . to ensure continuity , we must include an equation which equates an edge s acceleration vectors from the neighboring cells : @xmath239 where @xmath240 and @xmath241 are the edge accelerations as determined by eq . ( [ eq : gacc_phiz_phie ] ) for the left ( @xmath242 ) and right ( @xmath243 ) cells . while we could have chosen a simpler expression for eq . ( [ eq : gacc_continuity ] ) , the choice of sign and coefficients ensures that the eventual system of linear equations involves a symmetric positive - definite matrix . this enables the use of fast and standard iterative matrix inversion methods such as the conjugate - gradient method . the stencil for this equation involves the edge in question , the cell centers on either side , and the rest of the edges associated with these cells . at first glance , it seems that we need to invert the system of equations that includes the cell - centered equations , eq . ( [ eq : gacc_phiz_phie ] ) , and the edge - centered equations , eq . ( [ eq : gacc_continuity ] ) . instead , we use eq . ( [ eq : discretepoisson_div ] ) to express the cell - centered potentials in terms of edge - centered potentials and the rhs of the equation : @xmath244 where @xmath245 , and @xmath246 substituting eq . ( [ eq : phizeq ] ) into the expression for the edge acceleration , eq . ( [ eq : gacc_phiz_phie ] ) , @xmath247 where @xmath248 . finally , substituting eq . ( [ eq : gacc_phie ] ) into eq . ( [ eq : gacc_continuity ] ) gives the matrix equation for gravity with only edge - centered potential unknowns : @xmath249 the stencil for this equation involves the edge and all other edges associated with the zones on either side of the edge in question . taken together , the equations result in a set of linear equations for the edge - centered potentials , @xmath214 . since this class of matrix equations is ubiquitous , there exists many accurate and fast linear system solvers we can exploit . to do so , we recast our discrete form of poisson s equation as @xmath250 where @xmath251 is a vector of the edge - centered potentials , and @xmath252 is the corresponding vector of source terms . we solve the matrix equation using the conjugate - gradient method with a multigrid preconditioner . in particular , we use the algebraic multigrid package , amg1r6 for details on performance . ] @xcite . unfortunately , this discretization scheme does not adequately define the accelerations at the nodes where they are needed in eq . ( [ eq : mom_discrete2 ] ) . therefore , we use the least - squares minimization method to determine the gradient on the unstructured mesh . assuming a linear function for the potential in the neighborhood of node @xmath14 , @xmath253 , we seek to minimize the difference between @xmath254 and @xmath255 , where @xmath254 is the neighbor s value and @xmath255 is the evaluation of @xmath253 at the position of the neighbor , @xmath256 . more explicitly , we minimize the following equation : @xmath257 where the `` neighbors '' are the nearest cell - centers and edges to node @xmath14 and are denoted by @xmath258 , @xmath259 @xmath260 , @xmath261 , and @xmath262 . usually , minimization of eq . ( [ eq : least_gacc_eq1 ] ) leads to a matrix equation for two unknowns : @xmath263 and @xmath264 , the gradients for @xmath8 in the @xmath19 and @xmath20 directions , respectively ( [ section : swept ] ) . however , @xmath265 , the potential at the node is not defined and is a third unknown with which eq . ( [ eq : least_gacc_eq1 ] ) must be minimized . performing the least - squares minimization process with respect to the three unknowns , @xmath265 , @xmath263 , and @xmath264 , leads to the following set of linear equations for each node : @xmath266 where @xmath267 inversion of this linear system gives the potential and the gradient at the node . specifically , we use the adjoint matrix inversion method to find the inverse matrix and the three unknowns , including @xmath263 and @xmath264 . the outer boundary condition we employ for the poisson solver is a multipole expansion for the gravitational potential at a spherical outer boundary . assuming no material exists outside the calculational domain and that the potential asymptotes to zero , the potential at the outer boundary , @xmath268 , is given by @xmath269 where @xmath270 is the usual legendre polynomial . in discretized form , ( [ eq : boundary_eq1 ] ) becomes @xmath271 where @xmath272 is the order at which the multipole expansion is truncated . for 1d spherically symmetric simulations , calculation of gravity can be straightforward using @xmath273 where @xmath274 is all of the mass enclosed by point @xmath25 . instead , we solve poisson s equation for gravity using similar methods to those employed for the 2d gravity . by doing this , we remain consistent when comparing our 1d and 2d results . furthermore , we show that deriving a poisson solver based upon the support - operator method is equivalent to the more traditional form , eq . ( [ eq : gacc_enclosed ] ) . the inherent symmetries of 1d spherical simulations reduce eq . ( [ eq : ge_equation ] ) to the following form : @xmath275 unlike the full 2d version , one may eliminate the edge potentials in favor for the cell - centered potentials and the accelerations at the edges . since each edge @xmath207 will have an equation for its value associated with cells @xmath24 and @xmath276 , this leads to @xmath277 using the continuity of accelerations at the edges , @xmath278 the expression for gravitational acceleration at edge @xmath207 in terms of the cell - centered potentials is @xmath279 substituting this expression into @xmath280 leads to a tridiagonal matrix equation for the cell - centered potential , which can be inverted in o(n ) operations to give the cell - centered potentials . substituting these cell - centered potentials back into eq . ( [ eq : gacc_1dsph ] ) gives the gravitational acceleration needed for the momentum equation ( eqs . [ eq : mom_lag ] and [ eq : mom_discrete2 ] ) . it should be noted that eq . ( [ eq:1dgaussgravity ] ) can be rewritten as a recursion relation for the gravitational acceleration , @xmath281 , at node @xmath25 : @xmath282 where @xmath283 is the mass of the cell interior to point @xmath25 . with the boundary condition @xmath284 and then recursively solving eq . ( [ eq : recursive ] ) for @xmath281 from the center outward , the expression for the gravitational acceleration simplifies to eq . ( [ eq : gacc_enclosed ] ) . this is the traditional form by which the gravitational acceleration is calculated for 1d spherically symmetric simulations and for which the potential is usually not referenced . the beauty of our approach is that 1d and 2d simulations are consistent , and that it self - consistently gives the gravitational potential . for self - gravitating hydrodynamic systems , the total energy is @xmath285 where @xmath286 is the total internal energy , @xmath287 is the total kinetic energy , and @xmath288 is the total potential energy : @xmath289 an alternative form is possible with the substitution of eq . ( [ eq : poisson_div ] ) into eq . ( [ eq : totpot1 ] ) and integration by parts : @xmath290 \ , .\ ] ] the integral form of eq . ( [ eq : totalenergy ] ) is @xmath291 and total energy conservation is given by @xmath292 since the solution to poisson s equation for gravity , using green s function , is @xmath293 the time derivative of the total potential energy may take three forms : @xmath294 note that this is true only for the total potential energy and does not imply @xmath295 . using eq . ( [ eq : dudt ] ) and the eulerian form of the hydrodynamics equations ( see appendix [ section : eulerian_hydro ] ) , eq . ( [ eq : dedt ] ) becomes a surface integral : @xmath296 which simply states that energy is conserved in the absence of a flux of energy through the bounding surface . the discrete analogs of @xmath286 and @xmath287 are trivially obtained using @xmath297 and @xmath298 , respectively . the discrete total potential energy may take two forms , the analogs of eqs . ( [ eq : totpot1 ] ) and ( [ eq : totpot2 ] ) : @xmath299 \ , , \\ \end{array } \right .\ ] ] where @xmath300 indicates outer - boundary values . upon trying both forms , we get similar results . therefore , we use the simpler form involving the cell - center potential : @xmath301 . because our discrete hydrodynamics equations including gravity do not give strict energy conservation , we use the core - collapse simulation ( [ section : corecollapse ] ) to gauge how well energy is conserved . rather than measuring the relative error in total energy by @xmath302 , where @xmath303 , @xmath304 is the initial total energy , and @xmath305 is the total energy at timestep @xmath59 , we use @xmath306 . for stellar profiles , the kinetic energy is small and the internal and gravitational energies are nearly equal in magnitude , but opposite in sign . since the total energy is roughly zero in comparison to the primary constituents , @xmath286 and @xmath288 , of the total energy , we use the gravitational potential energy as a reference . for example , the internal and gravitational energies start at @xmath307 erg and reach @xmath308 ergs at the end of the run . however , the total energy is a small fraction of these energies initially , @xmath309 , and after core bounce , @xmath310 . hence , we measure the relative error in total energy as @xmath306 . the total energy for the 1d and 2d core - collapse simulations evolve similarly and is conserved quite well . for all times except a few milliseconds around bounce , total energy is conserved better than @xmath11 @xmath311 . during collapse , from 0 to 147 ms , the total energy deviates by only @xmath13 @xmath312 . over a span of 5 ms around bounce , @xmath313 ms , the total energy changes by @xmath12.2 @xmath314 . for 100s of milliseconds afterward , the relative error in the total energy reduces to @xmath17 @xmath312 . hence , for all but about 5 ms of a simulation that lasts many hundreds of milliseconds the relative error in total energy is better than @xmath11 @xmath315 . we include here several assessments of the 1d and 2d poisson solvers . first , we calculate , using the butterfly mesh , the potential for a homogeneous sphere and compare to the analytic solution . this tests overall accuracy and the ability of our solver to give spherically symmetric potentials when a non - spherical mesh is employed . similarly , we substantiate the algorithm s ability to produce aspherical potentials of homogeneous oblate spheroids . then , we verify that the hydrodynamics and gravity solvers give accurate results for stars in hydrostatic equilibrium , and for a dynamical problem , the goldreich - weber self - similar collapse @xcite . figure [ maclaurin_sphere ] compares the simulated and analytic solutions for a homogeneous sphere , having density @xmath316 g @xmath181 , and maximum radius @xmath317 cm . the analytic potential inside the sphere is @xmath318 , and the top panel of fig . [ maclaurin_sphere ] shows the relative difference of the analytic potential and the numerical solution , @xmath213 , as a function of radius . results for four resolutions of the butterfly mesh are shown : 2550 cells with an effective radial resolution @xmath319 0.02 cm ( blue ) ; 8750 cells with @xmath319 0.01 cm ( green ) ; 15,200 cells with @xmath319 0.006 cm ( yellow ) ; and 35,000 cells with @xmath320 0.005 cm ( red ) . two facts are obvious : 1 ) the solutions are accurate at a level of @xmath321 for @xmath322 and @xmath323 for @xmath324 ; and 2 ) the degree to which the solution is spherically symmetric is similarly a few times @xmath325 for @xmath322 and a few times @xmath326 for @xmath324 . plotting the minimum error as a function of the effective radial resolution , the bottom panel of fig . [ maclaurin_sphere ] verifies that the 2d poisson solver convergences with 2@xmath327-order accuracy ( solid line ) . next , we calculate the aspherical potential for an oblate spheroid . the homogeneous oblate spheroid has an elliptic meridional cross section , and the minor- and major - axes of the ellipse are @xmath328 and @xmath329 , where @xmath329 is the equatorial radius . thus , the eccentricity of the spheroid is @xmath330 . given a uniform density , @xmath331 , the potential for a spheroid is @xmath332 \right ) \ , , \ ] ] where , for oblate spheroids , @xmath333 \frac{(1 - e^2)^{1/2}}{e^2 } \ , , \ ] ] @xmath334 \frac{(1 - e^2)^{1/2}}{e^2 } \ , , \ ] ] and @xmath335 @xcite . numerical results with @xmath336 are shown in fig . [ maclaurin_spheroid ] . as with the sphere , @xmath337 g @xmath181 and @xmath317 cm . however , for the given eccentricity , the polar - axis radius , @xmath328 , is 0.6 cm . once again , the grid is a butterfly mesh , but this time the outer boundary follows the ellipse defining the surface of the spheroid . the top panel of fig . [ maclaurin_spheroid ] presents the spheroid s potential , and the degree of accuracy is presented in the bottom plot . with @xmath338 and @xmath322 , the relative error in the potential ranges from @xmath339 near the outer boundary to @xmath321 in interior regions . conspicuous are features in the relative error that track abrupt grid orientation changes in the mesh . fortunately , these features have magnitudes smaller than or similar to the relative error in the local region . the relative error in the gravitational acceleration magnitude , latexmath:[$(\|\vec{g}_p\| - @xmath1@xmath341 to @xmath1@xmath326 , where @xmath342 is the analytic acceleration and @xmath343 is the maximum magnitude on the grid . typical errors in the acceleration direction range from @xmath1@xmath341 to @xmath1@xmath326 radians with rare deviations as large as @xmath1@xmath344 radians near the axis and abrupt grid orientation changes . in fig . [ laneemden_plot ] , we demonstrate that the ale algorithm in combination with our gravity solver produces reasonably accurate hydrostatic equilibria . the grid is the butterfly mesh with 8750 zones , and the initial model is a lane - emden polytrope with @xmath345 , @xmath346 m@xmath347 , and @xmath348 cm . crosses show the density profile at @xmath349 s , while the solid line shows the maximum density as a function of time and that the star pulsates . these oscillations result from the slight difference between an analytic hydrostatic equilibrium structure and a discretized hydrostatic equilibrium structure , and with increasing resolution , they decrease in magnitude . interestingly , the oscillations continue for many cycles with very little attenuation . simulating the goldreich - weber self - similar collapse @xcite in 1d and 2d is a good test of dynamic simulations including gravity . the analytic profile is similar to a lane - emden polytrope with @xmath350 , and in fact , we use the gamma - law eos with @xmath351 . while the lane - emden polytropes are assumed to be in hydrostatic equilibrium , the goldreich - weber self - similar collapse has a homologous velocity profile and a self - similar density profile . the physical dimensions have been scaled so that @xmath352 m@xmath347 , the initial central density is @xmath353 g @xmath181 , and the maximum radius of the profile is @xmath354 cm . for the 1d simulation , we initiate the grid with 200 evenly - spaced zones , and for 2d , a butterfly mesh with 35,000 zones ( effectively with 200 radial zones ) is used to initiate the grid . subsequent evolution for both simulations uses the lagrangian configuration . figure [ gwplot ] shows snapshots of density vs. radius for 1d ( top panel ) and 2d ( bottom panel ) simulations at @xmath3550 , 20 , 40 , 60 , 80 , 100 , 120 , and 130 ms . both plots indicate that the simulations ( crosses ) track the analytic solution ( solid lines ) . quantitatively , we measure a relative difference , @xmath356 , for the reported times , where @xmath357 are the simulated cell - center densities and @xmath358 are the analytic values . consistently , the largest deviations are at the center and table [ table : gw ] gives these values for the 1d simulation ( 2@xmath0 column ) and the 2d simulation ( 3@xmath359 column ) . the relative differences range from @xmath360 at 0 ms to @xmath361 at 130 ms for the 1d simulation and @xmath362 at 0 ms to @xmath363 at 130 ms for the 2d simulation . at all times , the departure from spherical symmetry is no more than @xmath364 . boundary conditions are implemented in one of two ways . either an external pressure is specified or the velocities at the nodes are fixed . for external pressures , ghost cells are defined that have no true volume or mass associated with them . their only function is to apply an external force to the boundary cells equal to an external pressure times the boundary surface area . specifying nodal velocities on the boundary accomplishes the same task . in this case , there is an implied external force and pressure . in practice , generic dynamical boundaries use the external pressure boundary condition . on the other hand , pistons , reflecting walls , and the azimuthal axis have the velocity perpendicular to the boundary specified , while the parallel component executes unhindered hydrodynamic motions . to resolve shocks over just a few zones , we include artificial viscosity in the equations of hydrodynamics . for 1d simulations , we add a viscous - like term to the pressure @xcite . we denote this viscous pressure by @xmath365 . the original realization of @xmath365 employed one term proportional to @xmath366 , where @xmath367 is the difference in velocity from one zone to the next . while this form adequately resolved shocks , unphysical oscillations were observed in the post - shock flow . a second term , linear in @xmath367 , was then added that effectively damped these oscillations @xcite . another form which we employ for our 1d simulations is @xmath368 where @xmath369 is the sound speed , @xmath370 is the parameter associated with the linear term , and @xmath371 is the parameter associated with the quadratic term . this form has the appealing attribute that it is motivated by the expression for the shock - jump condition for pressure in an ideal gas @xcite . for 2d simulations , the artificial viscosity scheme we have settled upon is the tensor artificial viscosity algorithm of @xcite . here , we do not re - derive the artificial viscosity scheme , but highlight some of its salient features and practical implementations . the useful feature of the tensor algorithm is its ability to calculate artificial viscosity on an arbitrary grid , while suppressing artificial grid buckling when the flow is not aligned with the grid . in the strictest sense , the tensor artificial viscosity does not employ the simple viscous pressure described above . instead , @xcite assume that the artificial viscosity tensor is a combination of a scalar coefficient , @xmath372 and the gradient of the velocity tensor , @xmath373 . therefore , references to the parameters @xmath370 and @xmath371 in 1d and 2d refer to the same linear or quadratic dependence on @xmath374 . with this assumption , the momentum equation , eq . ( [ eq : mom_lag ] ) , becomes @xmath375 and the energy equation , eq . ( [ eq : ene_lag ] ) , is @xmath376 where @xmath377 , using the normal einstein summation convention . with the method of support operators , @xcite derive discrete forms of these equations . analogous to the gradient of the pressure term , the discrete artificial viscosity term in the momentum equation becomes a summation of corner forces , @xmath378 and the corresponding term in the energy equation is the usual force - dot - velocity summation : @xmath379 therefore , implementation of the artificial viscosity scheme is straightforward and similar to that for the pressure forces . in practice , this artificial viscosity scheme is formulated for cartesian coordinates , and to obtain the equivalent force for cylindrical coordinates , we multiply the cartesian subcell force by @xmath81 . in general , this works quite well , but sacrifices strict momentum conservation ( see [ section : momcons ] ) . a problem that can plague lagrangian codes using cells with 4 or more sides is the unphysical hourglass mode @xcite . to suppress this problem , we employ a modified version of the subcell pressure algorithm of @xcite . for all physical modes ( translation , shear , and extension / contraction ) the divergences of the velocity on the subcell and cell levels are equal , @xmath380 hence , as long as the calculation is initiated such that the subcell densities of a cell are equal to the cell - averaged density , then they should remain equal at all subsequent times . any deviation of the subcell density from the cell density , @xmath381 , is a direct result of the hourglass mode . the scheme of @xcite uses this deviation in subcell density to define a subcell pressure that is related to the deviation in subcell and cell density , @xmath382 , by @xmath383 with the corresponding deviation in pressure : @xmath384 . application of these pressures as subcell forces counteracts the hourglass distortion . however , the pressure throughout the cell is no longer uniform . as a result , the subcell pressures exert forces on the cell centers and mid - edge points . these locations in the grid are not subject to physical forces as are the nodes . instead , their movement is tied to the motions of the nodes . to conserve momentum , the forces acting on these enslaved points must be redistributed to the dynamical nodes . though the choice of redistribution is not unique , we follow the procedure established by @xcite for the form of the subcell force ( for the definitions of @xmath41 and @xmath43 , see [ section : coords_mesh ] and fig . [ fig : grid ] . ) : @xmath385 \ , .\ ] ] this particular scheme works quite well for lagrangian calculations . however , it can be incompatible with subcell remapping algorithms . the hourglass suppression scheme described above assumes that only hourglass motions result in nonzero values for @xmath382 . even in calculations that are completely free of hourglass motions , the subcell remapping scheme can and will produce subcell densities within a cell that are different from the cell density . since this difference of subcell and cell densities has nothing to do with hourglass motions , any subcell forces that arise introduce spurious motions that do nt correct the hourglass motions . investigating several alternative schemes , we settled on a modification of the @xcite approach . after each remap , we define a tracer density whose only purpose is to track the relative changes in the subcell and cell volumes during the subsequent hydrodynamic solve . for convenience , and to keep the magnitude of the subcell pressures about right , we set the subcell tracer density equal to the cell density . after replacing @xmath382 with the difference between the subcell tracer density and the cell density , @xmath386 , we proceed with the scheme prescribed by @xcite . we have found that the subcell forces of @xcite significantly resist hourglass motions only after @xmath382 achieves significant magnitude . for our new hourglass scheme , if a remap is implemented after each hydrodynamic solve , @xmath386 does not have a chance to achieve large values , and hence , the subsequent subcell forces are not very resistant to hourglass distortions . however , for many simulations , we have found that it is not necessary to remap after every lagrangian hydrodynamic solve . rather , the remap may be performed after @xmath387 timesteps . in these circumstances , the subcell tracer density is allowed to evolve continuously as determined by the hydrodynamic equations . with large @xmath387 , @xmath386 does develop significant amplitude and the subcell forces become effective in suppressing the hourglass modes . in fact , in the limit that @xmath388 , this scheme becomes the hourglass suppression scheme of @xcite . in practice , we multiply this subcell force by a scaling factor . to determine the appropriate magnitude of this scaling factor , we executed many test problems and found that problems that involved the perturbation of hydrostatic equilibrium provide a good test of the robustness of the hourglass fix . during the testing protocol , we ran a simulation for at least 200,000 timesteps , ensuring that the hourglass fix remains robust for long calculations . for lagrangian calculations , a scaling factor @xmath389 produced reasonable results . for runs in which remapping occurred after @xmath387 timesteps , we found that larger values of the scaling factor were required . however , for very large values of @xmath387 , such as 64 or greater , such large values compromised small structures in the flow . hence , we settled on scaling factors near 1 for large @xmath387 . for small @xmath387 , we found that the difference in tracer density and the cell - centered density had little time to build to significant amplitudes , so larger values of the scaling factor are required , but anything above 4 produced noticeable problems in flows with many timesteps . in summary , the scaling factor should be adjusted to be @xmath389 for large @xmath387 and @xmath390 for small @xmath387 . to demonstrate the need for an hourglass suppression scheme , and that our algorithm to address the hourglass distortions works , we show in fig . [ gw_fmerit_comp ] results for the goldreich - weber self - similar collapse in 2d ( [ section : gravity_tests ] ) , with and without the hourglass suppression ( [ section : hourglass ] ) . these results represent a lagrangian simulation of the self - similar collapse at @xmath391 ms . on the left ( `` @xmath392 '' ) the grid clearly shows problematic grid buckling when the hourglass suppression is turned off . the grid on the right ( `` @xmath393 '' ) shows the elimination of grid buckling with the use of the hourglass suppression scheme and a scaling factor of 2.0 . figure [ rt_stills_fmerit ] illustrates the problem of using the hourglass suppression scheme of @xcite in combination with the subcell remapping algorithm ( [ section : remap ] ) . all three panels show the results of the single - mode rayleigh - taylor instability at @xmath394 s. for the left panel , no hourglass suppression is employed . this panel represents our control . while the hourglass instability does not cause serious problems for the calculation , one can see evidence of slight hourglass patterns at the scale of the grid resolution . in particular , two blobs near the top and on the edges form distinct patterns . the central panel demonstrates the problem of using the subcell remapping scheme and the subcell pressure method of @xcite . on the other hand , the right panel demonstrates that the modified subpressure scheme we have developed suppresses the hourglass distortions , while preserving the expected flow . fluid flows with large vorticity quickly tangle a lagrangian mesh , presenting severe problems for lagrangian hydrodynamic codes . to avoid entangled grids , it is common to remap the state variables after each lagrangian hydrodynamic advance to another less tangled mesh . even flows that exhibit very little vorticity , but extreme compression or expansion along a particular direction , leading to very skewed cells , can limit the accuracy of lagrangian calculations . in this case as well , one can employ a remapping scheme which is designed to minimize the calculational error . to remap , a new grid must be established . one choice is to remap to the original grid , thereby effectively solving the hydrodynamics equations on an eulerian mesh . another is to use a reference jacobian matrix rezone strategy , which establishes a new mesh close to the lagrangian mesh , while reducing numerical error by reducing unnecessarily skewed zones . the arbitrary nature of the remapping algorithm even allows the freedom of choosing one rezoning scheme in one sector of the grid and another rezoning scheme in other sectors . for example , one could remap in an eulerian fashion in one sector , let the grid move in a lagrangian manner in another , and in the intervening region smoothly match these two regions . generally , there are two options for remapping schemes . one can remap from one arbitrary grid to another completely unrelated arbitrary grid . for these unrelated grids , one needs to determine the overlap of the zones of the first grid with the zones of the second . in general , this can be a very cumbersome and expensive process . as a result most ale codes , remap to an arbitrary grid that is not too different from the first . specifically , the usual stipulation employed is that the face of each cell does not traverse more than one cell during a timestep , and that the connectivity among nodes , faces , edges , and cells remains the same from the first grid to the remapped grid . the regions swept by the faces then contain the mass , momentum , or energy which is added to the cell on one side of the face and subtracted from the cell on the other side . it is in this context that we use the swept - region remapping algorithms of @xcite , @xcite , @xcite , and @xcite . this remapping scheme may be described by four stages : 1 . the first is the gathering stage . this is a stage in which subcell quantities of the mass , momenta , and energies are defined in preparation for the bulk of the remapping process . 2 . the second is to remap the subcell quantities from the lagrangian grid to the new rezoned grid using the swept - region approximation . in doing so , the remapping algorithm remains 2nd - order accurate , but avoids time intensive routines to calculate the overlap regions of the old and new grids . 3 . the third is to repair the subcell densities . because exact spatial integration is avoided with the swept - region approximation , local bounds of subcell densities may be violated . therefore , a repair algorithm which redistributes mass , momentum , and energies to preserve the local bounds is implemented . finally , there is the scattering stage , in which the primary variables of the hydrodynamics algorithm are recovered for the new rezoned grid . in order to remap the primary quantities using the subcell remapping algorithm , mass , momenta , internal energy , and kinetic energy must be defined at the subcell level . in the lagrangian stage , the subcell mass , @xmath47 , is already defined . the subcell density which is a consequence of the change in the subcell volume ( @xmath33 ) during the lagrangian hydro step is @xmath395 unfortunately , there is no equivalent quantity for the subcell internal energy . instead , we define the internal energy density for each zone , @xmath396 , fit a linear function for the internal energy density , @xmath397 , and integrate this function over the volume of each subcell to determine the total subcell internal energy and , consequently , the average subcell internal energy density : @xmath398 performing the integration of eq . ( [ eq : gather_energy ] ) involves a volume integral and volume integrals weighted by @xmath19 and @xmath20 in cartesian coordinates or @xmath21 and @xmath24 in cylindrical coordinates ( see appendix [ section : integrals ] for formulae calculating these discrete integrals ) . note , that , by construction , these newly defined internal energies satisfy conservation of internal energy for each zone , @xmath399 . similarly , there are no readily defined subcell - centered averages for the momenta , @xmath400 , or velocities , @xmath401 . however , they should be related to one another by @xmath402 if the gathering stage is to preserve momentum conservation , then the following equality must hold : @xmath403 by inspection , it might seem natural to set @xmath404 . however , we follow the more accurate suggestion made by @xcite to define the subcell - averaged velocity as follows : @xmath405 where @xmath406 is obtained by averaging over all nodal velocities associated with zone @xmath24 , and @xmath407 and @xmath408 are edge - averaged velocities given by @xmath409 and @xmath410 . substituting eq . ( [ eq : subcellvel ] ) into eq . ( [ eq : momconssubcell ] ) and rearranging terms , we get an expression for the subcell velocity that depends on known subcell masses and nodal velocities : @xmath411 we now rewrite eq . ( [ eq : vsubcellexplicit1 ] ) in a form that obviously makes it easy to write the gathering operation in matrix form : @xmath412 if @xmath413 is a vector of one component of all velocities associated with cell @xmath24 , and @xmath414 is the equivalent for subcell velocities , then @xmath415 where the matrix , @xmath416 , for each zone is given by the coefficients in eq . ( [ eq : vsubcellexplicit ] ) . to conserve total energy in the remap stage , we define the subcell kinetic energy and follow the same gather procedure as for the velocity by simply replacing velocity with the specific kinetic energy : @xmath417 we ensure conservation of total kinetic energy for a cell : @xmath418 where @xmath419 is the subcell - averaged specific kinetic energy . then , the transformation from nodal specific kinetic energies to subcell - averaged specific kinetic energies is given by @xmath420 where @xmath421 is a vector of the subcell - averaged specific kinetic energies , @xmath422 is a vector of the nodal specific kinetic energies , and @xmath416 is the same transformation matrix used for the velocity transformations . having found the specific kinetic energies , the calculation of the kinetic energy for each subcell is then straightforward : @xmath423 . upon completion of the gathering stage , each relevant quantity , @xmath47 , @xmath400 , @xmath424 , and @xmath425 , is expressed as a fundamentally conserved quantity for each subcell . from these conserved quantities and the subcell volumes , we have the corresponding densities , which , as is explained in [ section : swept ] , are important components of the remapping process . having gathered all relevant subcell quantities , we begin the bulk of the remapping process . since all variables are now expressed in terms of a conserved quantity ( subcell mass , energy , and momentum ) and a density ( mass density , energy density , and momentum density ) , for clarity of exposition , we focus on representative variables , subcell mass and density , to explain the generic remapping algorithm . when there are differences in the algorithm for the other variables , we note them . we have instituted the remapping algorithm of @xcite . to avoid the expensive process of finding the overlap of the lagrangian grid with the rezoned grid , this algorithm modifies the mass of each subcell with a swept - edge approximation for the rezoning process . as long as the connectivity and neighborhood for each cell remain the same throughout the calculation , time spent in connectivity overhead is greatly reduced . the objective of the remapping algorithm is then to find the amount of mass , @xmath426 , that each edge effectively sweeps up due to the rezoning process and to add or subtract this change in mass to the old subcell mass to obtain the new subcell mass , @xmath427 : @xmath428 of course , the change in mass is obtained by integration of a density function over the volume of the swept region : @xmath429 the density function , @xmath430 , may take on any functional form . to maintain second - order accuracy of the algorithm we use a linear density function ; @xmath431 where @xmath432 is the average of the nodal positions defining the subcell . in determining the gradient , @xmath433 , we employ one of two standard methods for calculating the gradient using the densities from subcell @xmath143 and its neighbors . the first method uses green s theorem to rewrite a bounded volume integral as a boundary integral around the same volume . we begin with a definition of the average gradient of @xmath6 for a given region . for each component , the average gradient is @xmath434 where @xmath435 and @xmath436 are the gradients in the x- and y - directions , respectively . the region over which the integrals are evaluated is defined by segments connecting the centers of the neighboring subcells . hence , the discrete form of the average gradient is @xmath437 while this method works reasonably well , an unfortunate consequence of the integration and division by volume is that the value of the gradient is directly influenced by the shape and size of the cell . an alternative method for determining gradients , which is not as easily influenced by the shape of the mesh , is a least - squares procedure . for simplicity of notation , we describe this method in the context of cells rather than subcells . again , assuming a linear form for @xmath438 , we seek to minimize the difference between @xmath439 and @xmath440 , where @xmath439 is the neighbor s value and @xmath440 is the extrapolation of cell @xmath24 s linear function to the position of the neighbor , @xmath441 . more explicitly , we wish to minimize the following equation : @xmath442 where the set of neighbors for cell @xmath24 is denoted by @xmath443 , @xmath444 @xmath445 and @xmath446 are the @xmath19 and @xmath20 components of the gradient , @xmath447 , @xmath448 , @xmath449 , and @xmath450 . minimizing eq . ( [ eq : least_eq1 ] ) with respect to the two unknowns , leads to the following set of linear equations : @xmath451 for each subcell , where @xmath452 solving this linear system with cramer s rule then gives the least - squares gradients . after calculating the gradient we ensure monotonicity using the barth - jespersen limiter @xcite . the gradient is limited by a scalar , @xmath213 , that has a range between 0 and 1 : @xmath453 first , we determine the minimum and maximum value of @xmath6 among @xmath357 and its neighbors @xmath439 : @xmath454 @xmath455 we satisfy the requirement that @xmath456 this is accomplished in the following way . @xmath457 where @xmath458 is a limiter associated with each node of the cell and is given by @xmath459 where @xmath460 is the unlimited linear function evaluated at the nodes . with the linear function @xmath461 defined for each subcell , we return to the task of determining the swept mass , @xmath426 . the specific linear function @xmath461 for density used in eq . ( [ eq : swept_eq2 ] ) depends upon which subcell the edge encroaches upon . in fact , it is quite possible that the volume swept by the edge intersects more than one immediate neighbor of the subcell in question . of course , an accurate swept - edge integration scheme would take into account all spatial functions in the relevant subcells . @xcite , and references therein , have noted that such an accurate scheme may be cumbersome and computational expensive . instead , they suggest an approximate swept region integration in which only the subcells to the left and right of the oriented edge need matter in determining the spatial function of density used for integration . to this end , an oriented volume integral is defined : @xmath462 whether this volume integral is negative or positive , the density function is taken from either the left or right subcell : @xmath463 while the above scheme is second - order accurate and mass conserving , the approximations made in the swept - edge algorithm can violate local bounds . in practice , the values of the remapped densities should be bounded by the neighborhood values of the original grid . computationally expensive schemes that find the overlap regions among the old and new grid can be made to preserve the bounds . however , the local bounds can be violated in swept - edge remapping . using the linearity - and - bound - preserving method of @xcite , we conservatively repair quantities by redistributing mass , energy , momenta , and number to neighbors satisfying the local bounds of the quantities on the previous lagrangian grid . however , the order in which one processes the cells and subcells influences the specific values that emerge from the repair process . therefore , we employ the order independent scheme of @xcite . these repair schemes are designed to repair subcell densities . for velocities and composition we wish to preserve the bounds of @xmath7 and @xmath13 . therefore , we have appropriately adjusted the scheme to conservatively repair momenta and particle number by preserving the local bounds of @xmath7 and @xmath13 , respectively . the final remap step is to recover the primary variables for the next hydrodynamic step . the new cell - centered and node - centered masses are @xmath464 and @xmath465 respectively . consequently , the subcell and cell - averaged densities are @xmath466 and @xmath467 respectively . recovering the velocities at the nodes is slightly more involved . first , the new subcell - averaged velocity is defined using the remapped momenta and masses : @xmath468 then , for each cell we invert the matrix equation that transforms node velocities into subcell velocities : @xmath469 in this scheme , each cell provides its own velocity for a node , which may be different from another cell s value for the same node . therefore , for each node velocity , we average node velocities resulting from the inversion of eq . ( [ eq : vmatrix ] ) : @xmath470 finally , we recover the specific internal energy . comparing the sum of the old internal energy and kinetic energy with the sum of remapped internal energy and a kinetic energy defined by the remapped node velocities will not necessarily ensure conservation of total energy . therefore , the old kinetic energy is remapped along with the node velocities . the discrepancy between these two kinetic energy representations is then added to the internal energy , ensuring conservation of total energy during the remap stage : @xmath471 \ , .\ ] ] the new cell - centered specific internal energy is then obtained using the modified internal energy , @xmath472 , and the remapped mass , @xmath473 : @xmath474 this ensures conservation of total energy at the expense of consistent remapping of internal energy . whichever is desirable depends upon the problem . for example , in flows with large kinetic energies and small internal energies , the discrepancy in kinetic energy remapping could substantially alter or even dominate the internal energy . consequently , we include a flag in bethe - hydro that determines whether the remapped kinetic energy differences are added to the internal energy . figure [ rhoremap_step ] demonstrates the basic effectiveness of the remapping algorithm . we remap a 1d step function in density many times . from @xmath475 to @xmath476 cm , the density is @xmath477 g @xmath181 , and from @xmath478 to @xmath479 cm , the density is @xmath480 g @xmath181 . this test has 50 cells and @xmath481 nodes . indexing each node by @xmath14 , the positions of the nodes are @xmath482 where @xmath483 and the remapping step is @xmath59 . as a result , the grid completes two full cycles in this remapping test . figure [ rhoremap_step ] shows the result of this remapping test where @xmath484 . the top panel displays the density profile for remapping steps 0 to @xmath485 , while the bottom panel shows the profile for steps @xmath485 to @xmath486 . comparison of the first half ( top panel ) with the second half ( bottom panel ) indicates that the remapping process diffuses the discontinuity over a small number ( @xmath14 ) of zones initially , and `` diffusion '' slows substantially after the initial phase , maintaining a somewhat consistent width in the discontinuity . barring any nuclear or chemical transformations , the composition equation , eq . ( [ eq : advec ] ) , states that @xmath13 is conserved . the lagrangian hydro portion of our algorithm will not change a cell s composition . all alterations in composition are , therefore , a result of advection , and in an ale code advection is handled by the remapping algorithm . equation ( [ eq : advec ] ) may be written in eulerian form : @xmath487 which implies a close dependence of the advection of @xmath13 on @xmath6 and @xmath7 . in practice , if the composition is not advected in a manner entirely consistent with the advection of mass , then the advection of composition will develop peculiarities . since the remapping of density is truly a remapping of the subcell masses , we have designed a scheme for the remapping of composition which operates on an equivalent subcell quantity . specifically , we define new lagrangian quantities for composition . they are the number of species @xmath14 in the cell , @xmath488 , and the number of species @xmath14 in the subcell , @xmath489 . analogous to the mass density is the number density , @xmath490 and @xmath491 . these new quantities are related to the previously defined quantities by : @xmath492 and @xmath493 or @xmath494 . the remap ( or advection ) of composition follows the scheme outlined for the advection of mass on the subcell level , with @xmath495 replacing @xmath6 and @xmath496 replacing mass . the new compositions are then determined by @xmath497 . the final minor , but crucial , difference between mass and composition remapping occurs during the repair process . while the repair process maintains the bounds on the number density , we also enforce bound preservation of the compositions . while we do not address specific rate equations in this work , transport and nuclear processes will require consideration of nonzero terms on the rhs of eqs . ( [ eq : advec ] ) and ( [ eq : advec_eulerian ] ) . since our division of composition into subcell compositions has consequences for discrete implementations with rates , we include here some discussion . normally , the discrete form of eq . ( [ eq : advec ] ) with a nonzero rhs would be designed to operate at the cell level . in other words , it would involve @xmath498 and @xmath499 . however , the fundamental lagrangian subunit is the subcell . therefore , we have developed an algorithm to handle the composition changes , @xmath500 , at the subcell level due to changes at the cell level . irrespective of the scheme employed , a condition which must be satisfied is that @xmath501 we d like to convert the statement of number conservation into the more useful relationship : @xmath502 when the change in composition due to the rates is applied in operator - split form to the discrete hydro equations , we have the following relations : @xmath503 and @xmath504 our dilemma is that the rate equations will determine @xmath505 , but they put no constraint on @xmath506 . a fairly natural and simple choice is @xmath507 , giving @xmath508 substituting eq . ( [ eq : deltansubcell ] ) into eq . ( [ eq : deltancell ] ) we have @xmath509 hence , by simply stating that @xmath507 , we have developed discrete rate equations that satisfy conservation of species number and operates at the subcell level . the remap of the angular momentum deserves special mention . as in mass remapping , we try to find new subcell angular momenta , @xmath510 , based upon an approximate swept - region approach . in other words , @xmath511 which parallels eq . ( [ eq : swept_eq1 ] ) in form . a notable difference , however , is the expression for the swept angular momentum . it no longer involves a simple volume integral , but an integral weighted by @xmath512 ( see appendix [ section : integrals ] for calculating the discrete analog of this integral ) : @xmath513 after finding the new subcell angular momentum we determine the new cell angular momentum , @xmath514 and node angular momentum , @xmath515 and in turn determine the new angular velocity ; @xmath516 in this section , we characterize bethe - hydro s performance using several test problems . first , we demonstrate that this code produces 2@xmath0-order accurate solutions for self - gravitating flows . to assess the accuracy of high mach - number simulations , we use the sod shock tube problem , sedov blast wave , and noh implosion problem , which all have analytic solutions . furthermore , we simulate the saltzman and dukowicz piston problems to test the code s ability to simulate piston - driven shocks using oblique meshes . in simulating two important hydrodynamic instabilities , the rayleigh - taylor and kelvin - helmholtz instabilities , we further demonstrate this code s strengths and limitations . demonstrating bethe - hydro s ability to simulate astrophysical phenomena , we conclude with a core - collapse supernova simulation . to verify the 2@xmath517-order character of bethe - hydro , a smooth hydrodynamics flow is required . the goldreich - weber self - similar collapse ( see [ section : gravity_tests ] and fig . [ gwplot ] ) satisfies this requirement and is ideal to check convergence of the hydrodynamic and gravity solvers and their coupling . we use @xmath518-norms of the error : @xmath519 where the error is @xmath520 , @xmath358 is the analytic density at @xmath521 ms , @xmath522 is the zone size , and @xmath523 is 1 for 1d and 2 for 2d . strictly speaking , the simulations are lagrangian and have time - varying zone sizes . fortunately , this collapse problem is self - similar , implying a direct correlation between the starting resolution and the resolution at a later time . therefore , we use the initial zone size in eq . ( [ eq : l1norm ] ) and in fig . [ gw_resstudy ] . the @xmath518-norm as a function of @xmath524 ( crosses ) is plotted in fig . [ gw_resstudy ] . as the figure demonstrates , both 1d simulations ( top panel ) and 2d simulations ( bottom panel ) converge with roughly 2@xmath517-order accuracy . the sod shock tube problem is a standard analytic test that assesses the code s ability to simulate three characteristic waves : a shock , a rarefaction wave , and a contact discontinuity . a simple gamma - law eos is employed with @xmath525 . initially , the computational domain is divided into left and right static regions with different densities and pressures . specifically , the left and right regions have densities of 1.0 g @xmath181 and @xmath526 g @xmath181 and pressures of 1.0 erg @xmath181 and 0.1 erg @xmath181 , respectively . this gives rise to a self - similar solution involving a shock propagating to the right , a rarefaction wave propagating to the left , and a contact discontinuity in between . in fig . [ sodplot ] , we display the sod shock tube test results for 1d lagrangian , 1d eulerian , 2d lagrangian , and 2d eulerian configurations . cell - centered densities and locations are marked with plus signs , while the analytic results are denoted by solid dark lines . the 1d lagrangian result shows the appropriate density profile , while the other profiles have been shifted vertically so that distinguishing features are more easily compared . in addition , the profiles are further distinguished by displaying them at different times : 1d lagrangian ( @xmath527 s ) , 1d eulerian ( @xmath528 s ) , 2d lagrangian ( @xmath529 s ) , and 2d eulerian ( @xmath530 s ) . the 1d calculations are resolved with 400 zones . similarly , the 2d tests are resolved along the direction of shock propagation with 400 zones , and they are resolved in the perpendicular direction by 10 zones . the overall features of the shock , post - shock material , contact discontinuity , and rarefaction wave are reproduced . for all scenarios , the shock is resolved within a few zones . for the 1d eulerian and lagrangian calculations the post - shock density is good to @xmath531% and @xmath532% , respectively . upon viewing the profiles in greater detail , there is a noticeable departure from the analytic solution at the tail of the rarefaction wave where the density dips below the expected value . this error in density at the tail of the rarefaction wave is @xmath533% for the lagrangian simulation and @xmath389% for the eulerian simulation . since the contact discontinuity moves with the flow speed , both lagrangian calculations resolve the contact discontinuity exactly from one zone to the next and maintain this resolution throughout the simulation . the eulerian calculations , not surprisingly , distribute the contact discontinuity over several ( @xmath14 ) zones . a classic test , the sedov blast wave problem provides a quantitative measure of a code s ability to simulate spherical explosions . initial conditions are set so that the total energy of the sedov blast is 0.244816 ergs , @xmath534 g @xmath181 , @xmath535 ergs g@xmath186 , and the gamma - law eos has @xmath345 . we compare the simulations with the analytic result ( solid lines ) for the following scenarios : 1d lagrangian , 1d eulerian , 2d lagrangian using the butterfly mesh , and 2d lagrangian using the spiderweb mesh . results of the sedov explosion are plotted in fig . [ sedovplot ] . the top panel compares the analytic and numerical density profiles , while the bottom panel shows the relative differences between the numerical solutions and the analytic profile . to easily distinguish different runs and details , we plot the 1d lagrangian calculation at @xmath536 s , the 1d eulerian at @xmath537 s , the 2d lagrangian using the butterfly mesh at @xmath538 s , and the 2d lagrangian using the spiderweb mesh at @xmath539 s. the 1d calculations are resolved with 400 zones . the spiderweb test has a total of 12,381 zones with 200 radial zones and a maximum of 64 angular zones , while the butterfly mesh has a total of 35,000 zones with effectively 200 radial and 200 angular zones . qualitatively , all simulations reproduce the overall structure and position of the shock and post - shock flows . there are some practical issues about formulating the sedov runs that are worth mentioning . for the 1d simulations , we found that simply placing the initial explosion energy in a small number of inner zones was adequate . on the other hand , the 2d simulations using non - spherical grids required more care . simply depositing all the energy within a small number of zones near the center led to severe grid tangling and distortions . we remedied this by initiating all profiles with the analytic solution at @xmath540 s. the bottom panel of fig . [ sedovplot ] emphasizes the quantitative accuracy of the post - shock solutions . plotted are the relative errors of the density , @xmath541 , versus radius , where @xmath6 is the simulated density profile , @xmath358 is the analytic profile , and @xmath542 is the maximum density of the analytic profile . the 1d lagrangian simulation gives the best results with a maximum relative error near the shock of @xmath543% . in comparison , the 1d eulerian test gives @xmath390% deviation near the shock . the 2d lagrangian simulation using the spiderweb mesh has a maximum deviation similar to the 1d lagrangian simulation , even though the 1d simulation has radial zones with half the zone size . the simulation using the spiderweb mesh shows some slight departure from spherical symmetry . specifically , for most of the post - shock region , the peak - to - peak variation of density is less than @xmath389% , and near the shock the peak - to - peak variation of density reaches @xmath544% . the 2d lagrangian simulation using the butterfly mesh has a maximum density deviation of @xmath390% near the shock and a deviation from spherical symmetry of similar magnitude . initial conditions for the noh problem are uniform density @xmath316 g @xmath181 , zero ( or very small ) internal energy @xmath545 ergs , a convergent velocity field with magnitude @xmath546 cm s@xmath186 , and a gamma - law eos with @xmath345 . subsequent evolution produces a symmetric self - similar flow including supersonic accretion , an accretion shock , and stationary post - shock matter in which kinetic energy has been converted into internal energy . the analytic solution for this problem is @xmath547 where @xmath548 is the velocity magnitude , the shock position is @xmath549 , the shock velocity is @xmath550 , and @xmath523 is 2 for 2d cartesian , 3 for 2d simulations using cylindrical coordinates , and 3 for 1d spherically symmetric simulations . in fig . [ noh1dres ] , we present the density profiles at @xmath527 s for 1d , spherically symmetric simulations of the noh problem using four resolutions . the initial grid spans the range @xmath551 cm and is evenly divided into 200 , 400 , 800 , and 1600 zones . the solid line is the analytic solution for @xmath552 . higher resolution simulations more accurately capture the shock position and the post - shock density profile . however , all resolutions depart significantly from the analytic solution near the center . this is `` wall heating '' and is a common problem for lagrangian schemes @xcite . the four panels of fig . [ nohplot_2d ] present the density profiles for 2d simulations . three of the four panels show the results using cartesian coordinates . the top - left panel shows the results using a cartesian grid with 100@xmath553200 zones , the top - right panel shows the results using the butterfly mesh with 22,400 zones , and the lower - left panel shows the results using the spiderweb mesh . the fourth panel , lower - right , shows the results using 2d cylindrical coordinates and a cartesian mesh with 100@xmath553200 zones . obvious is the fact that the mesh employed has consequences for the solution . both the top - left and bottom - right panels indicate that using the cartesian mesh for this problem produces fairly smooth results , with some asymmetry ( @xmath17% ) in the post - shock region . the lower - left panel shows that using the spiderweb mesh produces perfectly symmetric solutions except near the center where the deviation from symmetry is as large as @xmath125% . finally , simulations using the butterfly mesh , top - right panel , shows a similarly mixed capacity to preserve symmetry . in this light , it is important to choose a grid that minimizes numerical artifacts for the problem at hand . a task easily accomplished with the use of ale methods . a standard test for arbitrary grid codes , the saltzman piston problem @xcite addresses the ability of a code to simulate a simple piston - driven shock using a grid with mesh lines oblique to the shock normal . the top panel of fig . [ saltzman ] shows the grid with @xmath554 zones . for a grid with @xmath555 nodes within a domain defined by @xmath556 and @xmath557 , where @xmath558 and @xmath559 , the @xmath19 positions of the nodes are given by @xmath560 where @xmath561 and @xmath562 . at the piston , the left wall is moving at a constant velocity , 1.0 cm s@xmath186 , to the right . initially , the density and internal energy are set equal to 1.0 g @xmath181 and 0.0 ergs , respectively , and we use a gamma - law eos with @xmath345 . in the bottom panel of fig . [ saltzman ] , we show the grid and the density colormap at @xmath563 s after the shock has traversed the domain twice , reflecting off the right and left walls once . in this snapshot , the shock is traveling to the right . our results are to be compared with figs . 15 and 16 of @xcite . figure 15 of @xcite depicts a simulation with severe grid buckling , which we do not observe in our simulations . instead , our fig . [ saltzman ] shows reduced grid buckling and appropriate densities in accordance with the results of fig . 16 of @xcite . the dukowicz piston problem @xcite is another test using an oblique mesh . the initial setup involves two regions . region 1 has a density of 1 g @xmath181 and is resolved with @xmath564 zones . in the vertical domain , @xmath565 cm , and the mesh lines evenly partition the space into 120 sections . dividing region 1 horizontally involves mesh lines with changing orientation . the leftmost mesh line is parallel to the vertical , while the rightmost mesh line is oriented 60@xmath566 relative to the vertical . the mesh lines in between smoothly transition from 0@xmath566 to 60@xmath566 . region 2 has a density of 1.5 g @xmath181 and is gridded with a @xmath567 mesh , with the vertical mesh lines uniformly slanted at 60@xmath566 . including region 1 and region 2 , the bottom boundary spans the range @xmath568 cm and is evenly divided into 304 segments . initially , both regions are in equilibrium with @xmath569 erg @xmath181 . the top , bottom , and right boundaries are reflecting , and the left boundary is a piston with a velocity in the positive @xmath19 direction and a magnitude of 1.48 cm s@xmath186 ( see the top panel of fig . [ dukowicz ] for a low resolution example of this grid ) . a piston - driven shock travels from left to right , and encounters the density jump at an angle of 60@xmath566 , producing a rich set of phenomena . the incident shock continues into the lower density region , a transmitted / refracted shock propagates into the higher density region , a vortex sheet develops behind the transmitted shock , and a reflected shock propagates into the incident shock s post - shock flow ( see the labels in the bottom plot of fig . [ dukowicz ] for visual reference ) . the results are to be compared with the semi - analytic shock - polar analysis presented by @xcite ( see figure 13 and table i of @xcite ) . in their paper , angles subtended by the five regions are presented . in table [ table : dukowicz ] , we recast this information as the angles that the transmitted shock , vortex sheet , reflected shock , and incident shock have with @xmath19-axis . the first row gives the analytic values , which should be compared with the simulated orientations in the second row . strikingly , despite the fact that the incident shock has traversed a grid with an oblique mesh , the simulated and analytic orientations differ very little . the simulated and analytic reflected - shock orientations agree to within @xmath570% , and the orientations of the transmitted shock and vortex sheet agree to within 2% . here , we simulate the growth of a single mode of the rayleigh - taylor instability . a heavy fluid is placed on top of a lighter fluid in the presence of a constant gravitational acceleration . the top and bottom densities are @xmath571 g @xmath181 and @xmath572 g @xmath181 , respectively , and the gravitational acceleration points downward with magnitude @xmath573 cm s@xmath574 . the domain is a rectangle with @xmath575 cm and @xmath576 cm and @xmath577 grid zones . the top and bottom boundaries are reflecting while the left and right boundaries are periodic . for such configurations , small perturbations of the interface between the heavy and light fluids are unstable to exponential growth . assuming that the boundaries are far from the interface , the exponential growth rate of a perturbation with wavenumber @xmath578 is @xmath579 for all single - mode rayleigh - taylor simulations we initiate the perturbation by setting the vertical component of the velocity equal to @xmath580 , where the wavelength , @xmath581 , is @xmath582 cm . therefore , we simulate one wavelength of the mode , and the exponential growth rate ( @xmath583 ) should be 0.65 s@xmath186 . figure [ rt_stills_res ] shows the evolution of a single - mode rayleigh - taylor instability at @xmath394 s for four resolutions . from left to right , the grid sizes are @xmath584 , @xmath585 , @xmath577 , and @xmath586 . gross features compare well , with all the resolutions differing by only a few percent in the maximum and minimum position of the interface . however , higher resolution simulations manifest greater complexity for the kelvin - helmholtz rolls . one can compare the third panel to the results of fig . 4.5 of @xcite . in a general sense , these authors conclude that fewer features imply more dissipation . however , it could be that some of the kelvin - helmholtz rolls are seeded by grid noise in some schemes . next , we discuss the effects of artificial viscosity on the single - mode rayleigh - taylor flow . as stated in [ section : artificial_viscosity ] , artificial viscosity is a requirement for ale algorithms in order to simulate shocks . in the current formulation , there are two parameters of the artificial viscosity scheme . one parameter , @xmath587 , is the coefficient for @xmath588 , which is largest in shocks . effectively , this term provides shock resolution . the second parameter , @xmath370 , is the coefficient of the term that is proportional to @xmath589 and is designed to reduce the amount of post - shock ringing in the solution . typical values suggested for both are 1.0 @xcite . for the single - mode rayleigh - taylor test , the flows are subsonic , so the term multiplied by @xmath370 has the greatest impact on the magnitude of the artificial viscosity forces . each panel of fig . [ rt_stills_visc ] presents the nonlinear rayleigh - taylor flow after 12.75 s for different values of the artificial viscosity parameter , @xmath370 . clearly , @xmath590 ( left panel ) reproduces the expected results for this low - mach - number flow . on the other hand , the @xmath591 run ( center panel ) displays significant departures for the kelvin - helmholtz rolls , while the overall progression of the plumes remains similar . unfortunately , the model with @xmath592 ( right panel ) completely suppresses the kelvin - helmholtz rolls on these scales and plume progression is severely retarded . in fig . [ rt_linear ] , the interface perturbation amplitude is plotted as a function of time ( dashed lines ) . the solid line is the analytic exponential growth rate scaled to the simulation results . the dashed lines show the simulation results for viscosity parameters , @xmath370 , of 0.01 , 0.1 , and 1.0 . three distinct phases are apparent : an early transient phase , a phase in which the slope most closely matches the exponential growth rate , and the subsequent nonlinear phase . the simulations with @xmath590 and 0.1 manifest exponential growth for several e - folding times . the run with @xmath592 , on the other hand , seems to follow the linear phase for only 1 s ( @xmath1 1/2 e - folding ) , if at all . around 5 s , the evolution of the interface amplitude enters the nonlinear phase . for the artificial viscosity scheme that we employ , the rayleigh - taylor instability test indicates that lower values of @xmath370 are preferred . however , tests of the sod shock - tube problem including a parameter study of @xmath370 indicate that unwanted post - shock ringing is diminished only for values above @xmath593 . this represents the primary weakness of bethe - hydro to simulate flows with both shocks and rayleigh - taylor instabilities . to be clear , this does not represent a weakness of ale methods in general , but of the tensor artificial viscosity algorithm designed to mitigate the hourglass instability that we employ . other artificial viscosity schemes such as edge viscosity ( see @xcite for references ) allow for proper development of hydrodynamic instabilities , but do little to suppress the hourglass instability ( milan kucharik , private communication ) . of great interest to users of ale is a methodology that suppresses the hourglass instability and post - shock ringing while enabling proper evolution of hydrodynamic instabilities . another important phenomenon we explore is the kelvin - helmholtz shear instability . @xcite have shown that sph has trouble simulating the kelvin - helmholtz instability when extreme density contrasts are involved . we find that this instability is reasonably well handled in bethe - hydro , and that the evolution during the small amplitude regime is accurately characterized by analytic linear analysis . the calculational domain covers the square region bounded by @xmath594 cm and @xmath595 cm and has @xmath596 zones . the top and bottom boundaries are reflecting , while the left and right boundaries are periodic . for @xmath597 cm , @xmath598 g @xmath181 , and for @xmath599 cm , @xmath600 , where @xmath601 . these regions are in pressure equilibrium with @xmath602 erg @xmath181 . the shearing velocity , @xmath603 is scaled with respect to the slowest sound speed ( sound speed in the bottom region , @xmath604 ) . this relative shearing velocity is split between the top region , which flows to the left with speed @xmath605 , and the bottom region , which flows to the right with the same speed . for this test , the linear coefficient in the viscosity , @xmath370 , is set to 0.01 . a gamma - law eos is used with @xmath345 . the initial perturbation ( see agertz et al . 2007 ) is placed in a small band centered on the interface and is a perturbation in velocity given by @xmath606 where @xmath607 cm . it should be noted that this perturbation is not an eigenmode of the instability . therefore , simulations have a transient phase at the beginning in which this perturbation settles into one or more modes of the instability . we have found that using amplitudes suggested by @xcite , @xmath608 and 1/40 , produces strong transients that complicate the interpretation of the linear regime . to avoid strong long - lasting transients , we set @xmath609 . these initial conditions should produce a perturbation of the interface that grows exponentially in magnitude with an e - folding time given by @xmath610 for the simulations presented here , we consider two shearing velocities : @xmath611 cm s@xmath186 and @xmath612 cm s@xmath186 . hence , @xmath613 is 0.262 s and 0.523 s , respectively . the top panel of fig . [ kh_still ] shows the evolution at @xmath614 s for @xmath615 . the first set of nonlinear kelvin - helmholtz rolls have appeared . qualitatively , the morphology of the rolls is similar to the results presented in figure 13 . of @xcite labeled by `` grid 1 '' at @xmath616 ( there is a factor of @xmath617 difference in the definition of @xmath613 between our work and theirs ) . there are two main differences . for one , we simulate with a factor two larger wavelength to adequately resolve the wavelength for linear analysis . secondly , the time at which we present the results in the top panel of fig . [ kh_still ] corresponds to @xmath618 , not @xmath619 . we have noticed similar discrepancies in the time required to achieve similar evolution using @xmath620 and @xmath621 . additionally , we analyze the growth rate of the interface during the linear regime . we determine the interface amplitude by generating a contour for @xmath622 , measuring the peak to peak amplitude , and dividing by two . the bottom plot of fig . [ kh_still ] shows this amplitude ( solid line ) versus time and compares to the expected exponential growth ( dashed line ) for a simulation with @xmath623 ( green ) and @xmath624 ( blue ) . there are three distinct phases in the log - linear plot : an early transient phase , a phase in which the slope most closely matches the exponential growth rate , and the subsequent nonlinear phase . while the growth rate during the linear regime differs slightly ( @xmath110% ) from theory , the bottom panel of fig . [ kh_still ] demonstrates that the correct dependence of the growth rate on @xmath603 is obtained . incorporating all aspects of bethe - hydro , we simulate 1d and 2d hydrodynamic core - collapse of a 15 m@xmath347 star . we use the s15s7b2 model ( s15 ) of @xcite and the shen eos @xcite and initiate collapse using a @xmath625-@xmath6 parameterization @xcite . while there are no analytic solutions for such a test , we test to see whether the density profiles , timescales , and shock radii , etc . all match experience with other codes and results published @xcite . for 2d simulations , the grid is composed of a butterfly mesh in the interior with a minimum cell size of @xmath10.5 km and extends to 50 km where a spherical grid carries the domain out to 4000 km . in total , there are 23,750 cells , with an effective resolution of @xmath1250 radial and @xmath1100 angular zones . for the best comparison , the 1d grid has 250 zones , mimicking the effective radial resolution of the 2d simulation . figure [ cc_rhoplot ] depicts the density vs. radius for 1d ( lines ) and 2d ( crosses ) at times 0 , 70 , 110 , 130 , 140 , and 150 ms after the start of the calculation . core bounce occurs at 148 ms . other than a @xmath110% difference in the shock radii at 150 ms , 1d and 2d calculations track one another quite well . furthermore , fig . [ cc_gaccplot ] shows an exceptional correspondence between the 1d and 2d gravitational accelerations . using this core - collapse test to best represent the conditions during astrophysical simulations , we describe some timing results of this problem . a standard measure is the average cpu time spent per cell per cycle . using one core of a dual - core amd opteron2.8 ghz processor for this test , the average time spent is @xmath626 cpu seconds / cycle / cell . 80% of which is spent in amg1r6 , the multigrid solver . for this problem , the iterative multigrid solver usually takes @xmath15 to @xmath110 cycles to achieve a fractional residual of @xmath1@xmath627 to @xmath1@xmath628 , where the fractional residual of the linear system @xmath629 is @xmath630 . we re - simulate the collapse of the s15 model , but with an angular velocity profile of the form @xmath631 , where @xmath632 km and @xmath633 radians s@xmath186 . the @xmath188 vs. @xmath21 plot is shown in fig . [ cc_angvelplot ] . core bounce occurs at 151 ms . first of all , total angular momentum is conserved to machine accuracy . since there are no analytic descriptions for the evolution of angular velocity in core collapse scenarios , we can not validate the numerical results via analytic theory . however , the central angular velocity , @xmath634 is proportional to @xmath635 , where @xmath439 is the central density . since the central density compresses from @xmath110@xmath636 g @xmath181 at @xmath637 ms to @xmath12.2 @xmath638 g @xmath181 at @xmath639 ms , @xmath634 should be @xmath11600 radians s@xmath186 at @xmath639 ms . this is consistent with results shown in fig . [ cc_angvelplot ] . furthermore , the angular velocity evolves smoothly with no evidence of axis effects . on the other hand , there is a slight , but noticeable , glitch at the location of the shock . in this paper , we have presented the algorithms employed in bethe - hydro , a new code for 1d and 2d astrophysical hydrodynamic simulations . the hydrodynamic core is an ale algorithm , and its most striking feature is the ability to use arbitrary , unstructured grids . with finite - differencing based upon the support - operator method @xcite of @xcite and @xcite , energy is conserved to roundoff error in the absence of rotation and gravity , and momentum is strictly conserved using cartesian coordinates . for all other circumstances , energy and momentum are conserved accurately , if not precisely . we use a subcell remapping scheme that conservatively remaps mass , momentum , energy , and number density . for 2d calculations using cylindrical coordinates , we include rotational terms in the lagrangian solver , and develop a remapping algorithm that conservatively remaps angular momentum while minimizing unwanted features in the angular velocity near the axis . to provide shock resolution and grid stability , we use the tensor artificial viscosity of @xcite , and to minimize hourglass instabilities , we have developed a subcell pressure method that is a close derivative of the scheme developed by @xcite , but avoids pathological problems when used in conjunction with the subcell remapping algorithm . finally , we have developed a gravity solver for arbitrary grids that uses a support - operator technique for elliptic equations @xcite and an iterative multigrid - preconditioned conjugate - gradient method @xcite to solve the system of linear equations . overall , bethe - hydro offers many unique and useful features for astrophysical simulations . for one , by using ale techniques , the structure of bethe - hydro is straightforward , enabling simple inclusion of a variety of additional physics packages . examples , which will be discussed in future papers , are nuclear networks and time - dependent radiation transport . in contrast with other techniques , such as higher - order godunov methods , no assumptions are made in ale techniques about characteristic waves nor the relationships among important thermodynamic variables . hence , one of its useful features is an ability to use a general eos . most important among bethe - hydro s strengths is the ability to solve self - gravitating hydrodynamic flows on arbitrary grids . this is achieved primarily because the foundation of the lagrangian hydrodynamics solver is an arbitrary , unstructured polygonal grid . furthermore , grids may be time - dependent since the hydrodynamic flow from one timestep is remapped to another arbitrary grid for subsequent evolution . consequently , simulations may be executed using a purely lagrangian , purely eulerian , or an arbitrarily defined time - dependent grid . with bethe - hydro , simulations have great flexibility , tailoring the grid to minimize numerical error and suiting the grid to the computational challenge . ironically , this flexibility leads to bethe - hydro s most prominent weakness , which is shared among all ale codes , the hourglass instability and grid buckling . we use subcell pressure and tensor artificial viscosity algorithms to mitigate these numerical artifacts , but the resolution is imperfect . in [ section : hourglass ] , we present a subcell pressure algorithm that is compatible with the subcell remapping algorithm , but the efficacy of hourglass elimination is slightly compromised . although the tensor artificial viscosity does well to mitigate grid buckling , it can be more resistive to the proper development of hydrodynamic instabilities compared to other artificial viscosity schemes used in ale ( milan kucharik , private communication ) . hence , we diminish the effects of the hourglass instability , but with some unwanted side effects . while mitigation in 2d is tractable , the instability in 3d has many more modes and is not as easily eliminated ( guglielmo scovazzi , private communication ) , making long 3d ale simulations a challenge . throughout this paper , we have not only demonstrated bethe - hydro s flexibility , but have shown that it produces accurate and 2@xmath0-order convergent solutions . with density distributions having analytic potentials , we have shown that the 2d gravity solver gives accurate spherically symmetric potentials when a non - spherical grid is used , it produces accurate non - spherical potentials , and solutions converge with 2@xmath0-order accuracy . further tests demonstrated accurate solutions for self - gravitating hydrostatic and dynamic problems . we have shown that our hourglass elimination algorithm minimizes the hourglass instability and does not present problems when used in conjunction with subcell remapping . in addition , we have quantified the accuracy of hydrodynamic simulations by simulating problems with analytic solutions . simulating piston - driven shocks using oblique meshes , we confirmed we can obtain accurate solutions in the context of arbitrary grids . verifying the code s ability to capture basic hydrodynamic instabilities , we simulate the rayleigh - taylor and kelvin - helmholtz instabilities . concluding the tests , we simulated a supernova core collapse , which demonstrates the ability to simulate complex astrophysical phenomena . simulating hydrodynamic flow is fundamental to understanding most astrophysical objects , and despite the long tradition of hydrodynamic simulations many puzzles remain . this is due primarily to the need to address time - dependent gravitational potentials , complicated equations of state ( eoss ) , flexible grids , multi - d shock structures , and chaotic and turbulent flows . therefore , with bethe - hydro , we introduce a uniquely flexible and functional tool for advancing the theory of complex astrophysical phenomena . we would like to thank mikhail shashkov , raphael loubre , milan kucharik , and burton wendroff for valuable conversations about ale techniques . discussions concerning the 2d gravity solver with ivan hubeny , jim morel , david moulton , and david keyes of tops were very fruitful . we acknowledge support for this work from the scientific discovery through advanced computing ( scidac ) program of the doe , under grant numbers de - fc02 - 01er41184 and de - fc02 - 06er41452 , and from the nsf under grant number ast-0504947 . thanks the joint institute for nuclear astrophysics ( jina ) for support under nsf grant phy0216783 . we thank jeff fookson and neal lauber of the steward computer support group for their invaluable help with the local beowulf cluster , grendel . in eulerian form , the equations for conservation of mass , momentum , and energy are @xmath640 @xmath641 and @xmath642 = - \rho \vec{\nabla } \phi \cdot \vec{v } \ , .\ ] ] in this appendix , we give the exact analytic volumes and weighted - volume integrals in discrete form . since we are dealing with 2d simulations all integrals are 2d integrals , and can be further reduced using green s formula to 1d boundary integrals . for example , the volume of zone @xmath24 in cartesian coordinates is @xmath643 or alternatively , @xmath644 where the direction of the boundary integral is clockwise , @xmath207 indicates an edge of the cell , and subscripts 1 and 2 represent the endpoints of edge @xmath207 . in cylindrical coordinates , the volume integral is @xmath645 throughout bethe - hydro , integrals of linear functions produce volume integrals weighted by @xmath19 or @xmath20 . the discrete form of these integrals are @xmath646 and @xmath647 using cylindrical coordinates these are @xmath648 and @xmath649 ( z_2 - z_1 ) \ , . \end{array}\ ] ] , o. , moore , b. , stadel , j. , potter , d. , miniati , f. , read , j. , mayer , l. , gawryszczak , a. , kravtsov , a. , nordlund , . , pearce , f. , quilis , v. , rudd , d. , springel , v. , stone , j. , tasker , e. , teyssier , r. , wadsley , j. , & walder , r. 2007 , , 380 , 963 , g. , dearborn , d. s. p. , dossa , d. d. , eggleton , p. p. , taylor , a. , castor , j. i. , murray , s. , cook , k. h. , eltgroth , p. g. , cavallo , r. m. , turcotte , s. , keller , s. c. , & pudliner , b. s. 2003 , in astronomical society of the pacific conference series , vol . 293 , 3d stellar evolution , ed . s. turcotte , s. c. keller , & r. m. cavallo , 158381 lll 0 & @xmath360 & @xmath362 + 20 & @xmath669 & @xmath670 + 40 & @xmath671 & @xmath672 + 60 & @xmath673 & @xmath674 + 80 & @xmath675 & @xmath676 + 100 & @xmath677 & @xmath678 + 120 & @xmath679 & @xmath680 + 130 & @xmath681 & @xmath363 +
the study of the disruption of the ordering tendency in low - dimensional antiferromagnets at zero temperature and the subsequent emergence of unconventional phases is an active topic of research , on both theoretical or experimental fronts . frustrating interactions , quantum fluctuations or imposition of a magnetic field are among the main ingredients to destabilize the antiferromagnetic long - range order @xcite . considerable efforts have been devoted to studying and characterizing the resulting phases of matter , which may assume another type of non - magnetic ordering ( such as crystalline states of singlet valence bonds ) , or may not exhibit any kind of local order at all ( _ quantum spin liquids _ @xcite ) . two - dimensional non - bipartite lattice with antiferromagnetic interactions between spin-1/2 moments are amongst the most studied , as antiferromagnetic order is most severely challenged by quantum fluctuations there . the recent numerical findings of topological spin liquid phases in the realistic su(2 ) kagom - lattice antiferromagnet @xcite , or its xxz version @xcite provide examples of stabilization of such exotic phases of matter . previously , @xmath3 topological spin liquids were found in toy models , such as constrained quantum dimer models ( qdm ) @xcite on the frustrated triangular @xcite or kagom @xcite lattices . these simplified models advantageously allow for a better analytical handle and understanding of the physics of spin liquids , but their connection to su(2 ) microscopic hamiltonians is not always direct . another route to realizing exotic phases of matter is to start from a model with highly degenerate manifold of ground states ( _ ice manifold _ ) obeying local constraints and then derive an effective hamiltonian describing the emergent degrees of freedom in this manifold @xcite . following this strategy , balents , fisher and girvin ( bfg ) introduced a spin-1/2 xxz model with extended interactions on the kagom lattice @xcite . using a generalized qdm derived for low energies , these authors argued that this system hosted a topological gapped @xmath3 spin liquid phase @xcite , which was further confirmed numerically in subsequent works @xcite . one of the hallmarks of a @xmath3 liquid phase is the existence of a topological correction to entanglement entropy , which was shown to be present in the bfg model @xcite . in considering the nature of the ( gapped ) phases resulting from the destruction of antiferromagnetic order , one important guiding principle is provided by a general statement about the relation between the presence / absence of an excitation gap and commensurability . for instance , it can be rigorously shown @xcite that a given spin system ( in the absence of magnetic field ) is gapless when the value of total spin per unit cell is an half - odd - integer ( this is the case for , _ e.g. _ , the spin-1/2 heisenberg model on the kagom lattice ) and the ground state is non - degenerate . that is , if we have a featureless liquid - like ground state with an excitation gap , the state necessarily has a _ hidden _ degeneracy ( probably of topological origin ) . a simpler version of this argument @xcite applies to any spin-@xmath4 systems with u(1 ) symmetry and is of direct relevance to magnetization plateaux ( see @xcite for a review ) : a unique ground state with a gap is possible only when the number of spins @xmath5 within a lattice unit cell and the magnetization @xmath6 per site satisfy @xmath7 . recently , a field - theoretical meaning was given to this relation @xcite and it was predicted that when the @xmath8 is a simple ( non - integral ) rational number , we may have gapped featureless liquid phases ( _ i.e. _ , spin - liquid plateaus ) as well as more conventional plateaus with magnetic superstructures . the search for microscopic models that potentially host these spin - liquid plateaus formed in high magnetic fields is quite intriguing in the light of both the field control of spin liquids @xcite and a recent theoretical report @xcite . in this paper , we will investigate the ground - state phases of the bfg model in the presence of a magnetic field , which , as we will see , exhibits large magnetization plateaus at magnetization per site @xmath0 and @xmath1 as well as the one at @xmath9 ( the gapped spin liquid reported in ref . @xcite ) . on these plateaus , the low - energy physics is also well captured by effective constrained models on a triangular lattice : a generalized qdm for the @xmath9 plateau @xcite , the usual qdm for @xmath10 , and a quantum loop model ( qlm ) for @xmath0 . while much is known already about the nature of the @xmath10 plateau thanks to numerous extensive studies @xcite done for the qdm on the triangular lattice , the phase diagram of the qlm , which is of direct relevance to the @xmath0 plateau , is not known to our best knowledge . we will investigate it in details using simple analytical considerations and quantum monte carlo ( qmc ) simulations . the paper is organized as follows . in sec . [ sec : model ] , we present the spin 1/2 model considered in this study and provide the low - energy mapping to constrained models valid for the magnetization plateaus at @xmath0 ( qlm ) and @xmath1 ( qdm ) . we also review some key results obtained previously when no magnetic field is present . [ sec : loop_model ] is devoted to a complete mapping of the phase diagram of the qlm on the triangular lattice . for the qlm with only the kinetic term , that is relevant to the original spin model , we find that the ground state of the qlm displays crystalline behavior with a spontaneous breaking of six - fold rotation symmetry , but not of translation symmetry . we will then turn , in sec . [ sec : spin_model ] , to large - scale numerical simulations of the original microscopic spin model at @xmath0 using direct qmc simulations . computation of the diagonal spin structure factor confirms the existence of a stripe ordering of down spins with a moderately large correlation length . the presence of this crystalline phase is further corroborated using rnyi entanglement entropy . conclusions are given in sec . [ sec : conclusion ] . typical examples of the third - nearest - neighbor interaction between different hexagons [ that are not included in the hamiltonian ] are shown by green dashed lines . others are obtained by space - group operations . ] in the search for simple microscopic models ( e.g. with two - spin interactions ) hosting quantum spin liquid phases , bfg introduced a spin-1/2 xxz model on the kagom lattice @xcite with the aim to reproduce , in a certain regime of parameters , the physics of a ( generalized ) quantum dimer model , known for hosting a @xmath3 topological spin liquid phase . we reproduce here their construction for completeness and obtain the effective hamiltonians for the plateaus . the hamiltonian is defined on the hexagonal plaquettes of the kagom lattice ( see fig . [ fig : kagome_lattice ] ) and reads : @xmath11{hex } } } } \left [ j_{z}\sum_{\langle i , j \rangle\in { \raisebox{-0.3ex}{\includegraphics[scale=0.1]{hex } } } } s^z_i s^z_j - h\sum_{i \in { \raisebox{-0.3ex}{\includegraphics[scale=0.1]{hex } } } } s^z_i \right ] , \\ & h_{xy } = \sum_{{\raisebox{-0.3ex}{\includegraphics[scale=0.1]{hex } } } } j_{xy } \sum_{\langle i , j \rangle\in { \raisebox{-0.3ex}{\includegraphics[scale=0.1]{hex } } } } ( s^{+}_{i}s^{-}_{j } + s^{-}_{i}s^{+}_{j } ) , \end{split } \label{eq : bfg_model}\ ] ] where @xmath12 is the spin-1/2 operator . @xmath13 is the diagonal part of the hamiltonian made up of the ising interactions between first , second and third neighbors _ within _ each plaquette ( see fig . [ fig : kagome_lattice ] for all the corresponding links ) as well as the zeeman term for a magnetic field @xmath14 along the @xmath15-axis . note that the third - neighbor interactions between sites belonging to different hexagons , shown by the green dashed lines in fig . [ fig : kagome_lattice ] , are not included ( see sec . [ sec : mapping - to - qdm ] for the effect of these neglected interactions ) . the second part @xmath16 contains the spin - flip terms between sites within the same hexagon . originally , bfg considered all such terms on each plaquette , but , as will be clear below , setting @xmath17 for the second and third neighbors does not alter the physical behavior [ see eq . ] . compared to the original hamiltonian defined in ref . , we added the magnetic field term and the coefficient of the @xmath18 terms is modified by a factor of @xmath19 . the model can also be reformulated in terms of hardcore bosons , with the boson density playing the role of the magnetization . in this language , the plateaus at @xmath0 and @xmath10 that we will investigate correspond to @xmath1 and @xmath20 filling , respectively ( and their respective particle - hole values ) . and different values of @xmath21 , in the ground state ( @xmath22 is taken as the unit of energy ) . inset : magnetization curve for @xmath23 ( ising limit ) . the three plateaus at magnetizations @xmath24 are clearly visible for the lowest value of the off - diagonal coupling constant @xmath25 , and gradually disappear as @xmath26 gets larger . ] in the following we will be interested in the strongly anisotropic limit @xmath27 where we will see that the dimer physics emerges . since the diagonal part in can be written as : @xmath28{hex } } } } ( s^z_{{\raisebox{-0.3ex}{\includegraphics[scale=0.1]{hex}}}})^2 - h\sum_{{\raisebox{-0.3ex}{\includegraphics[scale=0.1]{hex } } } } s^z_{{\raisebox{-0.3ex}{\includegraphics[scale=0.1]{hex } } } } - \frac{3}{4 } n_{{\raisebox{-0.3ex}{\includegraphics[scale=0.1]{hex } } } } j_z , \label{eq : ice_limit}\ ] ] with @xmath29{hex}}}}$ ] the number of hexagons , every state satisfying the constraint @xmath30{hex}}}}=0,1,2,3 $ ] ( depending on the value of @xmath14 ) on all the plaquettes is a ground state of the classical part @xmath13 . first excited states are separated by a gap of magnitude @xmath18 . as a consequence , the magnetization curve @xmath31 ( with normalization @xmath32 ) displays plateaus of width @xmath18 at the values @xmath33 ( see the inset of fig . [ fig : mofh ] ) and saturation at @xmath19 ( @xmath34 where @xmath35 is the number of down spins per hexagon ) which are expected to survive for a finite value of @xmath21 . as an illustration , we present in fig . [ fig : mofh ] the magnetization curve in the ground state for a moderate system size @xmath36 and finite small values of @xmath21 , as obtained from qmc simulations ( see sec . [ sec : spin_model ] for details and a finite - size scaling analysis of the size of the @xmath0 plateau ) . the above construction leads to magnetization plateaus where the ground states degeneracies scale exponentially with the system size @xcite . dynamics induced by the off - diagonal terms will lift this degeneracy , as presented below . a simple way to visualize the local constraints @xmath30{hex}}}}=0,1,2,3 $ ] ( for all hexagons ) in the ising limit is to faithfully map the ground state configurations to those of dimers : draw a dimer on a link of the triangular lattice formed by the centers of hexagons when the middle of the link is occupied by a down spin , and leave the same link empty for an up spin . the magnetization plateaus @xmath34 are then reproduced by having a constraint such that every triangular lattice site must be touched by @xmath37 dimers . this procedure is illustrated in fig . [ fig : dimer_mapping ] for the @xmath0 ( @xmath38 ) plateau with two dimers per site , which allows us to visualize the ground state spin configurations of the diagonal part @xmath39 as fully - packed self - avoiding loops on the triangular lattice . and @xmath0 in terms of dimers ( red ) on the underlying triangular lattice ( blue ) . loop configurations are formed due to the constraint of two dimers per triangular lattice site . ] the generalized quantum dimer models appear when treating @xmath16 in degenerate perturbation theory . clearly , first - order processes do not contribute , since applying the spin - flip term only once produces configurations with two excited hexagons , thus outside of the ground state manifold . in second - order , two types of processes are allowed . the first process simply flips the same pair of spins twice , merely shifting the total energy by @xmath40{hex}}}}j_{xy}^2/j_z$ ] . the second process involves a flippable bow - tie plaquette , as pictured in green in fig . [ fig : dimer_hopping ] . the two pairs of antiparallel spins are flipped , generating another state fulfilling the ground state constraint . the second - order effective hamiltonian is thus given by the ring - exchange hamiltonian @xcite : @xmath41 where the sum runs over all flippable bow - tie plaquettes ( labels @xmath42 and @xmath43 denote the exterior sites of the bow - ties ) and the ring - exchange amplitude is given by @xmath44 . quite importantly , the sign of @xmath21 is immaterial which allows numerical simulations of the original microscopic hamiltonian ( [ eq : bfg_model ] ) with qmc techniques , by considering the model with a ferromagnetic @xmath45 coupling . also in order to obtain the ring - exchange hamiltonian , only the xy - interactions on the nearest - neighbor ( @xmath46 ) and the next - nearest - neighbor ( @xmath47 ) bonds are crucial : @xmath48 from this , one can see , as stated previously , that suppressing the further - neighbor xy interactions in @xmath16 [ eq . ] does not alter physics qualitatively and simply rescales @xmath49 by a factor 1/2 . we exploit this property in sec . [ sec : spin_model ] . on a flippable bow - tie plaquette ( green ) for @xmath50 . the antiparallel spins of two flippable pairs are exchanged , corresponding to a flip of two parallel dimers on the underlying triangular lattice ( blue ) . ] in the dimer language , the ring move on a bow - tie of the original lattice corresponds to the flip of two parallel dimers on a diamond of the dual triangular lattice ( fig . [ fig : dimer_hopping ] ) . this is nothing but the off - diagonal term of the quantum dimer model @xcite on a triangular lattice @xcite , which , as usual , is conveniently extended with a diagonal potential ( @xmath51 ) term which counts the number of flippable plaquettes ( diamonds ) : @xmath52{plaq } } } \left ( \| \raisebox{-0.65ex}{\includegraphics[scale=0.45]{dimer1 } } \rangle \langle \raisebox{-0.65ex}{\includegraphics[scale=0.45]{dimer2 } } \| + \| \raisebox{-0.65ex}{\includegraphics[scale=0.45]{dimer2 } } \rangle \langle \raisebox{-0.65ex}{\includegraphics[scale=0.45]{dimer1 } } \| \right ) \\ & + v \sum_{{\includegraphics[scale=0.25]{plaq } } } \left ( \| \raisebox{-0.65ex}{\includegraphics[scale=0.45]{dimer1 } } \rangle \langle \raisebox{-0.65ex}{\includegraphics[scale=0.45]{dimer1 } } \| + \| \raisebox{-0.65ex}{\includegraphics[scale=0.45]{dimer2 } } \rangle \langle \raisebox{-0.65ex}{\includegraphics[scale=0.45]{dimer2 } } \| \right ) . \end{split } \label{eq : qdm_triangular}\ ] ] in the ising limit @xmath53 , low - energy physics of the bfg model is expected to be described by this hamiltonian for @xmath54 . the following four - spin interaction @xcite @xmath55 ( @xmath56 label the four spins on a bow - tie in a anti - clockwise way ) reduces , in the same limit , to the @xmath51-term ( @xmath57 ) as it counts the number of flippable plaquettes . for the @xmath10 plateau corresponding to one dimer ( or , one @xmath58 spin ) per hexagon , one may use instead the _ inter - hexagon _ ( two - spin ) third - neighbor interaction @xmath59 that is not contained in model to realize the dimer interaction @xmath60 . thus , one sees that _ the effective model is identical for the three magnetization plateaux , the only difference coming from the constraint on the hilbert space_. for the constraint of one dimer per site ( corresponding to the plateau @xmath10 ) , we have the usual qdm @xcite , while in the case with two dimers per site ( the plateau at @xmath0 ) , a qlm is obtained as announced in the introduction . the original model derived by bfg @xcite describing the @xmath9 plateau corresponds to three dimers per site . table [ tab : effective - model - summary ] summarizes the nature of the effective models for the different plateaus . the mapping to dimer models [ including the diagonal @xmath51-term in eq . ] allows us to benefit from the accumulated knowledge on this family of models @xcite . qdms generically admit a rokhsar - kivelson @xcite ( rk ) point @xmath61 , where the ground - state wave - function is exactly known . for the qdm on the triangular lattice , the ground state is exactly shown to be a topological gapped @xmath3 liquid phase @xcite , which furthermore extends in the region @xmath62 . in the three - dimer model , bfg confirmed the presence of a topological phase at the rk point @xcite , showing in particular that visons are gapped , implying the presence of deconfined fractionalized spinons @xcite . the authors of ref . also speculated that the @xmath3 liquid phase could extend beyond the @xmath54 point corresponding to the original microscopic spin model , in contrast to the one - dimer case @xcite for which the liquid phase ends at @xmath63 @xcite . an exact diagonalization study @xcite of the hamiltonian ( [ eq : ring_hamiltonian ] ) shows the presence of a vison gap in a regime including the bfg point ( @xmath54 ) , thereby supporting the above suggestion . however , the presence of the spin liquid was only firmly established using qmc simulations on large systems . first , a complete numerical phase diagram of the model ( [ eq : bfg_model ] ) with ferromagnetic @xmath21 and @xmath64 was obtained @xcite : at strongly negative @xmath65 , a planar ( _ i.e. _ , superfluid in the bosonic language ) phase accompanied by the breaking of the u(1 ) symmetry is found , as expected . when the magnitude of the spin - flip term was increased , a continuous phase transition to an apparently featureless insulating phase was observed . the phase diagram was also extended to finite temperature @xcite . finally , a direct evidence of the topological nature of the insulator is obtained by computing the topological entanglement ( rnyi ) entropy @xcite , which was evaluated as @xmath66 as expected for a @xmath3 spin liquid phase . as a consequence of the condensation of fractionalized excitations , the transition between the planar phase and the featureless insulator is an exotic 3d @xmath67 quantum critical point @xcite . while the nature of the phases encountered at the @xmath9 and @xmath10 plateaus is well understood from this previous set of study , the ground state physics on the @xmath0 plateau has never been investigated to our best knowledge . in the following , we focus on this situation by performing numerical simulations first on the effective qlm ( sec . [ sec : loop_model ] ) , and then on the microscopic spin model ( sec . [ sec : spin_model ] ) . .[tab : effective - model - summary ] three plateaus of @xmath68 bfg model and the corresponding effective models . [ cols="<,^,^,^",options="header " , ] we consider here the effective model ( [ eq : qdm_triangular ] ) in the case of two dimers per site , where allowed states are represented by configurations of self - avoiding loops such as represented in fig . [ fig : dimer_mapping ] . motivated by the original spin model eq . at the @xmath0 plateau , we are primarily interested in the nature of the ground state of the effective qlm at @xmath69 , however we will also investigate the nature of the surrounding phases in the phase diagram . we first give , in sec . [ sec : loop_general ] , general arguments on the structure of the phase diagram as well as on the topological properties of the loop configurations , before complementing this analysis in sec . [ sec : loop_qmc ] with qmc simulations of the loop model . anticipating the results obtained , we present in fig . [ fig : loop_phase_diag ] the ground - state phase diagram of the qlm on the triangular lattice to illustrate the following discussion . when @xmath70 , the ground - state energy is minimized by configurations where loops form straight lines along one of the three lattice directions , as shown in the left part of the phase diagram presented in fig . [ fig : loop_phase_diag ] . as aligned up and down spins form alternating stripes in the corresponding configurations of the original spin models ( see the right panel of fig . [ fig : qmccorr ] ) , we call this a _ stripe phase_. in this gapped ordered phase , some of the rotations and reflections of the triangular lattice are spontaneously broken ( _ i.e. _ , the point group changes from @xmath71 to @xmath72 ) , leading to the three - fold degenerate ground states ( in the qdm , the corresponding columnar phase further breaks translation symmetry ) . at the rk point @xmath73 , the ground state is given by the equal - weighted sum of all fully - packed loop configurations on the triangular lattice . to our best knowledge , the equivalent classical problem has not been studied before . as the triangular lattice is non - bipartite , we expect that the loop segment correlations are short - ranged , indicating a _ liquid _ phase presumably with a gap , as also found for the rk point of the qdm @xcite . and @xmath74 . the high symmetry points @xmath75 , @xmath76 and @xmath77 required for the non - flippable states are represented . ] when @xmath78 , the ground states are readily found to have zero energy and correspond to non - flippable loop configurations ( _ i.e. _ for which the kinetic term vanishes ) , which is again similar to the staggered phase of the qdm . while the construction of those states is quite straightforward in the qdm , we have found that for @xmath79 systems , the qlm accepts two different families of non - flippable configurations , both of which are shown in the right part of the phase diagram of fig . [ fig : loop_phase_diag ] . the first family is obtained by starting from the staggered states of the qdm @xcite , which are 12-fold degenerate . on the left , one of those 12 states is represented in red , with only two of the three types of links ( corresponding to the three lattice directions 0 , 1 , and 2 in fig . [ fig : brillouin_zone ] ) occupied . a non - flippable loop configuration can then be generated by adding parallel ( blue ) dimers on half of the remaining type of links , for which there are two possibilities ( a one - step translation in the direction of these parallel dimers also creates a different non - flippable state ) . the same prescription can be applied starting from the other qdm staggered states , leading to a total degeneracy of @xmath80 . the second family is obtained by tiling the lattice with the shortest possible triangular loops , as depicted on the right part of fig . [ fig : loop_phase_diag ] . the degeneracy of this family is readily found to be 6 . considering the brillouin zone drawn in the right panel of fig . [ fig : brillouin_zone ] , the first family of non - flippable states requires the cluster to have both @xmath81 and @xmath82 points ( i.e. @xmath83 multiple of 4 ) , while the second one needs the @xmath84 point ( i.e. @xmath83 multiple of 3 ) . we finally remark that the nature of non - flippable states ( or absence thereof ) can be different for other types of clusters ( rectangular clusters for instance ) . from these considerations , the point @xmath54 lies between the two limiting cases @xmath85 ( stripe ) and @xmath86 ( liquid ) . as both the stripe and liquid phases are gapped , they should extend in a finite region of phase space close to these limits , and it is not clear _ a priori _ in which phase the point @xmath54 is located . of course , it is also possible that one or several intervening crystalline phases arise : natural candidates are equivalent of plaquette , or @xmath87 phases observed in qdms @xcite . this can only be determined by exact numerical calculations , which will be presented below in sec . [ sec : loop_qmc ] . invariants @xmath88 associated to the parity of the number of loop segments crossed by the line . in this example , the number of loop segments pointing in the vertical @xmath89 ( horizontal @xmath90 ) direction which cross the line is odd ( even ) , corresponding to @xmath91 ( @xmath92 ) . this number does not depend on the precise location of the line ( parallel lines shifted by one or several lattice units result in the same parity ) . the flipping terms of the hamiltonian , acting for instance on the shaded plaquette ( other flips are possible in this configuration ) , do not change these two parities . note that a loop segment located on a diagonal bond of a plaquette contributes to both @xmath93 and @xmath94 . ] it is interesting to note at this stage that on a manifold with non - zero genus ( cylinder , torus , etc ) , the qlm model possesses topological sectors which can be defined in the same way as in the qdm case . as illustrated in fig . [ fig : loop_winding ] for a torus , the parity of the number of loop segments crossing a dashed line in either the @xmath90 or @xmath89 directions of the torus is conserved by the dynamics of the hamiltonian . on the torus , this defines a pair of @xmath3 invariants @xmath95 and @xmath96 , and thus divides the full hilbert space into four distinct topological sectors , which we label by @xmath88 . the existence of these four topological sectors is the key to identify the @xmath3 spin liquid phase close to @xmath73 , as _ e.g. _ in the qdm on the triangular lattice @xcite . one may wonder if the @xmath3 spin liquid found here is the same as that in the usual triangular - lattice qdm since even ( odd ) number of dimers exiting from each site in the qlm ( qdm ) suggests a mapping to an even ( odd ) ising gauge theory @xcite . however , this difference is not important in considering the underlying nature of the spin liquids . in fact , by explicitly constructing the ground state of the toric code @xcite on a triangular lattice , one can show that both spin liquids share the same quasiparticle ( anyon ) contents and exhibit the same ground - state degeneracy originating from the @xmath2 flux , which we will use in the following . in the topological @xmath2 spin liquid phase , the four topological sectors ( on a torus ) corresponds to the four different ways of threading the @xmath2 flux through the two periods of the system that can not be eliminated by finite - order perturbations @xcite . therefore , the four sectors , in the topological spin liquid phase , are separated from each other by exponentially small gaps ( _ topological gaps _ ) and get degenerate in the thermodynamic limit . as was previously observed for the qdm @xcite , we expect the liquid phase to be detected by the closing of the above topological gaps . on the other hand , in the ordered stripe phase , topological sectors which do not contain the pure stripe configurations ( shown in the left part of fig . [ fig : loop_phase_diag ] ) can not optimize the attractive @xmath51-term and should have a higher energy : correspondingly , the topological gap should be finite even in the thermodynamic limit , providing us with a practical method to distinguish the two phases . one should however be careful in identifying the sectors to which the stripe configurations belong . indeed on a torus with @xmath97 , the qlm can be studied for even and odd @xmath83 ( in contrast to the qdm that can only host even-@xmath83 samples ) . when @xmath83 is odd , the three stripe configurations shown in fig . [ fig : loop_phase_diag ] are located in the three different sectors @xmath98 , @xmath99 and @xmath100 which are strictly identical ( the number of configurations is the same and the corresponding blocks of the hamiltonian are identical ) : we define the topological gap for odd-@xmath83 samples as @xmath101 , where @xmath102 is the ground state energy in the @xmath88 sector . for even @xmath83 , the three stripe configurations belong to the @xmath103 sector , and all other sectors are identical , and we use @xmath104 . we now turn to qmc simulations of the qlm on a torus ( for clusters with the geometry represented in fig . [ fig : brillouin_zone ] , with @xmath97 and @xmath105 sites ) , as defined in eq . . we use the reptation qmc @xcite , which projects a given initial configuration belonging to each topological sector onto the ground state in the same sector . we use a formulation of reptation qmc similar to the one presented in ref . : to accelerate the convergence , we work in a continuous - time formalism as well as with a guiding wave function that provides a finite fugacity for each flippable plaquette of the triangular lattice ( an optimized value for the fugacity is obtained from a short preliminary run ) . we simulated even-@xmath83 samples up to @xmath106 ( and one data point at @xmath107 ) , and odd-@xmath83 samples up to @xmath108 . as a function of @xmath109 , for different sample sizes @xmath83 . the finite - size effects are different for even ( @xmath110 ; left panel ) and odd ( @xmath111 ; right panel ) values of @xmath83 . in the thermodynamic limit , we expect that in both cases , @xmath112 if finite for the stripe phase @xmath113 , and vanishes for the @xmath2 spin liquid phase @xmath114 . left inset : finite - size scaling as a function of @xmath115 for the even-@xmath83 samples provides an estimate @xmath116 of the value where the gap first vanishes . right inset : zoom close to the rk point for odd-@xmath83 samples . ] we first present the results for the topological gap @xmath112 , as obtained by computing the ground state energy in each non - equivalent sector using the standard mixed estimator @xcite . [ fig : loop_gap ] shows @xmath112 as a function of @xmath109 for even ( left panel ) and odd ( right panel ) @xmath83 , for different system sizes . for even samples , the topological gap @xmath110 has a monotonous behavior : it is clearly finite for low values of @xmath109 and vanishes with system size close enough to the rk point @xmath73 . finite - size scaling as a function of @xmath115 ( left inset of fig . [ fig : loop_gap ] ) indicates that the gap vanishes around @xmath116 , separating the stripe phase for @xmath117 from the topological @xmath2 liquid phase for @xmath118 . for the largest samples , our data at @xmath119 and @xmath120 show that the ground state energies in different topological sectors are identical within error bars ( we can not resolve the exponentially small splitting in the liquid phase ) . the gap for the odd-@xmath83 samples displays an apparently different behavior , with the topological sector @xmath103 having a higher energy for low enough @xmath109 , resulting in @xmath121 as the @xmath51-term favors the sectors @xmath98 , @xmath99 and @xmath100 . on the other hand , when the system is close enough to the rk point , the sector @xmath103 has lower energy ( resulting in @xmath122 in our definition of @xmath111 ) for finite - size samples . if we assume that the ordering of the four levels in the spin liquid phase is the same as that for the even-@xmath83 samples , the sign change marks the transition . the value of @xmath109 at which @xmath123 varies with system size and appears to extrapolate to a consistent value for the critical point @xmath116 ( see right inset of fig . [ fig : loop_gap ] ) . here again , for the largest odd-@xmath83 samples , the energies in all topological sectors are indistinguishable within error bars for @xmath119 and @xmath120 , indicating the topological degeneracy . the topological gap for both odd and even sample is clearly finite for @xmath69 ( with a value @xmath124 , as estimated from the even-@xmath83 data , for which there is almost no size dependence ) , which indicates that this point is located in an ordered phase ( presumably , in the stripe phase ) . in order to confirm this , we computed the loop segment occupations ( using middle - slices in the reptile representation @xcite ) in the topological sector @xmath103 ( which hosts all three stripe configurations ) for the even-@xmath83 samples . this allows to determine the stripe structure factor , which we define as : @xmath125 where @xmath126 if there is a loop segment at site @xmath127 occupying a link in the direction @xmath128(@xmath129 ; the three values taken by @xmath128 correspond to the three different directions as represented in the left bottom part of fig . [ fig : brillouin_zone ] ) and @xmath130 otherwise . in the second line of eq . , we have introduced the notation @xmath131 for the total number of bonds in the direction @xmath128 . the structure factor divided by the sample size is expected to converge to the square of the stripe order parameter @xmath132 in the thermodynamic limit : @xmath133 , with a non - zero @xmath134 if the ground state exhibits the long - range stripe order . as a function of inverse system size , for different values of @xmath109 . ] the stripe structure factor @xmath135 is shown in fig . [ fig : loop_sf ] for different values of the coupling @xmath109 as a function of inverse size @xmath136 ( for @xmath137 ) . we expect that @xmath135 should saturate in an exponential way ( due to the finite correlation length ) to @xmath138 in the stripe phase , and to @xmath139 in a phase with no stripe order ( at the critical point , it should decay to @xmath139 with a non - trivial power ) . due to the moderate system sizes accessible , we can not expect to see the exponential decay : the extrapolation to the thermodynamic limit is indeed delicate for large values of @xmath109 . considering the data for @xmath140 ( see fig . [ fig : loop_sf ] ) , we observe that the structure factor extrapolates to a non - zero value . from this data set , we obtained an estimate for the critical point @xmath141 , which is less precise , though consistent , than the one determined above by the topological gap . of the quantifier of @xmath142 anisotropy of histograms of loop occupations , as a function of @xmath109 for three even lattices sizes @xmath137 . insets : color - coded normalized histograms of loop occupations numbers ( using a color interpolation scheme ) , deep in the stripe phase ( left inset , @xmath36 , @xmath143 ) and closer to the critical point ( right inset , @xmath36 , @xmath144 ) . vertical axis of histograms is @xmath145 ( ranging from @xmath146 to @xmath147 ) , horizontal axis is @xmath148 ( from @xmath149 to @xmath150 ) . the histograms are enclosed in an equilateral triangle pictured close to the right inset . ] note that , in analogy with the situation in the qdm , one may suspect that other types of symmetry - breaking order ( such as plaquette order ) could also exhibit a non - zero value of @xmath132 . to clarify this , it is useful to consider the histograms of the occupations of loop segments appearing in the monte carlo simulations , as was done in similar simulations for valence bond crystals @xcite . to have a two - dimensional representation of the histograms for the loop segments occupying one of the three lattice directions , we introduce @xmath151 and @xmath152 ( with @xmath153 being the total number of bonds in the direction @xmath128 ) . the three line configurations shown in the left part of fig . [ fig : loop_phase_diag ] ( having respectively @xmath154 or @xmath155 equal to @xmath156 ) occupy , in the @xmath157 representation , the three vertices @xmath158 of an equilateral triangle in which the histogram is enclosed , as represented in fig . [ fig : loop_cos ] . the left inset of fig . [ fig : loop_cos ] presents the histogram obtained for a moderate system size @xmath36 and @xmath143 , which is already located deep inside the stripe phase : we clearly observe a @xmath159-symmetric feature with three peaks ( corresponding to high occupations ) close to the corners of the triangle corresponding to the three line configurations . the distance of these peaks from the origin @xmath160 gives a measure of the order parameter @xmath132 . on the other hand , the _ angles _ @xmath161 associated to the three peaks in this two - dimensional representation also carry important information on the precise nature of the phase . for instance , the angles associated with the three stripe patterns are @xmath139 , @xmath162 and @xmath163 . other candidate phases ( such as the equivalent of plaquette phases ) would show up in the histograms as sharp peaks at different angles . we find that , as @xmath109 gets closer to the critical point @xmath164 , the typical extents ( radius ) of the histograms get smaller ( as expected from the decrease of the order parameter ) , and that at the same time the shape of the histograms change from a @xmath159-symmetric one to a circular - symmetric one as is exemplified in the right inset of fig . [ fig : loop_cos ] . this is reminiscent of the @xmath165 valence bond histograms observed in the vicinity of deconfined quantum critical points @xcite or even at moderate proximity of the rk point of the qdm on the square lattice @xcite . to quantify this effect , we calculated a measure of the @xmath159-symmetry of the histograms @xmath166 using formulations similar to those in refs . @xcite . for a pure stripe phase , we have @xmath167 , while for a circular histogram @xmath168 . the variation of @xmath169 as a function @xmath109 with different system sizes displayed in the main panel of fig . [ fig : loop_cos ] confirms the behavior observed visually in the histograms : for the region @xmath170 , @xmath169 tends to approach unity with increasing system sizes , while in the region @xmath171 , @xmath169 appears to vanish within the system sizes available to us . while therefore we can not exclude an intermediate different crystalline phase , it appears unlikely as there is no direct evidence either in the energy or in the structure factor of a phase transition ( see a similar situation for the square lattice qdm @xcite ) . thus , we expect that on larger systems @xmath169 would also tend to approach unity for @xmath172 . note that the computations of @xmath169 are difficult as we need long monte carlo runs to be fully ergodic ( _ i.e. , _ in order not to get locked in one type of line configurations , in particular in the stripe phase ) , and also to ensure that statistical fluctuations are not too important when the order parameter is small . we may speculate that these circular histograms are indications of the @xmath159 anisotropy being a dangerously irrelevant variable at the critical point @xmath164 . although the precise characterization of the transition between the @xmath3 spin liquid and the stripe phase is an interesting issue , it is beyond the scope of the present work . at the point @xmath69 of our main interest , our qmc data ( see figs . [ fig : loop_sf ] and [ fig : loop_cos ] ) clearly indicate that the ground state of the effective hamiltonian is located in the stripe phase . the resulting phase diagram obtained from these qmc results as well as the general considerations in sec . [ sec : loop_general ] is given in fig [ fig : loop_phase_diag ] . in this section , we present the results of the qmc simulations for the original spin-1/2 hamiltonian ( [ eq : bfg_model ] ) at the @xmath0 plateau . we used the stochastic series expansion ( sse ) algorithm @xcite combined with a plaquette generalization @xcite that is necessary to circumvent the freezing encountered in the bond formulation of the sse when @xmath173 is large @xcite . in addition to this difficulty , capturing ground state physics requires to reach a very low temperature - range characterized by the small energy scale @xmath174 of the effective model . nevertheless , we have been able to simulate @xmath175 systems , with @xmath176 sites , up to @xmath177 , or slightly beyond @xmath178 for some observables . for numerical ease , we restrict ourselves to the truncated model where only the nearest - neighbor interactions ( i.e. , the black bonds in fig . [ fig : kagome_lattice ] ) are retained in @xmath16 ( see the discussions in sec . [ sec : mapping - to - qdm ] ) . in the following , we set @xmath179 as the energy scale , and use ferromagnetic values @xmath180 ( for the qmc simulations to have no sign problem ) and @xmath181 for the magnetic field ( corresponding to the middle of the @xmath0 plateau ; see fig . [ fig : mofh ] ) . in order to locate the transition point out of the plateau phase as a function of @xmath21 , we performed a finite - size analysis of its width . resulting data are shown in fig . [ fig : plateau_size ] , and the transition point is estimated to be around @xmath182 . for values @xmath183 , the system is thus in a planar phase ( superfluid in the bosonic language ) , and in the following we focus on the plateau region @xmath184 . in sec [ sec : plateau_nature ] , we examine the nature of the plateau phase , and discuss what could be the scenarios for the transition to the planar phase . sec [ sec : plateau_renyi ] is devoted to an alternative characterization of the phase using the rnyi entanglement entropy . plateau width . the plateau is found to disappear around @xmath182 . ] for magnetization plateaus , as has been mentioned in the introduction , it is useful to consider the quantity @xmath8 , where @xmath5 is the number of sites per unit cell ( @xmath185 for the kagom lattice ) . indeed , it has been shown analytically @xcite that when this quantity is fractional , the ground state on this plateau must be degenerate , _ i.e. _ , either in a crystalline state with magnetic superstructures or in a topological state , while a non - degenerate featureless state with a gap is forbidden . if , on the other hand , it takes an integer value , which is the case here at the @xmath0 plateau ( see table [ tab : effective - model - summary ] ) , all these possibilities are allowed . guided by the results of the effective qlm in sec [ sec : loop_model ] , we expect the plateau phase to be in a ordered phase where the down spins form stripes . this is indeed evident when plotting the connected @xmath186-@xmath186 spin correlation function in real space , as shown in the left panel of fig . [ fig : qmccorr ] , where one readily sees the presence of stripes . the corresponding phase breaks the six - fold rotation symmetry , and is three - fold degenerate . cartoon pictures of the spin and dimer configurations in this phase are drawn on the right panel of fig . [ fig : qmccorr ] . in order to confirm the long - range nature of this order , we compute the fourier transform of the above spin - spin correlations , _ i.e. _ , the diagonal spin structure factor : @xmath187 where @xmath188 is the position of spin @xmath189 , @xmath176 , @xmath190 and the average local magnetization is @xmath191 . for an ordered phase at wave - vector @xmath192 , @xmath193 diverges as @xmath156 , and for the three - fold degenerate solid , the three bragg peaks are located at wave - vectors @xmath194 , @xmath195 and @xmath196 . note that this solid can be accommodated on any cluster used for the simulations . finite - size scaling of @xmath197 as a function of @xmath115 for several values of @xmath21 in the plateau phase clearly indicates the presence of the stripe phase ( see fig . [ fig : scalingsq ] ) . in agreement with the qlm study at @xmath69 , the order develops already for small system sizes . inside the magnetization plateau . despite the relatively small available sizes , we can reach the regime of saturation , confirming the striped solid anticipated from the qlm . ] two comments are in order here . first , @xmath193 used in fig . [ fig : scalingsq ] is the average of the values at the three wave - vectors @xmath198 , which is valid when the simulation is perfectly ergodic . however , as mentioned earlier , the large anisotropy @xmath199 in the parameter range of interest can create difficulties in the qmc simulations @xcite . by monitoring the three values @xmath200 separately , we have checked that the simulations are ergodic for the sizes @xmath201 considered here . however , for smaller values of @xmath202 or larger @xmath83 , we indeed noticed a quick loss of ergodicity , even with the plaquette algorithm , translating into very different values at the three wave - vectors . the same behavior was observed in ref . which used a similar qmc algorithm . second , we sometimes found a non - monotonic behavior of the structure factor @xmath197 for larger sizes ( see fig . [ fig : scalingsq ] ) . again , a similar behavior was reported in ref . where the increase of the order parameter was interpreted as the threshold where stripes start developing in the system . in practice , this complicates a proper extrapolation of the order parameter to its thermodynamic limit value . we can however compare the order of magnitude obtained with the value for a perfectly equal superposition of the three ordered states , for which a straightforward calculation yields @xmath203 . we conclude this section by a discussion on the possible scenarios for the quantum phase transition between the plateau ( stripe ) crystal and planar phases . since the stripe crystal breaks discrete lattice symmetries , one expects , on general grounds , that the transition will be of first order type , or that both phases coexist ( in a small region of the phase diagram ) . this latter scenario implies the existence of an intervening supersolid phase , whose presence was found in several similar models on other frustrated lattices @xcite . due to the limited systems sizes available to us , we have not been able to favor one of these two scenarios . we observed ( data not shown ) an apparent crossing in the spin stiffness , with critical exponents compatible with a continuous phase transition , and no sign of a double peak structure in the kinetic energy histograms . however , it has been shown in similar situations that the correct nature of the phase transition may be difficult to capture . in the ( @xmath21,@xmath14 ) plane , varying the xy interaction @xmath21 at the constant field @xmath181 corresponds to a transition point located close to the tip of plateau phase ( _ i.e. _ the tip of the insulator lobe in the bosonic picture ) , for which previous studies have revealed that such a crystal - planar phase transition may appear to be continuous while it is in fact _ weakly first - order _ @xcite . we believe that such a picture should also apply here , although proving this definitively would imply simulations on very large samples , which is currently out of reach for the present model . we have shown , using a conventional approach ( _ i.e. _ by guessing a broken symmetry and then measuring the corresponding structure factor ) , that the @xmath0 plateau corresponds to a stripe crystal shown in fig . [ fig : qmccorr ] . it is interesting to check whether other means could detect the nature of the ground state without a priori knowledge on the broken symmetry . an elegant approach to obtain the dominant order parameter in an unbiased way is provided by the correlation density matrix @xcite , but it is not easily accessible within qmc simulations . instead , we will focus here on using the entanglement entropy as a means to access the nature of the ground state . indeed , in the recent years , a large number of studies have shown how the scaling behavior of a block entanglement entropy can give access to some ground state properties , such as the central charge in one - dimension @xcite or the topological order underlying a given ground state wave function @xcite . generically , any entanglement entropy ( whether rnyi or von neumann ) will scale with the `` area '' of the boundary between the block and the rest of the system ( the so - called `` area - law '' ; see ref . for a review ) , possibly with interesting ( universal ) sub - leading terms : @xmath204 where @xmath5 is the rnyi index , @xmath205 the length of the boundary of block a and @xmath206 a constant for instance . topological phases can be detected using the constructions of levin - wen @xcite or kitaev - preskill @xcite ( see also ref . ) , that enables one to extract a negative constant term @xmath207 , independent of @xmath5 @xcite , where @xmath208 is the topological entanglement entropy . this was used for instance to conclude that the ground state of the bfg model at @xmath9 is indeed a @xmath3 spin liquid with @xmath209 @xcite . the constructions of refs . @xcite use a subtraction of terms which all scale like the area of the subsystems considered , which can be troublesome in qmc simulations which inevitably exhibit finite error bars . computing the scaling of @xmath210 for a single block a ( without using subtraction schemes ) , is in general computationally simpler . this is what we performed for the case of our discrete symmetry breaking ( @xmath211 ) on the @xmath0 plateau , taking for a the geometry of a cylinder block embedded in a torus sample ( see the inset of fig . [ fig : renyi_vs_l ] ) . we expect that for such a symmetry - broken state , the sub - leading term will be a _ positive _ constant @xmath212 , where @xmath213 is the number of degenerate ground states ( note the sign of @xmath214 ; see appendix [ sec : appendixa ] ) . we first remark that , for the simple type of order on the @xmath0 plateau , the precise block geometry is not very important , which allows us to consider the convenient cylinder block geometry . second , it is quite important for this result to hold that all @xmath213 degenerate states ( quasi degenerate on a finite system but with an exponentially small splitting ) are observed in the qmc simulations . hence , ergodicity must be ensured to extract the degeneracy , which we observe is the case if we are not too deep inside the stripe phase and system sizes are not too large . note that such a _ positive _ correction gives a vanishing contribution to the previously mentioned geometric constructions @xcite , as they are precisely built to solely capture long - range entanglement . ( @xmath215 ) on the @xmath0 plateau , as a function of @xmath83 , on various systems with @xmath176 sites . simulations parameters are @xmath216 , @xmath217 and @xmath181 . in each case , the a block corresponds to a half - system cylinder containing @xmath218 unit cells ( see inset ) . the positive correction @xmath219 is shown by a star . ] our numerical data are presented in fig . [ fig : renyi_vs_l ] . we have computed the rnyi entropies @xmath220 and @xmath221 on the @xmath0 plateau at low temperatures using the qmc extended ensemble method proposed in ref . ( with a small modification where the size of a is dynamically varied during the simulation ) . an important first observation is that if one considers only the smallest ( four or five ) system sizes , a naive fit leads to a negative intercept and thus to an apparent topological phase . one could perhaps interpret this unexpected finite - size effect in the following way : close enough to the transition to the superfluid phase , the superfluid correlation length can be of the order of , or larger than , the system size for small enough samples . effectively , the system acquires a superfluid component , which we expect to contribute to @xmath222 as @xmath223 . this form is derived in ref . for a system that spontaneously breaks a continuous symmetry , in which case the logarithm prefactor @xmath224 is universal ( positive and related to the number of goldstone modes ) , and @xmath225 a non - universal constant . assuming such a form is also valid in a system which has effectively both superfluid and crystalline correlations ( _ i.e. _ , supersolid ) , it could result in fits that display a non - universal , possibly negative ( for small enough systems , since otherwise @xmath226 for large enough @xmath83 ) , constant intercept . note that it would be interesting to check this behaviour using simpler bosonic models that exhibit supersolid behaviour . however , when the system is large enough ( beyond @xmath227 in our simulations ) , we observe a qualitative change of behavior in @xmath222 as a function of @xmath83 . note that this length scale coincides with the one where the stripe structure factor starts saturating , as observed in fig . [ fig : scalingsq ] . for these larger systems , entanglement entropies are compatible with a positive intercept @xmath228 , although the accuracy is not excellent . for instance , fitting our @xmath220 data using ( [ eq : sq ] ) and sizes with @xmath229 , we get @xmath230 , which deviates quite significantly from @xmath231 . indeed , as already advocated , a precise determination of @xmath206 is rather difficult due to the need of excellent ergodicity and hence large autocorrelation times for this quantity . typically , we need more than @xmath232 measurements to get reliable data . in search for magnetization plateaus of spin - liquid nature , we have considered the effect of a magnetic field on an anisotropic spin-1/2 model on the kagom lattice that was introduced in ref . . in the easy - axis limit , this model is designed to exhibit several magnetization plateaus , at magnetization per site @xmath233 ( plus saturation @xmath19 ) . the properties of these plateaus have been investigated first by mapping the original spin model to simpler constrained models on the triangular lattice . focusing on the @xmath0 plateau , we have investigated in detail the effective quantum loop model and mapped out its phase diagram using a reptation quantum monte - carlo algorithm . from this , we predict a 3-fold degenerate stripe phase , that breaks rotation symmetry , to appear on this magnetization plateau . we have also observed that the @xmath0 plateau with the stripe order turns into a topological @xmath2 spin - liquid plateau when an additional repulsion among the dimer segments is added . then , we have performed large - scale numerical simulations using the sse quantum monte - carlo algorithm on the original microscopic model . this has confirmed the existence of a stripe phase found in the effective loop model , using both standard structure - factor measurements and its signature in entanglement entropies . in particular , we have observed that , for length scales shorter than @xmath234 , the scaling of the block entanglement entropy may be misleadingly interpreted as giving a _ negative _ intercept @xmath206 , while it becomes clearly _ positive _ for larger systems . this should be used as a caveat in other situations . when the xy - interaction @xmath26 increases , there is a quantum phase transition to a planar phase ( superfluid of the equivalent hardcore bosons ) . while it appears to be continuous in our simulations , we believe that it should appear weakly first - order on larger lattices , as observed for instance in similar physical situations @xcite . we have not investigated in details the @xmath10 plateau , where the effective model is given by the standard quantum dimer model ( _ i.e. _ a single dimer per site ) on the triangular lattice , that has been widely studied in the past @xcite . in fact , the expected crystal in this phase would have a @xmath87 unit cell and an extremely small order parameter @xcite . given that numerical simulations of the microscopic spin model are more involved ( larger systems , many different energy scales etc . ) , it remains challenging to detect such a weak order in the direct simulations for the original bfg spin model . before concluding , let us mention possible extensions of the model ( [ eq : bfg_model ] ) . in sec . [ sec : loop_model ] , we have seen that there exists a gapped @xmath2 spin liquid phase around the rk point of the qlm ( see fig . [ fig : loop_winding ] ) . therefore , it would be interesting to drive the system toward the spin - liquid phase . in fact , it is possible to consider additional interactions which would result in a non - zero interaction term @xmath51 in the effective constrained model ( [ eq : qdm_triangular ] ) . the four - spin interaction @xmath235 [ eq . ] or the inter - hexagon third - neighbor interaction @xmath236 [ eq . ] , that is not included in the model , should do the job . this could allow us to reach the @xmath3 liquid state at both @xmath10 and @xmath0 magnetization plateaus when @xmath237 , while the transition between the crystalline and topological phases could in principle be monitored by examining the rnyi entropy behavior . this extension of the bfg model would certainly be a very interesting playground to investigate such topological phases , their detections and their transitions to conventional symmetry - breaking states . also our construction is not restricted to spin-1/2 systems ( or to the equivalent hardcore - boson systems ) and applies to higher - spin systems as well . in fact , we can obtain the same effective models ( qdm and qlm ) for the spin-1 version of the model with an additional single - ion anisotropy @xmath238 . as a final remark , let us briefly comment on possible experimental realization of the bfg model . because of the artificial interactions required to build the ice manifold of the bfg model , it is difficult to make direct connections between the model we considered and the existing kagome compounds @xcite . however , it remains an interesting prospect to investigate whether such toy models could be realized experimentally in tunable artificial systems . indeed , it was suggested very recently that the bfg interactions and hexagonal plaquettes could be reproduced using a 2d cold ion crystal , with an example given for one and two hexagons @xcite . _ note added : _ while finishing this manuscript , we became aware of the recent preprint by roychowdhury et al . @xcite who recently derived the same loop model for the equivalent hardcore - boson hamiltonian , and further studied the phase diagram of the loop model with a potential term . while we agree on the phase structure of the loop model , the position of the critical point separating the stripe from the @xmath3 liquid phase is noticeably different [ @xmath239 versus @xmath240 ] . we acknowledge useful discussions with g. misguich , r. melko , and d. schwandt . we also thank f. pollmann for the helpful communication on their results . this work was performed using hpc resources from genci ( grants x2014050225 and x2015050225 ) and calmip ( grants 2014-p0677 and 2015-p0677 ) . our qmc sse simulations are based on the code from the alps libraries @xcite . one of the authors ( k.t . ) was supported in part by jsps kakenhi grant no . 24540402 and no . 15k05211 let us consider @xmath213 states @xmath241 , each corresponding to one of the degenerate states in the thermodynamic limit . this would be for instance @xmath242 and @xmath243 in an ising phase , or the three states with down spins occupying one of the three kagom sublattices ( and up spins on the other sites , see fig . [ fig : qmccorr ] ) , in the stripe crystal phase discussed in the main text . these @xmath213 states are always orthogonal in the thermodynamic limit . since they are simple product states , one can readily show , for any block @xmath81 of the system which complement scales with system size , that the reduced density matrix corresponding to a cat superposition of these states , @xmath244 , is given by : @xmath245 note that this equality only holds for the reduced density matrix , and not the whole density matrix . on any finite system , since these @xmath213 states are exponentially close in energy , and a qmc simulation would sample each @xmath246 with equal probabilities . hence , one would compute a reduced density matrix corresponding to a mixed state : @xmath247 which is precisely the same object as eq . ( [ eq : cat ] ) . so our qmc simulation is able to access the reduced density matrix of the cat state , provided it is ergodic . now , each @xmath248 is excessively simple for each product state , as its spectrum contains only one non - zero eigenvalue @xmath249 since they are not entangled . moreover , they can be diagonalized simultaneously ( they are diagonal in the @xmath186 basis in our example ) , which leads to an entanglement spectrum with non - zero eigenvalues @xmath250 with degeneracy @xmath213 . from it , the entanglement entropy ( von neumann or renyi ) is simply @xmath251 . this results looks quite intuitive indeed and was conjectured in ref . it was already noticed in ref . @xcite for a cat state made of @xmath213 degenerate ising ground states . we emphasize that the same result holds for the mixed state obtained in our qmc simulation . for more realistic wave - functions , the fluctuations around the ideal product state gives and area law , but the constant term will be unchanged : @xmath252 as a last remark , we would like to emphasize that in other more complex cases ( e.g. for states which are not one - site product states , such as valence - bond columnar states ) , the discussion is more involved as the constant term can now depend explicitly on the form of the cut as well as on the rnyi index @xmath5 . 74ifxundefined [ 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 , , and , eds . , link:\doibase 10.1007/978 - 3 - 642 - 10589 - 0 [ _ _ ] , ed . ( , ) link:\doibase 10.1038/nature08917 [ * * , ( ) ] link:\doibase 10.1126/science.1201080 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.067201 [ * * , ( ) ] link:\doibase 10.1038/nphys2465 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.114.037201 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.61.2376 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.86.1881 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.89.137202 [ * * , ( ) ] link:\doibase 10.1103/physrevb.65.224412 [ * * , ( ) ] link:\doibase 10.1103/physrevb.66.205104 [ * * , ( ) ] http://link.aps.org/abstract/prb/v69/e064404 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.94.146805 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.97.207204 [ * * , ( ) ] link:\doibase 10.1126/science.1212207 [ * * , ( ) ] link:\doibase 10.1038/nphys2036 [ * * , ( ) ] link:\doibase 10.1103/physrevb.69.104431 [ * * , ( ) ] link:\doibase 10.1007/s00220 - 007 - 0342-z [ * * , ( ) ] link:\doibase 10.1103/physrevlett.84.1535 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.064412 [ * * , ( ) ] http://dx.doi.org/10.1038/nature09910 [ * * , ( ) ] link:\doibase 10.1038/ncomms3287 [ * * , ( ) ] link:\doibase 10.1103/physrevb.71.224109 [ * * , ( ) ] link:\doibase 10.1103/physrevb.71.020401 [ * * , ( ) ] link:\doibase 10.1103/physrevb.74.134301 [ * * , ( ) ] link:\doibase 10.1103/physrevb.74.054402 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.127202 [ * * , ( ) ] link:\doibase 10.1103/physrevb.77.134421 [ * * , ( ) ] http://stacks.iop.org/1367-2630/13/i=8/a=085001 [ * * , ( ) ] link:\doibase 10.1103/physrevb.66.174405 [ * * , ( ) ] link:\doibase 10.1103/physrevb.66.214513 [ * * , ( ) ] link:\doibase 10.1103/physrevb.62.7850 [ * * , ( ) ] link:\doibase 10.1103/physrevb.76.224431 [ * * , ( ) ] link:\doibase 10.1103/physrevb.81.205102 [ * * , ( ) ] link:\doibase 10.1103/physrevb.73.245105 [ * * , ( ) ] link:\doibase 10.1103/physrevb.65.024504 [ * * , ( ) ] http://www.sciencedirect.com/science/article/b6wb1-47hjx86-1/2/45c94ccbbe3ebdfc944a7aa4ac9c932f [ * * , ( ) ] link:\doibase 10.1103/physrevb.44.2664 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.82.4745 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.227202 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.180414 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.024406 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.134407 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.087203 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.212406 [ * * , ( ) ] link:\doibase 10.1103/physrevb.90.245143 [ * * , ( ) ] @noop ( ) link:\doibase 10.1103/physrevlett.99.207203 [ * * , ( ) ] link:\doibase 10.1103/physrevb.43.5950 [ * * , ( ) ] link:\doibase 10.1103/physreve.66.046701 [ * * , ( ) ] link:\doibase 10.1103/physrevb.70.100410 [ * * , ( ) ] link:\doibase 10.1103/physrevb.74.214517 [ * * , ( ) ] http://stacks.iop.org/0953-8984/19/i=14/a=145203 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.95.127205 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.95.127206 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.95.127207 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.97.147202 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.97.067202 [ * * , ( ) ] link:\doibase 10.1103/physrevb.76.235107 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.212402 [ * * , ( ) ] http://stacks.iop.org/1742-5468/2004/i=06/a=p06002 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.96.110405 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.96.110404 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.82.277 [ * * , ( ) ] link:\doibase 10.1103/physrevb.75.214407 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.103.261601 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.235116 [ * * , ( ) ] @noop ( ) , `` , '' ( , ) chap . , p. link:\doibase 10.1088/1367 - 2630/17/6/065018 [ * * , ( ) ] link:\doibase 10.1103/physrevb.92.075141 [ * * , ( ) ] http://stacks.iop.org/1742-5468/2011/i=05/a=p05001 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.024424 [ * * , ( ) ]
partial orders are a principle tool for modelling `` true concurrency '' semantics of concurrent systems ( cf . @xcite ) . they are utilized to develop powerful partial - order based automatic verification techniques , e.g. , _ partial order reduction _ for model checking concurrent software ( see , e.g. , ( * ? ? ? * chapter 10 ) and @xcite ) . partial orders are also equipped with _ traces _ , their powerful formal language counterpart , proposed by mazurkiewicz @xcite . in _ the book of traces _ @xcite , trace theory has been used to tackle problems from diverse areas including formal language theory , combinatorics , graph theory , algebra , logic , and _ concurrency theory_. however , while partial orders and traces can sufficiently model the earlier than " relationship , janicki and koutny argued that it is problematic to use a single partial order to specify both the earlier than " and the not later than " relationships @xcite . this motivates them to develop the theory of _ relational structures _ , where a pair of relations is used to capture concurrent behaviors . the most well - known among the classes of relational structures proposed by janicki and koutny is the class of _ stratified order structures _ ( _ so - structures _ ) @xcite . a so - structure is a triple @xmath0 , where @xmath1 and @xmath2 are binary relations on @xmath3 . they were invented to model both the earlier than " ( the relation @xmath1 ) and not later than " ( the relation @xmath2 ) relationships , under the assumption that system runs are described by _ stratified partial orders _ , i.e. , step sequences . they have been successfully applied to model inhibitor and priority systems , asynchronous races , synthesis problems , etc . ( see for example @xcite and others ) . _ com__bined _ trace _ ( _ comtrace _ ) notion , introduced by janicki and koutny @xcite , generalizes the trace notion by utilizing step sequences instead of words . first the set of all possible steps that generates step sequences are identified by a relation @xmath4 , which is called _ simultaneity_. second a congruence relation is determined by a relation @xmath5 , which is called _ serializability _ and in general _ not _ symmetric . then a comtrace is defined as a finite set of congruent step sequences . comtraces were introduced as a formal language representation of so - structures to provide an operational semantics for petri nets with inhibitor arcs . unfortunately , comtraces have been less often known and applied than so - structures , even though in many cases they appear to be more natural . we believe one reason is that the comtrace notion was too succinctly discussed in @xcite without a full treatment dedicated to comtrace theory . motivated by this , janicki and the author have devoted our recent effort on the study of comtraces @xcite , yet there are too many different aspects to explore and the truth is we can barely scratch the surface . in particular , the huge amount of results from trace theory ( e.g. , from @xcite ) desperately needs to be generalized to comtraces . these tasks are often non - trivial since we are required to develop intuition and novel techniques to deal with the complex interactions of the `` earlier than '' and `` not later than '' relations . this paper gives a novel characterization of comtraces using labeled so - structures . such definition is interesting for the following reasons . first , it defines exactly the class of labeled so - structures that can be represented by comtraces . it is worth noting that this point is particularly important . even though it was shown in @xcite that every comtrace can be represented by a labeled so - structure , the converse could not be shown because a class of labeled so - structures that defines precisely the class of comtraces was not known . the closest to our characterization is the _ combined dependency graph _ ( _ cd - graph _ ) notion ( analogous to _ dependence graph _ representation of traces ) introduced recently by kleijn and koutny @xcite , but again a theorem showing that combined dependency graphs can be represented by comtraces was not given . our approach is quite different and based on some new ideas discussed in section 4 of this paper . second , even though the step sequence definition of comtraces is more suitable when dealing with formal language aspects of comtraces , the labeled so - structure representation is more suitable for a variety of powerful order - theoretic results and techniques available to us ( cf . @xcite ) . finally , the labeled so - structure definition of comtrace can be easily extended to _ infinite comtraces _ , which describe nonterminating concurrent processes . the labeled poset representation of infinite traces is already successfully applied in both theory and practice , e.g. , @xcite . although such definition is equivalent to the one using quotient monoid over infinite words @xcite , we believe that infinite labeled posets are sometimes simpler . indeed the celebrated work by thiagarajan and walukiewicz ( cf . @xcite ) on linear temporal logic for traces utilizes the labeled poset characterization of infinite traces , where _ configurations _ of a trace are conveniently defined as _ finite downward closed subsets _ of the labeled poset representation . we will not analyze infinite comtraces or logics for comtraces in this paper , but these are fruitful directions to explore using the results from this paper . the paper is organized as follows . in section 2 , we recall some preliminary definitions and notations . in section 3 , we give a concise exposition of the theory of so - structures and comtraces by janicki and koutny @xcite . in section 4 , we give our definition of comtraces using labeled so - structure and some remarks on how we arrived at such definition . in section 5 , we prove a representation theorem showing that our comtrace definition and the one by janicki and koutny are indeed equivalent ; then using this theorem , we prove another representation theorem showing that our definition is also equivalent to the cd - graph definition from @xcite . in section 6 , we define _ composition _ operators for our comtrace representation and for cd - graphs . finally , in section 7 , some final remarks and future works are presented . the _ powerset _ of a set @xmath3 will be denoted by @xmath6 , i.e. @xmath7 . the set of all _ non - empty _ subsets of @xmath3 will be denoted by @xmath8 . in other words , we let @xmath10 denote the _ identity relation _ on a set @xmath3 . if @xmath11 and @xmath12 are binary relations on a set @xmath3 ( i.e. , @xmath13 ) , then their _ composition _ @xmath14 is defined as @xmath15 . we also define @xmath16 the relations @xmath17 and @xmath18 are called the _ ( irreflexive ) transitive closure _ and _ reflexive transitive closure _ of @xmath11 respectively . a binary relation @xmath19 is an _ equivalence relation _ relation on @xmath3 if and only if ( iff ) @xmath11 is _ reflexive _ , _ symmetric _ and _ transitive_. if @xmath11 is an equivalence relation , then for every @xmath20 , the set @xmath21_r { \triangleq}{\{y\;\|\;y\;r\;x \wedge y\in x\}}$ ] is the equivalence class of @xmath22 with respect to @xmath11 . we also define @xmath23_r\mid x\in x\}}$ ] , i.e. , the set of all equivalence classes of @xmath3 under @xmath11 . we drop the subscript and write @xmath21 $ ] when @xmath11 is clear from the context . a binary relation @xmath24 is a _ partial order _ iff @xmath11 is _ irreflexive _ and _ transitive_. the pair @xmath25 in this case is called a _ partially ordered set _ ( _ poset _ ) . the pair @xmath25 is called a _ finite poset _ if @xmath3 is finite . for convenience , we define : @xmath26 a poset @xmath25 is _ total _ iff @xmath27 is empty ; and _ stratified _ iff @xmath28 is an equivalence relation . evidently every total order is stratified . for every finite set @xmath3 , a set @xmath29 can be seen as an alphabet . the elements of @xmath30 are called _ steps _ and the elements of @xmath31 are called _ step sequences_. for example , if the set of possible steps is @xmath32 , then @xmath33 is a step sequence . the triple @xmath34 , where @xmath35 denotes the step sequence concatenation operator ( usually omitted ) and @xmath36 denotes the empty step sequence , is a monoid . let @xmath37 be a step sequence . we define @xmath38 , the number of occurrences of an event @xmath39 in @xmath40 , as @xmath41 , where @xmath42 denotes the cardinality of the set @xmath3 . then we can construct its unique _ enumerated step sequence _ @xmath43 as + @xmath44 + we will call such @xmath45 an _ event occurrence _ of @xmath46 . for instance , if we let @xmath47 , then @xmath48 we let @xmath49 denote the set of all event occurrences in all steps of @xmath50 . for example , when @xmath47 , @xmath51 we also define @xmath52 to be the function that returns the _ label _ of @xmath53 for each @xmath54 . for example , if @xmath55 , then @xmath56 . hence , from an enumerated step sequence @xmath57 , we can uniquely reconstruct its step sequence @xmath58 for each @xmath59 , we let @xmath60 denote the consecutive number of a step where @xmath53 belongs , i.e. , if @xmath61 then @xmath62 . for our example , @xmath63 , @xmath64 , etc . it is important to observe that step sequences and stratified orders are interchangeable concepts . given a step sequence @xmath65 , define the binary relation @xmath66 on @xmath67 as + @xmath68 intuitively , @xmath69 simply means @xmath53 occurs before @xmath70 on the step sequence @xmath65 . thus , @xmath71 iff @xmath72 ; and @xmath73 iff @xmath74 . obviously , the relation @xmath66 is a stratified order and we will call it the stratified order _ generated by the step sequence _ @xmath65 . conversely , let @xmath75 be a stratified order on a set @xmath76 . the set @xmath76 can be represented as a sequence of equivalence classes @xmath77 ( @xmath78 ) such that @xmath79 the sequence @xmath80 is called the step sequence _ representing _ @xmath75 . a detailed discussion on this connection between stratified orders and step sequences can be found in @xcite . in this section , we review the janicki koutny theory of stratified order structures and comtraces from @xcite . the reader is also referred to @xcite for an excellent introductory survey on the subject with many motivating examples . a _ relational structure _ is a triple @xmath81 , where @xmath3 is a set and @xmath82 , @xmath83 are binary relations on @xmath3 . a relational structure @xmath84 is an _ extension _ of @xmath85 , denoted as @xmath86 , iff @xmath87 , @xmath88 and @xmath89 . a _ stratified order structure _ ( _ so - structure _ ) is a relational structure @xmath90 such that for all @xmath91 , the following hold : @xmath92 when @xmath3 is finite , @xmath12 is called a _ finite so - structure_. [ def : sos ] the axioms imply that @xmath93 is a partial order and @xmath94 the axioms and imply @xmath95 is a _ strict preorder_. the relation @xmath1 is called _ causality _ and represents the earlier than " relationship while the relation @xmath2 is called _ weak causality _ and represents the not later than " relationship . the axioms model the mutual relationship between earlier than " and not later than " relations , provided that _ the system runs are stratified orders_. historically , the name `` stratified order structure '' came from the fact that stratified orders can be seen as a special kind of so - structures . for every stratified poset @xmath96 , the triple @xmath97 is a so - structure.[prop : soss ] we next recall the notion of _ stratified order extension_. this concept is extremely important since the relationship between stratified orders and so - structures is exactly analogous to the one between total orders and partial orders . let @xmath98 be a so - structure . a _ stratified _ order @xmath75 on @xmath3 is a _ stratified extension _ of @xmath12 if and only if @xmath99 . the set of all stratified extensions of @xmath12 is denoted as @xmath100 . [ def : extsos ] szpilrajn s theorem @xcite states that every poset can be reconstructed by taking the intersection of all of its total order extensions . janicki and koutny showed that a similar result holds for so - structures and stratified extensions : let @xmath101 be a so - structure . then + @xmath102 [ theo : szpstrat ] using this theorem , we can show the following properties relating so - structures and their stratified extensions . for every so - structure @xmath101 , 1 . @xmath103 2 . @xmath104 [ cor : szpstrat ] * 1 . * see ( * ? ? ? * theorem 3.6 ) . * 2 . * follows from * 1 . * and theorem [ theo : szpstrat ] . _ comtraces _ were introduced in @xcite as a generalization of traces to represent so - structures . the _ comtrace congruence _ is defined via two relations _ simultaneity _ and _ serializability_. let @xmath105 be a finite set ( of events ) and let @xmath106 be two relations called _ serializability _ and _ simultaneity _ respectively and the relation @xmath4 is irreflexive and symmetric . the triple @xmath107 is called a _ comtrace alphabet_. [ def : comalpha ] intuitively , if @xmath108 then @xmath39 and @xmath109 can occur simultaneously ( or be a part of a _ synchronous _ occurrence in the sense of @xcite ) , while @xmath110 means that @xmath39 and @xmath109 may occur simultaneously or @xmath39 may occur before @xmath109 . we define @xmath111 , the set of all possible _ steps _ , to be the set of all cliques of the graph @xmath112 , i.e. , + @xmath113 for a comtrace alphabet @xmath114 , we define @xmath115 to be the relation comprising all pairs @xmath116 of step sequences such that + @xmath117 + where @xmath118 and @xmath119 , @xmath120 , @xmath121 are steps satisfying @xmath122 and @xmath123 . we define _ comtrace congruence _ @xmath124 . we define the comtrace concatenation operator @xmath125 as @xmath126\circledast[t ] { \triangleq}[r\ast t]$ ] . the quotient monoid @xmath127)$ ] is called the monoid of _ comtraces _ over @xmath128 . [ def : commonoid ] note that since @xmath5 is irreflexive , @xmath129 implies that @xmath130 . we will omit the subscript @xmath128 from the comtrace congruence @xmath131 , and write @xmath132 and @xmath133 when it causes no ambiguity . to shorten our notations , we often write @xmath134_{\theta}$ ] or @xmath134 $ ] instead of @xmath134_{{\equiv}_{\theta}}$ ] to denote the comtrace generated by the step sequence @xmath135 over @xmath128 . let @xmath136 where @xmath39 , @xmath109 and @xmath137 are three atomic operations , where + @xmath138 + assume simultaneous reading is allowed . then only @xmath109 and @xmath137 can be performed simultaneously , and the simultaneous execution of @xmath109 and @xmath137 gives the same outcome as executing @xmath109 followed by @xmath137 . we can then define the comtrace alphabet @xmath114 , where @xmath139 and @xmath140 . this yields @xmath141 . thus , @xmath142 = { \bigl\ { \{a\}\{b , c\},\{a\}\{b\}\{c\}\bigr\}}$ ] is a comtrace . but @xmath143 . [ ex : comtrace1 ] even though traces are quotient monoids over sequences and comtraces are quotient monoids over step sequences , traces can be regarded as special kinds of comtraces when the relation @xmath144 . for a more detailed discussion on this connection between traces and comtraces , the reader is referred to @xcite . let @xmath145 . we define the relations @xmath146 as : 1 . @xmath147 2 . @xmath148 . [ def : s2inv ] it is worth noting that the structure @xmath149 is exactly the _ cd - graph _ ( cf . definition [ def : comdag ] ) that represents the comtrace @xmath150 $ ] . this gives us some intuition on how koutny and kleijn constructed the cd - graph definition in @xcite . we also observe that @xmath151 is usually _ not _ a so - structure since @xmath152 and @xmath153 describe only basic `` local '' causality and weak causality invariants of the event occurrences of @xmath65 by considering pairwise serializable relationships of event occurrences . hence , @xmath152 and @xmath153 might not capture `` global '' invariants that can be inferred from of definition [ def : sos ] . to ensure all invariants are included , we need the following @xmath154-closure operator . for every relational structure @xmath155 we define @xmath156 as + @xmath157 [ def : so - cl ] intuitively @xmath154-closure is a generalization of transitive closure for relations to relational structures . the motivation is that for appropriate relations @xmath82 and @xmath83 ( see assertion ( 3 ) of proposition [ prop : so - cl ] ) , the relational structure @xmath158 is a so - structure . the @xmath154-closure operator satisfies the following properties : let @xmath155 be a relational structure . 1 . if @xmath83 is irreflexive then @xmath159 . 2 . @xmath160 . @xmath156 is a so - structure if and only if @xmath161 is irreflexive . 4 . if @xmath12 is a so - structure then @xmath162 . 5 . if @xmath12 be a so - structure and @xmath163 , then @xmath164 and @xmath165 is a so - structure . [ prop : so - cl ] given a step sequence @xmath145 and its respective comtrace @xmath166\in { \mathbb{s}}_{\theta}^/\!\equiv$ ] , we define the relational structures @xmath167 as : + @xmath168.[def : s2sos ] the relational structure @xmath167 is called the _ so - structure defined by the comtrace _ @xmath166 $ ] , where @xmath169 , @xmath170 and @xmath171 are used to denote the event occurrence set , causality relation and weak causality relation induced by the comtrace @xmath172 respectively . the following nontrivial theorem and its corollary justifies the name by showing that step sequences in a comtrace @xmath172 are exactly stratified extension of the so - structure @xmath173 , and that @xmath167 is uniquely defined for the comtrace @xmath172 regardless of the choice of @xmath174 . for each @xmath175 , the relational structure @xmath167 is a so - structure and @xmath176 . [ theo : com2sos ] for all @xmath177 , 1 . @xmath178 2 . @xmath179 [ cor : com2sos ] even though theorem [ theo : com2sos ] shows that each comtrace can be represented uniquely by a labeled so - structure , it does not give us an explicit definition of how these labeled so - structures look like . in this section , we will give an exact definition of labeled so - structures that represent comtraces . to provide us with more intuition , we first recall how mazurkiewicz traces can be characterized as labeled posets . a _ trace concurrent alphabet _ is a pair @xmath180 , where @xmath181 is a symmetric irreflexive binary relation on the finite set @xmath105 . a _ trace congruence _ @xmath182 can then be defined as the smallest equivalence relation such that for all sequences @xmath183 , if @xmath184 , then @xmath185 . the elements of @xmath186 are called _ traces_. traces can also be defined alternatively as posets whose elements are labeled with symbols of a concurrent alphabet @xmath180 satisfying certain conditions . given a binary relation @xmath187 , the _ covering relation _ of @xmath11 is defined as @xmath188 . an alternative definition of mazurkiewicz trace is : a trace over a concurrent alphabet @xmath180 is a finite labeled poset @xmath189 , where @xmath190 is a labeling function , such that for all @xmath191 , 1 . @xmath192 , and 2 . @xmath193 . [ def : ltraces ] a trace in this definition is only identified unique up to _ label - preserving isomorphism_. the first condition says that immediately causally related event occurrences must be labeled with dependent events . the second condition ensures that any two event occurrences with dependent labels must be causally related . the first condition is particularly important since two immediately causally related event occurrences will occur next to each other in at least one of its linear extensions . this is the key to relate definition [ def : ltraces ] with quotient monoid definition of traces . thus , we would like to establish a similar relationship for comtraces . an immediate technical difficulty is that weak causality might be cyclic , so the notion of `` immediate weak causality '' does not make sense . however , we can still deal with cycles of a so - structure by taking advantage of the following simple fact : _ the weak causality relation is a strict preorder_. let @xmath101 be a so - structure . we define the relation @xmath194 as + @xmath195 since @xmath95 is a strict preorder , it follows that @xmath196 is an equivalence relation . the relation @xmath196 will be called the _ @xmath95-cycle equivalence relation _ and an element of the quotient set @xmath197 will be called a _ @xmath95-cycle equivalence class_. we then define the following binary relations @xmath198 and @xmath199 on the quotient set @xmath197 as @xmath200 \widehat{{\prec } } [ \beta ] { \stackrel{\textit{\scriptsize{df}}}{\iff}\ } ( [ \alpha]\times[\beta])\;\cap { \prec}\not=\emptyset \hspace{3mm}\text { and } \hspace{3 mm } [ \alpha ] \widehat{{\sqsubset } } [ \beta ] { \stackrel{\textit{\scriptsize{df}}}{\iff}\ } ( [ \alpha]\times[\beta])\;\cap { \sqsubset}\not=\emptyset \label{qsos}\end{aligned}\ ] ] using this quotient construction , every so - structure , whose weak causality relation might be cyclic , can be uniquely represented by an _ quotient so - structure . the relational structure @xmath201 is a so - structure , the relation @xmath199 is a partial order , and for all @xmath202 , 1 . @xmath203 \widehat{{\prec } } [ \beta]$ ] 2 . @xmath204 \widehat{{\sqsubset } } [ \beta ] \vee ( \alpha\not=\beta \wedge [ \alpha]=[\beta])$ ] follows from definition [ def : sos ] . using ( [ qsos ] ) and theorem [ theo : szpstrat ] , it is not hard to prove the following simple yet useful properties of @xmath95-cycle equivalence classes . let @xmath205 be a so - structure . we use @xmath65 and @xmath206 to denote some step sequences over @xmath8 . then for all @xmath191 , 1 . @xmath207=[\beta ] \iff \forall \lhd \in ext(s),\ ; \alpha \simeq_{\lhd } \beta$ ] 2 . @xmath208v$ ] 3 . @xmath207 { { \hat{{\sqsubset}}}^{\mathsf{cov}\ ; } } [ \beta ] \implies \exists\lhd \in ext(s),\;\omega_{\lhd } = u[\alpha][\beta]v$ ] [ prop : covlsos ] each @xmath95-cycle equivalence class is what juhs , lorenz and mauser called a _ synchronous step _ they also used equivalence classes to capture synchronous steps but only for the special class of _ synchronous closed _ so - structures , where @xmath209 is an equivalence relation . we extend their ideas by using @xmath95-cycle equivalence classes to capture what we will call _ non - serializable sets _ in arbitrary so - structures . the name is justified in assertion ( 1 ) of proposition [ prop : covlsos ] stating that two elements belong to the same non - serializable set of a so - structure @xmath12 iff they must be executed simultaneously in every stratified extension of @xmath12 . furthermore , we show in assertion ( 2 ) that all elements of a non - serializable set must occur together as a single step in at least one stratified extension of @xmath12 . assertion ( 3 ) gives a sufficient condition for two non - serializable sets to occur as consecutive steps in at least one stratified extension of @xmath12 . before we proceed to define comtrace using labeled so - structure , we need to define _ label - preserving isomorphisms _ for labeled so - structures more formally . a tuple @xmath210 is a _ labeled relational structure _ iff @xmath211 is a relational structure and @xmath212 is a function with domain @xmath3 . if @xmath211 is a so - structure , then @xmath85 is a _ labeled so - structure_. given two labeled relational structures @xmath213 and @xmath214 , we write @xmath215 to denote that @xmath216 and @xmath217 are _ label - preserving isomorphic _ ( _ lp - isomorphic _ ) . in other words , there is a bijection @xmath218 such that for all @xmath219 , 1 . @xmath220 2 . @xmath221 3 . @xmath222 such function @xmath223 is called a _ label - preserving isomorphism _ ( _ lp - isomorphism _ ) . note that all notations , definitions and results for so - structures are applicable to labeled so - structures . we also write @xmath224 $ ] or @xmath225 } $ ] to denote the lp - isomorphic class of a labeled relational structure @xmath226 . we will not distinguish an lp - isomorphic class @xmath227 } $ ] with a single labeled relational structure @xmath85 when it does not cause ambiguity . we are now ready to give an alternative definition for comtraces . to avoid confusion with the comtrace notion by janicki and koutny in @xcite , we will use the term _ lsos - comtrace _ to denote a comtrace defined using our definition . given a comtrace alphabet @xmath114 , a _ lsos - comtrace _ over @xmath128 is ( an lp - isomorphic class of ) a finite labeled so - structure @xmath228 } $ ] such that @xmath190 and for all @xmath191 , 1 . @xmath207 ( { { \hat{{\sqsubset}}}^{\mathsf{cov}\;}}\cap \hat{{\prec } } ) [ \beta ] \implies \lambda([\alpha])\times \lambda([\beta ] ) \nsubseteq ser$ ] 2 . @xmath207 ( { { \hat{{\sqsubset}}}^{\mathsf{cov}\;}}\setminus \hat{{\prec } } ) [ \beta ] \implies \lambda([\beta])\times \lambda([\alpha ] ) \nsubseteq ser$ ] 3 . @xmath229)},\ ; a\cup b = [ \alpha ] \implies \lambda(a)\times \lambda(b ) \not \subseteq ser$ ] 4 . @xmath230 5 . @xmath231 we write @xmath232 to denote the class of all lsos - comtraces over @xmath128 . [ def : lcomtrace ] let @xmath136 , @xmath233 and @xmath234 . then we have @xmath235 . the lp - isomorphic class of the labeled so - structure @xmath236 depicted in figure [ fig : f1 ] ( the dotted edges denote @xmath95 relation and the solid edges denote both @xmath93 and @xmath95 relations ) is a lsos - comtrace . the graph in figure [ fig : f2 ] represents the labeled quotient so - structure @xmath237 of @xmath85 , where we define @xmath238 . @xmath239[f-]{a } \ar@{-->}[dr]\ar@/^/[rr ] \ar@/_1.5pc/ [ ddrr ] & & + [ o][f-]{c}\ar@/_/ @{-->}[dd]\\ & * + [ o][f-]{c } \ar[ur]\ar[dr ] & \\ * + [ o][f-]{b } \ar[ur]\ar@/_/[rr]\ar@/^1.5pc/ [ uurr ] & & * + [ o][f-]{b}\ar@/_/ @{-->}[uu ] } $ ] @xmath240{a } \ar@{-->}[dr]\ar@/^1pc/[drr ] & & \\ & * + [ f-]{c } \ar[r ] & * + [ f-]{b , c}\\ * + [ f-]{b } \ar[ur]\ar@/_1pc/[urr ] & & } $ ] the lsos - comtrace @xmath224 $ ] actually corresponds to the comtrace @xmath241 $ ] , and we will show this relationship formally in section [ sec : representation ] . [ ex : comtrace2 ] definition [ def : lcomtrace ] can be extended to define _ infinite comtrace _ as follows . instead of asking @xmath3 to be finite , we require a labeled so - structure to be _ initially finite _ @xcite ) , i.e. , @xmath242 is finite for all @xmath243 . the initially - finiteness not only gives us a sensible interpretation that every event only causually depends on finitely many events , but also guarantees that the covering relations of @xmath244 and @xmath245 are well - defined . since each lsos - comtrace is defined as a class of lp - isomorphic labeled so - structures , dealing with lsos - comtrace might seem tricky . the _ no autoconcurrency _ property , i.e. , the relation @xmath5 is irreflexive , gives us a _ canonical _ way to enumerate the events of a lsos - comtrace very similar to how the events of a comtrace are enumerated . given a step sequence @xmath246 and any function @xmath223 defined on @xmath247 , we define @xmath248 , i.e. , the step sequence derived from @xmath135 by applying the function @xmath223 successively on each @xmath249 . note that @xmath250 denotes the _ image _ of @xmath251 under @xmath223 . given a lsos - comtrace @xmath252 } $ ] over a comtrace alphabet @xmath114 , a stratified order @xmath253 can be seen as a step sequence @xmath254 . 1 . for every @xmath255 ( @xmath256 ) , @xmath257 2 . @xmath258 . [ prop : validss ] proposition [ prop : validss ] ensures that @xmath259 is a valid step sequence over @xmath128 . recall that @xmath260 denotes the enumerated step sequence of @xmath65 and @xmath261 denotes the set of event occurrences . bijection _ @xmath262 as + @xmath263 by proposition [ prop : validss ] , the function @xmath264 is well - defined . moreover , we can show that @xmath264 is uniquely determined by @xmath85 regardless of the choice of @xmath253 . given @xmath265 , let @xmath266 and @xmath267 . then @xmath268 . [ prop : uniquexi ] henceforth , we will ignore subscripts and reserve the notation @xmath269 to denote the kind of mappings as defined above . we then define the _ enumerated so - structure _ of @xmath85 to be the labeled so - structure @xmath270 , where @xmath271 for @xmath259 and @xmath253 ; and the relations @xmath272 are defined as + @xmath273 clearly , the enumerated so - structure @xmath274 can be uniquely determined from @xmath85 using the preceding definition . from our construction , we can easily show the following important relationships : 1 . @xmath275 and @xmath85 are lp - isomorphic under the mapping @xmath269 . 2 . the labeled so - structures @xmath276 and @xmath277 are lp - isomorphic under the mapping @xmath269 and @xmath278 . [ prop : isoext ] in other words , the mapping @xmath279 plays the role of both the lp - isomorphism from @xmath275 to @xmath85 and the lp - isomorphism from the stratified extension @xmath280 of @xmath275 to the stratified extension @xmath96 of @xmath85 . these relationships can be best captured using the commutative diagram on the right . @xmath281_{\displaystyle \xi } \ar@{^{(}->}[d]_{\displaystyle id_{\sigma } } & ( x,{\prec},{\sqsubset},\lambda ) \ar@{^{(}->}[d]^{\displaystyle id_{x } } \\ ( \sigma,\lhd_{u},\lhd_{u}^{\frown},\ell ) \ar@{->}[r]^{\displaystyle \xi } & ( x,\lhd,\lhd^{\frown},\lambda)}$ ] we can even observe further that two lsos - comtraces are identical if and only if they define the same enumerated so - structure . henceforth , we will call an enumerated so - structure defined by a lsos - comtrace @xmath85 _ the canonical representation _ of @xmath85 . recently , inspired by the dependency graph notion for mazurkiewicz traces ( cf . * chapter 2 ) ) , kleijn and koutny claimed without proof that their _ combined dependency graph _ notion is another alternative way to define comtraces @xcite . in section [ sec : representation ] , we will give a detailed proof of their claim . given an comtrace alphabet @xmath282 , a _ combined dependency graph _ ( _ cd - graph _ ) over @xmath128 is ( a lp - isomorphic class of ) a finite labeled relational structure @xmath283 } $ ] such that @xmath190 , the relations @xmath284 are irreflexive , @xmath285 is a so - structure , and for all @xmath286 , @xmath287 @xmath288 @xmath289 @xmath290 + we will write @xmath291 to denote the class of all cd - graphs over @xmath128 . [ def : comdag ] cd - graphs can be seen as reduced graph - theoretic representations for lsos - comtraces , where some arcs that can be recovered using @xmath154-closure are omitted . it is interesting to observe that the non - serializable sets of a cd - graph are exactly the _ strongly connected components _ of the directed graph @xmath292 and can easily be found in time @xmath293 using any standard algorithm ( cf . * section 22.5 ) ) . + cd - graphs were called _ dependence comdags _ in @xcite . but this name could be misleading since the directed graph @xmath292 is not necessarily acyclic . for example , the graph on the right is the cd - graph that corresponds to the lsos - comtrace from figure [ fig : f1 ] , but it is not acyclic . ( here , we use the dotted edges to denote @xmath294 and the solid edges to denote _ only _ @xmath295 . ) thus , we use the name `` combined dependency graph '' instead . @xmath296[f-]{a } \ar@{-->}[dr ] \ar@/_1.5pc/ [ ddrr ] & & * + [ o][f-]{c}\ar@/_/ @{-->}[dd]\\ & * + [ o][f-]{c } \ar[ur]\ar[dr ] & \\ * + [ o][f-]{b } \ar[ur]\ar@/_/[rr]\ar@/^1.5pc/ [ uurr ] & & * + [ o][f-]{b}\ar@/_/ @{-->}[uu ] } $ ] this section contains the main technical contribution of this paper by showing that for a given comtrace alphabet @xmath128 , @xmath297 , @xmath232 and @xmath291 are three equivalent ways of talking about the same class of objects . we will next prove the first representation theorem which establishes the representation mappings between @xmath297 and @xmath232 . let @xmath298 and @xmath299 be stratified order structures such that @xmath300 . then @xmath301 . [ prop : stratsubset ] follows from theorem [ theo : szpstrat ] . for the next two lemmata , we let @xmath85 be a lsos - comtrace over a comtrace alphabet @xmath114 . let @xmath302 be the canonical representation of @xmath85 . let @xmath303 and @xmath304 . since @xmath65 is a valid step sequence in @xmath31 ( by proposition [ prop : validss ] ) , we can construct @xmath305 } = ( \sigma_{u},{\prec}_{[u]},{\sqsubset}_{[u]})$ ] from definition [ def : s2sos ] . our goal is to show that the stratified order @xmath305}$ ] defined by the comtrace @xmath150 $ ] is exactly @xmath306 . @xmath305 } \subseteq ( \sigma_{u},{\prec}_{0},{\sqsubset}_{0})$ ] . [ lem : l1 ] by proposition [ prop : so - cl ] , to show @xmath305 } = ( \sigma_{u},{\prec}_{u},{\sqsubset}_{u})^{\lozenge } \subseteq ( \sigma_{u},{\prec}_{0},{\sqsubset}_{0})$ ] , it suffices to show that @xmath307 . since @xmath274 is the canonical representation of @xmath85 , it is important to observe that @xmath308 . ( @xmath309 ) : assume @xmath310 . then from definition [ def : s2inv ] , @xmath311 . since @xmath312 , it follows from definition [ def : lcomtrace ] that @xmath313 or @xmath314 . suppose for a contradiction that @xmath314 , then by theorem [ theo : szpstrat ] , @xmath315 . but since we assume that @xmath303 , it follows that @xmath316 and @xmath317 , a contradiction . hence , we have shown @xmath313 . ( @xmath318 ) : can be shown in a similar way . @xmath305 } \supseteq ( \sigma_{u},{\prec}_{0},{\sqsubset}_{0})$ ] . [ lem : l2 ] in this proof , we will include subscripts for equivalence classes to avoid confusing the elements from quotient set @xmath319 with the elements from the quotient comtrace monoid @xmath320 . in other words , we write @xmath207_{\equiv_{{\sqsubset}_{0}}}$ ] to denote an element of the quotient set @xmath319 , and write @xmath150_{\theta}$ ] to denote the comtrace generated by @xmath65 . let @xmath321 . to show @xmath305 } \supseteq s'$ ] , by proposition [ prop : stratsubset ] , it suffices to show @xmath322 } ) \subseteq ext(s')$ ] . from theorem [ theo : com2sos ] , we know that @xmath322_{\theta } } ) = { \{\lhd_w\mid w\in [ u]_{\theta}\}}$ ] . thus we only need to show that for all @xmath323_{\theta}$ ] , @xmath324 . we observe that from @xmath65 , by definition [ def : commonoid ] , we can generate all the step sequences in the comtrace @xmath150_{\theta}$ ] in stages using the following recursive definition : @xmath325 since the set @xmath150_{\theta}$ ] is finite , @xmath150_{\theta}=d^n(u)$ ] for some stage @xmath326 . for the rest of the proof , we will prove by induction on @xmath327 that for all @xmath328 , if @xmath329 then @xmath330 . + * base case : * when @xmath331 , @xmath332 . since @xmath333 , it follows from proposition [ prop : isoext ] that @xmath334 . + * inductive case : * when @xmath335 , let @xmath40 be an element of @xmath336 . then either @xmath337 or @xmath338 . for the former case , by inductive hypothesis , @xmath339 . for the latter case , there must be some element @xmath340 such that @xmath341 or @xmath342 . by induction hypothesis , we already known @xmath343 . we want to show that @xmath324 . there are two cases to consider : + _ * case ( i ) : * _ + when @xmath341 , by definition [ def : commonoid ] , there are some @xmath344 and steps @xmath345 such that @xmath346 and @xmath347 where @xmath119 , @xmath120 , @xmath121 satisfy @xmath348 and @xmath349 and @xmath129 . let @xmath350 and @xmath351 be enumerated step sequences of @xmath206 and @xmath40 respectively . suppose for a contradiction that @xmath352 . by definition [ def : extsos ] , there are @xmath353 and @xmath354 such that @xmath355 . we now consider the quotient set @xmath356 . by proposition [ prop : covlsos ] ( 1 ) , @xmath357 . since @xmath355 , it follows that @xmath207_{\equiv_{{\sqsubset}_{0}}}\hat{{\sqsubset}}_{0 } [ \beta]_{\equiv_{{\sqsubset}_{0}}}$ ] . thus , from the fact that @xmath358 is partial order , there must exists a chain @xmath200_{\equiv_{{\sqsubset}_{0}}}=\ ; [ \gamma_1]_{\equiv_{{\sqsubset}_{0 } } } \;{{\hat{{\sqsubset}}_{0}}^{\mathsf{cov}\;}}\ ; [ \gamma_2]_{\equiv_{{\sqsubset}_{0}}}\;{{\hat{{\sqsubset}}_{0}}^{\mathsf{cov}\;}}\ ; \ldots\;{{\hat{{\sqsubset}}_{0}}^{\mathsf{cov}\;}}\;[\gamma_k]_{\equiv_{{\sqsubset}_{0}}}\;= [ \beta]_{\equiv_{{\sqsubset}_{0 } } } \label{eq : chain1}\end{aligned}\ ] ] then by theorem [ theo : szpstrat ] and the fact that @xmath343 , we know that @xmath359 for all @xmath255 . in other words , since the chain implies that every @xmath360 must always occur between @xmath53 and @xmath70 in all stratified extensions of @xmath361 and @xmath362 , we also have @xmath359 . hence , by proposition [ prop : covlsos ] ( 1 ) , we have @xmath363_{\equiv_{{\sqsubset}_{0}}}\subseteq { \overline{a}}$ ] for all @xmath255 , @xmath256 . also from of definition [ def : lcomtrace ] and that @xmath364 , we know that for each @xmath360 , either @xmath363_{\equiv_{{\sqsubset}_{0}}}\subseteq { \overline{b}}$ ] or @xmath363_{\equiv_{{\sqsubset}_{0}}}\subseteq { \overline{c}}$ ] . now we note that the first element on the chain @xmath365_{\equiv_{{\sqsubset}_{0}}}=[\alpha]_{\equiv_{{\sqsubset}_{0}}}\subseteq { \overline{c}}$ ] and the last element on the chain @xmath366_{\equiv_{{\sqsubset}_{0}}}=[\beta]_{\equiv_{{\sqsubset}_{0}}}\subseteq { \overline{b}}$ ] . thus , there exist two consecutive elements @xmath363_{\equiv_{{\sqsubset}_{0}}}$ ] and @xmath367_{\equiv_{{\sqsubset}_{0}}}$ ] on the chain such that @xmath363_{\equiv_{{\sqsubset}_{0}}}\subseteq { \overline{c}}$ ] and @xmath367_{\equiv_{{\sqsubset}_{0}}}\subseteq { \overline{b}}$ ] . but then it follows that * @xmath367_{\equiv_{{\sqsubset}_{0}}}\times [ \gamma_i]_{\equiv_{{\sqsubset}_{0}}}\subseteq ser$ ] and @xmath363_{\equiv_{{\sqsubset}_{0 } } } { { \hat{{\sqsubset}}_{0}}^{\mathsf{cov}\ ; } } [ \gamma_{i+1}]_{\equiv_{{\sqsubset}_{0}}}$ ] * @xmath368_{\equiv_{{\sqsubset}_{0 } } } \hat{{\prec}}_{0 } [ \gamma_{i+1}]_{\equiv_{{\sqsubset}_{0}}}\bigr)$ ] since @xmath343 and @xmath369 these contradict of definition [ def : lcomtrace ] since @xmath274 is a lsos - comtrace . + _ * case ( ii ) : * _ + when @xmath342 , by definition [ def : commonoid ] , there are some @xmath344 and steps @xmath345 such that @xmath370 and @xmath371 where @xmath119 , @xmath120 , @xmath121 satisfy @xmath348 and @xmath349 and @xmath129 . let @xmath372 and @xmath373 be enumerated step sequences of @xmath206 and @xmath40 respectively . suppose for a contradiction that @xmath352 . by definition [ def : extsos ] , there are @xmath374 and @xmath375 such that @xmath313 . by proposition [ prop : covlsos ] ( 1 ) , @xmath357 . thus , using a dual argument to the proof of case ( i ) , we can build a chain @xmath200_{\equiv_{{\sqsubset}_{0}}}=\ ; [ \gamma_1]_{\equiv_{{\sqsubset}_{0 } } } \;{{\hat{{\sqsubset}}_{0}}^{\mathsf{cov}\;}}\ ; [ \gamma_2]_{\equiv_{{\sqsubset}_{0}}}\;{{\hat{{\sqsubset}}_{0}}^{\mathsf{cov}\;}}\ ; \ldots\;{{\hat{{\sqsubset}}_{0}}^{\mathsf{cov}\;}}\;[\gamma_k]_{\equiv_{{\sqsubset}_{0}}}\;= [ \beta]_{\equiv_{{\sqsubset}_{0 } } } \end{aligned}\ ] ] we then argue that there are two consecutive elements on the chain such that @xmath363_{\equiv_{{\sqsubset}_{0}}}\subseteq { \overline{b}}$ ] and @xmath367_{\equiv_{{\sqsubset}_{0}}}\subseteq { \overline{c}}$ ] , which implies 1 . @xmath376_{\equiv_{{\sqsubset}_{0}}}\times [ \gamma_{i+1}]_{\equiv_{{\sqsubset}_{0}}}\subseteq ser$ ] and @xmath363_{\equiv_{{\sqsubset}_{0 } } } { { \hat{{\sqsubset}}_{0}}^{\mathsf{cov}\ ; } } [ \gamma_{i+1}]_{\equiv_{{\sqsubset}_{0}}}$ ] 2 . @xmath363_{\equiv_{{\sqsubset}_{0 } } } \hat{{\prec}}_{0 } [ \gamma_{i+1}]_{\equiv_{{\sqsubset}_{0}}}$ ] since @xmath343 and @xmath377 these contradict of definition [ def : lcomtrace ] . we also need to show that the labeled so - structure defined from each comtrace is indeed a lsos - comtrace . in other words , we need to show the following lemma . let @xmath114 be a comtrace alphabet . given a step sequence @xmath378 , the lp - isomorphic class @xmath379},{\prec}_{[u]},{\sqsubset}_{[u]},\ell \right ] } $ ] is a lsos - comtrace over @xmath128 . [ lem : l3 ] the proof of this lemma is straightforward by checking that @xmath379},{\prec}_{[u]},{\sqsubset}_{[u]},\ell \right ] } $ ] satisfies all conditions . let @xmath128 be a comtrace alphabet . 1 . the mapping @xmath380 is defined as + @xmath381 } , $ ] + where the function @xmath382 is defined in section [ sec : steps ] and @xmath383 is the so - structure defined by the comtrace @xmath172 from definition [ def : s2sos ] . the mapping @xmath384 is defined as + @xmath385 [ def : repmaps ] intuitively , the mapping @xmath386 is used to convert a comtrace to lsos - comtrace while the mapping @xmath387 is used to transform a lsos - comtrace into a comtrace . the fact that @xmath386 and @xmath387 are valid representation mappings for @xmath388 and @xmath232 will be shown in the following theorem . let @xmath128 be a comtrace alphabet . 1 . for every @xmath389 , @xmath390 . 2 . for every @xmath391 , @xmath392 . [ theo : rep ] * 1 . * the fact that @xmath393 follows from lemma [ lem : l3 ] . now for a given @xmath389 , we have @xmath394 . thus , it follows that @xmath395 * 2 . * assume @xmath270 is the canonical representation of @xmath85 . observe that since @xmath396 , we have @xmath397 let @xmath398 . we will next show that @xmath399 and @xmath400 } $ ] . fix an arbitrary @xmath401 , from lemmas [ lem : l1 ] and [ lem : l2 ] , @xmath305 } = ( \sigma,{\prec}_{0},{\sqsubset}_{0})$ ] . from theorem [ theo : com2sos ] , @xmath402})\bigr\ } } = [ u]$ ] . and the rest follows . the theorem says that the mappings @xmath386 and @xmath387 are inverses of each other and hence are both _ bijective_. using theorem [ theo : rep ] , we are going to show that the _ combined dependency graph _ notion proposed in @xcite is another correct alternative definition for comtraces . first we need to define several representation mappings that are needed for our proofs . let @xmath128 be a comtrace alphabet . 1 . the mapping @xmath403 is defined as + @xmath404 + where @xmath65 is any step sequence in @xmath172 and @xmath405 and @xmath406 are defined as in definition [ def : s2inv ] . the mapping @xmath407 is defined as @xmath408 . the mapping @xmath409 is defined as + @xmath410 . [ def : drepmaps ] before proceeding futher , we want to make sure that : 1 . @xmath407 is a well - defined function . 2 . @xmath403 is a well - defined function . given a cd - graph @xmath411 } \in { \mathsf{cdg}}(\theta)$ ] , let @xmath412 } = d_1^{\lozenge}$ ] . we know that @xmath85 is uniquely defined , since by definition [ def : comdag ] , @xmath413 is a so - structure , and so - structures are fixed points of @xmath154-closure ( by proposition [ prop : so - cl ] ( 4 ) ) . we will next show that @xmath85 is a lsos - comtrace by verifying the conditions of definition [ def : lcomtrace ] . conditions and are exactly and . : suppose for contradiction that there exist two distinct non - serializable sets @xmath207,[\beta]\subset x$ ] such that @xmath207({{\hat{{\sqsubset}}}^{\mathsf{cov}\;}}\cap \hat{{\prec}})[\beta]$ ] and @xmath414)\times \lambda([\beta ] ) \subseteq ser$ ] . clearly , this implies that @xmath415 , and thus by the @xmath154-closure definition , @xmath70 is reachable from @xmath53 on the directed graph @xmath416 , where @xmath417 . now we consider a shortest path @xmath418 + @xmath419 + on @xmath420 that connects @xmath53 to @xmath70 . we will prove by induction on @xmath421 that there exist two consecutive @xmath422 and @xmath423 on @xmath418 such that @xmath424 $ ] and @xmath425 $ ] and @xmath426 , which contradicts with @xmath414)\times \lambda([\beta ] ) \subseteq ser$ ] . + * base case : * when @xmath427 , then @xmath428 . since @xmath207({{\hat{{\sqsubset}}}^{\mathsf{cov}\;}}\cap \hat{{\prec}})[\beta]$ ] , we have @xmath429 , which by implies @xmath430 . + * inductive case : * when @xmath431 , we consider @xmath432 and @xmath433 . if @xmath434 $ ] and @xmath435 $ ] , then by @xmath207({{\hat{{\sqsubset}}}^{\mathsf{cov}\;}}\cap \hat{{\prec}})[\beta]$ ] , we have @xmath436 , which immediately yields @xmath437 . otherwise , we have @xmath438\cup [ \beta]$ ] or @xmath439 $ ] . for the first case , we get @xmath207\hat{{\sqsubset}}[\delta_2]\hat{{\sqsubset}}[\beta]$ ] , which contradicts that @xmath207{{\hat{{\sqsubset}}}^{\mathsf{cov}\;}}[\beta]$ ] . for the latter case , we can apply induction hypothesis on the path @xmath440 . and can also be shown similarly using `` shortest path '' argument as above . these proofs are easier since we only need to consider paths with edges in @xmath294 . * 2 . * by the proof of ( * lemma 4.7 ) , for any two step sequences @xmath50 and @xmath65 in @xmath441 , we have @xmath442 iff @xmath443 ) = { \mathsf{ct2dep}}([t])$ ] . thus the mapping @xmath444 is well - defined . the mapping @xmath407 is injective . [ lem : dlinj ] assume that @xmath445 , such that @xmath446 $ ] . since @xmath154-closure operator does not change the labeling function , we can assume that @xmath447 $ ] and @xmath448 . we will next show that @xmath449 . ( @xmath450 ) : let @xmath286 such that @xmath451 . suppose for a contradiction that @xmath452 . since @xmath451 , by , @xmath453 . thus , by , @xmath454 . but since @xmath448 , it follows that @xmath455 ( by proposition [ prop : so - cl ] ) . thus , @xmath456 and @xmath457 , a contradiction . ( @xmath458 ) : can be proved similarly . by reversing the role of @xmath459 and @xmath460 , we have @xmath461 . thus , we conclude @xmath462 . we are now ready to show the following representation theorem which ensures that @xmath463 and @xmath464 are valid representation mappings for @xmath232 and @xmath291 . let @xmath128 be a comtrace alphabet . 1 . for every @xmath465 , @xmath466 . 2 . for every @xmath391 , @xmath467 . [ theo : deprep ] * 1 . * let @xmath468 and let @xmath469 . suppose for a contradiction that @xmath470 and @xmath471 . from how @xmath386 is defined , @xmath472 . thus , it follows that @xmath473 . but this contradicts the injectivity of @xmath464 from lemma [ lem : dlinj ] . let @xmath391 and let @xmath474 . suppose for a contradiction that @xmath475 and @xmath476 . since @xmath477 , if we let @xmath478 , then @xmath479 . thus , we have shown that @xmath480 and @xmath481 , contradicting theorem [ theo : rep ] ( 2 ) . this theorem shows that lsos - comtraces and cd - graphs are equivalent representations for comtraces . the main advantage of cd - graph definition is its simplicity while the lsos - comtrace definition is stronger and more convenient to prove properties about labeled so - structures that represent comtraces . we do not need to prove another representation theorem for cd - graphs and comtraces since their representation mappings are simply the composition of the representation mappings from theorems [ theo : rep ] and [ theo : deprep ] . recall for a comtrace monoid @xmath127)$ ] , the comtrace operator @xmath125 is defined as @xmath126\circledast[t ] = [ r\ast t]$ ] . we will construct analogous composition operators for lsos - comtraces and cd - graphs . we will then show that lsos - comtraces ( cd - graphs ) over a comtrace alphabet @xmath128 together with its composition operator form a monoid isomorphic to the comtrace monoid @xmath127)$ ] . given two sets @xmath482 and @xmath483 , we write @xmath484 to denote the _ disjoint union _ of @xmath482 and @xmath483 . such disjoint union can be easily obtained by renaming the elements in @xmath482 and @xmath483 so that @xmath485 . we define the lsos - comtrace composition operator as follows . let @xmath486 and @xmath487 be lsos - comtraces over an alphabet @xmath114 , where @xmath488 } $ ] . the _ composition _ @xmath489 of @xmath486 and @xmath487 is defined as ( a lp - isomorphic class of ) a labeled so - structure @xmath490 } $ ] such that @xmath491 , @xmath492 , and @xmath493 , where + @xmath494 + @xmath495 observe that the operator is well - defined since we can easily check that : for every @xmath496 , @xmath497 . we will next show that this composition operator @xmath498 properly corresponds to the operator @xmath125 of the comtrace monoid over @xmath128 . let @xmath128 be a comtrace alphabet . then 1 . for every @xmath499 , @xmath500 2 . for every @xmath501 , @xmath502 [ prop : hom1 ] * 1 . * assume @xmath503 } $ ] , @xmath504 } $ ] and @xmath505 } $ ] . we can pick @xmath506 and @xmath507 . then observe that a stratified order @xmath75 satisfying @xmath508 is an extension of @xmath509 . thus , by theorem [ theo : rep ] , we have @xmath510\circledast [ { \mathsf{map}}(\lambda_2,\lhd_2 ) ] = [ { \mathsf{map}}(\lambda,\lhd ) ] = { \mathsf{lct2ct}}(q)$ ] as desired . * 2 . * without loss of generality , we can assume that @xmath511 $ ] , @xmath512 $ ] and @xmath513 = \mathbf{r}\circledast \mathbf{t}$ ] , where @xmath514 . by appropriate reindexing , we can also assume that @xmath515 . under these assumptions , let @xmath516 } $ ] , @xmath517 } $ ] and @xmath518 } $ ] , where @xmath519 is simply the standard labeling functions . it will now suffice to show that @xmath520 . ( @xmath521 ) : let @xmath522 . by definitions [ def : s2inv ] and [ def : s2sos ] , we have + @xmath523 + @xmath524 + thus , by proposition [ prop : so - cl ] ( 5 ) , we have @xmath525 as desired . furthermore , by proposition [ prop : so - cl ] ( 5 ) , @xmath526 is a so - structure . ( @xmath527 ) : by definitions [ def : s2inv ] and [ def : s2sos ] , we have @xmath528 and @xmath529 . since we already know @xmath526 is a so - structure , it follows from proposition [ prop : so - cl ] ( 5 ) that + @xmath530 let @xmath531 denote the lp - isomorphic class @xmath532 $ ] . then we observe that @xmath533 ) = \mathbb{i}$ ] and @xmath534 $ ] . by proposition [ prop : hom1 ] and theorem [ theo : rep ] , the structure @xmath535 is isomorphic to the monoid @xmath536)$ ] under the isomorphisms @xmath380 and @xmath384 . thus , the triple @xmath537 is also a monoid . we can summarize these facts in the following theorem : the mappings @xmath386 and @xmath387 are monoid isomorphisms between two monoids @xmath536)$ ] and @xmath535 . similarly , we can also define a composition operator for cd - graphs . let @xmath538 and @xmath539 be cd - graphs over an alphabet @xmath114 , where @xmath540 } $ ] . the _ composition _ @xmath541 of @xmath538 and @xmath539 is defined as ( a lp - isomorphic class of ) a labeled so - structure @xmath542 } $ ] such that @xmath491 , @xmath492 , and + @xmath543 + @xmath544 from this definition , it is straightforward to show the following propositions , which we will state without proofs . for every @xmath545 , @xmath546 . let @xmath128 be a comtrace alphabet . then 1 . for every @xmath499 , @xmath547 2 . for every @xmath548 , @xmath549 [ prop : hom2 ] putting the two preceding propositions and theorem [ theo : deprep ] together , we conclude : the mappings @xmath463 and @xmath464 are monoid isomorphisms between two monoids @xmath535 and @xmath550 . the simple yet useful construction we used extensively in this paper is to build a quotient so - structure modulo the @xmath95-cycle equivalence relation . intuitively , each @xmath95-cycle equivalence class consists of all the events that must be executed simultaneously with one another and hence can be seen as a single `` composite event '' . the resulting quotient so - structure is technically easier to handle since both relations of the quotient so - structure are acyclic . from this construction , we were able to give a labeled so - structure definition for comtraces similar to the labeled poset definition for traces . this quotient construction also explicitly reveals the following connection : a step on a step sequence @xmath135 is not serializable with respect to the relation @xmath5 of a comtrace alphabet if and only if it corresponds to a @xmath95-cycle equivalence class of the lsos - comtrace representing the comtrace @xmath134 $ ] ( cf . proposition [ prop : covlsos ] ) . we have also formally shown that the quotient monoid of comtraces , the monoid of lsos - comtraces and the monoid of cd - graphs _ over the same comtrace alphabet _ are indeed isomorphic by establishing monoid isomorphisms between them . these three models are formal linguistic , order - theoretic , and graph - theoretic respectively , which allows us to apply a variety of tools and techniques . an immediate future task is to develop a framework similar to the one in this paper for _ generalized comtraces _ , proposed and developed in @xcite . generalized comtraces extend comtraces with the ability to model events that can be executed _ earlier than or later than but never simultaneously_. another direction is to define and analyze infinite comtraces ( and generalized comtraces ) in a spirit similar to the works on infinite traces , e.g. , @xcite . it is also promising to use infinite lsos - comtraces and cd - graphs to develop logics for comtraces similarly to what have been done for traces ( cf . @xcite ) . i am grateful to prof . ryszard janicki for introducing me comtrace theory . i also thank the mathematics institute of warsaw university and the theoretical computer science group of jagiellonian university for their supports during my visits . it was during these visits that the ideas from this paper emerge . this work is financially supported by the ontario graduate scholarship and the natural sciences and engineering research council of canada . the anonymous referees are thanked for their valuable comments that help improving the readability of this paper . 19 e. clarke , o. grumberg and d. peled , _ model checking _ , mit press , cambridge , 1999 . t. h. cormen , c. e. leiserson and r. l. rivest , _ introduction to algorithms _ , second edition , mit press , 2001 . b. a. davey and h. a. priestley , , cambridge university press 2002 . v. diekert , on the concatenation of infinite traces , proc . of stacs , _ lncs _ 480 ( 1991 ) , 105117 . v. diekert and g. rozenberg ( eds . ) , . world scientific 1995 . v. diekert and y mtivier , , _ handbook of formal languages , vol . 3 : beyond words _ , pp 457 - 533 , springer 1997 . v. diekert , m. horsch , m. kufleitner , on first - order fragments for mazurkiewicz traces , _ fundam . _ 80(1 - 3 ) : 1 - 29 , 2007 . j. esparza and k. heljanko , _ unfoldings a partial - order approach to model checking _ , springer 2008 . a. farzan and p. madhusudan , causal dataflow analysis for concurrent programs , proc . of cav , _ lncs _ 4144 ( 2006 ) : 315328 . a. farzan and p. madhusudan , causal atomicity , proc . of tacas 2007 , _ lncs _ 4424 ( 2007 ) , 102116 . p. c. fishburn , _ interval orders and interval graphs _ , j. wiley 1985 , new york . h. gaifman and v. pratt , partial order models of concurrency and the computation of function , _ proc . of lics87 _ , pp . 7285 . p. gastin , infinite traces , proc . of semantics of systems of concurrent processes , _ lncs _ 469 ( 1990 ) , 277308 . t. gazagnaire , b. genest , l. hlout , p. s. thiagarajan , s. yang , causal message sequence charts , _ theor . 410(41 ) : 40944110 , 2009 . r. janicki , relational structures model of concurrency . , 45(4 ) : 279320 , 2008 . r. janicki and m. koutny , invariants and paradigms of concurrency theory , proc . of _ parle _ 91 , _ lncs _ 506 , springer 1991 , pp . 5974 . r. janicki and m. koutny , structure of concurrency , , 112(1 ) : 552 , 1993 . r. janicki and m. koutny , semantics of inhibitor nets , _ information and computation _ , 123(1 ) : 116 , 1995 . r. janicki and m. koutny , fundamentals of modelling concurrency using discrete relational structures , _ acta informatica _ , 34 : 367388 , 1997 . r. janicki and m. koutny , on causality semantics of nets with priorities , _ fundamenta informaticae _ 34 : 222255 , 1999 . r. janicki and d. t. m. l , modelling concurrency with quotient monoids , proc of petri nets 2008 , _ lncs _ 5062 , springer 2008 , pp . 251269 . r. janicki and d. t. m. l , modelling concurrency with comtraces and generalized comtraces , submitted in 2009 . available at : http://arxiv.org/abs/0907.1722 g. juhs , r. lorenz , s. mauser , causal semantics of algebraic petri nets distinguishing concurrency and synchronicity , _ fundamenta informatica _ 86(3 ) : 255 - 298 , 2008 . g. juhs , r. lorenz , s. mauser , synchronous + concurrent + sequential = earlier than + not later than , proc . of acsd06 , turku , finland 2006 , pp . 261 - 272 , ieee press . h. c. m. kleijn and m. koutny , process semantics of general inhibitor nets , _ information and computation _ , 190:1869 , 2004 . j. kleijn and m. koutny , formal languages and concurrent behaviour , _ studies in computational intelligence _ , 113:125 - 182 , 2008 . d. t. m. l , studies in comtrace monoids , master thesis , dept . of computing and software , mcmaster university , canada , august 2008 . a. mazurkiewicz , concurrent program schemes and their interpretation , tr daimi pb-78 , comp . science depart . , aarhus university , 1977 . v. pratt , modeling concurrency with partial orders , _ international journal of parallel programming _ , 15(1):3371 , 1986 . e. szpilrajn , sur lextension de lordre partiel , _ fund . mathematicae _ 16 , 386389 , 1930 . p. s. thiagarajan and i. walukiewicz , an expressively complete linear time temporal logic for mazurkiewicz traces , _ inf . comput . _ 179(2 ) : 230249 , 2002 . * 1 . * ( @xmath551 ) : since @xmath207=[\beta]$ ] , we know that @xmath552 or @xmath553 . the former case is trivial . for the latter case , by theorem [ theo : szpstrat ] , we have @xmath554 and @xmath555 . but this implies that @xmath556 . ( @xmath557 ) : the case when @xmath552 is trivial . assume that @xmath558 and @xmath556 . thus , by theorem [ theo : szpstrat ] , @xmath559 and @xmath457 . but this means @xmath53 and @xmath70 belong to the same equivalence class . * suppose for a contradiction that all @xmath560 can not be written in the form of @xmath561v$ ] . this implies that there exists some @xmath562 $ ] such that for all @xmath560 , @xmath563 . but by theorem [ theo : szpstrat ] , this yields @xmath564 and @xmath565 , contradicting with @xmath566 $ ] . assume @xmath207 { { \hat{{\sqsubset}}}^{\mathsf{cov}\ ; } } [ \beta]$ ] . suppose for a contradiction that there does not exist @xmath560 such that @xmath561[\beta]v$ ] for some step sequences @xmath65 and @xmath206 . then , by theorem [ theo : szpstrat ] , there must exist some @xmath567\cup[\beta])$ ] , such that @xmath568 . since @xmath569\cup[\beta]$ ] , this yields @xmath207{\sqsubset}[\gamma ] { \sqsubset}[\beta]$ ] , which contradicts that @xmath207 { { \hat{{\sqsubset}}}^{\mathsf{cov}\ ; } } [ \beta]$ ] . assume @xmath570 and @xmath558 . thus , @xmath571 . thus , by corollary [ cor : szpstrat ] ( 2 ) , we have @xmath572 . hence , by of definition [ def : lcomtrace ] , @xmath573 . since @xmath4 is irreflexive , this also shows that any two distinct @xmath53 and @xmath70 in @xmath251 have different labels . thus , @xmath257 for all @xmath255 . * 2 . * from the proof of * 1 . * , we know that @xmath570 and @xmath558 implies @xmath573 . thus , @xmath574 for all @xmath255 . observe that from proposition [ prop : validss ] , we have @xmath575 . it remains to show that @xmath576 for all @xmath577 . suppose for a contradiction that @xmath578 for some @xmath577 . from the definition above , there are two distinct elements @xmath202 , such that @xmath579 and @xmath580 and @xmath581 . since @xmath5 is irreflexive , @xmath582 . thus , by of definition [ def : lcomtrace ] , @xmath583 or @xmath584 . without loss of generality , we assume @xmath583 and that @xmath585 for some event @xmath586 . again by of definition [ def : lcomtrace ] , we know that elements having the same label are totally ordered by @xmath93 . thus , if @xmath587 is the number of elements in @xmath3 labeled by @xmath39 , then we have @xmath588 and @xmath589 . but then @xmath590 implies that @xmath591 , while @xmath592 implies that @xmath593 , which is absurd . let @xmath594},{\prec}_{[u]},{\sqsubset}_{[u]},l \bigr)$ ] . from theorem [ theo : com2sos ] , @xmath85 is a labeled so - structure . it only remains to show that @xmath85 satisfies conditions of definition [ def : lcomtrace ] . : assume @xmath207 ( { { \hat{{\sqsubset}}}^{\mathsf{cov}\;}}\cap \hat{{\prec } } ) [ \beta]$ ] and suppose for a contradiction that @xmath414)\times \lambda([\beta ] ) \subseteq ser$ ] . then from proposition [ prop : covlsos ] ( 3 ) , there exists @xmath595 such that @xmath596[\beta]w$ ] . from theorem [ theo : com2sos ] , since we have @xmath597\bigr\ } } = ext(s_{[u]})$ ] , it follows that @xmath598[\beta]w ) \in [ u]$ ] . but @xmath414)\times \lambda([\beta ] ) \subseteq ser$ ] implies @xmath599\cup [ \beta]v ) \in [ u]$ ] . hence , @xmath600\cup [ \beta]v$ ] is also a stratified extension of @xmath85 , which contradicts that @xmath207 \hat{{\prec } } [ \beta]$ ] . using a similar argument , we can show using proposition [ prop : covlsos ] ( 1,4 ) and using proposition [ prop : covlsos ] ( 1,2 ) . : since @xmath601 , it follows from corollary [ cor : szpstrat ] that there exists @xmath595 where @xmath602 . since @xmath603\bigr\ } } = ext(s_{[u]})$ ] , there exists a sequence @xmath604 $ ] such that @xmath605 . this implies @xmath53 and @xmath70 belong to the same step in @xmath606 . thus , we have @xmath573 .
when extending the symmetries of the standard model of particle physics ( sm ) @xcite to include supersymmetry @xcite , the yukawa couplings are fixed by the renormalisable superpotential @xcite @xmath4 w_{p_6 } & = & \epsilon_{ab } \left(h^e_{ij } l_i^a h_d^b { \bar e}_j + h^d_{ij } q_i^a h_d^b { \bar d}_j + h^u_{ij } q_i^a h_u^b { \bar u}_j + \mu h_d^ah_u^b \right)\ , , \label{superpot - p6 } \\[1 mm ] w_{\not{p}_6}^{\not l } & = & \epsilon_{ab}\left(\frac{1}{2}\lam_{ijk } l_i^al_j^b{\bar e}_k + \lam_{ijk}^\prime l_i^aq_j^b{\bar d}_k + \kap_i l_i^ah_u^b\right ) \ , , \label{superpot - lv } \\[1 mm ] w_{\not{p}_6}^{\not b } & = & \frac{1}{2 } \epsilon_{rst}\lam_{ijk}^ { \prime\prime}{\bar u}_i^r{\bar d}_j^s { \bar d}_k^t \ , . \label{superpot - bv}\end{aligned}\ ] ] here , @xmath5 are generation indices , @xmath6 are su(2 ) and @xmath7 are su(3 ) indices . @xmath8 denote the lepton doublet and singlet left - chiral superfields ; @xmath9 denote the quark doublet and singlet superfields , respectively . @xmath10 are dimensionless coupling constants and @xmath11 are mass mixing parameters . together the operators in @xmath12 and @xmath13 lead to rapid proton decay in disagreement with the experimental lower bounds on the proton lifetime @xcite . a possible solution to this problem is to introduce the discrete @xmath14 symmetry , proton hexality , @xmath15 @xcite , which prohibits both @xmath16 and @xmath17 , as well the dangerous dimension-5 proton decay operators @xcite ; this is the minimal supersymmetric standard model ( mssm ) @xcite . ( note that the widely used discrete @xmath18 symmetry r - parity does not prohibit the dimension - five proton decay operators . ) however , in order to stabilize the proton it is sufficient to prohibit either the superpotential @xmath19 via baryon - triality @xcite or the superpotential @xmath20 via lepton - parity @xcite . lepton - parity is not discrete gauge anomaly - free @xcite and we thus disregard it in the following . baryon - parity has the further advantage of allowing for non - zero neutrino masses without the need for right - handed neutrinos . we thus consider here the total superpotential given by @xmath21 we shall focus exclusively on the tri - linear couplings . at any given scale the bi - linear terms @xmath22 can be rotated away through a basis redefinition @xcite . this is not true , when embedding the theory in a more unified model , supergravity @xcite . however , at @xmath23 the natural value is @xmath24 @xcite , which leads to @xmath25 at low - energy . thus the bi - linear terms are mainly relevant for neutrino masses , see for example @xcite , and we shall neglect them in the following . the tri - linear operators in @xmath12 , lead to novel supersymmetric collider signatures beyond those of the mssm @xcite . in particular , the operators in @xmath26 induce lepton flavour violation ( lfv ) as well as lepton number violation , neither of which has been observed @xcite . there is an extensive literature on the resulting bounds on the operators @xmath27 from indirect processes , see _ e.g. _ refs . @xcite , including also several overviews @xcite . however , due to the improved data in particular on @xmath3-meson and @xmath0 decays , it is the purpose of this paper to present a systematic update of the bounds resulting from lepton decays as well as leptonic and semi - leptonic decays of mesons . in the process , we have found several new bounds resulting from @xmath0 , @xmath2 and kaon decays . we have also found a need to correct some results in the literature with respect to @xmath3-meson decays . our paper is organized as follows . in sect . [ analytic - expressionsa ] we start from an effective lagrangian , where the supersymmetric scalar fermions have been integrated out and then present the treatment of the qcd bound state in sect . [ analytic - expressionsb ] . general analytic expressions for the decay rates of the various lepton and meson decays are shown in sect . [ analytic - expressionsc ] . in sect . [ numerical - results ] , we insert the present experimental results into the analytical expressions to obtain our new bounds . these are summarized in tables [ new_bounds_highlights_table ] - [ lambdaprime - lambdaprime_table3 ] . in sect . [ discussion ] , we discuss the implications of our results . formul for the meson decay constants and the general lepton and meson decay matrix elements are collected in the appendices . because the sfermions are constrained to be heavy , @xmath28 gev @xmath29 ( this work does not consider the decays of particles heavier than @xmath3 mesons ) , we approximate their propagators as static @xmath30 . this is equivalent to integrating out the sfermionic degrees of freedom to obtain an effective interaction lagrangian @xcite and taking only the leading term in an expansion in inverse sfermion mass , @xmath31 \label{eq : leffb } \\ & & \hspace{2.9 mm } - \left [ { \frac{1}{{2 m_{{{\tilde{u}}_{g}}}^{2 } } } } { { \lam } ' } _ { ign } { { \lam}'}_{kgm}^{{\ast } } ( { \bar{d}}^{n } { \gamma}^{{\mu } } p_{r } d^{m } ) ( { \bar{l}}^{k } { \gamma}_{{\mu } } p_{l } l^{i } ) + { \mbox { h.c . } } \right ] \label{eq : leffc } \\ & & \hspace{2.9 mm } + \left [ { \frac{1}{{2 m_{{{\tilde{d}}_{g}}}^{2 } } } } { { \lam } ' } _ { img } { { \lam}'}_{kng}^{{\ast } } ( { \bar{u}}^{n } { \gamma}^{{\mu } } p_{l } u^{m } ) ( { \bar{l}}^{k } { \gamma}_{{\mu } } p_{l } l^{i } ) + { \mbox { h.c . } } \right ] \bigg\}\ , , \label{eq : leffd } \label{effective_lagrangian_equation}\end{aligned}\ ] ] where ( [ eq : leffa ] ) and ( [ eq : leffb ] ) result from integrating out the sneutrino fields , and ( [ eq : leffc ] ) and ( [ eq : leffd ] ) result from integrating out the up type and down type squark fields respectively , using some fierz identities . the index @xmath32 denotes the generation . there are additional terms in the effective lagrangian which arise when integrating out the charged sleptons , which we do not consider here : for the product of two @xmath33 or an @xmath33 and an @xmath34 operator , these lead to neutrinos in the final state . thus lepton flavour violation is not observable in the resulting lepton or meson decays ; for the product of two @xmath34 operators , the resulting meson decays are purely hadronic . in the following , we shall assume that the decay is dominated by the exchange of a sfermion of a single generation , either because it is lighter than the others or because it has a larger product of couplings ( the double coupling dominance convention ) . subsequent expressions with an index @xmath32 are thus implicitly for only one value of @xmath32 , though one may always deduce the general result by replacing expressions like @xmath35 with @xmath36 etc . it is also assumed in this paper that the sneutrino higgsino and squark mixing can be neglected . such mixings just add to the notational burden . if one insists on accounting for mixing , one can make the replacement @xmath37 @xmath38 @xmath39 etc . for squark mixing matrices @xmath40 , with a similar expression for mixing between the sneutrinos and higgses . the decay constant , @xmath41 , of a vector meson @xmath42 with momentum @xmath43 is defined as @xmath44 where @xmath45 is the polarization vector of @xmath42 , @xmath46 is the vector meson mass , and @xmath47 is the coefficient of @xmath48 in the quark model wavefunction of the meson , @xmath49 , @xmath50 . for a pseudoscalar meson @xmath51 , we use the pcac condition @xcite and define the decay constant @xmath52 through the axial vector matrix element @xmath53 where @xmath54 is the analogue of @xmath47 . as described in appendix [ app : qcd ] , the equation of motion for the quark fields can be used to derive the pseudoscalar matrix element from the axial vector matrix element ( [ pcac_condition_equationi ] ) . we find @xmath55 the factor @xmath56 is proportional to the sum of current quark masses @xmath57 and @xmath58 , @xmath59 and @xmath60 . for the proper definition of @xmath56 , a list of the coefficients @xmath61 and more details see appendix [ app : qcd ] . the feynman graphs and matrix elements for the various decays considered in this paper are given in appendix [ app : me ] . upon squaring the matrix elements , summing over the final spin states and averaging over the initial spin states , we arrive at the following expressions for the decay widths : @xmath62 for a heavy lepton @xmath63 decaying into leptons @xmath64 and @xmath65 and an anti lepton @xmath66 , @xmath67 where we approximate the final state ( anti)leptons as massless . for a heavy lepton @xmath68 decaying into a lepton @xmath69 and a vector meson consisting of valence quark @xmath70 and anti quark @xmath71 , there are two cases : up type squark mediated : @xmath72 or down type squark mediated : @xmath73 where we have introduced the notation @xmath74 to mean only summing over the down type quarks in the meson and @xmath75 to mean summing over the up type quarks . for a heavy lepton @xmath68 decaying into a lepton @xmath69 and a pseudoscalar meson consisting of valence quark @xmath70 and anti quark @xmath71 , there are three cases : up type squark mediated : @xmath76 down type squark mediated : @xmath77 or sneutrino mediated : @xmath78 \times { \frac { ( m_{{l^{i}}}^{2 } - m_{p}^{2})^{2}}{128 { \pi } m_{{\tilde{{\nu}}}^{g}}^{4 } } } { \frac{\| f_{p}\|^{2 } m_{p}^{4}}{m_{{l^{i}}}^{3 } } } \left ( 1 + { \mathcal{o}}\left ( { \frac{{m_{{l^{k}}}}}{{m_{{l^{i } } } } } } \right ) \right)\ , . \label{gamma_to_p_snu_equation}\end{gathered}\ ] ] for a vector meson @xmath42 decaying into a lepton of generation @xmath79 and an anti lepton of generation @xmath80 , there are again two cases : up type squark mediated : @xmath81 \times \left ( 1 + { \mathcal{o}}\left ( { \frac{{m_{{l^{{k'}}}}}}{{m_{v } } } } \right ) \right ) \label{gamma_v_to_su_equation}\end{gathered}\ ] ] or down type squark mediated : @xmath82 \times \left ( 1 + { \mathcal{o}}\left ( { \frac{{m_{{l^{{k'}}}}}}{{m_{{l^{{i ' } } } } } } } \right ) \right)\ , . \label{gamma_v_to_sd_equation}\end{gathered}\ ] ] for a pseudoscalar meson @xmath51 decaying into a lepton of generation @xmath79 and an anti lepton of generation @xmath80 , there are again three cases : up type squark mediated : @xmath83 down type squark mediated : @xmath84 or sneutrino mediated : @xmath85 \times { \frac { ( m_{p}^{2 } - m_{{l^{{i'}}}}^{2 } ) ^{2}}{64 { \pi } m_{{\tilde{{\nu}}}^{g}}^{4 } } } \left ( 1 + { \mathcal{o}}\left ( { \frac{{m_{{l^{{k'}}}}}}{{m_{p } } } } \right ) \right)\ , . \label{gamma_p_to_snu_equation}\end{gathered}\ ] ] we assume for simplicity the _ double coupling dominance hypothesis _ , that the bounds from any one experimental result are applied to only one product of couplings . the input values for the various fermion and meson masses and decay constants are listed in table [ input_data_table ] . all the @xmath52 values and masses were taken from the 2006 edition of _ review of particle physics _ by the particle data group ( pdg ) @xcite . the @xmath86 values were calculated from @xmath87 according to @xmath88 where @xmath89 are factors determined by the electric charge of the quarks that built up the meson @xcite . the experimental results on lifetimes , decay widths and branching fractions are also taken from the 2006 review of the pdg @xcite . in tables [ new_bounds_highlights_table ] , [ very_improved_bounds_highlights_table ] and [ combined_better_than_product_bounds_highlights_table ] we present what may be considered the most interesting results of our analysis . the coupling combinations which had no bounds previously are collected in table [ new_bounds_highlights_table ] . those combinations which have improved by a factor of @xmath90 or more are presented in table [ very_improved_bounds_highlights_table ] , and the cases where the new combined bound is better than the previously published product of individual bounds are presented in table [ combined_better_than_product_bounds_highlights_table ] . here and in the following , the symbol @xmath91 $ ] denotes @xmath92 , the sfermion mass in units of @xmath93 gev . this also indicates the mediating sfermion for the decay . the superscript @xmath94 in table [ new_bounds_highlights_table ] indicates that this bound comes from a decay which involves a difference of couplings , so there could be a cancellation which would lead to the double coupling dominance hypothesis giving an excessively tight bound . while we also include very loose bounds in our listings , we note that couplings @xmath95 would imply a breakdown of our perturbative analysis . in tables [ lambda - lambda_table ] to [ lambdaprime - lambdaprime_table6 ] , we collect all our bounds on the products of couplings @xmath96 . the results have been arranged so that the number made from reading off the indices of the couplings to make a six digit number @xmath97 ascends . in the rightmost columns of tables [ lambda - lambda_table ] to [ lambdaprime - lambdaprime_table6 ] , `` new '' indicates a previously unpublished result ( see also table [ new_bounds_highlights_table ] ) , `` upd . '' indicates that the bound has been updated and tightened in this paper , `` agr . '' indicates that the bound has not changed and we agree with the previously published result @xcite , and `` unimp . '' indicates that our bound from decay data is less strong than the previously published result , which in these cases is from a different experimental source ( the non observation of @xmath98 in @xmath99 gives a better bound on @xmath100 than that of @xmath101 ) . `` corr . '' indicates that we disagree with the previously published result @xcite , `` corr.(@xmath102 ) '' indicating that our result is stronger than the incorrect previous bound and `` corr.(@xmath103 ) '' indicating that our result is less strong . the reference in this column gives the previous published bound . where two references are given , the comparison is between our bound on a product of two couplings and the product of the bounds on individual couplings . note that the @xmath104 decays can proceed through standard model interactions @xcite . however , the sm contribution is suppressed by a small ckm matrix element and by the decay only arising at one loop level and has thus been neglected in our analysis . the bounds presented here generally update those presented in the literature , with the noted disagreement with some of the bounds coming from the @xmath3 meson data . many bounds have been improved , some through tighter experimental decay bounds like those from @xmath105 decays , others through using @xmath106 instead of @xmath107 , which also leads to some previously unpublished bounds . the @xmath108 decay data was also previously unpublished , but does not seem particularly useful , with bounds of order @xmath109 . the decay @xmath110 seems to give previously unpublished bounds too , which are more stringent . the decay @xmath111 leads to bounds which are less strong than those from @xmath110 . note , however , that @xmath112 is free from potential interference effects induced by the coupling of the mediating squark to both down and strange quarks . assuming that the sfermion masses are of order @xmath93 gev and taking the square root of the bound on a coupling product to be a rough guide to the bound on each coupling gives an estimate of the couplings @xmath113 being of order @xmath114 , apart from the very tight bounds from the non observation of @xmath115 . the bounds on the couplings @xmath116 vary considerably , though those involving a third generation quark are consistently of order @xmath114 . since these come from @xmath3 meson decays , they are likely to become even tighter with more data from @xmath3 factories . mass + meson & ( in gev ) & ( in gev ) & fermion & ( in gev ) + @xmath117 & @xmath118 & @xmath119 & @xmath120 & @xmath121 + @xmath122 & @xmath123 & @xmath124 & @xmath125 & @xmath126 + @xmath127 & @xmath123 & @xmath124 & @xmath105 & @xmath128 + @xmath108 & @xmath129 & @xmath119 & @xmath130 & @xmath131 + @xmath132 & @xmath133 & @xmath134 & @xmath135 & @xmath136 + @xmath137 & @xmath138 & @xmath139 & @xmath140 & @xmath141 + @xmath142 & @xmath143 & @xmath144 & @xmath65 & @xmath145 + @xmath146 & @xmath147 & @xmath144 & @xmath64 & @xmath148 + vector & mass & @xmath86 & + meson & ( in gev ) & ( in gev ) & + @xmath149 & @xmath150 & 0.22 & + @xmath151 & @xmath152 & 0.23 & + @xmath153 & @xmath154 & 0.23 & + @xmath155 & @xmath156 & 0.41 & + & decay + @xmath157 & 2.1 & @xmath158^{2}$ ] & @xmath159 + @xmath160 & & & + @xmath161 & 9.7 & @xmath162 @xmath158^{2}$ ] & @xmath163 + @xmath164 & & & + @xmath165 & 1.0 & @xmath166 @xmath158^{2}$ ] & @xmath167 + @xmath168 & & & + @xmath169 & 2.3 & @xmath166 @xmath170^{2}$ ] & @xmath163 + @xmath171 & 1.5 & @xmath172 @xmath170^{2}$ ] & @xmath159 + @xmath173 @xmath94 & 1.2 & @xmath166 @xmath170^{2}$ ] & @xmath174 + @xmath173 & 3.4 & @xmath166 @xmath170^{2}$ ] & @xmath175 + @xmath176 & 2.4 & @xmath166 @xmath170^{2}$ ] & @xmath167 + @xmath177 @xmath94 & 1.6 & @xmath166 @xmath170^{2}$ ] & @xmath178 + @xmath177 & 3.4 & @xmath166 @xmath170^{2}$ ] & @xmath179 + coupling & & + combination & & decay & & decay & key + @xmath180 @xmath181 & 4.6 & @xmath162 @xmath158^{2}$ ] & @xmath174 & 1.6 & @xmath182 @xmath158^{2}$ ] & @xmath174 & upd . @xcite + @xmath183 @xmath181 & & & & & & & + @xmath180 @xmath184 & 9.7 & @xmath162 @xmath158^{2}$ ] & @xmath163 & 8.5 & @xmath182 @xmath158^{2}$ ] & @xmath185 & upd . @xcite + @xmath183 @xmath186 & & & & & & & + @xmath187 @xmath184 & 1.0 & @xmath166 @xmath158^{2}$ ] & @xmath167 & 7.6 & @xmath182 @xmath158^{2}$ ] & @xmath188 & upd . @xcite + @xmath189 @xmath186 & & & & & & & + @xmath189 @xmath190 & 6.7 & @xmath191 @xmath158^{2}$ ] & @xmath178 & 1.7 & @xmath166 @xmath158^{2}$ ] & @xmath178 & upd . @xcite + @xmath187 @xmath181 & 3.7 & @xmath162 @xmath158^{2}$ ] & @xmath178 & 1.7 & @xmath182 @xmath158^{2}$ ] & @xmath178 & upd . @xcite + @xmath189 @xmath181 & & & & & & & + coupling & & + combination & & decay & & decay & key + @xmath192 @xmath193 & 1.2 & @xmath166 @xmath194^{2}$ ] & @xmath195 & 0.02 & @xmath194 $ ] & apv in cs & upd . @xcite , + & & & & 0.12 & @xmath194 $ ] & @xmath196 & @xcite + @xmath197 @xmath198 & 9.0 & @xmath166 @xmath194^{2}$ ] & @xmath199 & 0.21 & @xmath194 $ ] & @xmath200 & upd . @xcite , + & & & & 5.9 & @xmath201 @xmath194 $ ] & @xmath202 & @xcite + @xmath203 @xmath204 & & + @xmath205 @xmath206 & & decay & & key + @xmath207 @xmath208 & 7.0 & @xmath162 @xmath209^{2}$ ] & @xmath210 & 2.1 & @xmath166 @xmath209^{2}$ ] & upd . @xcite + @xmath207 @xmath211 & 6.8 & @xmath162 @xmath209^{2}$ ] & @xmath212 & 2.0 & @xmath166 @xmath209^{2}$ ] & upd . @xcite + @xmath207 @xmath213 & 5.6 & @xmath162 @xmath209^{2}$ ] & @xmath214 & 1.9 & @xmath166 @xmath209^{2}$ ] & upd . @xcite + @xmath215 @xmath208 & 6.8 & @xmath162 @xmath209^{2}$ ] & @xmath216 & 2.2 & @xmath166 @xmath209^{2}$ ] & upd . @xcite + @xmath215 @xmath211 & 7.0 & @xmath162 @xmath209^{2}$ ] & @xmath210 & 2.1 & @xmath166 @xmath209^{2}$ ] & upd . @xcite + @xmath215 @xmath213 & 6.8 & @xmath162 @xmath209^{2}$ ] & @xmath216 & 2.2 & @xmath166 @xmath209^{2}$ ] & upd . @xcite + @xmath217 @xmath218 & 6.6 & @xmath219 @xmath220^{2}$ ] & @xmath115 & 6.6 & @xmath219 @xmath220^{2}$ ] & agr . @xcite + @xmath217 @xmath221 & 7.0 & @xmath162 @xmath220^{2}$ ] & @xmath222 & 2.7 & @xmath166 @xmath220^{2}$ ] & upd . @xcite + @xmath217 @xmath223 & 7.0 & @xmath162 @xmath220^{2}$ ] & @xmath222 & 2.7 & @xmath166 @xmath220^{2}$ ] & upd . @xcite + @xmath217 @xmath224 & 6.8 & @xmath162 @xmath220^{2}$ ] & @xmath212 & 2.0 & @xmath166 @xmath220^{2}$ ] & upd . @xcite + @xmath218 @xmath221 & 6.8 & @xmath162 @xmath220^{2}$ ] & @xmath212 & 2.0 & @xmath166 @xmath220^{2}$ ] & upd . @xcite + @xmath218 @xmath223 & 5.2 & @xmath162 @xmath220^{2}$ ] & @xmath225 & 1.9 & @xmath166 @xmath220^{2}$ ] & upd . @xcite + @xmath218 @xmath224 & 7.0 & @xmath162 @xmath220^{2}$ ] & @xmath210 & 2.1 & @xmath166 @xmath220^{2}$ ] & upd . @xcite + @xmath226 @xmath227 & 6.6 & @xmath219 @xmath228^{2}$ ] & @xmath115 & 6.6 & @xmath219 @xmath220^{2}$ ] & agr . @xcite + @xmath226 @xmath229 & 7.0 & @xmath162 @xmath228^{2}$ ] & @xmath222 & 2.7 & @xmath166 @xmath228^{2}$ ] & upd . @xcite + @xmath226 @xmath230 & 6.6 & @xmath219 @xmath228^{2}$ ] & @xmath115 & 6.6 & @xmath219 @xmath220^{2}$ ] & agr . @xcite + @xmath226 @xmath231 & 6.8 & @xmath162 @xmath228^{2}$ ] & @xmath212 & 2.0 & @xmath166 @xmath228^{2}$ ] & upd . @xcite + @xmath227 @xmath229 & 6.8 & @xmath162 @xmath228^{2}$ ] & @xmath212 & 2.0 & @xmath166 @xmath228^{2}$ ] & upd . @xcite + @xmath227 @xmath231 & 5.6 & @xmath162 @xmath228^{2}$ ] & @xmath214 & 1.9 & @xmath166 @xmath228^{2}$ ] & upd . @xcite + @xmath229 @xmath230 & 5.2 & @xmath162 @xmath228^{2}$ ] & @xmath225 & 1.9 & @xmath166 @xmath228^{2}$ ] & upd . @xcite + @xmath229 @xmath232 & 7.0 & @xmath162 @xmath228^{2}$ ] & @xmath210 & 2.1 & @xmath166 @xmath228^{2}$ ] & upd . @xcite + @xmath230 @xmath231 & 7.0 & @xmath162 @xmath228^{2}$ ] & @xmath210 & 2.1 & @xmath166 @xmath228^{2}$ ] & upd . @xcite + @xmath232 @xmath231 & 6.8 & @xmath162 @xmath228^{2}$ ] & @xmath216 & 2.2 & @xmath166 @xmath228^{2}$ ] & upd . @xcite + @xmath203 @xmath233 & & + @xmath205 @xmath206 & & decay & & decay & key + @xmath207 @xmath234 & 1.2 & @xmath182 @xmath209^{2}$ ] & @xmath235 & 2.1 & @xmath236 @xmath209^{2}$ ] & @xmath237 in @xmath238ti & unimp . @xcite + & 0.39 & @xmath209^{2}$ ] & @xmath159 & & & & + & 0.41 & @xmath209^{2}$ ] & @xmath239 & & & & + & 16 & @xmath209^{2}$ ] & @xmath240 & & & & + @xmath207 @xmath241 & 6.7 & @xmath242 @xmath209^{2}$ ] & @xmath243 & 6 & @xmath242 @xmath209^{2}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath207 @xmath244 & 1.3 & @xmath191 @xmath209^{2}$ ] & @xmath245 & 2.3 & @xmath191 @xmath209^{2}$ ] & @xmath245 & corr.(@xmath102 ) @xcite + @xmath207 @xmath207 & 6.7 & @xmath242 @xmath209^{2}$ ] & @xmath243 & 6 & @xmath242 @xmath209^{2}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath207 @xmath215 & 2.1 & @xmath209^{2}$ ] & @xmath159 & & none & n / a & new + & 3.6 & @xmath246 @xmath209^{2}$ ] & @xmath240 & & & & + @xmath207 @xmath208 & 7.6 & @xmath191 @xmath209^{2}$ ] & @xmath247 & 4.7 & @xmath191 @xmath209^{2}$ ] & @xmath247 & corr.(@xmath103 ) @xcite + @xmath207 @xmath211 & 1.3 & @xmath191 @xmath209^{2}$ ] & @xmath248 & 2.3 & @xmath191 @xmath209^{2}$ ] & @xmath248 & corr.(@xmath102 ) @xcite + @xmath207 @xmath213 & 7.6 & @xmath191 @xmath209^{2}$ ] & @xmath249 & 4.7 & @xmath191 @xmath209^{2}$ ] & @xmath249 & corr.(@xmath103 ) @xcite + @xmath215 @xmath244 & 6.2 & @xmath250 @xmath209^{2}$ ] & @xmath251 & 1.5 & @xmath191 @xmath209^{2}$ ] & @xmath251 & corr.(@xmath102 ) @xcite + @xmath215 @xmath208 & 1.2 & @xmath191 @xmath209^{2}$ ] & @xmath252 & 1.7 & @xmath191 @xmath209^{2}$ ] & @xmath253 & upd . @xcite + @xmath215 @xmath211 & 6.2 & @xmath250 @xmath209^{2}$ ] & @xmath251 & 1.5 & @xmath191 @xmath209^{2}$ ] & @xmath251 & corr.(@xmath102 ) @xcite + @xmath215 @xmath213 & 1.2 & @xmath191 @xmath209^{2}$ ] & @xmath252 & 1.8 & @xmath191 @xmath209^{2}$ ] & @xmath253 & upd . @xcite + @xmath208 @xmath234 & 6.7 & @xmath191 @xmath209^{2}$ ] & @xmath178 & 1.7 & @xmath166 @xmath209^{2}$ ] & @xmath178 & upd . @xcite + & 1.0 & @xmath166 @xmath209^{2}$ ] & @xmath254 & & & & + @xmath208 @xmath241 & 1.0 & @xmath166 @xmath209^{2}$ ] & @xmath167 & & none & n / a & new + @xmath208 @xmath244 & 2.2 & @xmath162 @xmath209^{2}$ ] & @xmath255 & 6.2 & @xmath162 @xmath209^{2}$ ] & @xmath255 & corr.(@xmath102 ) @xcite + @xmath208 @xmath207 & 1.0 & @xmath166 @xmath209^{2}$ ] & @xmath167 & 7.6 & @xmath182 @xmath209^{2}$ ] & @xmath188 & upd . @xcite + @xmath208 @xmath215 & 3.7 & @xmath162 @xmath209^{2}$ ] & @xmath178 & 1.7 & @xmath182 @xmath209^{2}$ ] & @xmath178 & upd . @xcite + @xmath208 @xmath211 & 2.2 & @xmath162 @xmath209^{2}$ ] & @xmath256 & 6.2 & @xmath162 @xmath209^{2}$ ] & @xmath255 & corr.(@xmath102 ) @xcite + @xmath211 @xmath234 & 8.5 & @xmath191 @xmath209^{2}$ ] & @xmath174 & 1.6 & @xmath166 @xmath209^{2}$ ] & @xmath174 & upd . @xcite + & 7.1 & @xmath162 @xmath209^{2}$ ] & @xmath195 & & & & + @xmath211 @xmath241 & 9.7 & @xmath162 @xmath209^{2}$ ] & @xmath163 & 8.5 & @xmath182 @xmath209^{2}$ ] & @xmath185 & upd . @xcite + @xmath211 @xmath244 & 3.7 & @xmath162 @xmath209^{2}$ ] & @xmath257 & 4.9 & @xmath162 @xmath209^{2}$ ] & @xmath257 & corr.(@xmath102 ) @xcite + @xmath211 @xmath207 & 9.7 & @xmath162 @xmath209^{2}$ ] & @xmath163 & & none & n / a & new + @xmath211 @xmath215 & 4.6 & @xmath162 @xmath209^{2}$ ] & @xmath174 & 1.6 & @xmath182 @xmath209^{2}$ ] & @xmath174 & upd . @xcite + @xmath211 @xmath211 & 3.7 & @xmath162 @xmath209^{2}$ ] & @xmath258 & 4.9 & @xmath162 @xmath209^{2}$ ] & @xmath258 & corr.(@xmath102 ) @xcite + @xmath203 @xmath233 & & + @xmath205 @xmath206 & & decay & & decay & key + @xmath213 @xmath234 & 6.7 & @xmath191 @xmath209^{2}$ ] & @xmath178 & 1.7 & @xmath166 @xmath209^{2}$ ] & @xmath178 & upd . @xcite + & 1.0 & @xmath166 @xmath209^{2}$ ] & @xmath254 & & & & + @xmath213 @xmath241 & 1.0 & @xmath166 @xmath209^{2}$ ] & @xmath167 & 7.6 & @xmath182 @xmath209^{2}$ ] & @xmath188 & upd . @xcite + @xmath213 @xmath244 & 2.2 & @xmath162 @xmath209^{2}$ ] & @xmath256 & 6.2 & @xmath162 @xmath209^{2}$ ] & @xmath255 & corr.(@xmath102 ) @xcite + @xmath213 @xmath207 & 1.0 & @xmath166 @xmath209^{2}$ ] & @xmath167 & & none & n / a & new + @xmath213 @xmath215 & 3.7 & @xmath162 @xmath209^{2}$ ] & @xmath178 & 1.7 & @xmath182 @xmath209^{2}$ ] & @xmath178 & upd . @xcite + @xmath213 @xmath211 & 2.2 & @xmath162 @xmath209^{2}$ ] & @xmath255 & 6.2 & @xmath162 @xmath209^{2}$ ] & @xmath255 & corr.(@xmath102 ) @xcite + @xmath217 @xmath221 & 4.1 & @xmath191 @xmath220^{2}$ ] & @xmath259 & 1.7 & @xmath191 @xmath220^{2}$ ] & @xmath259 & corr.(@xmath103 ) @xcite + @xmath217 @xmath260 & 2.3 & @xmath162 @xmath220^{2}$ ] & @xmath261 & 1.4 & @xmath162 @xmath220^{2}$ ] & @xmath262 & unimp . @xcite + @xmath217 @xmath223 & 4.1 & @xmath191 @xmath220^{2}$ ] & @xmath259 & 1.7 & @xmath191 @xmath220^{2}$ ] & @xmath259 & corr.(@xmath103 ) @xcite + @xmath217 @xmath224 & 2.3 & @xmath162 @xmath220^{2}$ ] & @xmath261 & 2.3 & @xmath191 @xmath220^{2}$ ] & @xmath262 & unimp . @xcite + @xmath218 @xmath217 & 1.2 & @xmath182 @xmath220^{2}$ ] & @xmath235 & 2.1 & @xmath236 @xmath220^{2}$ ] & @xmath237 in @xmath238ti & unimp . @xcite + & 0.38 & @xmath220^{2}$ ] & @xmath159 & & & & + & 0.41 & @xmath220^{2}$ ] & @xmath239 & & & & + & 16 & @xmath220^{2}$ ] & @xmath240 & & & & + @xmath218 @xmath218 & 6.7 & @xmath242 @xmath220^{2}$ ] & @xmath243 & 6 & @xmath242 @xmath220^{2}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath218 @xmath221 & 1.3 & @xmath191 @xmath220^{2}$ ] & @xmath248 & 2.3 & @xmath191 @xmath220^{2}$ ] & @xmath248 & corr.(@xmath103 ) @xcite + @xmath218 @xmath263 & 6.7 & @xmath242 @xmath220^{2}$ ] & @xmath243 & 6 & @xmath242 @xmath220^{2}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath218 @xmath264 & 2.1 & @xmath220^{2}$ ] & @xmath159 & & none & n / a & new + & 3.6 & @xmath246 @xmath220^{2}$ ] & @xmath240 & & & & + @xmath218 @xmath260 & 7.6 & @xmath191 @xmath220^{2}$ ] & @xmath249 & 4.7 & @xmath191 @xmath220^{2}$ ] & @xmath249 & corr.(@xmath103 ) @xcite + @xmath218 @xmath223 & 1.3 & @xmath191 @xmath220^{2}$ ] & @xmath245 & 2.3 & @xmath191 @xmath220^{2}$ ] & @xmath245 & corr.(@xmath102 ) @xcite + @xmath218 @xmath224 & 7.6 & @xmath191 @xmath220^{2}$ ] & @xmath247 & 4.7 & @xmath191 @xmath220^{2}$ ] & @xmath249 & corr.(@xmath103 ) @xcite + @xmath221 @xmath217 & 8.5 & @xmath191 @xmath220^{2}$ ] & @xmath174 & 1.6 & @xmath166 @xmath220^{2}$ ] & @xmath174 & upd . @xcite + & 7.1 & @xmath162 @xmath220^{2}$ ] & @xmath195 & & & & + @xmath221 @xmath218 & 9.7 & @xmath162 @xmath220^{2}$ ] & @xmath163 & & none & n / a & new + @xmath221 @xmath221 & 3.7 & @xmath162 @xmath220^{2}$ ] & @xmath258 & 4.9 & @xmath162 @xmath220^{2}$ ] & @xmath258 & corr.(@xmath102 ) @xcite + @xmath221 @xmath263 & 9.7 & @xmath162 @xmath220^{2}$ ] & @xmath163 & 8.5 & @xmath182 @xmath220^{2}$ ] & @xmath185 & upd . @xcite + @xmath221 @xmath264 & 4.6 & @xmath162 @xmath220^{2}$ ] & @xmath174 & 1.6 & @xmath182 @xmath220^{2}$ ] & @xmath174 & upd . @xcite + @xmath221 @xmath223 & 3.7 & @xmath162 @xmath220^{2}$ ] & @xmath257 & 4.9 & @xmath162 @xmath220^{2}$ ] & @xmath257 & corr.(@xmath102 ) @xcite + @xmath203 @xmath233 & & + @xmath205 @xmath206 & & decay & & decay & key + @xmath223 @xmath217 & 8.5 & @xmath191 @xmath220^{2}$ ] & @xmath174 & 1.6 & @xmath166 @xmath220^{2}$ ] & @xmath174 & upd . @xcite + & 7.1 & @xmath162 @xmath220^{2}$ ] & @xmath195 & & & & + @xmath223 @xmath218 & 9.7 & @xmath162 @xmath220^{2}$ ] & @xmath163 & 8.5 & @xmath182 @xmath220^{2}$ ] & @xmath185 & upd . @xcite + @xmath223 @xmath221 & 3.7 & @xmath162 @xmath220^{2}$ ] & @xmath257 & 4.9 & @xmath162 @xmath220^{2}$ ] & @xmath257 & corr.(@xmath102 ) @xcite + @xmath223 @xmath263 & 9.7 & @xmath162 @xmath220^{2}$ ] & @xmath163 & & none & n / a & new + @xmath223 @xmath264 & 4.6 & @xmath162 @xmath220^{2}$ ] & @xmath174 & 1.6 & @xmath182 @xmath220^{2}$ ] & @xmath174 & upd . @xcite + @xmath223 @xmath223 & 3.7 & @xmath162 @xmath220^{2}$ ] & @xmath258 & 4.9 & @xmath162 @xmath220^{2}$ ] & @xmath258 & corr.(@xmath102 ) @xcite + @xmath224 @xmath217 & 6.7 & @xmath191 @xmath220^{2}$ ] & @xmath178 & 1.7 & @xmath166 @xmath220^{2}$ ] & @xmath178 & upd . @xcite + & 1.0 & @xmath166 @xmath220^{2}$ ] & @xmath254 & & & & + @xmath224 @xmath218 & 1.0 & @xmath166 @xmath220^{2}$ ] & @xmath167 & 7.6 & @xmath182 @xmath220^{2}$ ] & @xmath188 & upd . @xcite + @xmath224 @xmath221 & 2.2 & @xmath162 @xmath220^{2}$ ] & @xmath256 & 6.2 & @xmath162 @xmath220^{2}$ ] & @xmath255 & corr.(@xmath102 ) @xcite + @xmath224 @xmath263 & 1.0 & @xmath166 @xmath220^{2}$ ] & @xmath167 & & none & n / a & new + @xmath224 @xmath264 & 3.7 & @xmath162 @xmath220^{2}$ ] & @xmath178 & 1.7 & @xmath182 @xmath220^{2}$ ] & @xmath178 & upd . @xcite + @xmath224 @xmath223 & 2.2 & @xmath162 @xmath220^{2}$ ] & @xmath255 & 6.2 & @xmath162 @xmath220^{2}$ ] & @xmath255 & corr.(@xmath102 ) @xcite + @xmath226 @xmath229 & 4.1 & @xmath191 @xmath228^{2}$ ] & @xmath259 & 1.7 & @xmath191 @xmath220^{3}$ ] & @xmath259 & corr.(@xmath103 ) @xcite + @xmath226 @xmath231 & 2.3 & @xmath162 @xmath228^{2}$ ] & @xmath261 & 2.3 & @xmath191 @xmath228^{2}$ ] & @xmath262 & unimp . @xcite + @xmath226 @xmath265 & 4.1 & @xmath191 @xmath228^{2}$ ] & @xmath259 & 1.7 & @xmath191 @xmath228^{2}$ ] & @xmath259 & corr.(@xmath103 ) @xcite + @xmath226 @xmath266 & 2.3 & @xmath162 @xmath228^{2}$ ] & @xmath261 & 2.3 & @xmath191 @xmath228^{2}$ ] & @xmath262 & unimp . @xcite + @xmath227 @xmath226 & 1.2 & @xmath182 @xmath228^{2}$ ] & @xmath235 & 2.1 & @xmath236 @xmath228^{2}$ ] & @xmath237 in @xmath238ti & unimp . @xcite + & 0.38 & @xmath228^{2}$ ] & @xmath159 & & & & + & 0.41 & @xmath228^{2}$ ] & @xmath239 & & & & + & 16 & @xmath228^{2}$ ] & @xmath240 & & & & + @xmath227 @xmath227 & 6.7 & @xmath242 @xmath228^{2}$ ] & @xmath243 & 6 & @xmath242 @xmath209^{3}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath227 @xmath229 & 1.3 & @xmath191 @xmath228^{2}$ ] & @xmath248 & 2.3 & @xmath191 @xmath228^{2}$ ] & @xmath248 & corr.(@xmath102 ) @xcite + @xmath227 @xmath230 & 6.7 & @xmath242 @xmath228^{2}$ ] & @xmath243 & 6 & @xmath242 @xmath209^{3}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath227 @xmath232 & 2.1 & @xmath228^{2}$ ] & @xmath159 & & none & n / a & new + & 3.6 & @xmath246 @xmath228^{2}$ ] & @xmath240 & & & & + @xmath227 @xmath231 & 7.6 & @xmath191 @xmath228^{2}$ ] & @xmath249 & 4.7 & @xmath191 @xmath228^{2}$ ] & @xmath249 & corr.(@xmath103 ) @xcite + @xmath227 @xmath265 & 1.3 & @xmath191 @xmath228^{2}$ ] & @xmath245 & 2.3 & @xmath191 @xmath228^{2}$ ] & @xmath245 & corr.(@xmath102 ) @xcite + @xmath227 @xmath266 & 7.6 & @xmath191 @xmath228^{2}$ ] & @xmath247 & 4.7 & @xmath191 @xmath228^{2}$ ] & @xmath247 & corr.(@xmath103 ) @xcite + @xmath203 @xmath233 & & + @xmath205 @xmath206 & & decay & & decay & key + @xmath229 @xmath226 & 8.5 & @xmath191 @xmath228^{2}$ ] & @xmath174 & 1.6 & @xmath166 @xmath209^{2}$ ] & @xmath174 & upd . @xcite + & 7.1 & @xmath162 @xmath228^{2}$ ] & @xmath195 & & & & + @xmath229 @xmath227 & 9.7 & @xmath162 @xmath228^{2}$ ] & @xmath163 & & none & n / a & new + @xmath229 @xmath229 & 3.7 & @xmath162 @xmath228^{2}$ ] & @xmath258 & 4.9 & @xmath162 @xmath228^{2}$ ] & @xmath258 & corr.(@xmath102 ) @xcite + @xmath229 @xmath230 & 9.7 & @xmath162 @xmath228^{2}$ ] & @xmath163 & 8.5 & @xmath182 @xmath228^{2}$ ] & @xmath185 & upd . @xcite + @xmath229 @xmath232 & 4.6 & @xmath162 @xmath228^{2}$ ] & @xmath174 & 1.6 & @xmath182 @xmath228^{2}$ ] & @xmath174 & upd . @xcite + @xmath229 @xmath265 & 3.7 & @xmath162 @xmath228^{2}$ ] & @xmath257 & 4.9 & @xmath162 @xmath228^{2}$ ] & @xmath257 & corr.(@xmath102 ) @xcite + @xmath230 @xmath226 & 1.2 & @xmath182 @xmath228^{2}$ ] & @xmath235 & 2.1 & @xmath236 @xmath228^{2}$ ] & @xmath237 in @xmath238ti & unimp . @xcite + & 0.38 & @xmath228^{2}$ ] & @xmath159 & & & & + & 0.41 & @xmath228^{2}$ ] & @xmath239 & & & & + @xmath230 @xmath226 & 16 & @xmath228^{2}$ ] & @xmath240 & & & & + @xmath230 @xmath227 & 6.7 & @xmath242 @xmath228^{2}$ ] & @xmath243 & 6 & @xmath242 @xmath209^{3}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath230 @xmath229 & 1.3 & @xmath191 @xmath228^{2}$ ] & @xmath245 & 2.3 & @xmath191 @xmath228^{2}$ ] & @xmath245 & corr.(@xmath102 ) @xcite + @xmath230 @xmath230 & 6.7 & @xmath242 @xmath228^{2}$ ] & @xmath243 & 6 & @xmath242 @xmath209^{3}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath230 @xmath232 & 2.1 & @xmath228^{2}$ ] & @xmath159 & & none & n / a & new + & 3.6 & @xmath246 @xmath228^{2}$ ] & @xmath240 & & & & + @xmath230 @xmath231 & 7.6 & @xmath191 @xmath228^{2}$ ] & @xmath247 & 4.7 & @xmath191 @xmath228^{2}$ ] & @xmath247 & corr.(@xmath103 ) @xcite + @xmath230 @xmath265 & 1.3 & @xmath191 @xmath228^{2}$ ] & @xmath248 & 2.3 & @xmath191 @xmath228^{2}$ ] & @xmath248 & corr.(@xmath102 ) @xcite + @xmath230 @xmath266 & 7.6 & @xmath191 @xmath228^{2}$ ] & @xmath249 & 4.7 & @xmath191 @xmath228^{2}$ ] & @xmath249 & corr.(@xmath103 ) @xcite + @xmath232 @xmath229 & 6.2 & @xmath250 @xmath228^{2}$ ] & @xmath251 & 1.5 & @xmath191 @xmath228^{2}$ ] & @xmath251 & corr.(@xmath102 ) @xcite + @xmath232 @xmath231 & 1.2 & @xmath191 @xmath228^{2}$ ] & @xmath252 & 1.7 & @xmath191 @xmath228^{2}$ ] & @xmath253 & upd . @xcite + @xmath232 @xmath265 & 6.2 & @xmath250 @xmath228^{2}$ ] & @xmath251 & 1.5 & @xmath191 @xmath228^{2}$ ] & @xmath251 & corr.(@xmath102 ) @xcite + @xmath232 @xmath266 & 1.2 & @xmath191 @xmath228^{2}$ ] & @xmath252 & 1.8 & @xmath191 @xmath209^{3}$ ] & @xmath253 & upd . @xcite + @xmath231 @xmath226 & 6.7 & @xmath191 @xmath228^{2}$ ] & @xmath178 & 1.7 & @xmath166 @xmath228^{2}$ ] & @xmath178 & upd . @xcite + & 1.0 & @xmath166 @xmath228^{2}$ ] & @xmath254 & & & & + @xmath231 @xmath227 & 1.0 & @xmath166 @xmath228^{2}$ ] & @xmath167 & & none & n / a & new + @xmath231 @xmath229 & 2.2 & @xmath162 @xmath228^{2}$ ] & @xmath255 & 6.2 & @xmath162 @xmath228^{2}$ ] & @xmath255 & corr.(@xmath102 ) @xcite + @xmath231 @xmath230 & 1.0 & @xmath166 @xmath228^{2}$ ] & @xmath167 & 7.6 & @xmath182 @xmath228^{2}$ ] & @xmath188 & upd . @xcite + @xmath231 @xmath232 & 3.7 & @xmath162 @xmath228^{2}$ ] & @xmath178 & 1.7 & @xmath182 @xmath228^{2}$ ] & @xmath178 & upd . @xcite + @xmath231 @xmath265 & 2.2 & @xmath162 @xmath228^{2}$ ] & @xmath256 & 6.2 & @xmath162 @xmath228^{2}$ ] & @xmath255 & corr.(@xmath102 ) @xcite + @xmath267 @xmath233 & & + @xmath205 @xmath206 & & decay & & decay & key + @xmath234 @xmath244 & 2.6 & @xmath182 @xmath268^{2}$ ] & @xmath259 & 0.03 & @xmath268 $ ] & apv in cs & unimp . @xcite , + & & & & 0.18 & @xmath268 $ ] & @xmath269 & @xcite + @xmath234 @xmath217 & 0.36 & @xmath270^{2}$ ] & @xmath235 & 4.5 & @xmath236 @xmath270^{2}$ ] & @xmath237 in @xmath238ti & unimp . @xcite + & 11 & @xmath270^{2}$ ] & @xmath239 & & & & + & 1.5 & @xmath172 @xmath270^{2}$ ] & @xmath159 & & & & + & 1.9 & @xmath246 @xmath270^{2}$ ] & @xmath240 & & & & + @xmath234 @xmath217 & 0.36 & @xmath268^{2}$ ] & @xmath235 & 4.3 & @xmath236 @xmath268^{2}$ ] & @xmath237 in @xmath238ti & unimp . @xcite + & 11 & @xmath268^{2}$ ] & @xmath239 & & & & + & 1.5 & @xmath172 @xmath268^{2}$ ] & @xmath159 & & & & + & 1.9 & @xmath246 @xmath268^{2}$ ] & @xmath271 & & & & + @xmath234 @xmath218 & 2.7 & @xmath219 @xmath268^{2}$ ] & @xmath243 & 3 & @xmath219 @xmath268^{2}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath234 @xmath221 & 1.6 & @xmath166 @xmath268^{2}$ ] & @xmath248 & 4.7 & @xmath166 @xmath268^{2}$ ] & @xmath248 & upd . @xcite + @xmath234 @xmath263 & 2.8 & @xmath182 @xmath270^{2}$ ] & @xmath272 & 0.02 & @xmath270 $ ] & apv in cs & unimp . @xcite , + & & & & 0.21 & @xmath270 $ ] & @xmath196 & @xcite + @xmath234 @xmath226 & 1.2 & @xmath166 @xmath270^{2}$ ] & @xmath195 & 0.02 & @xmath270 $ ] & apv in cs & upd . @xcite , + & & & & 0.12 & @xmath270 $ ] & @xmath196 & @xcite + & 2.0 & @xmath166 @xmath270^{2}$ ] & @xmath174 & & & & + & 2.4 & @xmath166 @xmath270^{2}$ ] & @xmath273 & & & & + @xmath234 @xmath226 & 1.2 & @xmath166 @xmath268^{2}$ ] & @xmath195 & 2.4 & @xmath166 @xmath268^{2}$ ] & @xmath273 & upd . @xcite + & 2.0 & @xmath166 @xmath268^{2}$ ] & @xmath174 & & & & @xmath94 + & 2.4 & @xmath166 @xmath268^{2}$ ] & @xmath273 & & & & + @xmath234 @xmath227 & 2.3 & @xmath166 @xmath268^{2}$ ] & @xmath163 & & none & n / a & new + & 3.6 & @xmath166 @xmath268^{2}$ ] & @xmath274 & & & & + @xmath234 @xmath229 & 2.7 & @xmath166 @xmath268^{2}$ ] & @xmath258 & 5.9 & @xmath166 @xmath268^{2}$ ] & @xmath258 & upd . @xcite + @xmath241 @xmath244 & 9.3 & @xmath268^{2}$ ] & @xmath261 & 4.3 & @xmath162 @xmath268^{2}$ ] & @xmath275 & unimp . @xcite + @xmath241 @xmath217 & 2.7 & @xmath219 @xmath268^{2}$ ] & @xmath243 & 3 & @xmath219 @xmath268^{2}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath241 @xmath218 & 0.36 & @xmath276^{2}$ ] & @xmath235 & 4.5 & @xmath236 @xmath276^{2}$ ] & @xmath237 in @xmath238ti & unimp . @xcite + & 1.1 & @xmath276^{2}$ ] & @xmath239 & & & & + & 1.6 & @xmath172 @xmath276^{2}$ ] & @xmath159 & & & & + & 1.9 & @xmath246 @xmath276^{2}$ ] & @xmath240 & & & & + @xmath267 @xmath233 & & + @xmath205 @xmath206 & & decay & & decay & key + @xmath241 @xmath218 & 76 & @xmath268^{2}$ ] & @xmath159 & & none & n / a & new + & 1.1 & @xmath277 @xmath268^{2}$ ] & @xmath271 & & & & + @xmath241 @xmath221 & 9.4 & @xmath166 @xmath268^{2}$ ] & @xmath249 & 2.7 & @xmath162 @xmath268^{2}$ ] & @xmath278 & unimp . @xcite + @xmath241 @xmath264 & 2.8 & @xmath182 @xmath276^{2}$ ] & @xmath272 & 0.02 & @xmath276 $ ] & apv in cs & unimp . @xcite , + & & & & 0.21 & @xmath276 $ ] & @xmath196 & @xcite + @xmath241 @xmath226 & 2.3 & @xmath166 @xmath268^{2}$ ] & @xmath163 & 2.7 & @xmath166 @xmath268^{2}$ ] & @xmath279 & upd . @xcite + & 2.9 & @xmath166 @xmath268^{2}$ ] & @xmath280 & & & & + @xmath241 @xmath227 & 1.2 & @xmath166 @xmath276^{2}$ ] & @xmath195 & 0.02 & @xmath276 $ ] & apv in cs & upd . @xcite , + & & & & 0.12 & @xmath276 $ ] & @xmath196 & @xcite + & 2.0 & @xmath166 @xmath276^{2}$ ] & @xmath174 & & & & + & 2.4 & @xmath166 @xmath276^{2}$ ] & @xmath273 & & & & + @xmath241 @xmath227 & 1.2 & @xmath166 @xmath268^{2}$ ] & @xmath174 & & none & n / a & @xmath94 new + & 3.4 & @xmath166 @xmath268^{2}$ ] & @xmath175 & & & & + @xmath244 @xmath217 & 1.6 & @xmath166 @xmath268^{2}$ ] & @xmath245 & 4.7 & @xmath166 @xmath268^{2}$ ] & @xmath245 & upd . @xcite + @xmath244 @xmath218 & 9.4 & @xmath166 @xmath268^{2}$ ] & @xmath247 & 2.7 & @xmath162 @xmath268^{2}$ ] & @xmath281 & unimp . @xcite + @xmath244 @xmath221 & 0.36 & @xmath282^{2}$ ] & @xmath235 & 4.5 & @xmath236 @xmath282^{2}$ ] & @xmath237 in @xmath238ti & unimp . @xcite + & 11 & @xmath282^{2}$ ] & @xmath239 & & & & + & 1.5 & @xmath172 @xmath282^{2}$ ] & @xmath159 & & & & + & 1.9 & @xmath246 @xmath282^{2}$ ] & @xmath240 & & & & + @xmath244 @xmath260 & 2.8 & @xmath182 @xmath282^{2}$ ] & @xmath272 & 0.02 & @xmath282 $ ] & apv in cs & unimp . @xcite , + & & & & 0.21 & @xmath282 $ ] & @xmath196 & @xcite + @xmath244 @xmath226 & 2.7 & @xmath166 @xmath268^{2}$ ] & @xmath257 & 5.9 & @xmath166 @xmath268^{2}$ ] & @xmath257 & upd . @xcite + @xmath244 @xmath229 & 1.2 & @xmath166 @xmath282^{2}$ ] & @xmath195 & 0.02 & @xmath282 $ ] & apv in cs & upd . @xcite , + & & & & 0.12 & @xmath282 $ ] & @xmath196 & @xcite + & 2.0 & @xmath166 @xmath282^{2}$ ] & @xmath174 & & & & + & 2.4 & @xmath166 @xmath282^{2}$ ] & @xmath273 & & & & + @xmath267 @xmath233 & & + @xmath205 @xmath206 & & decay & & decay & key + @xmath207 @xmath208 & 2.6 & @xmath182 @xmath283^{2}$ ] & @xmath259 & 0.03 & @xmath283 $ ] & apv in cs & unimp . @xcite , + & & & & 0.18 & @xmath283 $ ] & @xmath269 & @xcite + @xmath207 @xmath217 & 9.0 & @xmath166 @xmath270^{2}$ ] & @xmath199 & 0.21 & @xmath270 $ ] & @xmath200 & upd . @xcite , + & & & & 5.9 & @xmath201 @xmath270 $ ] & @xmath202 & @xcite + @xmath207 @xmath263 & 1.6 & @xmath270^{2}$ ] & @xmath284 & 0.21 & @xmath270 $ ] & @xmath200 & unimp . @xcite , + & & & & 0.21 & @xmath270 $ ] & @xmath285 & @xcite + @xmath207 @xmath263 & 0.36 & @xmath283^{2}$ ] & @xmath235 & 4.3 & @xmath236 @xmath283^{2}$ ] & @xmath237 in @xmath238ti & unimp . @xcite + & 11 & @xmath283^{2}$ ] & @xmath239 & & & & + & 1.5 & @xmath172 @xmath283^{2}$ ] & @xmath159 & & & & + & 1.9 & @xmath246 @xmath283^{2}$ ] & @xmath271 & & & & + @xmath207 @xmath264 & 2.7 & @xmath219 @xmath283^{2}$ ] & @xmath243 & 3 & @xmath219 @xmath283^{2}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath207 @xmath260 & 1.6 & @xmath166 @xmath283^{2}$ ] & @xmath248 & 4.7 & @xmath166 @xmath283^{2}$ ] & @xmath248 & upd . @xcite + @xmath207 @xmath230 & 5.9 & @xmath270^{2}$ ] & @xmath286 & 0.21 & @xmath270 $ ] & @xmath200 & unimp . @xcite , + & & & & 0.52 & @xmath270 $ ] & @xmath287 & @xcite + @xmath207 @xmath230 & 1.2 & @xmath166 @xmath283^{2}$ ] & @xmath195 & 2.4 & @xmath166 @xmath283^{2}$ ] & @xmath273 & upd . @xcite + & 2.0 & @xmath166 @xmath283^{2}$ ] & @xmath174 & & & & @xmath94 + & 2.4 & @xmath166 @xmath283^{2}$ ] & @xmath273 & & & & + @xmath207 @xmath232 & 2.3 & @xmath166 @xmath283^{2}$ ] & @xmath163 & & none & n / a & new + & 3.6 & @xmath166 @xmath283^{2}$ ] & @xmath274 & & & & + @xmath207 @xmath231 & 2.7 & @xmath166 @xmath283^{2}$ ] & @xmath258 & 5.9 & @xmath166 @xmath283^{2}$ ] & @xmath258 & upd . @xcite + @xmath215 @xmath208 & 4.1 & @xmath283^{2}$ ] & @xmath261 & 4.3 & @xmath162 @xmath283^{2}$ ] & @xmath275 & unimp . @xcite + @xmath215 @xmath218 & 9.0 & @xmath166 @xmath276^{2}$ ] & @xmath199 & 0.21 & @xmath276 $ ] & @xmath200 & upd . @xcite , + & & & & 5.9 & @xmath201 @xmath276 $ ] & @xmath202 & @xcite + @xmath215 @xmath263 & 2.7 & @xmath219 @xmath283^{2}$ ] & @xmath243 & 3 & @xmath219 @xmath283^{2}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath215 @xmath264 & 1.6 & @xmath276^{2}$ ] & @xmath284 & 0.21 & @xmath270 $ ] & @xmath200 & unimp . @xcite , + & & & & 0.21 & @xmath270 $ ] & @xmath285 & @xcite + @xmath215 @xmath264 & 76 & @xmath283^{2}$ ] & @xmath159 & & none & n / a & new + & 1.1 & @xmath277 @xmath283^{2}$ ] & @xmath271 & & & & + @xmath267 @xmath233 & & + @xmath205 @xmath206 & & decay & & decay & key + @xmath215 @xmath260 & 9.4 & @xmath166 @xmath283^{2}$ ] & @xmath249 & 2.7 & @xmath162 @xmath283^{2}$ ] & @xmath278 & unimp . @xcite + @xmath215 @xmath230 & 2.3 & @xmath166 @xmath283^{2}$ ] & @xmath163 & 2.7 & @xmath166 @xmath283^{2}$ ] & @xmath279 & upd . @xcite + & 2.9 & @xmath166 @xmath283^{2}$ ] & @xmath280 & & & & + @xmath215 @xmath232 & 5.9 & @xmath276^{2}$ ] & @xmath286 & 0.21 & @xmath276 $ ] & @xmath200 & unimp . @xcite , + & & & & 0.52 & @xmath276 $ ] & @xmath287 & @xcite + @xmath215 @xmath232 & 1.2 & @xmath166 @xmath283^{2}$ ] & @xmath174 & & none & n / a & @xmath94 new + & 3.4 & @xmath166 @xmath283^{2}$ ] & @xmath175 & & & & + @xmath208 @xmath221 & 9.0 & @xmath166 @xmath282^{2}$ ] & @xmath199 & 0.21 & @xmath282 $ ] & @xmath200 & upd . @xcite , + & & & & 5.9 & @xmath201 @xmath282 $ ] & @xmath202 & @xcite + @xmath208 @xmath263 & 1.6 & @xmath166 @xmath283^{2}$ ] & @xmath245 & 4.7 & @xmath166 @xmath283^{2}$ ] & @xmath245 & upd . @xcite + @xmath208 @xmath264 & 9.4 & @xmath166 @xmath283^{2}$ ] & @xmath247 & 2.7 & @xmath162 @xmath283^{2}$ ] & @xmath281 & unimp . @xcite + @xmath208 @xmath260 & 1.6 & @xmath282^{2}$ ] & @xmath284 & 0.21 & @xmath270 $ ] & @xmath200 & unimp . @xcite , + & & & & 0.21 & @xmath270 $ ] & @xmath285 & @xcite + @xmath208 @xmath230 & 2.7 & @xmath166 @xmath283^{2}$ ] & @xmath257 & 5.9 & @xmath166 @xmath283^{2}$ ] & @xmath257 & upd . @xcite + @xmath208 @xmath231 & 5.9 & @xmath282^{2}$ ] & @xmath286 & 0.21 & @xmath282 $ ] & @xmath200 & unimp . @xcite , + & & & & 0.52 & @xmath282 $ ] & @xmath287 & @xcite + @xmath211 @xmath288 & 2.6 & @xmath182 @xmath289^{2}$ ] & @xmath259 & 0.03 & @xmath289 $ ] & apv in cs & unimp . @xcite , + & & & & 0.18 & @xmath289 $ ] & @xmath269 & @xcite + @xmath211 @xmath223 & 0.36 & @xmath289^{2}$ ] & @xmath235 & 4.3 & @xmath236 @xmath289^{2}$ ] & @xmath237 in @xmath238ti & unimp . @xcite + & 11 & @xmath289^{2}$ ] & @xmath239 & & & & + & 1.5 & @xmath172 @xmath289^{2}$ ] & @xmath159 & & & & + & 1.9 & @xmath246 @xmath289^{2}$ ] & @xmath271 & & & & + @xmath211 @xmath224 & 2.7 & @xmath219 @xmath289^{2}$ ] & @xmath243 & 3 & @xmath219 @xmath289^{2}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath211 @xmath290 & 1.6 & @xmath166 @xmath289^{2}$ ] & @xmath248 & 4.7 & @xmath166 @xmath289^{2}$ ] & @xmath248 & upd . @xcite + @xmath211 @xmath265 & 1.2 & @xmath166 @xmath289^{2}$ ] & @xmath195 & 2.4 & @xmath166 @xmath289^{2}$ ] & @xmath273 & upd . @xcite + & 2.0 & @xmath166 @xmath289^{2}$ ] & @xmath174 & & & & @xmath94 + & 2.4 & @xmath166 @xmath289^{2}$ ] & @xmath273 & & & & + @xmath211 @xmath266 & 2.3 & @xmath166 @xmath289^{2}$ ] & @xmath163 & & none & n / a & new + & 3.6 & @xmath166 @xmath289^{2}$ ] & @xmath274 & & & & + @xmath211 @xmath291 & 2.7 & @xmath166 @xmath289^{2}$ ] & @xmath258 & 5.9 & @xmath166 @xmath289^{2}$ ] & @xmath258 & upd . @xcite + @xmath213 @xmath288 & 4.1 & @xmath289^{2}$ ] & @xmath261 & 4.3 & @xmath162 @xmath289^{2}$ ] & @xmath275 & unimp . @xcite + @xmath267 @xmath233 & & + @xmath205 @xmath206 & & decay & & decay & key + @xmath213 @xmath223 & 2.7 & @xmath219 @xmath289^{2}$ ] & @xmath243 & 3 & @xmath219 @xmath289^{2}$ ] & @xmath243 & @xmath94 agr . @xcite + @xmath213 @xmath224 & 76 & @xmath289^{2}$ ] & @xmath159 & & none & n / a & new + & 1.1 & @xmath277 @xmath289^{2}$ ] & @xmath271 & & & & + @xmath213 @xmath290 & 9.4 & @xmath166 @xmath289^{2}$ ] & @xmath249 & 2.7 & @xmath162 @xmath289^{2}$ ] & @xmath278 & unimp . @xcite + @xmath213 @xmath265 & 2.3 & @xmath166 @xmath289^{2}$ ] & @xmath163 & 2.7 & @xmath166 @xmath289^{2}$ ] & @xmath279 & upd . @xcite + & 2.9 & @xmath166 @xmath289^{2}$ ] & @xmath280 & & & & + @xmath213 @xmath266 & 1.2 & @xmath166 @xmath289^{2}$ ] & @xmath174 & & none & n / a & @xmath94 new + & 3.4 & @xmath166 @xmath289^{2}$ ] & @xmath175 & & & & + @xmath288 @xmath223 & 1.6 & @xmath166 @xmath289^{2}$ ] & @xmath245 & 4.7 & @xmath166 @xmath289^{2}$ ] & @xmath245 & upd . @xcite + @xmath288 @xmath224 & 9.4 & @xmath166 @xmath289^{2}$ ] & @xmath247 & 2.7 & @xmath162 @xmath289^{2}$ ] & @xmath281 & unimp . @xcite + @xmath288 @xmath265 & 2.7 & @xmath166 @xmath289^{2}$ ] & @xmath257 & 5.9 & @xmath166 @xmath289^{2}$ ] & @xmath257 & upd . @xcite + @xmath217 @xmath221 & 5.4 & @xmath162 @xmath268^{2}$ ] & @xmath251 & 2.1 & @xmath166 @xmath268^{2}$ ] & @xmath251 & upd . @xcite + @xmath217 @xmath226 & 1.6 & @xmath166 @xmath270^{2}$ ] & @xmath178 & 4.4 & @xmath166 @xmath270^{2}$ ] & @xmath292 & upd . @xcite + & 1.8 & @xmath166 @xmath270^{2}$ ] & @xmath254 & & & & + & 4.3 & @xmath166 @xmath270^{2}$ ] & @xmath292 & & & & + @xmath217 @xmath226 & 1.6 & @xmath166 @xmath268^{2}$ ] & @xmath178 & 4.4 & @xmath166 @xmath268^{2}$ ] & @xmath292 & @xmath94 upd . @xcite + & 1.8 & @xmath166 @xmath268^{2}$ ] & @xmath254 & & & & + & 4.3 & @xmath166 @xmath268^{2}$ ] & @xmath292 & & & & + @xmath217 @xmath227 & 2.4 & @xmath166 @xmath268^{2}$ ] & @xmath167 & & none & n / a & new + & 3.6 & @xmath166 @xmath268^{2}$ ] & @xmath293 & & & & + @xmath217 @xmath229 & 1.6 & @xmath166 @xmath268^{2}$ ] & @xmath255 & 7.3 & @xmath166 @xmath268^{2}$ ] & @xmath255 & upd . @xcite + @xmath218 @xmath221 & 1.0 & @xmath166 @xmath268^{2}$ ] & @xmath252 & 4.6 & @xmath191 @xmath268^{2}$ ] & @xmath253 & unimp . @xcite + @xmath218 @xmath226 & 2.4 & @xmath166 @xmath268^{2}$ ] & @xmath167 & 3.4 & @xmath166 @xmath268^{2}$ ] & @xmath294 & upd . @xcite + & 3.6 & @xmath166 @xmath268^{2}$ ] & @xmath295 & & & & + @xmath218 @xmath227 & 1.6 & @xmath166 @xmath276^{2}$ ] & @xmath178 & 4.4 & @xmath166 @xmath276^{2}$ ] & @xmath292 & upd . @xcite + & 1.8 & @xmath166 @xmath276^{2}$ ] & @xmath254 & & & & + & 4.3 & @xmath166 @xmath276^{2}$ ] & @xmath292 & & & & + @xmath218 @xmath227 & 9.2 & @xmath162 @xmath268^{2}$ ] & @xmath178 & & none & n / a & @xmath94 new + & 3.4 & @xmath166 @xmath268^{2}$ ] & @xmath179 & & & & + @xmath221 @xmath226 & 1.6 & @xmath166 @xmath268^{2}$ ] & @xmath256 & 7.3 & @xmath166 @xmath268^{2}$ ] & @xmath256 & upd . @xcite + @xmath221 @xmath229 & 1.6 & @xmath166 @xmath282^{2}$ ] & @xmath178 & 4.4 & @xmath166 @xmath282^{2}$ ] & @xmath292 & upd . @xcite + & 1.8 & @xmath166 @xmath282^{2}$ ] & @xmath254 & & & & + & 4.3 & @xmath166 @xmath282^{2}$ ] & @xmath292 & & & & + @xmath267 @xmath233 & & + @xmath205 @xmath206 & & decay & & decay & key + @xmath263 @xmath260 & 5.4 & @xmath162 @xmath283^{2}$ ] & @xmath251 & 2.1 & @xmath166 @xmath283^{2}$ ] & @xmath251 & upd . @xcite + @xmath263 @xmath230 & 2.9 & @xmath270^{2}$ ] & @xmath296 & 0.21 & @xmath270 $ ] & @xmath285 & unimp . @xcite , + & & & & 0.52 & @xmath270 $ ] & @xmath287 & @xcite + @xmath263 @xmath230 & 1.6 & @xmath166 @xmath283^{2}$ ] & @xmath178 & 4.4 & @xmath166 @xmath283^{2}$ ] & @xmath292 & @xmath94 upd . @xcite + & 1.8 & @xmath166 @xmath283^{2}$ ] & @xmath254 & & & & + & 4.3 & @xmath166 @xmath283^{2}$ ] & @xmath292 & & & & + @xmath263 @xmath232 & 2.4 & @xmath166 @xmath283^{2}$ ] & @xmath167 & & none & n / a & new + & 3.6 & @xmath166 @xmath283^{2}$ ] & @xmath293 & & & & + @xmath263 @xmath231 & 1.6 & @xmath166 @xmath283^{2}$ ] & @xmath255 & 7.3 & @xmath166 @xmath283^{2}$ ] & @xmath255 & upd . @xcite + @xmath264 @xmath260 & 1.0 & @xmath166 @xmath283^{2}$ ] & @xmath252 & 4.6 & @xmath191 @xmath283^{2}$ ] & @xmath253 & unimp . @xcite + @xmath264 @xmath230 & 2.4 & @xmath166 @xmath283^{2}$ ] & @xmath167 & 3.4 & @xmath166 @xmath283^{2}$ ] & @xmath294 & upd . @xcite + & 3.6 & @xmath166 @xmath283^{2}$ ] & @xmath295 & & & & + @xmath264 @xmath232 & 2.9 & @xmath276^{2}$ ] & @xmath296 & 0.21 & @xmath276 $ ] & @xmath285 & unimp . @xcite , + & & & & 0.52 & @xmath276 $ ] & @xmath287 & @xcite + @xmath264 @xmath232 & 9.2 & @xmath162 @xmath283^{2}$ ] & @xmath178 & & none & n / a & @xmath94 new + & 3.4 & @xmath166 @xmath283^{2}$ ] & @xmath179 & & & & + @xmath260 @xmath230 & 1.6 & @xmath166 @xmath283^{2}$ ] & @xmath256 & 7.3 & @xmath166 @xmath283^{2}$ ] & @xmath256 & upd . @xcite + @xmath260 @xmath231 & 2.9 & @xmath282^{2}$ ] & @xmath296 & 0.21 & @xmath282 $ ] & @xmath285 & unimp . @xcite , + & & & & 0.52 & @xmath282 $ ] & @xmath287 & @xcite + @xmath223 @xmath290 & 5.4 & @xmath162 @xmath289^{2}$ ] & @xmath251 & 2.1 & @xmath166 @xmath289^{2}$ ] & @xmath251 & upd . @xcite + @xmath223 @xmath265 & 1.6 & @xmath166 @xmath289^{2}$ ] & @xmath178 & 4.4 & @xmath166 @xmath289^{2}$ ] & @xmath292 & @xmath94 upd . @xcite + & 1.8 & @xmath166 @xmath289^{2}$ ] & @xmath254 & & & & + & 4.3 & @xmath166 @xmath289^{2}$ ] & @xmath292 & & & & + @xmath223 @xmath266 & 2.4 & @xmath166 @xmath289^{2}$ ] & @xmath167 & & none & n / a & new + & 3.6 & @xmath166 @xmath289^{2}$ ] & @xmath293 & & & & + @xmath223 @xmath291 & 1.6 & @xmath166 @xmath289^{2}$ ] & @xmath255 & 7.3 & @xmath166 @xmath289^{2}$ ] & @xmath255 & upd . @xcite + @xmath224 @xmath290 & 1.0 & @xmath166 @xmath289^{2}$ ] & @xmath252 & 4.6 & @xmath191 @xmath289^{2}$ ] & @xmath253 & unimp . @xcite + @xmath224 @xmath265 & 2.4 & @xmath166 @xmath289^{2}$ ] & @xmath167 & 3.4 & @xmath166 @xmath289^{2}$ ] & @xmath294 & upd . @xcite + & 3.6 & @xmath166 @xmath289^{2}$ ] & @xmath295 & & & & + @xmath224 @xmath266 & 9.2 & @xmath162 @xmath289^{2}$ ] & @xmath178 & & none & n / a & @xmath94 new + & 3.4 & @xmath166 @xmath289^{2}$ ] & @xmath179 & & & & + @xmath290 @xmath265 & 1.6 & @xmath166 @xmath289^{2}$ ] & @xmath256 & 7.3 & @xmath166 @xmath289^{2}$ ] & @xmath256 & upd . @xcite + we have defined the decay constants of vector and pseudoscalar mesons through @xmath297 where @xmath298 is the coefficient of @xmath48 in the quark model wavefunction of the meson . as @xmath298 is not standard notation we shall describe it in some detail . firstly , it is only of relevance to the light mesons composed of @xmath299-quarks , as it is assumed that the charmed and bottom meson wavefunctions consist entirely of one quark bilinear , @xmath300 is entirely @xmath301 , so @xmath302 and all other @xmath303 . hence for mesons which are not part of the light su(3)@xmath304 octet or singlet , @xmath305 for the relevant @xmath306 and @xmath307 . similarly for the charged light mesons , @xmath308 is entirely @xmath309 , hence @xmath310 . for the neutral light mesons , we obtain @xmath54 from the standard pdg @xcite definition of the pseudoscalar decay constant @xmath311 where @xmath312 is the vector @xmath313 , and @xmath314 are su(3)-flavour indices . @xmath315 denotes a basis vector of the eight - dimensional representation of flavour su(3 ) , and @xmath316 are the gell - mann matrices ( normalized such that @xmath317 ; also here @xmath318 is defined as @xmath319 times the three by three identity matrix ) . to relate ( [ pcac_condition_equation ] ) and ( [ pseudoscalar_meson_decay_constant_definition_equation ] ) we note that the quark bilinears @xmath320 can be written as linear combinations of @xmath321 , so that @xmath322 expressing the physical meson states @xmath323 in terms of the basis states @xmath324 , we arrive at the generic equation ( [ pcac_condition_equation ] ) , where the coefficients @xmath325 are given as @xmath326 . let us consider a specific example and determine @xmath327 . we find @xmath328 and hence @xmath329 . note that with our definition ( [ pseudoscalar_meson_decay_constant_definition_equation ] ) @xmath330 mev . in our numerical analysis we take into account @xmath331-@xmath332 mixing , so @xmath108 and @xmath132 are not exactly @xmath333 and @xmath334 , but mixtures with a mixing angle @xmath335 radians @xcite , @xmath336 p^{{\mu}}$ ] . for @xmath153 and @xmath337 we assume ideal mixing , so that @xmath338 and @xmath339 . the non - trivial coefficients @xmath54 can be read off the quark bilinear coefficients listed in table [ quark_bilinear_coefficients_table ] . the @xmath47 are defined to be the same as the @xmath54 for their pseudoscalar counterparts . .non trivial quark bilinear coefficients . [ cols="<,<",options="header " , ] let us now discuss the derivation of the pseudoscalar matrix element from the axial vector matrix element ( [ pcac_condition_equation ] ) in its general form @xmath340 with @xmath341 . applying @xmath342 to both sides leads to @xmath343 now @xmath344 assuming that the quark fields satisfy the dirac equation , and @xmath345 here is defined as @xmath346 combining this result with equation ( [ differentiated_pseudoscalar_meson_decay_constant_definition_equation ] ) at @xmath347 leads to @xmath348 since @xmath349 where @xmath350 is defined such that @xmath351 , we arrive at @xmath352 by comparison with equation ( [ pseudoscalar_meson_approximation_equationi ] ) , @xmath56 is identified as @xmath353 take the neutral pion as an example : @xmath354 so that @xmath355 and analogously @xmath356 . we note that this result is in disagreement with @xcite . in this appendix we present the feynman graphs and matrix elements of the various decays . [ [ charged - lepton - decaying - into - two - charged - leptons - and - one - charged - anti - lepton ] ] charged lepton decaying into two charged leptons and one charged anti - lepton ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ( 256 , 100 ) ( 1 , 1 ) ( 0 , 15 ) ( 10 , 15 ) ( 0 , ) /straight1 ( 5 , 16 ) [ bl]@xmath359 ( 10 , 15 ) ( 25 , 20 ) ( 0 , ) /straight1 ( 15 , 18 ) [ bl]@xmath360 ( 10 , 15 ) ( 15 , 5 ) ( 0 , ) /scalardash1 ( 10 , 7)[bl]@xmath357 ( 15 , 5 ) ( 25 , 10 ) ( 0 , ) /straight1 ( 20 , 9 ) [ bl]@xmath361 ( 15 , 5 ) ( 25,0 ) ( 0 , ) /straight-1 ( 18 , 0 ) [ bl]@xmath362 ( 256 , 100 ) ( 1 , 1 ) ( 0 , 15 ) ( 10 , 15 ) ( 0 , ) /straight1 ( 5 , 16 ) [ bl]@xmath359 ( 10 , 15 ) ( 25 , 20 ) ( 0 , ) /straight1 ( 15 , 18 ) [ bl]@xmath360 ( 10 , 15 ) ( 15 , 5 ) ( 0 , ) /scalardash-1 ( 10 , 7 ) [ bl]@xmath363 ( 15 , 5 ) ( 25 , 10 ) ( 0 , ) /straight1 ( 20 , 9 ) [ bl]@xmath361 ( 15 , 5 ) ( 25 , 0 ) ( 0 , ) /straight-1 ( 18 , 0 ) [ bl]@xmath362 + the matrix element for this decay is given by @xmath364 [ { \bar{l}}^{l } i { \lam}_{mln}^{{\ast } } p_{r } l^{n } { \tilde{{\nu}}}^{m{\ast } } ] \| l^{a } ( p_{{l^{a } } } ) { \rangle } \nonumber\\ & = & { \frac{i}{{m_{{\tilde{{\nu}}}^{g}}^{2 } } } } \bigg ( [ { \bar{u } } ( p_{{l^{b } } } ) { \lam}_{gba}^{{\ast } } p_{r } u ( p_{{l^{a } } } ) ] [ { \bar{u } } ( p_{{l^{c } } } ) { \lam}_{gdc } p_{l } v ( p_{{{\bar{l}}^{d } } } ) ] \nonumber\\ & & \hspace{6 mm } - [ { \bar{u } } ( p_{{l^{b } } } ) { \lam}_{gdb } p_{l } v ( p_{{{\bar{l}}^{d } } } ) ] [ { \bar{u } } ( p_{{l^{c } } } ) { \lam}_{gca}^{{\ast } } p_{r } u ( p_{{l^{a } } } ) ] \nonumber\\ & & \hspace{6 mm } - [ { \bar{u } } ( p_{{l^{c } } } ) { \lam}_{gac } p_{l } u ( p_{{l^{a } } } ) ] [ { \bar{u } } ( p_{{l^{b } } } ) { \lam}_{gbd}^{{\ast } } p_{r } v ( p_{{{\bar{l}}^{d } } } ) ] \nonumber\\ & & \hspace{6 mm } + [ { \bar{u } } ( p_{{l^{b } } } ) { \lam}_{gab } p_{l } u ( p_{{l^{a } } } ) ] [ { \bar{u } } ( p_{{l^{c } } } ) { \lam}_{gcd}^{{\ast } } p_{r } v ( p_{{{\bar{l}}^{d } } } ) ] \bigg ) \label{m_a_to_bcd_equation}\end{aligned}\ ] ] [ [ charged - lepton - decays - into - a - charged - lepton - and - a - neutral - vector - meson ] ] charged lepton decays into a charged lepton and a neutral vector meson ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ( 256 , 100 ) ( 1 , 1 ) ( 0 , 15 ) ( 10 , 15 ) ( 0 , ) /straight1 ( 5 , 16 ) [ bl]@xmath365 ( 10 , 15 ) ( 19 , 20 ) ( 0 , ) /straight1 ( 13 , 19 ) [ bl]@xmath366 ( 10 , 15 ) ( 15 , 5 ) ( 0 , ) /scalardash0 ( 10 , 6 ) [ bl]@xmath367 ( 15 , 5 ) ( 19 , 10 ) ( 0 , ) /straight-1 ( 15 , 8 ) [ bl]@xmath368 ( 15 , 5 ) ( 25 , 0 ) ( 0 , ) /straight1 ( 18 , 0 ) [ bl]@xmath369 ( 18 , 20 ) ( 20 , 20 ) ( -0.8,)/straight0 ( 20 , 20 ) ( 20 , 10 ) ( -0.1,)/straight0 ( 20 , 10 ) ( 18 , 10 ) ( -0.8,)/straight0 ( 18 , 10 ) ( 18 , 20 ) ( -0.1,)/straight0 ( 19 , 15 ) ( 25 , 15 ) ( 0 , ) /straight1 ( 22 , 16 ) [ bl]@xmath370 ( 256 , 100 ) ( 1 , 1 ) ( 0 , 15 ) ( 10 , 15 ) ( 0 , ) /straight1 ( 5 , 16 ) [ bl]@xmath365 ( 10 , 15 ) ( 19 , 20 ) ( 0 , ) /straight-1 ( 13 , 19 ) [ bl]@xmath371 ( 10 , 15 ) ( 15 , 5 ) ( 0 , ) /scalardash0 ( 10 , 7 ) [ bl]@xmath372 ( 15 , 5 ) ( 19 , 10 ) ( 0 , ) /straight1 ( 15 , 8 ) [ bl]@xmath373 ( 15 , 5 ) ( 25 , 0 ) ( 0 , ) /straight1 ( 18 , 0 ) [ bl]@xmath369 ( 18 , 20 ) ( 20 , 20 ) ( -0.8,)/straight0 ( 20 , 20 ) ( 20 , 10 ) ( -0.1,)/straight0 ( 20 , 10 ) ( 18 , 10 ) ( -0.8,)/straight0 ( 18 , 10 ) ( 18 , 20 ) ( -0.1,)/straight0 ( 19 , 15 ) ( 25 , 15 ) ( 0 , ) /straight1 ( 22 , 16 ) [ bl]@xmath370 + the matrix element for this process is given by @xmath374 [ -i { { \lam}'}_{ktm}^ { { \ast } } { \bar{l}}_{l}^{k } d_{r}^{m } { \tilde{u}}_{l}^{t{\ast } } ] + [ -i { { \lam}'}_{imj } l_{l}^{i } u_{l}^{m } { \tilde{d}}_{r}^{j } ] [ -i { { \lam}'}_{knt}^{{\ast } } { \bar{u}}_{l}^{n } { \bar{l}}_{l}^{k } { \tilde{d}}_{r}^ { t{\ast } } ] \big ) \| { \mbox{in states } } { \rangle } \label{m_to_v_before_fierz_equation}\end{gathered}\ ] ] after some use of fierz identities ( and dropping the left / right subscripts on the squarks , as only `` left handed '' up type squarks or `` right handed '' down type squarks appear ) we find @xmath375 [ { \bar{l}}^{k } p_{r } { \gamma}_{{\mu } } l^{i } ] \| { \mbox{in states } } { \rangle } \label{m_to_v_after_fierz_equation}\end{aligned}\ ] ] contracting the meson state with the quark bilinear results in @xmath376 [ { \bar{l}}^{k } p_{r } { \gamma}_{{\mu } } l^{i } ] \| { \mbox{leptons in } } { \rangle } f_{v}^{{\ast } } m_{v } { \epsilon}_{v}^{{\mu}{\ast } } \nonumber\\ & = & { \frac{1}{4 } } \left [ { \sum_{{\mbox{\scriptsize d -- type } } } } { { \lam}'}_{ign } { { \lam}'}_{kgm}^{{\ast } } h_{v}^{mn{\ast } } \left ( { \frac{-i}{{m_{{\tilde{u}}^{g}}^{2 } } } } \right ) - { \sum_{{\mbox{\scriptsize u -- type } } } } { { \lam}'}_{img } { { \lam}'}_{kng}^{{\ast } } h_{v}^{mn{\ast } } \left ( { \frac{-i } { { m_{{\tilde{d}}^{g}}^{2 } } } } \right ) \right ] \nonumber \\[2 mm ] & & \times [ { \bar{u } } ( p_{{l^{k } } } ) p_{r } { \gamma}^{{\mu } } u ( p_{{l^{i } } } ) ] f_{v}^{{\ast } } m_{v } { \epsilon}_{v{\mu}}^{{\ast } } \label{m_to_v_final_equation}\end{aligned}\ ] ] where we have introduced the notation @xmath74 to mean only summing over the down type quarks in the meson and @xmath75 to mean summing over the up type quarks . the @xmath71 and @xmath70 in @xmath377 are the generation indices of the appropriate quarks and can be used in standard summation convention with the @xmath71 and @xmath70 appearing in the coupling indices . for example , for the decay @xmath378 , we have @xmath379 [ { \bar{u } } ( p_{e } ) p_{r } { \gamma}^{{\mu } } u ( p_{{\tau } } ) ] f_{{{\rho}^{0}}}^{{\ast } } m_{{{\rho}^{0 } } } { \epsilon}_{{{\rho}^{0}}{\mu}}^{{\ast}}\\ = { \frac{-1}{4 } } \left [ ( { { \lam}'}_{3g1 } { { \lam}'}_{1g1}^{{\ast } } h_{{{\rho}^{0}}}^{dd{\ast } } + { { \lam}'}_{3g1 } { { \lam}'}_{1g2}^{{\ast } } h_{{{\rho}^{0}}}^{ds{\ast } } + { { \lam}'}_{3g1 } { { \lam}'}_{1g3}^{{\ast } } h_{{{\rho}^{0}}}^{db{\ast } } + { { \lam}'}_{3g2 } { { \lam}'}_{1g1}^{{\ast } } h_{{{\rho}^{0}}}^{sd{\ast } } + { \mathellipsis } ) { \frac{i}{{m_{{\tilde{u}}^{g}}^{2 } } } } \right . \\ - \left . ( { { \lam}'}_{31 g } { { \lam}'}_{11g}^{{\ast } } h_{{{\rho}^{0}}}^ { uu{\ast } } + { { \lam}'}_{31 g } { { \lam}'}_{12g}^{{\ast } } h_{{{\rho}^{0}}}^ { uc{\ast } } + { \mathellipsis } ) { \frac{i}{{m_{{\tilde{d}}^{g}}^{2 } } } } \right ] [ { \bar{u } } ( p_{e } ) p_{r } { \gamma}^{{\mu } } u ( p_{{\tau } } ) ] f_{{{\rho}^{0}}}^ { { \ast } } m_{{{\rho}^{0 } } } { \epsilon}_{{{\rho}^{0}}{\mu}}^{{\ast } } \\ = { \frac{-1}{4 } } \left [ ( { { \lam}'}_{3g1 } { { \lam}'}_{1g1}^ { { \ast } } { \frac{-1}{{\sqrt{2 } } } } + { { \lam}'}_{3g1 } { { \lam}'}_{1g2}^ { { \ast } } * 0 + { { \lam}'}_{3g1 } { { \lam}'}_{1g3}^{{\ast } } * 0 + { { \lam } ' } _ { 3g2 } { { \lam}'}_{1g1}^{{\ast } } * 0 + { \mathellipsis } ) { \frac{i}{{m_{{\tilde{u}}^ { g}}^{2 } } } } \right . \\ - \left . ( { { \lam}'}_{31 g } { { \lam}'}_{11g}^{{\ast } } * { \frac{1 } { { \sqrt{2 } } } } + { { \lam}'}_{31 g } { { \lam}'}_{12g}^{{\ast } } * 0 + { \mathellipsis } ) { \frac{i}{{m_{{\tilde{d}}^{g}}^{2 } } } } \right ] [ { \bar{u } } ( p_{e } ) p_{r } { \gamma}^ { { \mu } } u ( p_{{\tau } } ) ] f_{{{\rho}^{0}}}^{{\ast } } m_{{{\rho}^{0 } } } { \epsilon}_{{{\rho}^{0}}{\mu}}^{{\ast } } \label{tau_to_e_rho_example_m_equation}\end{gathered}\ ] ] [ [ charged - lepton - decays - into - a - charged - lepton - and - a - neutral - pseudoscalar - meson ] ] charged lepton decays into a charged lepton and a neutral pseudoscalar meson ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ in addition to the two diagrams above , which can lead to pseudoscalar mesons as well as vector mesons , there is a further diagram for the decay into pseudoscalar mesons that is mediated by a sneutrino : ( 0 , 15 ) ( 10 , 15 ) ( 0 , ) /straight1 ( 5 , 16 ) [ bl]@xmath365 ( 10 , 15 ) ( 25 , 20 ) ( 0 , ) /straight1 ( 15 , 19 ) [ bl]@xmath369 ( 10 , 15 ) ( 15 , 5 ) ( 0 , ) /scalardash0 ( 10 , 7 ) [ bl]@xmath357 ( 15 , 5 ) ( 20 , 10 ) ( 0 , ) /straight1 ( 15 , 8 ) [ bl]@xmath366 ( 15 , 5 ) ( 20 , 0 ) ( 0 , ) /straight-1 ( 13 , 2 ) [ bl]@xmath368 ( 19 , 10 ) ( 21 , 10 ) ( -0.8,)/straight0 ( 21 , 10 ) ( 21 , 0 ) ( -0.1,)/straight0 ( 21 , 0 ) ( 19 , 0 ) ( -0.8,)/straight0 ( 19 , 0 ) ( 19 , 10 ) ( -0.1,)/straight0 ( 20 , 5 ) ( 25 , 5 ) ( 0 , ) /straight1 ( 22 , 6 ) [ bl]@xmath51 the contribution from the squark mediated diagrams is given by @xmath380 [ { \bar{l}}^{k } p_{r } { \gamma}_{{\mu } } l^{i } ] \| { \mbox{in states } } { \rangle } \label{m_to_p_squ_before_contraction_equation}\end{aligned}\ ] ] contracting the meson state with the quark bilinear one finds @xmath381 [ { \bar{l}}^{k } p_{r } { \gamma}_{{\mu } } l^{i } ] \| { \mbox{lepton in states } } { \rangle } f_{p}^{{\ast } } p_{p}^{{\mu } } \nonumber\\ & = & { \frac{1}{4 } } \left [ { \sum_{{\mbox{\scriptsize d -- type } } } } { { \lam}'}_{ign } { { \lam}'}_{kgm}^{{\ast } } h_{p}^{mn{\ast } } \left ( { \frac{-i}{{m_{{\tilde{u}}^{g}}^{2 } } } } \right ) - { \sum_{{\mbox{\scriptsize u -- type } } } } { { \lam}'}_{img}{{\lam}'}_{kng}^{{\ast } } h_{p}^{mn{\ast } } \left ( { \frac{-i}{{m_{{\tilde{d}}^{g}}^{2 } } } } \right ) \right ] \nonumber\\[2 mm ] & & { \times } [ { \bar{u } } ( p_{{l^{k } } } ) p_{r } { \gamma}_{{\mu } } u ( p_{{l^{i } } } ) ] f_{p}^{{\ast } } p_{p}^{{\mu } } \label{m_to_p_squ_final_equation}\end{aligned}\ ] ] the contribution from the sneutrino mediated diagrams is given by @xmath382 \\ { \times } [ { \bar{d}}^{n } i ( { { \lam}'}_{tmn } p_{l } { \tilde{{\nu}}}^{t } + { { \lam}'}_{tnm}^{{\ast } } p_{r } { \tilde{{\nu}}}^{t{\ast } } ) d^{m } ] \| { \mbox{in states } } { \rangle } \\ = { \frac{-i}{2 m_{{\tilde{{\nu}}}^{g}}^{2 } } } { \sum_{{\mbox{\scriptsize d -- type } } } } [ { \bar{u } } ( p_{{l^{k } } } ) { \lam}_{gik } p_{l } u ( p_{{l^{i } } } ) { { \lam}'}_{gnm}^{{\ast } } - { \bar{u } } ( p_{{l^{k } } } ) { \lam}_{gki}^{{\ast } } p_{r } u ( p_{{l^{i } } } ) { { \lam}'}_{gmn } ] \left [ { \frac{{h_{p}^{mn } f_{p } m_{p}^{2}}}{{{\mu}_{p}^{mn } } } } \right]^{{\ast } } \label{m_to_p_snu_equation}\end{gathered}\ ] ] noting that the sneutrino does not couple to up type quarks . s. l. glashow , nucl . * 22 * ( 1961 ) 579 . s. weinberg , phys . * 19 * ( 1967 ) 1264 . j. wess and b. zumino , nucl . b * 70 * ( 1974 ) 39 . n. sakai and t. yanagida , nucl . b * 197 * ( 1982 ) 533 . s. weinberg , phys . d * 26 * ( 1982 ) 287 . m. shiozawa _ et al . _ [ super - kamiokande collaboration ] , phys . * 81 * ( 1998 ) 3319 [ arxiv : hep - ex/9806014 ] . h. k. dreiner , c. luhn and m. thormeier , phys . d * 73 * ( 2006 ) 075007 [ arxiv : hep - ph/0512163 ] . s. dimopoulos , s. raby and f. wilczek , phys . b * 112 * ( 1982 ) 133 . for an introduction , see : s. p. martin , arxiv : hep - ph/9709356 ; i. j. r. aitchison , arxiv : hep - ph/0505105 ; m. drees , arxiv : hep - ph/9611409 . l. e. ibanez and g. g. ross , nucl . b * 368 * ( 1992 ) 3 . h. k. dreiner , c. luhn , h. murayama and m. thormeier , arxiv : hep - ph/0610026 . l. m. krauss and f. wilczek , phys . * 62 * ( 1989 ) 1221 ; l. e. ibanez and g. g. ross , phys . b * 260 * ( 1991 ) 291 . l. j. hall and m. suzuki , nucl . b * 231 * ( 1984 ) 419 . e. nardi , phys . d * 55 * ( 1997 ) 5772 [ arxiv : hep - ph/9610540 ] . b. c. allanach , a. dedes and h. k. dreiner , phys . d * 69 * ( 2004 ) 115002 [ erratum - ibid . d * 72 * ( 2005 ) 079902 ] [ arxiv : hep - ph/0309196 ] . m. hirsch , m. a. diaz , w. porod , j. c. romao and j. w. f. valle , phys . d * 62 * ( 2000 ) 113008 [ erratum - ibid . d * 65 * ( 2002 ) 119901 ] [ arxiv : hep - ph/0004115 ] ; a. abada , g. bhattacharyya and m. losada , phys . d * 66 * ( 2002 ) 071701 [ arxiv : hep - ph/0208009 ] ; e. j. chun , d. w. jung and j. d. park , phys . lett . b * 557 * ( 2003 ) 233 [ arxiv : hep - ph/0211310 ] ; d. e. kaplan and a. e. nelson , jhep * 0001 * ( 2000 ) 033 [ arxiv : hep - ph/9901254 ] ; r. hempfling , nucl . b * 478 * ( 1996 ) 3 [ arxiv : hep - ph/9511288 ] . s. dimopoulos , r. esmailzadeh , l. j. hall and g. d. starkman , phys . d * 41 * ( 1990 ) 2099 ; h. k. dreiner and g. g. ross , nucl . b * 365 * ( 1991 ) 597 . w. m. yao _ et al . _ [ particle data group ] , j. phys . g * 33 * ( 2006 ) 1 . g. bhattacharyya , j. r. ellis and k. sridhar , mod . lett . a * 10 * ( 1995 ) 1583 [ arxiv : hep - ph/9503264 ] . k. agashe and m. graesser , phys . d * 54 * ( 1996 ) 4445 [ arxiv : hep - ph/9510439 ] . d. choudhury and p. roy , phys . b * 378 * ( 1996 ) 153 [ arxiv : hep - ph/9603363 ] . j. h. jang , j. k. kim and j. s. lee , phys . d * 55 * ( 1997 ) 7296 [ arxiv : hep - ph/9701283 ] . j. e. kim , p. ko and d. g. lee , phys . d * 56 * ( 1997 ) 100 [ arxiv : hep - ph/9701381 ] . g. bhattacharyya and a. raychaudhuri , phys . d * 57 * ( 1998 ) 3837 [ arxiv : hep - ph/9712245 ] . o. lebedev , w. loinaz and t. takeuchi , phys . d * 61 * ( 2000 ) 115005 [ arxiv : hep - ph/9910435 ] ; o. lebedev , w. loinaz and t. takeuchi , phys . d * 62 * ( 2000 ) 015003 [ arxiv : hep - ph/9911479 ] ; s. a. abel , a. dedes and h. k. dreiner , jhep * 0005 * ( 2000 ) 013 [ arxiv : hep - ph/9912429 ] . h. k. dreiner , g. polesello and m. thormeier , phys . d * 65 * ( 2002 ) 115006 [ arxiv : hep - ph/0112228 ] . j. p. saha and a. kundu , phys . d * 66 * ( 2002 ) 054021 [ arxiv : hep - ph/0205046 ] . j. h. jang , y. g. kim and j. s. lee , phys . d * 58 * ( 1998 ) 035006 [ arxiv : hep - ph/9711504 ] . h. dreiner , g. polesello and m. thormeier , arxiv : hep - ph/0207160 . y. g. xu , r. m. wang and y. d. yang , arxiv : hep - ph/0610338 . v. d. barger , g. f. giudice and t. han , phys . rev . d * 40 * ( 1989 ) 2987 . s. davidson , d. c. bailey and b. a. campbell , z. phys . c * 61 * ( 1994 ) 613 [ arxiv : hep - ph/9309310 ] . h. k. dreiner , arxiv : hep - ph/9707435 . g. bhattacharyya , arxiv : hep - ph/9709395 . b. c. allanach , a. dedes and h. k. dreiner , phys . d * 60 * ( 1999 ) 075014 [ arxiv : hep - ph/9906209 ] . m. herz , diploma thesis , arxiv : hep - ph/0301079 . m. chemtob , prog . nucl . phys . * 54 * ( 2005 ) 71 [ arxiv : hep - ph/0406029 ] . r. barbier _ et al . _ , phys . * 420 * ( 2005 ) 1 [ arxiv : hep - ph/0406039 ] . our effective lagrangian is to be compared with those in @xcite , @xcite , @xcite , @xcite and @xcite ( noting that both @xcite and @xcite use the convention where there is _ no _ factor of @xmath383 before the @xmath113 terms ) . we disagree with the form of the effective lagrangians in @xcite and @xcite , and agree with those in @xcite , @xcite and @xcite . we note , however , that through projection onto vector or pseudoscalar quark bilinears the difference with @xcite reduces to a simple overall sign error of the matrix element , which is then eliminated by squaring . accordingly , this has no ill effects in the quadratic coupling dominance convention . also , in the case of @xcite we note that it is merely that the wrong coupling in the second term of their equation ( 7 ) has the @xmath384 denoting complex conjugation . m. neubert and b. stech , adv . high energy phys . * 15 * ( 1998 ) 294 [ arxiv : hep - ph/9705292 ] . in all these cases , it turns out we are agreeing with @xcite , though @xcite only gives the bounds to the first significant figure , and we assume that the differences ( @xmath385 compared to @xmath386 and @xmath387 compared to @xmath388 ) arise from rounding errors . we do not agree with the bounds presented in @xcite , which are just those taken from @xcite . we believe that @xcite presented incorrect bounds , and that this arises from their equation ( 13 ) for the decay width @xmath389 . we find that the ratio of the decay width ( 13 ) in @xcite over our decay width ( [ gamma_p_to_snu_equation ] ) is @xmath390 . while @xmath391 for heavy mesons , we believe that equation ( 13 ) in @xcite misses a factor @xmath392 which results in too tight bounds . however , as the experimental bounds for many of the rare @xmath3 decay branching ratios have improved , we still obtain bounds for the couplings associated with these decays tighter than in @xcite . note that we agree with the corresponding equation ( 8) in @xcite ( taking into account that @xcite defines the couplings @xmath113 such that the superpotential does _ not _ have the factor of @xmath383 before the trilinear lepton term ) and with the generic result in @xcite . @xcite updating bounds given in @xcite with data from sindrum ii collaboration , talk given at 14th international conference on particles and nuclei ( panic ) , williamsburg , virginia , usa , 22 - 28 may 1996 . presented by p. wintz . published by world scientific publishing co. pte . ltd . 1997 , ed . carlson and j.j . domingo , 458 .
in a small metallic island where the charging energy @xmath5 exceeds the temperature @xmath6 ( we use the unit @xmath7 ) , the coulomb interaction affects transport properties through the island . the resulting phenomenon is called the coulomb blockade ( cb ) and such system is named the single - electron transistor ( set ) . cb has been attracted much attention in the last decade @xcite and the nature of the transport properties of set has been clarified . set is interesting because it is regarded as one of the most simple examples of strongly correlated system which can be brought into the non - equilibrium state by applied bias voltage . early investigations consider the case where the tunneling conductance is so small that the higher order quantum fluctuation of island charge is negligible . recently , the quantum fluctuation in set has attracted much attention as one of the basic problems in this field . the quantum fluctuation is quantitatively characterized by the dimensionless parallel conductance : @xmath8 where @xmath9 is the quantum resistance and @xmath10 is the parallel tunneling resistance of the source and the drain junctions . there has been much development on theoretical investigation in the whole range of @xmath11 . especially in the _ weak tunneling regime _ ( @xmath12 ) , the life - time broadening of a charge state level is much smaller than the typical level spacing of charge states , and thus the effective two - state model , which is equivalent to the multichannel anisotropic kondo model in the equilibrium state@xcite , well describes the low - energy physics . with this model , it is predicted that the quantum fluctuation of charge causes the renormalization of the conductance and the charging energy below the kondo temperature " @xmath13 @xcite . the renormalization of the conductance is confirmed experimentally as @xmath14 dependence of the conductance peak at low temperature@xcite . it is also predicted that in the non - equilibrium state , the dissipative charge fluctuation causes the life - time broadening of a charge state level and smears the structures of @xmath15-@xmath16 characteristic@xcite . though investigations have revealed much about the quantum fluctuation , most of them have been limited to averaged quantities . in order to understand the nature of the quantum fluctuation , investigations on the higher - order correlation function of fluctuation operators are required . a good starting point may be the investigation on the second moment of fluctuation operators , i.e. , the noise@xcite . the charge noise and the current noise in the weak tunneling regime is also important for practical applications , because it determines the performance of set electrometers @xcite . the current noise is defined by the auto - correlation function of the current fluctuation operator @xmath17 as @xmath18 where @xmath19 means the statistical average . until recently , investigations on the noise have been done using the framework of the orthodox `` theory@xcite , which takes account of the lowest order quantum fluctuation , namely the sequential tunneling ( st ) process . recently , several authors@xcite discussed the higher order quantum fluctuation in cb regime within the co - tunneling theory''@xcite . however , there is no approximation covering both st and cb regime . the aim of the present work is to construct a theoretical framework , which covers both of these regimes for arbitrary @xmath11 and clarify how the quantum fluctuation affects the noise . the keldysh formalism@xcite has been one of the most powerful methods to study the non - equilibrium properties of mesoscopic systems . however to apply this method to set in the two - state limit , one must overcome a technical difficulty : the spin-1/2 operator , which is introduced to restrict charge number states by the strong coulomb interaction , prevents one from utilizing wick s theorem . the most successful treatment to overcome this problem is given in ref . @xcite , in which a formulation of perturbative expansion for the reduced density matrix in the real time domain is developed and the inelastic resonant tunneling processes is treated . the method of ref . @xcite enables one to classify various tunneling processes using diagrammatic techniques , and can be also applied to other systems with the strong local correlation , such as quantum dot@xcite . in spite of these successes , it seems to be still difficult to apply this method for the calculation of higher order correlation functions , since this method requires to solve a special integro - differential equation even for the calculation of the average in the presence of large quantum fluctuation@xcite . in this paper , we investigate the current noise and the charge noise in the regime of large quantum fluctuation of charge out of equilibrium . we adopt the schwinger - keldysh approach and the drone - fermion representation of the effective spin-1/2 operator@xcite . schwinger - keldysh approach enables us to calculate any order moment systematically by the functional derivative technique@xcite satisfying the charge conservation@xcite , and it helps us in manipulating many complicated terms . the drone - fermion representation allows us to utilize the fermionic wick s theorem and to take effects of the strong correlation into account . with the help of this technique , we can extensively take account of the higher order processes of tunneling . we will show that our approximation reproduces the resonant tunneling approximation ( rta)@xcite as for the average current and the average charge . the outline of this paper is as follows . in sec . [ sec : keldysh ] , we briefly summarize the keldysh formalism and introduce an approximate generating functional . we also show that the average and the noise expressions can be derived using the functional derivation . in sec . [ sec : main ] , we actually calculate the average current , the average charge , the current noise and the charge noise . in sec . [ sec : results ] we show numerical results for the noise and give some discussions on the non - equilibrium fluctuation , the thermal fluctuation and the renormalization effect . section [ sec : summary ] summarizes our results . in this section , we give preliminary definitions of the schwinger - keldysh approach and we summarize three useful representations : the _ closed time - path _ , the _ single time _ and the _ physical _ representations ( sec . 2 of ref . @xcite ) . for simplicity , we consider the following action of a free fermion with a linear source term to explain the basic formalism : @xmath20 where the closed time path @xmath21 consists of the forward branch @xmath22 , the backward branch @xmath23 and the imaginary time - path @xmath24 as shown in fig . [ fig : closedtimepath]@xcite . @xmath25 is a grassmann variable satisfying the anti - periodic boundary condition @xmath26 . the complex variable @xmath27 is defined only on the forward and backward branch @xmath28 . the generating functional for the connected closed time - path green function ( gf ) is defined as , @xmath29 \exp \left ( -s/{i}\hbar \right ) . \nonumber\ ] ] gf in the closed time - path representation is obtained by the second derivative of the generating functional with respect to @xmath27 ( @xmath30 ) : @xmath31 hereafter , we use arguments @xmath32 instead of @xmath33 for short . though the closed time - path representation makes the formulation compact , in order to obtain the physical quantities , we sometimes need the single time representation in which the time on @xmath21 is projected onto the real axis . in this representation the degrees of freedoms of fields are doubled which we denote as , @xmath34 etc . here @xmath35 is defined on @xmath36 and @xmath37 is the real time . in the same way as @xmath38 , gf is transformed into @xmath39 matrix in the keldysh space : @xmath40 here , arguments @xmath41 and @xmath42 are the real time and each component is defined with the statistical average in the path integral representation @xmath43 a \exp ( -s/{i}\hbar)/z \right\|_{j=0}$ ] as @xmath44 . diagonal components @xmath45 and @xmath46 are the causal and the anti - causal gf , respectively . off - diagonal components are correlation functions , which are written in the operator representation as @xmath47 and @xmath48 @xcite . here the statistical average is defined as @xmath49/z_0 $ ] where the hamiltonian operator @xmath50 is @xmath51 and the partition function @xmath52 is given as @xmath53 $ ] . it is known that four components of eq . ( [ eqn : ctpgfsingle ] ) are not independent . this redundancy is removed by the keldysh rotation@xcite . after the rotation , we obtain the so - called physical representation : @xmath54 gfs denoted by superscripts , @xmath55 , @xmath56 and @xmath57 are advanced , retarded and keldysh components , respectively . in the practical calculations , instead of @xmath58 and @xmath59 , the center - of - mass coordinate @xmath60 and the relative coordinate @xmath61 , @xmath62 are used in most cases . there are the following relations between components of gf : @xmath63 \rangle = g^c(1,2 ) . \label{eqn : comute}\end{aligned}\ ] ] here we introduce a notation @xmath64 whose physical meaning is the spectral density in the energy space . equation ( [ eqn : anti - comute ] ) is derived from the normalization of the step function ( see eq . ( 2.67 ) in ref . @xcite ) . the normalization condition of the density matrix results in an important equation @xmath65 ( see sec . 2.4 of ref . @xcite)@xcite which is equivalent to the following equations : @xmath66 in this section , we introduce our model hamiltonian and derive the generating functional for set , based on which we construct the perturbation theory . figure [ fig : system ] shows the equivalent circuit of a set . a metallic island exchanges electrons with the left ( right ) lead via a tunnel junction characterized by the tunneling matrix element @xmath67 . the island is coupled to leads and a gate via capacitors @xmath68 , @xmath69 and @xmath70 . we consider in the weak tunneling regime @xmath71 and use the two - state model . in this paper , we limit ourselves to the symmetric case : @xmath72 and @xmath73 . the total hamiltonian consists of the unperturbed part @xmath50 and the tunneling part @xmath74 . omitting a trivial constant , the unperturbed part is given as @xmath75 where @xmath76 is the annihilation operator of an electron with wave vector @xmath77 in the left ( right ) lead ( @xmath78 ) or in the island ( @xmath79 ) . the subscript @xmath80 numbers the transverse channels including spin degree of freedom . the density of states is considered as constant in each region : @xmath81 ( @xmath82 ) . the second term is the charging energy and the effective spin-1/2 operator @xmath83 acts on the lowest two charge states . the energy difference between two charge states is given by @xmath84 where @xmath85 is the gate charge . the tunneling hamiltonian @xmath86 describes the electron tunneling across the junctions and simultaneous change of the charge state of the island . @xmath87 is the phase difference between the left and the right leads and parameters @xmath88 characterizes the voltage drop between the left ( right ) lead and the island . the tunneling hamiltonian is adiabatically turned on in the remote past and off in the distant future . it is the widely adopted procedure , which ensures the time translational invariance to describe a stationary state . in order to utilize wick s theorem for fermions , we employ the mapping of the effective spin-1/2 operator onto two fermion operators @xmath89 and @xmath90 @xcite : @xmath91 , where @xmath92 is a majorana fermion operator ( @xmath93 ) . this representation is called drone - fermion representation @xcite , because @xmath94 is a drone " whose only job is to make spin-1/2 operators of different spins commute , rather than anti - commute@xcite . employing the hamiltonian operator in the drone - fermion representation and following the standard manner to introduce a path integral @xcite , we obtain the generating functional in the path integral representation : @xmath95 \exp \left ( -\frac{s}{{i}\hbar } \right ) , \label{eqn : z}\end{aligned}\ ] ] where all field variables are grassmann variables satisfying the anti - periodic boundary condition . the action is given by @xmath96 where , @xmath97 . ( [ eqn : action0 ] ) we introduced auxiliary source fields , @xmath98 and @xmath99 , in order to calculate the average and the noise by the functional derivation . it is noticed that the degrees of freedoms are doubled , as shown in eq . ( [ eqn : fieldsingletime ] ) . after the derivation , these variables are put as @xmath100 and @xmath101 to be related with the parameters of the actual system . by introducing a linear source term @xmath102 , where @xmath38 is a grassmann variable , all fields can be traced out@xcite . in the limit of large transverse - channel number , @xmath103 is expressed as@xcite @xmath104 \right ) \nonumber \\ & \times & \left . \exp \left ( -\frac{1}{2 { i}\hbar } \int_c { { \rm d}}1 { { \rm d}}2 \ , j(1 ) g_{\phi}(1,2 ) j(2 ) \right ) \right\|_{j=0 } \nonumber \\ & \times & 2 { \rm e}^{{\rm tr } \left [ \ln g_c^{-1 } \right ] } , \label{eqn : startz}\end{aligned}\ ] ] where we omitted the partition function of noninteracting electrons . the trace @xmath105 and the products represent the integration along @xmath21 as follows , @xmath106 = \int_c { { \rm d}}1 { { \rm d}}2 \ , g_c(1,2 ) j(2 ) g_{\phi}(2,1 ) . \nonumber\ ] ] the particle - hole gf @xmath107 , in the closed time - path form is written as @xmath108 where @xmath109 is the number of the transverse channels . the gf for free electron in the lead @xmath110 and the island @xmath111 , @xmath112 , is given as @xmath113 which satisfies the anti - periodic boundary condition : @xmath114 . the @xmath115-field and the @xmath116-field gfs defined by @xmath117 also satisfy the anti - periodic boundary condition . ( [ eqn : startz ] ) , we construct a systematic perturbation expansion in terms of @xmath118 . for example , we show the diagrammatic representation of the zeroth ( @xmath119 ) , the first ( @xmath120 ) and the second order contribution ( @xmath121 ) to the generating functional @xmath122 in fig . [ fig : diagram1 ] . here , solid lines , dotted lines and wavy lines represent gfs for @xmath115-field , @xmath116-field and particle - hole , respectively . practical forms are given as follows : @xmath123 \label{eqn : w0 } , \\ w^{(1 ) } & = & { i}\hbar \ , { \rm tr } \left [ g_c \sigma_c \right ] \label{eqn : w1 } , \\ w^{(2)}_{\text{$c$-field } } & = & \frac{{i}\hbar}{2 } { \rm tr } \left[(g_c\sigma_c)^2 \right ] \label{eqn : w2c},\end{aligned}\ ] ] where @xmath124 is the term corresponding to the first diagram in fig . [ fig : diagram1 ] ( c ) , which we call the _ @xmath115-field correction_. the other four diagrams can be written in the same way . here , the self - energy of the @xmath115-field @xmath125 is defined as @xmath126 in a previous paper@xcite , we have shown that the first order contribution causes the divergence at the degeneracy point @xmath127 for average charge . in order to regularize the divergence , we proposed an approximate generating functional obtained by summing up @xmath115-field corrections @xmath128 to infinite order : @xmath129 \label{eqn : wrta } \\ & = & - { i}\hbar \left ( { \rm tr}\left[g_c^{-1}\right ] - \sum_{n=1}^{\infty } \frac{1}{n } { \rm tr } \left[(g_c \sigma_c)^n \right ] \right ) . \nonumber\end{aligned}\ ] ] figure [ fig : diagram2 ] ( a ) shows the diagrammatic representation of @xmath130 . the circle represents the self - energy of @xmath115-field and the thick line represents the full @xmath115-field gf defined by the dyson equation @xmath131 whose diagrammatic representation is shown in fig . [ fig : diagram2 ] ( b ) . in sec . [ sec : reformulation ] we will show that @xmath130 reproduces results of rta for the normal metal island . in sec . [ sec : mainbody ] , we calculate the charge noise and the current noise based on @xmath130 . in this section , we summarize expressions for the average and the noise on the basis of functional derivative . relations between physical quantities in the generating functional representation and those in the operator representation are demonstrated in appendix [ appendix : expressions ] . we also show that the gauge invariance of generating functional leads to the charge conservation law . the exact average current expression is given by the functional derivative of exact generating functional @xmath122 with respect to the phase difference@xcite : @xmath132 the center - of - mass coordinate of the phase difference is determined by the josephson relation@xcite . the relative coordinates are put to zero because they are the fictitious variables . from now on , we suppress the equations in the subscript after the vertical bar for short . the average charge is calculated by the functional derivation in terms of the scalar potential of @xmath115-field as @xmath133 the current noise and the charge noise defined as eq . ( [ eqn : siidef ] ) are given by the second derivative with respect to @xmath134 and @xmath135 @xcite : @xmath136 where @xmath137 is obtained from the first term by replacing subscripts @xmath138 with @xmath115 . it is convenient to rewrite eq . ( [ eqn : i - exact ] ) in the form of the linear combination of the tunneling current at the left junction @xmath139 and that at the right junction @xmath140 as @xmath141 . in the same way , ( [ eqn : sii ] ) is written in terms of the correlation function of @xmath142 and @xmath143 ( @xmath144 ) , which we denote by @xmath145 , as @xmath146 here , @xmath147 and @xmath145 are written as @xmath148 by regarding @xmath149 and @xmath150 as formally independent variables . the generating functional eq . ( [ eqn : z ] ) is invariant under the gauge transformation , i.e. , the phase transformation of the @xmath115-field and the change of the @xmath115-field scalar potential : @xmath151 where @xmath152 is defined on @xmath28 . the relation between the gauge invariance and the charge conservation in the non - equilibrium state has been analyzed in ref . for our system , the following expressions of the current continuity and the charge conservation for correlation functions @xmath153 can be proved ( appendix . [ appendix : conservationlaw ] ) . in this section , we derive approximate current noise and charge noise expressions , which is main purpose of this paper . we summarize gfs in the physical representation in sec . [ sec : freegreenfunctions ] which are needed for practical calculations . in sec . [ sec : reformulation ] we calculate the average charge and the average current and show that our approximation completely reproduces the results of rta . we also give some notes on the diagrammatic rule suitable for the functional derivative technique . in sec . [ sec : mainbody ] , we calculate the current noise and the charge noise based on the diagrammatic rule . we will check that our results satisfy the charge conservation law and the fluctuation - dissipation theorem . we summarize the fourier transformation of the retarded and the keldysh component of gfs . hereafter in this subsection , we put the auxiliary source fields as shown in the subscript of eq . ( [ eqn : i - exact ] ) . the solutions of differential equations ( [ eqn : gfc ] ) and ( [ eqn : gfd ] ) imposing anti - periodic boundary condition are@xcite @xmath154 @xmath155 where @xmath156 is a positive infinitesimal number and the @xmath157-function in the energy space is defined as @xmath158 the advanced component is the complex conjugate of the retarded component : @xmath159 . the two components of particle - hole gf are given by , @xmath160 where @xmath161 ( appendix . [ appendix : loop ] ) . @xmath162 is the dimensionless conductance for tunnel junction @xmath163 written in terms of the tunnel resistance @xmath164 as @xmath165 . the spectral density of the particle - hole propagator is given by @xmath166 where the lorentzian cut - off function is introduced@xcite . the keldysh component and the component defined by eq . ( [ eqn : comute ] ) for the @xmath115-field self - energy eq . ( [ eqn : partselfenergy ] ) and those for the particle - hole gf are related to each other : @xmath167 the retarded component of the self - energy is given as @xmath168 where @xmath169 is the digamma function . the full @xmath115-field gf is obtained by solving the dyson equation in the closed time - path representation eq . ( [ eqn : fullgc])@xcite : @xmath170 where we used the definition of @xmath157-function eq . ( [ eqn : deltae ] ) . here we remark the following : equation ( [ eqn : cg ] ) shows that in the limit of @xmath171 , the charge states are independent of the initial equilibrium distribution , because the keldysh component of @xmath115-field gf represents the distribution of charge states . this fact suggests that @xmath130 describes a physically reasonable non - equilibrium stationary state , which should not depend on any initial state . we transform above expressions into single - time representation . employing eqs . ( [ eqn : anti - comute ] ) and ( [ eqn : comute ] ) we obtain @xmath172 here @xmath173 and @xmath174 are written with the fermi function @xmath175 and the bose distribution function @xmath176 . functions @xmath177 and @xmath178 are given by @xmath179 and @xmath180 , respectively . the reason why we adopt the generating functional approach is that once an approximate generating functional is obtained , one can calculate any order moment systematically by the functional derivation . in the following sections we perform practical calculations , employing @xmath130 introduced in sec . [ subsec : w ] . detailed discussions on our approximation are retained in appendix [ appendix : perturbartion ] . the average current is calculated by substituting @xmath130 into eq . ( [ eqn : ir])@xcite : @xmath181 \right\| \label{eqn : trace1}.\end{aligned}\ ] ] here , we used the fact that the self - energy eq . ( [ eqn : partselfenergy ] ) includes the phase factor through the particle - hole gf eq . ( [ eqn : particlehole ] ) . in the language of feynmann diagrams , eq . ( [ eqn : trace1 ] ) can be rewritten in a compact form : @xmath182{diagram11a.eps } \end{array } - \begin{array}{c } \includegraphics[width=7ex]{diagram11b.eps } \end{array } \right ] \right\| \label{eqn : i1}.\end{aligned}\ ] ] solid dots with @xmath37 represents @xmath183 . the circle with @xmath163 is the partial self - energy defined by eq . ( [ eqn : partselfenergy ] ) . here we obtain the diagrammatic rule similar to that of refs . @xcite . ( i ) : : the diagrams corresponding to the functional derivative with respect to @xmath184 is obtained by a series of the following operations ; to put a solid dot onto all possible positions of vertices in a closed diagram , to assign @xmath163 on a circle connected to the solid dot , to multiple @xmath185 , and to assign minus sign if a solid line comes into the solid dot . next we project the fictitious time on @xmath21 to the real axis . as the tunneling hamiltonian is zero on @xmath24 , @xmath186 is zero for @xmath187 or @xmath188 . hence , the diagrams in eq . ( [ eqn : i1 ] ) are rewritten as , @xmath189 \right\| , \label{eqn : trace2}\end{aligned}\ ] ] where we performed the keldysh rotation . @xmath190 is the pauli matrix in the keldysh space @xcite . the trace is carried out over the keldysh space and products represents the integration along the real - time as @xmath191 = \int_{-\infty}^{\infty } { { \rm d}}1 { { \rm d}}2 \ , { \rm tr } \left [ \tilde{g}_{\phi}(1,2 ) \tilde{\varphi}(2 ) \tilde{g}_{\phi}(2,1 ) \right ] . \nonumber\ ] ] the phase in the physical representation @xmath192 is written as @xmath193 which leads to a useful relation for the functional derivative technique : @xmath194 by using the property of the gf in the physical representation @xmath195 , and that of a pauli matrix @xmath196 , we can see that the second term of eq . ( [ eqn : trace2 ] ) is minus the complex conjugate of the first term . to generalize this property , we obtain a helpful rule to reduce the number of diagrams : a diagram which is complex conjugate of a certain diagram is obtained by changing the direction of all lines and putting minus sign if the diagram includes odd numbers of vertices without solid dot . by performing the fourier transformation , we rewrite eq . ( [ eqn : i1 ] ) as : @xmath197{diagram11a.eps } \end{array } \right\| = 2 e{\rm re } \int \frac{{{\rm d}}\varepsilon}{h } { \rm tr } \left [ \tilde{g}_c(\varepsilon ) \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{\sigma}_{\rm r}(\varepsilon ) { \mbox{\boldmath$\tau$}}^1 \right ] \nonumber \\ \label{eqn : diagram - trace } \\ & = & e \int \frac{{{\rm d}}\varepsilon}{h } \frac { \sigma_{\rm r}^c(\varepsilon ) g_c^k(\varepsilon ) } { 2 } - ( c \leftrightarrow k ) . \label{eqn : iphysical}\end{aligned}\ ] ] the second term @xmath198 is obtained from the first term by swapping superscripts @xmath57 and @xmath21 . by generalizing eq . ( [ eqn : diagram - trace ] ) , we obtain the rule to calculate the diagram : ( ii ) : : put @xmath39 matrix gfs in the physical representation to the corresponding lines and circles , put @xmath199 or @xmath200 to a vertex with or without solid dot . subsequently , carry out the trace over keldysh space and the integration over the frequency @xmath201 . employing eqs . ( [ eqn : anti - comute ] ) and ( [ eqn : comute ] ) , eq . ( [ eqn : iphysical ] ) is transformed into the single time representation : @xmath202 by using eqs . ( [ eqn : correlationalpha ] ) , ( [ eqn : correlationself ] ) and ( [ eqn : correlatonfullgc1 ] ) , and noting @xmath203 , we obtain the final form of the current expression which has the same form as the landauer formula : @xmath204 where , @xmath205 is the effective transmission probability of lead electrons thorough the island : @xmath206 from the second form , we can see our result is equivalent to that of rta@xcite . in the same way as the average current , the average charge is evaluated by substitute @xmath130 into eq . ( [ eqn : q - exact ] ) @xcite , @xmath207{diagram12.eps } \end{array } \right\| , \nonumber\end{aligned}\ ] ] where the solid dot with @xmath37 corresponds to @xmath208 whose practical expression is equal to the right - hand side of eq . ( [ eqn : deltaphysical1r ] ) in the physical representation . the diagrammatic rule of functional derivation in terms of the scalar potential is given as follows : ( i ) : : the diagrams corresponding to the functional derivative with respect to @xmath98 is obtained by inserting a solid dot into all possible positions of @xmath115-field gf . following the rule * ( ii ) * , the diagram for average charge can be also calculated : @xmath209 \nonumber \\ & = & \frac{1}{2}+\int \frac{{{\rm d}}\varepsilon}{\pi } \frac{\alpha^r(\varepsilon)}{\alpha^k(\varepsilon ) } { \rm i m } g_c^r ( \varepsilon ) \label{eqn : qrta}.\end{aligned}\ ] ] the imaginary part of @xmath210 , which has a peak at @xmath211 , describes the excitation property of the charge state . when the broadening of the peak is sufficiently small and @xmath212 , @xmath210 is approximately given by @xmath213 here @xmath214 is the renormalization factor @xmath215 . the low energy cut - off @xmath216 is @xmath217 where one of three parameters must be much larger than the other two@xcite . the imaginary part of self - energy @xmath218 represents the life - time broadening effect . when the renormalization effect is negligible @xmath219 , it is written as @xmath220 , where @xmath221 and @xmath222 @xmath223 ( @xmath224 ) is equal to the tunneling rate into ( out of ) the island through the junction r estimated by fermi s golden rule . it is noticed that the condition @xmath225 is enough to neglect the broadening of the peak . by using the approximation eq . ( [ eqn : approxg ] ) , eq . ( [ eqn : qrta ] ) for equilibrium state reproduces the result of rta@xcite , @xmath226 . according to ref . @xcite , eq . ( [ eqn : irta ] ) is consistent with the co - tunneling theory and orthodox theory for two - state limit@xcite . we can also confirm that eq . ( [ eqn : qrta ] ) is consistent with the orthodox theory in the following way : in the limit of @xmath227 with keeping @xmath228 or @xmath6 finite , the renormalization factor approaches unity and the imaginary part of eq . ( [ eqn : approxg ] ) reduces to the @xmath157-function @xmath229 . thus , eq . ( [ eqn : qrta ] ) reproduces the result of the orthodox theory @xmath230 where @xmath231 . here , we calculate the current noise and the charge noise based on the reformulated rta . in the following discussions we limit ourselves to zero frequency component @xmath232 where we used a fact that at a stationary state , the time translational invariance is satisfied and the correlation function depends only on the difference of @xmath37 and @xmath233 . from the definition eq . ( [ eqn : sii ] ) , one obtain the current noise by applying the rule * ( i ) * twice . reducing the number of diagrams and using the rule * ( ii ) * , which can be also applied to the calculation of zero frequency noise diagram ( appendix . [ appendix : noiserule ] ) , we obtain @xmath234{diagram13a.eps } \end{array } \delta_{\rm r , r ' } + \begin{array}{c } \includegraphics[width=8ex]{diagram13b.eps } \end{array } - \begin{array}{c } \includegraphics[width=8ex]{diagram13c.eps } \end{array } \right ] \right\| - \frac{(\delta \rightarrow c)}{4 } \nonumber \\ & = & 2 \int \frac{{{\rm d}}\varepsilon}{h } { \rm tr } \left [ \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{\sigma}_{\rm r}(\varepsilon ) \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{g}_c(\varepsilon ) \ , \delta_{\rm r , r ' } + \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{\sigma}_{\rm r}(\varepsilon ) { \mbox{\boldmath$\tau$}}^1 \tilde{g}_c(\varepsilon ) { \mbox{\boldmath$\tau$}}^1 \tilde{\sigma}_{\rm r'}(\varepsilon ) \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{g}_c(\varepsilon ) - \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{\sigma}_{\rm r}(\varepsilon ) { \mbox{\boldmath$\tau$}}^1 \tilde{g}_c(\varepsilon ) \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{\sigma}_{\rm r'}(\varepsilon ) { \mbox{\boldmath$\tau$}}^1 \tilde{g}_c(\varepsilon ) \right ] \nonumber \\ & - & ( { \mbox{\boldmath$\tau$}}^0 \rightarrow { \mbox{\boldmath$\tau$}}^1 ) , \nonumber\end{aligned}\ ] ] where we used a fact that the second term of eq . ( [ eqn : sii ] ) is obtained from the first term by changing @xmath235 to @xmath200 in the physical representation ( appendix . [ appendix : noiserule ] ) . after some straightforward calculations , we obtain the following expression : @xmath236 \label{eqn : siirta } \\ & = & \frac{2}{r_{\rm k } } \int { { \rm d}}\varepsilon [ t^f(\varepsilon ) \ { f^{-}_{\rm l}(\varepsilon ) f^{+}_{\rm r}(\varepsilon ) + f^{+}_{\rm l}(\varepsilon ) f^{-}_{\rm r}(\varepsilon ) \ } - t^f(\varepsilon)^2 \ { f^{-}_{\rm l}(\varepsilon ) - f^{-}_{\rm r}(\varepsilon ) \}^2 ] \label{eqn : siirta_2},\end{aligned}\ ] ] where @xmath237 is the other side of @xmath163 ( for example , when the index @xmath238 is @xmath239 the index of the other side @xmath240 is @xmath241 ) . the charge noise is evaluated by adopting the definition eq . ( [ eqn : sqq ] ) and by applying rules * ( i ) * and * ( ii ) * : @xmath242{diagram14.eps } \end{array } \right\| -\frac{(\delta \leftrightarrow c)}{4 } = e^2 \hbar^2 \int \frac{{{\rm d}}\varepsilon}{h } { \rm tr } \left [ \tilde{g}_c(\varepsilon ) \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{g}_c(\varepsilon ) \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \right ] - ( { \mbox{\boldmath$\tau$}}^0 \rightarrow { \mbox{\boldmath$\tau$}}^1 ) \nonumber \\ & = & e^2 \hbar^2 \int \frac{{{\rm d}}\varepsilon}{h } g_c^{-+}(\varepsilon ) g_c^{+-}(\varepsilon ) = \sum_{\rm r , r'=l , r } \frac{e^4 r_{\rm k}}{2 \pi^2 } \int { { \rm d}}\varepsilon \left\| g_c^r(\varepsilon ) \right\|^4 \alpha^k_{\rm r}(\varepsilon ) \alpha^k_{\rm r'}(\varepsilon ) f^{+}_{\rm r}(\varepsilon ) f^{-}_{\rm r'}(\varepsilon ) , \label{eqn : sqqrta}\end{aligned}\ ] ] where we used eqs . ( [ eqn : correlationalpha ] ) and ( [ eqn : correlatonfullgc1 ] ) . equation ( [ eqn : siirta_2 ] ) has the same form as the current noise expression of a point contact without coulomb interaction@xcite . this result is anticipated , because the tunneling current is expressed in the same form as the landauer formula . however there is an important difference as mentioned in ref . @xcite : the effective transmission probability includes the coulomb correlation and the inelastic relaxation effect . here we note the following points . first , our approximation satisfies the charge conservation law : as the approximate generating functional @xmath130 is invariant under the gauge transformation eq . ( [ eqn : gauge_t ] ) , we can show the conservation law , eqs . ( [ eqn : iconserve ] ) and ( [ eqn : sconserve ] ) , for the approximate expressions . especially for zero frequency component , they reduce to equations for the current conservation law , @xmath243 and @xmath244 . it is worth noticing that the gauge invariance is automatically satisfied when @xmath130 consists of closed diagrams . secondly , we can show that our result satisfies the fluctuation - dissipation theorem : at @xmath245 , the current noise expression is reduced to the johnson - nyquist formula @xmath246 where @xmath247 is the conductance expressed as @xmath248 . we discuss the current noise expression in the regime where @xmath225 and the renormalization effect is negligible , @xmath219 . using eqs . ( [ eqn : transmission ] ) and ( [ eqn : approxg ] ) , the first term of eq . ( [ eqn : siirta ] ) is approximately given by @xmath249 where @xmath250 for @xmath251 , @xmath252 is the co - tunneling current from lead r/@xmath253 to lead @xmath253/r , and eq . ( [ eqn : siicotwidth ] ) reproduces the co - tunneling theory @xcite in the two - state limit . equation ( [ eqn : siicotwidth ] ) , which we call the _ co - tunneling theory with life - time broadening _ , is equivalent to the previously proposed equation in ref . @xcite in the limit of @xmath254 . the validity of this approximation is discussed in the next section from the point of view of the numerical results . by taking the limit @xmath227 , with paying attention to the relation @xmath255 \}^2 = \delta(\varepsilon-\delta_0 ) $ ] , we obtain the result of the orthodox theory in the two - state limit : @xmath256 = \alpha_0^{-1 } 2 e \left ( i_{+}- \frac{2 i_{-}^2}{e \gamma } \right ) \label{eqn : siiorth},\ ] ] where @xmath257 . @xmath258 is the average current of the orthodox theory . the second term is related to the part of eq . ( [ eqn : siirta ] ) proportional to @xmath259 , which represents the coulomb correlation and reduces the current noise from the poissonian value . it is noticed that even at small tunneling conductance , the second term is important in st regime . in the same way as above discussions , for @xmath260 and @xmath219 , we obtain the co - tunneling theory@xcite with life - time broadening from eq . ( [ eqn : sqqrta ] ) : @xmath261 in the limit of @xmath227 , we can confirm that our result reproduces the orthodox theory@xcite : @xmath262 = \alpha_0 4 e^2 \gamma_{+ } \gamma_{-}/ \gamma^3 \label{eqn : sqqorth}.\ ] ] in this section we present results obtained by the numerical calculation of eqs . ( [ eqn : siirta ] ) and ( [ eqn : sqqrta])@xcite , based on which we discuss the non - equilibrium fluctuation in sec . [ sec : discussion1 ] and that of thermal fluctuation in sec . [ sec : discussion2 ] . we also discuss the effect of renormalization on the noise at low temperature caused by quantum fluctuation in sec . [ sec : discussion3 ] . in order to clarify the nature of non - equilibrium current fluctuation , we consider the case of zero temperature , where there is no thermal fluctuation . furthermore , we limit ourselves to the condition of the high bias voltage , @xmath263 , where the renormalization effect is negligible . figure [ fig : s ] shows the current noise ( the top panel ) and the charge noise ( the bottom panel ) as a function of the excitation energy @xmath264 for small @xmath11 ( ( a-1 ) and ( b-1 ) ) and those for large @xmath11 ( ( a-2 ) and ( b-2 ) ) . plots are normalized by values of the orthodox theory at @xmath127 . solid lines , dashed lines and dotted lines show results of our approximation ( eqs . ( [ eqn : siirta ] ) and ( [ eqn : sqqrta ] ) ) , the orthodox theory ( eqs . ( [ eqn : siiorth ] ) and ( [ eqn : sqqorth ] ) ) and the co - tunneling theory ( eqs . ( [ eqn : siicotwidth ] ) and ( [ eqn : sqqcotwidth ] ) with @xmath251 ) , respectively . when @xmath11 is small , our results well reproduce the orthodox theory ( ( a-1 ) and ( b-1 ) ) . for large @xmath11 ( ( a-2 ) and ( b-2 ) ) and in cb regime ( @xmath265 ) , they agree well with the co - tunneling theory . figure [ fig : in ] shows the average current ( a ) and the average charge ( b ) estimated by our approximation ( solid lines ) and the orthodox theory ( dashed lines ) . in cb regime , we can see both the average value and the noise are enhanced by the quantum fluctuation . around @xmath127 , the average current and the current noise are strongly suppressed due to the life - time broadening effect @xcite . next , we discuss the validity of the co - tunneling theory with life - time broadening ( see eqs . ( [ eqn : siicotwidth ] ) and ( [ eqn : sqqcotwidth ] ) ; dot - dashed lines in fig . [ fig : s ] ) . as for the charge noise in the limit of @xmath227 , it reproduces the result of the orthodox theory as well as our approximation ( fig . [ fig : s ] ( b-1 ) : the solid line and the dot - dashed line almost overlap each other ) . in st regime it overestimates the current noise ( fig . [ fig : s ] ( a-1 ) ) , because it does not take the coulomb correlation effect into account as mentioned before . we want to stress again that our result reproduces the orthodox theory in the limit of @xmath227 , which can not be achieved by the co - tunneling theory with life - time broadening . the physical picture of the non - equilibrium current fluctuation is understood more clearly with the help of the fano factor defined by @xmath266 . the fano factor is unity when the tunneling event of electrons is poissonian process and is suppressed when tunneling events are correlated . the orthodox theory predicts the sub - poissonian behavior in st regime because of the coulomb correlation : suppose one electron tunnels into the island through one junction , the next tunneling event must be the out going process of another electron through the other junction . in cb regime , the co - tunneling theory predict @xmath267 . it means that co - tunneling events , viz . the simultaneous tunneling events of two electrons through the two junctions , occur randomly . the figure [ fig : fano ] shows the excitation energy dependence of the fano factor obtained by our approximation . in the small @xmath11 limit , our approximation reproduces the orthodox theory in st regime and the co - tunneling theory in cb regime and smoothly interpolate two theories ( the dotted line actually almost coincides with the result of orthodox theory in st regime ) . for larger @xmath11 the fano factor is further suppressed ( the dashed and the solid line ) . especially , our result predicts the value smaller than @xmath268 at the degeneracy point . it is a distinctive result because the orthodox theory predicts the inequality @xmath269 ( p. 137 @xcite ) . next we consider the physical meaning of our result . in cb regime and near the threshold voltage , the origin of the suppression of fano factor is considered to be the enhancement of the effective transmission probability @xmath205 , because the increase in the current noise is much larger then that in the charge noise ( see figs . [ fig : s ] ( a-2 ) and [ fig : s ] ( b-2 ) ) . as the fano factor of the shot noise is approximately given by @xmath270@xcite , the enhancement of the transmission probability results in the suppression of the fano factor . around the degeneracy point , the origin of the suppression of fano factor is different , because the normalized current and the current noise are suppressed as shown in fig . [ fig : in ] ( a ) and fig . [ fig : s ] ( a-2 ) . we consider that the fano factor is suppressed because of the dissipation , i.e. the life - time broadening effect : rta takes account of the dissipation process which is the leak of an electron from the island while another electron tunnels into the island and relaxes to the local equilibrium state of the island . the suppression of the fano factor by the dissipation was previously predicted for the 1d electron channel coupled with a boson bath@xcite . next we discuss the effect of the thermal fluctuation . figure [ fig : nyquist ] shows the temperature dependence of the current noise ( a ) , the average current ( b ) and the fano factor ( c ) at a threshold for various @xmath11 . they are normalized by the value of the orthodox theory at @xmath271 . as the temperature increases , the average current and the current noise increase because of the thermal fluctuation . at sufficiently high temperature @xmath272 , orthodox theory predicts that the average current saturates at @xmath273 and the thermal fluctuation dominates the current noise , i.e. @xmath274 which is the similar form as the johnson - nyquist noise for the ohmic resistance@xcite ( the plot for @xmath275 almost coincides with the orthodox theory ) . our result further shows that the average current and the current noise are suppressed as @xmath11 increases ( panels ( a ) and ( b ) ) . it is considered to be attributed to the higher order tunneling effect : the lifetime broadening caused by the thermal fluctuation is enhanced for the large tunnel conductance . the panel ( c ) is the fano factor versus temperature plot . the fano factor is independent of @xmath11 , which means that the correlation between tunneling events does not depend on @xmath11 . next we consider the renormalization effect at low bias voltage and temperature : @xmath276 . figures [ fig : sqq - low ] ( a ) and ( b ) show the charge noise normalized by @xmath277 at @xmath127 as a function of the temperature and the bias voltage , respectively . the charge noise is suppressed for large @xmath11 , which is attributed to the renormalization of the system parameters . in the regime @xmath278 , where the life - time broadening effect is negligible ( @xmath225 ) , we can approximate eq . ( [ eqn : sqqrta ] ) as @xmath279 instead of eq . ( [ eqn : sqqorth ] ) . here @xmath280 and @xmath281 are tunneling rate eq . ( [ eqn : tunnelingrate ] ) written by using renormalized parameters such as @xmath282 where the renormalization factor @xmath214 is @xmath283 . the lower cut - off energy @xmath216 is @xmath284 for the panel ( a ) and @xmath285 for the panel ( b ) . it is natural to interpret eq . ( [ eqn : chargernormal ] ) that the charge of a carrier is modified as @xmath286 by the renormalization effect . the interpretation is similar to that of the doubling of shot noise at the normal - metal(n)-superconductor(s ) interface . since at ns interface the carrier is @xmath287-charged particle , viz . a cooper pair , the shot noise is twice as large as that at nn interface @xcite . though the normalized charge noise is suppressed with increasing @xmath11 , one see that the charge noise always diverges at @xmath288 for arbitrary @xmath11 in the weak tunneling regime . since the charge noise is related to the charge susceptibility `` for excitation energy @xmath264 , the divergence means that the number of charge changes by one '' at the degeneracy point when we sweep the excitation energy . it is confirmed by the fact that the slope of excitation energy dependence of the average charge ( an inset of fig . [ fig : sqq - low ] ( b ) ) diverges at the degeneracy point . next we discuss the renormalization effect on the fano factor . figure [ fig : fano3 ] shows the excitation energy dependence of the fano factor for various bias voltage . we can see at small bias voltage where the charging energy renormalization is pronounced @xmath289 , the valley structures of curves are widened . the same behavior can be seen in the differential conductance shown in refs . we also see that the fano factor is suppressed with increasing bias voltage at @xmath127 . this suppression is also attributed to the dissipation as discussed in sec . [ sec : discussion1 ] : as the bias voltage increases , the dissipative charge fluctuation is enhanced and thus the fano factor is suppressed . by using the drone - fermion representation and the schwinger - keldysh approach , we have calculated the current noise and the charge noise in the regime of large quantum fluctuations of charge out of equilibrium . we have reformulated and extended rta in a charge conserving way . our approximation interpolates previous theories , the orthodox theory and the co - tunneling theory : our result coincides with the orthodox theory in the limit of @xmath227 and is consistent with the co - tunneling theory in cb regime . the approximation is verified from the fact that the result satisfies the fluctuation - dissipation theorem . in previous papers , we also checked numerically that the energy sensitivity does not exceed the quantum limit@xcite . we showed that at zero temperature and @xmath263 , the life - time broadening caused by non - equilibrium dissipative charge fluctuation suppresses the current noise in st regime . it also suppresses the fano factor more than the coulomb correlation does . especially the fano factor is suppressed below the minimum value predicted by the orthodox theory @xmath268 around @xmath127 . the origin of the suppression is attributed to the charge fluctuation which appears as the enhancement of the transmission probability in cb regime and the dissipation in st regime . at @xmath290 , we showed that the average current and the current noise deviate from the predictions of the orthodox theory with increasing @xmath11 . however , the fano factor is independent of @xmath11 and is proportional to the temperature . it means that the current noise is dominated by the thermal fluctuation and the correlation between the tunneling events does not depends on @xmath11 . at small bias voltage and temperature @xmath276 , the charge noise is suppressed as compared with the prediction of the orthodox theory . we showed that it can be interpreted as the renormalization for the unit of island charge . although the charge is renormalized , the charge noise diverges at @xmath288 for arbitrary @xmath11 in the weak tunneling regime . it means that the quantum fluctuation does not wash out the charge quantization . in this paper , we have limited ourselves to the discussions on the second moment and the zero frequency component , because we think them primitive . the investigation on the frequency dependence of noise will be important to estimate the performance of high - speed set electrometer completely@xcite . the investigation on the higher order moment and the full counting statistics@xcite will help us to understand carriers of strongly correlated system out of equilibrium . we would like to thank y. isawa , j. martinek , yu . v. nazarov and g. johansson for variable discussions and comments . this work was supported by a grant - in - aid for scientific research ( c ) , no . 14540321 from mext . was supported by mext , grant - in - aid for encouragement of young scientists , no . in this appendix , we show the relation between expressions for average and noise in the operator representation and those in the generating functional representation . the variation of the exact action eq . ( [ eqn : action0 ] ) accompanied by the infinitesimal variation of @xmath291 is given as the twisted " combination ( sec . 9.3.2 in ref . @xcite ) of @xmath292 and @xmath291 , @xmath293 where the center - of - mass coordinate of charge is @xmath294 . employing this source term , we can show that the eq . ( [ eqn : q - exact ] ) is equivalent to @xmath295 where @xmath296 : @xmath297 where we used eq . ( [ eqn : anti - comute ] ) to obtain the final form . the second derivative @xmath298 is calculated as @xmath299 here the first term includes the correlation function of 4 field variables on a same branch , @xmath300 , which is not well defined at @xmath301 . usually , an additional operation to determine the order of field variables is required to remove the uncertainty . alternatively , we subtract a term @xmath302 from eq . ( [ eqn : sqq_k ] ) as shown in our definition eq . ( [ eqn : sqq ] ) . here we used the normalization eq . ( [ eqn : normalization ] ) , @xmath303 . as a result , the first term in eq . ( [ eqn : sqq_k ] ) is replaced by @xmath304 , which does not include the uncertainty . ( [ eqn : anti - comute ] ) , we show that our definition is equal to the standard charge noise expression @xmath305 we should stress that our definition does not change the final result because eq . ( [ eqn : sqq2nd ] ) is 0 from the normalization . the exact current expression is obtained in the same way . the source term corresponding to eq . ( [ eqn : ds ] ) is given by @xmath306 which leads to relations @xmath307 where the current operator at junction @xmath163 is defined as @xmath308 in this appendix , we demonstrate the charge conservation law . as @xmath122 is invariant under the transformation eq . ( [ eqn : gauge_t ] ) , we obtain an identity @xmath309 we can derive the other equation , which is obtained from above equation by replacing @xmath138 with @xmath115 . however , the latter equation is not important in the following discussions . by putting the auxiliary source fields as the values given in the subscripts of eq . ( [ eqn : i - exact ] ) , and employing eqs . ( [ eqn : i - exact ] ) and ( [ eqn : q - exact ] ) , we obtain the current continuity equation , eq . ( [ eqn : iconserve ] ) . next we demonstrate the charge conservation law for correlation functions . by acting the operator @xmath310 or @xmath311 on eq . ( [ eqn : identityw1 ] ) , we obtain following two equations : @xmath312 by comparing the left - hand side of the former equation and the right - hand side of the latter equation , by setting the auxiliary source fields as the values given in the subscripts of eq . ( [ eqn : i - exact ] ) , and by using eqs . ( [ eqn : sii ] ) and ( [ eqn : sqq ] ) , we obtain the charge conservation law for correlation functions , eq . ( [ eqn : sconserve ] ) . in this appendix , we calculate the particle - hole gf . we begin with the tunneling action for the large transverse channel obtained from eq . ( [ eqn : action0 ] ) by tracing out the electron degrees of freedom@xcite : @xmath313 . in the physical representation , it is rewritten as @xmath314 , where the vector field @xmath315 is defined in the same way as eq . ( [ eqn : fieldphysical ] ) . each component of @xmath316 can be calculated by utilizing the functional derivation . for example , ( 1,2)-component is @xmath317 . \nonumber \\ \label{eqn : tracea1}\end{aligned}\ ] ] as the functional derivative is @xmath318 in the physical representation , the trace yields @xmath319/2 $ ] . here we put @xmath320 and @xmath321 . the other components are evaluated in the same way . then four components are given as @xmath322 where we omitted arguments and coefficients . to obtain this form , we used the normalization eq . ( [ eqn : normalization ] ) or the relation @xmath323 ( see eqs . ( 2.64 ) and ( 2.65 ) in ref . @xcite ) . the following calculations are same as those in ref . @xcite . employing the fourier transforms of the gf defined in eq . ( [ eqn : gfl ] ) , @xmath324 and @xmath325 , we obtain eq . ( [ eqn : alphark ] ) . another loop diagram , the self - energy eq . ( [ eqn : partselfenergy ] ) , can be calculated in the same way . four components are given in the same form as those of eq . ( [ eqn : ph ] ) . by using eqs . ( [ eqn : drk ] ) and ( [ eqn : alphark ] ) , we obtain @xmath326 from these equations , eqs . ( [ eqn : selfenergy ] ) and ( [ eqn : selfenergyr ] ) can be derived . in this appendix , we describe some results of finite order perturbation theory and explain why we introduced eq . ( [ eqn : wrta ] ) which takes account of all orders for @xmath115-field corrections . first , we calculate the first order contribution of average charge by employing @xmath120 , eq . ( [ eqn : q - exact ] ) and rules * ( i ) * and * ( ii). @xmath327{diagram15.eps } \end{array } \right\| = -\frac{{i}}{2 \pi } \int { { \rm d}}\varepsilon { \rm tr } \left [ \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{g}_c(\varepsilon ) { \mbox{\boldmath$\tau$}}^1 \tilde{\sigma}_c(\varepsilon ) { \mbox{\boldmath$\tau$}}^1 \tilde{g}_c(\varepsilon ) \right ] \nonumber \\ & = & -{i}\int \frac{{{\rm d}}\varepsilon}{8 \pi } ( g_c^k(\varepsilon ) g_c^p(\varepsilon ) \sigma_c^p(\varepsilon ) + g_c^k(\varepsilon ) g_c^c(\varepsilon ) \sigma_c^c(\varepsilon)+ 2 \|g_c^r(\varepsilon)\|^2 \sigma_c^k(\varepsilon ) ) , \nonumber\end{aligned}\ ] ] where gf denoted with superscript @xmath328 is given as @xmath329 etc . in the equilibrium state , the second and the third terms of the second line are negligibly small @xmath330 and the first term is simplified to @xmath331 where we utilized the relation @xmath332 . in the limit of zero temperature , eq . ( [ eqn : q2nd ] ) leads to the log - divergence@xcite as @xmath333 . above result suggests that the @xmath115-field correction is responsible for the divergence . it is further confirmed by calculating second order contribution of average current generated from @xmath115-field correction @xmath334 . employing rules ( i ) * and * ( ii ) * , we obtain @xmath335{diagram16.eps } \end{array } \right\| \nonumber \\ & = & 2 e { \rm re } \int \frac{{{\rm d}}\varepsilon}{h } { \rm tr } \left [ \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{\sigma}_{\rm r}(\varepsilon ) { \mbox{\boldmath$\tau$}}^1 \tilde{g}_c(\varepsilon ) { \mbox{\boldmath$\tau$}}^1 \tilde{\sigma}_c(\varepsilon ) { \mbox{\boldmath$\tau$}}^1 \tilde{g}_c(\varepsilon ) \right ] \nonumber \\ & = & e \int \frac{{{\rm d}}\varepsilon}{2 h } \sigma^k_{\rm \bar{r}}(\varepsilon ) \sigma^c_{\rm r}(\varepsilon)-(c \leftrightarrow k ) + \delta i^{(2)}_{\rm r } \nonumber \\ & = & \gamma^{+}_{\rm r}(\eta ) - \gamma^{-}_{\rm r}(\eta ) + \delta i^{(2)}_{\rm r}. \label{eqn : i2}\end{aligned}\ ] ] the first and the second terms of eq . ( [ eqn : i2 ] ) , which are consistent of the expression for co - tunneling current @xcite , also diverge at the degeneracy point . from above discussions , we can deduce that the most simple way to regulate the divergence is to sum up @xmath115-field corrections @xmath128 up to infinite @xmath80 as shown in eq . ( [ eqn : wrta ] ) . it should be noted that the correction term @xmath336 diverges in the limit of @xmath171 . this divergence disappears when we consider eq . ( [ eqn : wrta ] ) as discussed in sec . [ sec : reformulation ] and sec . [ sec : mainbody ] . in this appendix , we demonstrate that the rule * ( ii ) * can be also applied to the calculation of zero - frequency noise . we also demonstrate the rule to calculate the second term of the noise expression eq . ( [ eqn : sii ] ) or eq . ( [ eqn : sqq ] ) . for example , we consider the current noise related to @xmath120 . though we consider the simple case , the following discussions can be generalized . from the definition eq . ( [ eqn : siipart ] ) , the noise diagrams is obtained as : @xmath337{diagram17.eps } \end{array } \right\| \delta_{\rm r , r ' } - \frac{(\delta \rightarrow c)}{4}. \label{eqn : sii_1}\ ] ] the integration in terms of @xmath37 in the first term is calculated as @xmath338{diagram17.eps } \end{array } \right\| = \int { { \rm d}}t { \rm tr } \left . \left [ \frac{\delta \varphi_{\rm r}}{\delta \varphi_{{\rm r } \ , \delta}(t ) } \sigma_{\rm r } \frac{\delta \varphi_{\rm r'}}{\delta \varphi_{{\rm r ' } \ , \delta}(t ' ) } g_c \right ] \right\| \nonumber \\ & = & \int \frac{{{\rm d}}\varepsilon}{h } { \rm tr } \left [ \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{\sigma}_{\rm r}(\varepsilon ) \frac{{\mbox{\boldmath$\tau$}}^0}{2 } \tilde{g}_c(\varepsilon ) \right ] . \label{eqn : trace4}\end{aligned}\ ] ] and thus we can see that the rule * ( ii ) * can be applied for the calculation of the zero - frequency current noise . as for the second term , the derivation @xmath339 appears instead of eq . ( [ eqn : deltaphysical1r ] ) . thus we can derive the rule that the second term is obtained from the first term by replacing @xmath235 with @xmath200 as @xmath340/h $ ] @xcite . as a result , ( [ eqn : sii_1 ] ) is expressed as @xmath341 \delta_{\rm r , r'}/(2 h ) - ( { \mbox{\boldmath$\tau$}}^0 \rightarrow { \mbox{\boldmath$\tau$}}^1 ) . \nonumber\ ] ] as for the charge noise , we can repeat the same discussions as above .
a real in the unit interval is _ computably enumerable _ ( _ c.e . _ ) if it is the limit of a computable , increasing sequence of rationals . we identify a real with its infinite binary expansion . in contrast with the case of a computable real , whose bits are given by a computable function , during the process of approximation of a c.e . real one may never know how close one is to the limit . a real is ( algorithmic ) random if its binary expansion is an algorithmic random ( infinite ) sequence @xcite . a prefix - free machine is a turing machine , shortly , _ machine _ , from strings to strings whose domain is a prefix - free set . a machine is universal if it can simulate every machine . chaitin @xcite introduced the halting probability @xmath1 of a universal machine @xmath0 , chaitin s omega number @xmath2 and proved that _ @xmath1 is c.e . and random_. as shown by calude , hertling , khoussainov , wang @xcite and kuera , slaman @xcite , ( see also @xcite ) there are no other c.e . random reals : [ representation : thm]the set of c.e . random reals coincides with the set of halting probabilities of all universal machines . random reals have been intensively studied in recent years , with many results summarised in @xcite . [ gregzfc ] assume that zfc ( zermelo - fraenkel set theory with choice ) is arithmetically sound ( that is , any theorem of arithmetic proved by zfc is true ) . then , for every universal machine @xmath0 , zfc can determine the value of only finitely many bits of @xmath1 , and one can calculate a bound on the number of bits of @xmath1 which zfc can determine . the real @xmath1 depends on @xmath0 , and so by tuning this choice one gets : [ solovay ] we can chose a universal machine @xmath0 so that zfc ( if arithmetically sound ) can not determine any bit of @xmath1 . this result was generalised as follows : [ crisincompl ] assume that zfc is arithmetically sound . let @xmath3 and consider the c.e . random real @xmath4 then , we can effectively construct a universal machine @xmath0 ( depending upon zfc and @xmath5 ) such that pa ( peano arithmetic ) proves the universality of @xmath0 , zfc can determine at most @xmath6 initial bits of @xmath1 and @xmath7 . the proof of theorem [ crisincompl ] in @xcite starts by fixing a universal machine @xmath8 such that the universality of @xmath8 is provable in pa and @xmath9 . solovay @xcite observed that `` it is by no means evident that there is a universal prefix - free machine whose universality is provable in pa and whose halting probability is @xmath5 '' . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ let @xmath10 be c.e . and random . is there any _ representation _ of @xmath5 for which pa can _ prove _ that @xmath11 c.e . and random ? _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ we give an affirmative answer to this question . a major step in the proof is to show that theorem [ representation : thm ] can be proved in pa . our proof , which is simpler than the standard one , can be used also for the original theorem . the paper also includes a sharper form of the kraft - chaitin theorem , as well as a formal proof of this theorem written with the proof assistant isabelle . in what follows proofs will be written in solovay s style @xcite . all necessary steps are presented in sufficient detail to leave the remaining formalisation routine . the formalisation of the kraft - chaitin theorem is presented with full details , and then a sketch of the formal proof in isabelle is discussed . the paper is organised as follows . sections 2 and 3 present all facts on formal provability and algorithmic information theory needed for this paper . the kraft - chaitin theorem is presented in section 4 . section 5 presents three ways to prove randomness , using martin - lf tests , prefix - free complexity , and solovay representation formula . in section 6 we revisit chaitin s theorem on the randomness of the halting probability of a universal machine . in section 7 we prove that a real @xmath12 is provably chaitin - random iff it is provable that @xmath7 for some provably universal machine @xmath0 ( see theorem [ thm : representation ] ) . in section 8 we prove our main theorem : every c.e . random real is provably random ( theorem [ thm : cerandprov ] ) . in section 9 we construct a universal machine @xmath0 based on which pa can not prove the randomness of its halting probability . section 10 presents a formal proof of the kraft - chaitin theorem written with isabelle . the final section 11 includes a few general remarks . by @xmath13 we denote the first - order language of arithmetic whose non - logical symbols consist of the constant symbols 0 and 1 , the binary relation symbol @xmath14 and two binary function symbols @xmath15 ( addition ) and @xmath16 ( multiplication ) . peano arithmetic ( see @xcite , shortly , pa ) is the first - order theory given by a set of 15 axioms defining discretely ordered rings , together with induction axioms for each formula @xmath17 in @xmath13 : @xmath18 01 this schema retains as much as possible from the second - order induction axiom : @xmath19 where @xmath20 ranges over all subsets of the domain , and @xmath21 range over elements of this domain , but avoids quantification over sets of natural numbers , which is impossible in first - order logic . in pa it is not possible to say that any set of natural numbers containing 0 and closed under successor is the entire set of natural numbers , but that any definable set of natural numbers has this property . the induction schema includes one instance of the induction axiom for every definition of a subset of the naturals . the structure * n * whose domain is the set of naturals @xmath22 , where the symbols in @xmath13 have the obvious interpretation , satisfies the axioms of pa ; this is the standard model for pa . there are non - standard models of pa that are not isomorphic to * n. if @xmath23 is a structure for @xmath13 and @xmath24 is an @xmath13-formula with free - variables @xmath25 and @xmath26 , then we write @xmath27 to mean that `` @xmath28 is true in @xmath23 when each variable @xmath29 is interpreted by @xmath30 '' . we blur the distinction between @xmath31 and the closed term of @xmath13 , @xmath32 , ( @xmath31 times ) . a formula @xmath33 of @xmath13 is @xmath34 if all its quantifiers are bounded . a formula @xmath35 of @xmath13 is @xmath36 if it is of the form @xmath37 with @xmath38 ; @xmath35 of @xmath13 is @xmath39 if it is of the form @xmath40 with @xmath38 . by pa @xmath41 we mean `` there is a proof in pa for @xmath42 '' . it is useful to know that pa proves the _ least number principle _ : @xmath43 for each formula @xmath44 of @xmath13 . an important link between computability and provability is given by the following results . a partial function from @xmath45 to @xmath45 is partial computable iff its graph is equivalent to a @xmath36 @xmath13-formula . [ cesets]a set @xmath46 is computably enumerable ( c.e . ) if there is a @xmath36 @xmath13-formula @xmath24 such that for all @xmath47 , @xmath48 iff n * @xmath49 . a total function @xmath50 is represented in pa if there is an @xmath13-formula @xmath51 such that for all @xmath52 : + 1 . pa @xmath53 , and 2 . if @xmath54 then pa @xmath55 . + ( here @xmath56 means `` there exists a unique '' . ) one can show that every total computable function is represented by a @xmath36-formula of pa @xcite . a function @xmath57 is _ provably computable _ @xcite if there exists a @xmath36-formula of pa @xmath58 such that : + 1 . @xmath59 , 2 . pa @xmath60 . + in view of corollary [ cesets ] , any provably computable function has a c.e . graph , so it is total and computable . these functions can be viewed as computable functions whose totality is proved by pa . every primitive recursive function is provably computable , but there exist computable functions which are not provably computable in pa . if @xmath61 is computable but not provably computable in pa , then the statement `` @xmath61 is total '' is true but unprovable in pa . in contrast with the case of computable functions , _ c.e . sets are provably enumerable _ @xcite ( because every non - empty c.e . set can be enumerated by a primitive recursive function , @xcite , p. 138 ) . in what follows all computations will be implemented by primitive recursive functions . hence , we will work with a special type of @xmath36 formulae . by abuse of language we say that a formula of pa is @xmath62 if it has the form @xmath63 , for some primitive recursive predicate @xmath64 . a formula of pa is @xmath65 if it has the form @xmath66 , for some primitive recursive predicate @xmath67 . our metatheory is zfc . we fix a ( relative ) interpretation of pa in zfc . each formula of @xmath13 has a translation into a formula of zfc determined by the interpretation of pa in zfc . by abuse of language we shall use the phrase `` sentence of arithmetic '' to mean a formula with no free variables of zfc that is the translation of some formula of pa . we assume that zfc is 1consistent , that is , if it proves a @xmath62 sentence then that sentence is true ( in the standard model of pa ) . [ solovay @xcite ] every @xmath65 sentence proved by zfc is true . as a consequence , it follows that _ if @xmath0 is a machine which _ pa _ can prove universal and zfc can prove the sentence `` the i - th digit of @xmath68 is @xmath69 '' , then the sentence is true_. whenever we talk about the provability of a sentence of arithmetic we mean that pa proves its corresponding translation formula . if there is a proof in pa for statement @xmath70 we say that @xmath70 provable in pa . we say that @xmath70 is provably @xmath71 ( where @xmath71 is a property ) if the statement `` @xmath70 has @xmath71 '' is provable in pa . _ all reals are in the unit interval . _ a c.e . real @xmath5 is represented by an increasing computable sequence of rationals converging to @xmath5 . we blur the distinction between the real @xmath5 and the infinite base - two expansion of @xmath5 , i.e. the infinite sequence @xmath72 ( @xmath73 such that @xmath74 by @xmath75 we denote the string of length @xmath31 , @xmath76 . the set of ( bit ) strings is denoted by @xmath77 ; @xmath78 denotes the empty string . if @xmath79 is a string then @xmath80 denotes the length of @xmath79 . we import the theory of computability from natural numbers to strings by fixing the canonical bijection between @xmath77 and @xmath45 induced by the linear order @xmath81 if @xmath82 or @xmath83 and @xmath79 lexicographically precedes @xmath84 . a machine @xmath0 is _ universal _ if for every machine @xmath8 there is a constant @xmath85 ( depending upon @xmath0 and @xmath8 ) such that for all strings @xmath86 , if @xmath87 , then @xmath88 for some string @xmath89 of length @xmath90 . the domain of @xmath0 is the set @xmath91 . the omega number @xmath92 is halting probability of @xmath0 . the _ prefix - free complexity _ of the string @xmath93 ( relatively to the machine @xmath94 ) is @xmath95 ( @xmath96 ) . if @xmath0 is a universal machine , then for every machine we can effectively construct a constant @xmath85 ( depending on @xmath0 and @xmath94 ) such that @xmath97 , for all @xmath98 . a real @xmath5 is _ chaitin - random _ if there exists a universal machine @xmath0 and constant @xmath85 such that for all @xmath99 , @xmath100 . open set is a c.e . union of intervals with rationals endpoints @xmath101 and @xmath102 is lebesgue measure . if @xmath103 is a prefix - free set , then @xmath104 denotes the lebesgue measure of the cylinder denoted by @xmath103 , i.e. all reals whose infinite binary expansions have a prefix in @xmath103 . to the string @xmath98 we associate the interval @xmath105 of measure @xmath106 . a martin - lf test ( shortly , ml test ) @xmath70 is a uniformly c.e . sequence of c.e . open sets @xmath107 such that for all @xmath108 , @xmath109 . a real @xmath5 is _ martin - lf - random _ ( shortly , _ ml - random _ ) if for every ml test @xmath70 there exists an @xmath110 such that @xmath111 . a classical theorem states that _ a real is chaitin - random iff it is ml - random _ @xcite . note that chaitin and martin - lf definitions apply to any real . in the special case of c.e . reals the following solovay representation formula stated in @xcite is used : a real @xmath5 is _ c.e . and random _ if there exists a universal machine @xmath0 , an integer @xmath112 and a c.e . real @xmath113 such that @xmath114 ( see lemma [ lem : solovayrepresentation ] ) . we start by showing that pa can prove the kraft - chaitin theorem @xcite . [ thm : kc ] suppose @xmath115 is a primitive recursive enumeration of `` requests '' which provably satisfies @xmath116 then there exists a provably prefix - free machine @xmath23 and a primitive recursive enumeration @xmath117 of @xmath118 such that the following is provable in pa : + 1 . @xmath119 , 2 . @xmath120 for all @xmath121 , 3 . @xmath122 for all @xmath121 . algorithm [ alg : kc ] below enumerates the graph of @xmath23 . intuitively , @xmath123 keeps track of the tree of prefixes we havent allocated yet . to start with we have allocated nothing , so @xmath124 . at each step we want a string ( node ) of a given length ( depth ) @xmath125 . the program selects the deepest leaf it can , then creates the smallest number of new leaves to create the node we need . + @xmath124 , @xmath126 , @xmath127 , @xmath128 . let @xmath129 be the longest element of @xmath103 of length at most @xmath125 . if no such string exists , terminate . define @xmath132 . @xmath135 . examining algorithm [ alg : kc ] , it is clear that the sequence @xmath136 is a primitive recursive enumeration of @xmath118 , and whenever @xmath137 is defined we have @xmath122 and @xmath120 . it remains to show that @xmath137 is defined for all @xmath121 ( i.e. the program never terminates ) , that @xmath118 is prefix - free , and @xmath138 . it suffices to establish , for all @xmath110 , the following invariants : + 1 . @xmath139 is prefix - free ( which implies that @xmath123 and @xmath140 individually are prefix - free ) , 2 . @xmath141 , 3 . @xmath142 , 4 . @xmath143 , 5 . if @xmath144 , then @xmath123 contains a string of length at most @xmath31 ( equivalently , @xmath123 contains strings of distinct length ) . the base case is trivial . for the inductive step , first observe that line 3 of algorithm [ alg : kc ] does nt terminate since @xmath145 by invariant 5 . we see that @xmath146 which is prefix - free establishing invariant 1 . from this we can see invariant 2 holds : @xmath147 next observe invariant 3 holds too : @xmath148 which implies that @xmath149 from this follows invariant 4 : @xmath150 finally , since @xmath129 is the longest string of length at most @xmath125 in @xmath151 , and we add strings of distinct length between @xmath152 and @xmath125 to @xmath123 to form @xmath153 , we see that @xmath153 consists of strings of distinct lengths . this establishes invariant 5 . in this section we discuss three forms of provability for randomness . there are two ways to represent a c.e . real number @xmath12 in pa : 1 ) by giving an increasing 1 - 1 primitive recursive function that enumerates a c.e . prefix - free set of strings @xmath154 such that @xmath155 , 2 ) by giving an increasing primitive recursive sequence @xmath156 of rationals in the unit interval whose limit is @xmath5 . it is clear that given the representation 1 ) one can effectively get the representation 2 ) . the converse is also true . [ lem : repce ] let @xmath5 be a c.e . real defined by the increasing primitive recursive sequence @xmath156 of rationals . then there is a primitive recursive sequence @xmath157 of natural numbers such that pa proves @xmath158 without loss of generality assume @xmath159 . define the primitive recursive sequences @xmath160 and @xmath161 by @xmath162 and for @xmath163 by @xmath164 since @xmath165 is strictly increasing we can establish by induction the inequality @xmath166 for all @xmath110 , making the logarithm well - defined . by construction we have @xmath167 define @xmath168 . since @xmath169 we have @xmath170 taking the limit we see that @xmath171 establishing our result . let @xmath5 be a c.e . real defined by the increasing primitive recursive sequence @xmath156 of rationals . then there is a machine @xmath23 such that pa proves that @xmath172 . use lemma [ lem : repce ] and theorem [ thm : kc ] . in what follows a c.e . real is given by one of the above representations . real @xmath5 is _ provably chaitin - random _ if there exists a provably universal machine @xmath0 and constant @xmath85 such that pa proves that for all @xmath99 , @xmath100 . real @xmath5 is _ provably ml - random _ if for every set @xmath70 which pa proves to be a ml test and pa proves that there exists an @xmath110 such that @xmath111 . the classical theorem that states that _ a real is chaitin - random iff it is ml - random _ is provable in pa . however , for the goal of this paper only one implication is needed : [ thm : chaitintoml ] every c.e . provably chaitin - random real is provably ml - random . take a c.e . real @xmath5 , a machine @xmath0 which is provably universal and a natural @xmath173 such that pa proves that for all @xmath99 , @xmath174 . we wish to prove that for every @xmath175 which pa proves to be a ml test there exists an @xmath110 such that pa proves that @xmath111 . following the proof of proposition 6.3.4 in @xcite it follows that pa proves the existence of a c.e . set @xmath176 such that each @xmath177 is prefix - free , and by taking @xmath178 we get @xmath179 . let @xmath180 be a 1 - 1 primitive recursive enumeration of the graph of @xmath0 . denote by @xmath181 for @xmath182 the projection functions and @xmath183 is a 1 - 1 primitive recursive enumeration of @xmath184 . note that @xmath185 can be expressed in pa : @xmath186 , we have : @xmath187 we can now use theorem [ thm : kc ] : there exists a provably prefix - free machine @xmath23 such that : @xmath188 , @xmath189 since @xmath0 is provably universal there is a constant @xmath190 such that for all strings @xmath191 , so in particular , pa proves that for all @xmath192 , if @xmath193 , then @xmath194 we are now in a position to find a natural @xmath192 such that pa proves that @xmath195 showing that @xmath5 is provably ml - random . note that for @xmath192 , pa proves @xmath195 iff for all @xmath196 , pa proves that @xmath197 . for all @xmath196 , pa proves that @xmath198 implies @xmath199 hence , for @xmath200 , pa proves that @xmath201 implies @xmath197 , so because @xmath5 is provably chaitin - random pa proves that @xmath195 . * comment * the above proof shows that that @xmath202 , where @xmath0 is a provably universal machine , is a provably ml test such that for all @xmath192 and provably ml test @xmath70 there exists @xmath203 such that pa proves the inclusion @xmath204 , i.e. @xmath205 is a provably universal ml test . to be able to complete our program we need to choose a specific representation for a c.e . and random real which can be `` understood '' by pa and , even more importantly , pa can extract from it a proof of the randomness of the real ( c.e . is obvious ) . first we work with solovay representation formula discussed at the end of section [ ait ] . a real @xmath5 is _ c.e . and provably random _ if there exists a representation of @xmath5 in the form @xmath206 where @xmath8 is a provably universal machine , @xmath207 is an integer and @xmath113 is a provably c.e . theorem [ lem : solovayrepresentation ] shows that all c.e . random reals have a representation of this form . in detail , pa receives an algorithm for a machine @xmath8 , a proof that @xmath8 is prefix - free and universal , an integer @xmath207 and a computable increasing sequence of rational converging to a real @xmath208 . the goal is to prove that pa can use this information to prove that @xmath209 is c.e . and random . chaitin @xcite proved that the halting probability of a universal machine is chaitin - random . this theorem is provable in pa : [ thm : chaitin ] suppose @xmath0 is provably universal . then @xmath210 is provably chaitin - random . let @xmath180 be a 1 - 1 primitive recursive enumeration of the graph of @xmath0 . denote by @xmath211 for @xmath182 the projection functions and @xmath183 is a 1 - 1 primitive recursive enumeration of @xmath184 . recall that @xmath185 can be expressed in pa . 01 @xmath212 @xmath213 define the machine @xmath23 by @xmath214 . since @xmath0 is provably universal , there is a @xmath85 such that for all @xmath98 , @xmath215 this shows that @xmath185 is provably total and @xmath0 is provably onto . define the primitive recursive sequence of rationals @xmath216 and notice that @xmath217 is provably strictly increasing ; @xmath1 is , by definition , the limit of this sequence . define @xmath218 if there exist @xmath84,@xmath219 such that + 1 . @xmath220 ( i.e. @xmath221 ) , 2 . @xmath84 is the least such that @xmath222 ( i.e. @xmath223 ) , 3 . @xmath224 is the lexicographically least string such that @xmath225 for all @xmath226 . this defines a provably prefix - free machine . observe that if @xmath227 is defined and @xmath228 then @xmath229 . from this we can establish that whenever @xmath227 is defined we have @xmath230 as @xmath0 is provably universal , there exists an @xmath231 such that for all @xmath232 , @xmath233 is provable in pa . denote by @xmath234 the @xmath110th digit of @xmath1 . since @xmath0 is provable onto , for each @xmath31 there exists a string @xmath235 such that @xmath236 since @xmath237 we know that @xmath238 is defined . let @xmath84 be the least natural ( found when evaluating @xmath238 ) such that @xmath239 the inequality @xmath240 is easy consequence , so for all @xmath241 we have @xmath242 . since @xmath238 equals @xmath243 for some @xmath241 by construction , we have that for all @xmath31 01 @xmath244 @xmath245 is provable in pa . that is , @xmath246 is provably chaitin - random . from theorem [ thm : chaitin ] we deduce that pa can prove the implication : `` if @xmath0 is a provably universal machine , then @xmath68 is chaitin - random . '' we know that every c.e . and random real is the halting probability of a universal machine , but we need more : _ can any c.e . and random real be represented as the halting probability of a provably universal machine ? _ first we have to check whether every universal machine is provably universal . [ thm : nonprovunivmachine ] there exist a provably universal machine and a universal machine that is not provably universal . the set of all provably prefix - free machines is c.e . , so if @xmath247 is a computably enumeration of provably prefix - free machines , then the machine @xmath0 defined by @xmath248 is a provably universal machine . let @xmath249 be a c.e . enumeration of all primitive recursive functions @xmath250 and @xmath251 a c.e . enumeration of all machines . fix a universal machine @xmath0 and consider the computable function @xmath252 such that : @xmath253 for every @xmath110 , @xmath254 is a universal machine iff @xmath255 is infinite ( if @xmath255 is finite , then so is @xmath254 ) . since the set of all indices of primitive recursive functions with infinite range is not c.e.it follows that there is an @xmath110 such that pa can not prove that @xmath256 is universal . theorem [ thm : nonprovunivmachine ] does not imply a negative answer for the previous question ; in fact , corollary [ cor : secondrep ] shows that the answer is affirmative . theorem [ thm : nonprovunivmachine ] produces examples of true and unprovable ( in pa ) statements of the form `` @xmath8 is universal '' . in this section we sharpen theorem [ representation : thm ] by proving that a real is provably c.e . and chaitin - random iff it is provable that the real is the halting probability of a provably universal machine . 01 [ lem : repce ] let @xmath5 be a c.e . real defined by the increasing primitive recursive sequence @xmath156 of rationals . then there is a primitive recursive sequence @xmath157 of natural numbers such that pa proves @xmath257 without loss of generality assume @xmath159 . define the primitive recursive sequences @xmath160 and @xmath161 by @xmath162 and for @xmath163 by @xmath258 since @xmath165 is strictly increasing we can establish by induction @xmath166 for all @xmath110 , making the logarithm well - defined . by construction we have @xmath259 define @xmath168 . since @xmath260 we have @xmath261 taking the limit we see that @xmath171 establishing our result . let @xmath5 be a c.e . real defined by the increasing primitive recursive sequence @xmath156 of rationals . then there is a machine @xmath23 such that pa proves that @xmath172 . according to solovay @xcite a c.e . real @xmath5 _ solovay dominates _ a c.e . real @xmath262 ( we write @xmath263 ) if there are two computable , increasing sequences @xmath264 and @xmath265 of rationals and a constant @xmath85 with @xmath266 , @xmath267 , and @xmath268 , for all @xmath31 . for c.e . reals @xmath269 , pa proves @xmath263 if there are two primitive recursive , increasing sequences @xmath264 and @xmath265 of rationals and a constant @xmath85 such that pa proves @xmath266 , @xmath267 , and @xmath268 , for all @xmath31 . [ thm : dominates ] if @xmath5 is c.e . and provably ml - random , and @xmath262 is c.e . , then @xmath270 is provable in pa . let @xmath264 and @xmath265 be primitive recursive sequences of rationals with limits @xmath5 and @xmath262 respectively . let @xmath271 . for each @xmath31 , for @xmath163 if @xmath272 $ ] then define @xmath273 = [ a_i , a_i + 2^{-n } ( b_i - b_{s^{n } } ) ) , $ ] where @xmath274 \neq \emptyset\}$ ] is the most recent non - empty stage , or @xmath275 if this is the first non - empty stage . otherwise define @xmath276 = \emptyset$ ] . let @xmath277 denote the @xmath219th non - empty stage , wherever that is well - defined , and let @xmath278 . observe that @xmath279 = \bigcup_{j\ge1 } [ a_{s^{n}_j } , a_{s^{n}_j } + 2^{-n } ( b_{s^{n}_j } - b_{s^{n}_{j-1}}))\ ] ] and that all the sets in the above union are disjoint by construction . as a result @xmath280 so pa proves that @xmath281 is a ml - test . because @xmath5 is provably ml - random , pa proves that there exists an @xmath282 such that @xmath283 , so for all @xmath284 we know that @xmath285 is well - defined . by construction we have the inequality @xmath286 which implies that @xmath287 defining @xmath288 and @xmath289 , we have for all @xmath284 that @xmath290 where @xmath291 and @xmath292 are primitive recursive sequences of rationals which provably converge to @xmath5 and @xmath262 respectively . so pa proves @xmath270 . 01 * comment * with the notation @xmath293 as below , let @xmath294 = [ a'_i , a'_i + 2^{-n } ( b'_{i } - b'_{i-1})),\ ] ] and observe that @xmath295 $ ] implies that @xmath296 which contradicts the construction of @xmath297 $ ] by virtue of which for all @xmath298 we have @xmath299 $ ] , hence @xmath300 and @xmath301 . consequently , @xmath302 , so @xmath270 with constant @xmath303 . [ cor : dominates ] if @xmath5 is c.e . and provably chaitin - random and @xmath262 is c.e . , then @xmath270 is provable in pa . see theorems [ thm : chaitintoml ] and [ thm : dominates ] . 01 [ cor : dominates ] if @xmath12 is c.e . , ml - random , and @xmath304 is c.e . , then @xmath270 . [ thm : omegarep ] suppose @xmath8 is a provably universal machine , @xmath5 is c.e . , and @xmath305 is provable in pa . then there exists a provably universal machine @xmath0 such that @xmath306 is provable in pa . since @xmath305 , there exist primitive recursive increasing sequences @xmath264 and @xmath265 of rationals , with limits @xmath5 and @xmath307 respectively , and a constant @xmath207 such that for all @xmath31 @xmath308 define @xmath271 . form the real @xmath309 by equation ( [ th2:ss ] ) the terms of the sum are positive , so @xmath310 is c.e . applying lemma [ lem : repce ] to @xmath311 we get a primitive recursive sequence @xmath312 of natural numbers such that @xmath313 let @xmath314 be a 1 - 1 primitive recursive enumeration of @xmath315 , and define the sequence of requests @xmath316 where @xmath224 is an arbitrarily fixed element in @xmath315 . by theorem [ thm : kc ] we get a provably prefix - free machine @xmath23 and a primitive recursive enumeration @xmath117 of @xmath118 such that the following three statements are provable : 1 ) @xmath119 , 2 ) @xmath120 for all @xmath110 , 3 ) @xmath122 for all @xmath110 . consider the machine @xmath317 . the machine @xmath0 is provably universal . indeed , @xmath318 and @xmath319 , by construction of @xmath23 . finally , it is provable that @xmath320 using all results above we obtain : [ thm : representation ] a c.e . real @xmath5 is provably chaitin - random iff it is provable that @xmath7 for some provably universal machine @xmath0 . suppose @xmath5 is provably c.e . and chaitin - random . by theorem [ thm : chaitintoml ] , @xmath5 it is provably ml - random . take a provably universal machine @xmath8 ( theorem [ thm : nonprovunivmachine ] ) . from theorem [ thm : dominates ] we see that @xmath305 is provable in pa . by theorem [ thm : omegarep ] we effectively get a @xmath0 which is provably universal and prefix - free such that @xmath7 is provable in pa . the converse is exactly theorem [ thm : chaitin ] . [ cor : provchatinrandimpliesrand ] every provably c.e . and chaitin - random real is provably random . if @xmath5 is provably chaitin - random and c.e . then by theorem [ thm : representation ] , @xmath7 for some provably universal machine @xmath0 , so @xmath5 satisfies solovay s formula ( [ solovayformula ] ) with @xmath321 . this section proves its title . we start with the following result by solovay @xcite : [ lem : solovayrepresentation ] let @xmath8 be a universal machine . if @xmath5 is c.e . and ml - random , then there exists an integer @xmath207 and a c.e . real @xmath208 such that ( [ solovayformula ] ) is satisfied . 01 @xmath322 using the proof of theorem [ thm : dominates ] , we deduce that @xmath323 ( because @xmath5 is c.e . and ml - random ) . consequently , we can consider the primitive recursive increasing sequences @xmath264 and @xmath265 of rationals , with @xmath271 and converging to @xmath5 and @xmath307 respectively , and a constant @xmath207 such that for all @xmath31 , @xmath324 the c.e . real @xmath325 is positive and @xmath326 it is not difficult to see that the converse implication in lemma [ lem : solovayrepresentation ] is also true . in fact , a sharper result can be proved : [ thm : provablyrandom ] let @xmath8 be provably universal , @xmath207 be an integer , @xmath311 a positive c.e . then @xmath327 is provably chaitin - random ( ml - random ) . let @xmath328 be a primitive recursive enumeration of the domain of @xmath8 and @xmath329 be a primitive recursive increasing sequence with limit @xmath311 . the sequence of rationals @xmath330 is primitive recursive , increasing and converges to @xmath5 . take @xmath331 and observe that for all @xmath31 , @xmath332 hence pa proves that @xmath333 . using theorem [ thm : omegarep ] we can find a provably universal machine @xmath0 such that @xmath306 is provable in pa . by theorem [ thm : chaitin ] , @xmath5 is provably chaitin - random and by theorem [ thm : chaitintoml ] , @xmath5 is provably ml - random . we can now state our main result : [ thm : cerandprov ] every c.e . and random real is provably c.e . and chaitin - random ( ml - random ) , hence provably c.e . and random . start with a provably universal machine @xmath8 ( theorem [ thm : nonprovunivmachine ] ) . by lemma [ lem : solovayrepresentation ] there exist @xmath85 and @xmath311 defining the representation ( [ solovayformula ] ) for @xmath5 : @xmath334 since @xmath8 is provably universal , theorem [ thm : provablyrandom ] shows that that @xmath335 is provably chaitin - random ( ml - random ) . therefore @xmath5 is provably chaitin - random ( ml - random ) . finally use corollary [ cor : provchatinrandimpliesrand ] to deduce that @xmath5 is provably random . theorem [ thm : representation ] can now be stated in the form : [ thm : representationfinal ] a real @xmath5 is provably c.e . and random iff it is provable that @xmath7 for some provably universal machine @xmath0 . use theorem [ thm : cerandprov ] and corollary [ cor : provchatinrandimpliesrand ] . [ cor : uu ] for every universal machine @xmath0 there exists a provably universal machine @xmath336 such that @xmath337 . since @xmath68 is c.e . and random , by theorem [ thm : cerandprov ] we deduce that @xmath68 is provably chaitin - random , so by theorem [ thm : representation ] we get a provably universal machine @xmath336 such that @xmath337 . [ cor : secondrep ] every c.e . and random real can be written as the halting probability of a provably universal machine . use theorems [ thm : cerandprov ] and [ thm : representationfinal ] . from the previous two sections we know that every c.e . random real can be written as the halting probability of a provably universal machine , so it is provable random . does there exist a universal machine whose halting probability is not provable random ? by theorem [ thm : chaitin ] such a machine should not be provably universal ( and such machines exist by theorem [ thm : nonprovunivmachine ] ) . we answer in the affirmative this question . to this aim we fix an effective enumeration of all c.e . reals in ( 0,1 ) @xmath338 ( for example , by enumerating all increasing primitive recursive sequences of rationals in ( 0,1 ) ) and define the set @xmath339 . a set @xmath340 is called c.e . if the set @xmath341 is c.e . note that in @xmath70 we enumerate all indices for all elements in @xmath70 . [ upp ] @xcite if @xmath340 is c.e . , then for all c.e . reals @xmath342 and @xmath343 we have @xmath344 . let @xmath345 be a c.e . not computable set of natural numbers enumerated by a primitive recursive function @xmath346 , and for each @xmath31 let @xmath347 be a primitive recursive increasing sequence of rationals in ( 0,1 ) such that @xmath348 . let @xmath349 and define the function @xmath350 because @xmath343 there exists a natural @xmath351 such that @xmath352 . if @xmath353 , then there exists an @xmath282 such that @xmath354 , hence @xmath355 , for @xmath356 , so @xmath357 . if @xmath358 , then for all @xmath110 , @xmath359 , so @xmath360 . because of the uniform definition of @xmath361 we can construct a computable function @xmath61 such that @xmath362 . finally , let s assume by absurdity that @xmath363 . the set @xmath364 is c.e . because @xmath70 is c.e . , but in view of the definition of @xmath365 , @xmath366 , a non c.e . let @xmath367 be a c.e . enumeration of all universal machines . consider now the sets @xmath368 and @xmath369 by enumerating proofs in pa we deduce that @xmath370 is c.e . , so @xmath371 , for some primitive recursive function @xmath61 . we have : @xmath372 . is @xmath373 c.e . ? the answer is negative : [ thm : haltprobnotprov ] there exists a universal machine @xmath374 such that @xmath375 for all @xmath110 . take a universal machine @xmath0 such that @xmath376 and construct the c.e . real @xmath377 , where @xmath378 . as @xmath379 and @xmath262 is not random , @xmath380 , for all @xmath110 , so by lemma [ upp ] , @xmath381 is not c.e . , hence the theorem is proved . * comment * there is no contradiction between corollary [ cor : uu ] and theorem [ thm : haltprobnotprov ] : there exist a universal machine @xmath374 and a provably universal machine @xmath382 such that @xmath383 and @xmath384 for all @xmath110 : pa can not prove the randomness of @xmath385 based on @xmath374 , but can prove the randomness of @xmath383 based on @xmath382 . in the above we gave proofs that various statements , once suitably formalised in the language of first order logic , were derivable from the axioms of pa . in principle , but for lack of space and patience , we could have presented complete pa derivations of each statement proved . instead , as is common practice for all but the simplest of results , we sketched constructions which leave the actual derivations implicit . recent advances in theorem proving computer programs , such as the proof assistant isabelle @xcite , have allowed complete formal derivations of nontrivial mathematical results . in such systems , humans write a sequence of proof commands , and the computer system searches for a complete derivation , if one exists . essentially , the human user gives a sequence of intermediate lemma with proof directions , and the computer interpolates the full derivation . ( for a recent perspective on the importance of formalising mathematics see @xcite . ) using isabelle , we formalised and proved the kraft - chaitin theorem ( theorem [ thm : kc ] ) , a key result in our above proof . to keep our presentation self - contained , we begin by showing how to formalise and prove a simple result about strings ; for a full introduction to the isabelle system see @xcite . we follow with a formalisation of the kraft - chaitin theorem , then sketch its formal proof . the full proof script is available online @xcite . to illustrate isabelle and its use , we will formalise and prove the following simple property of strings : given strings @xmath386 , if @xmath98 extends @xmath232 then @xmath387 extends @xmath232 . strings are naturally represented by the isabelle list data - type . here represents the empty list , and y#ys represents the list formed by concatenating the element y with the list ys . for example , the string @xmath388 is represented by 0 # 0 # 1 # [ ] ( or for short ) . the following code inductively defines whether the list a extends b , denoted extends a b : .... fun extends : : " ' a list = > ' a list = > bool " where " extends [ ] [ ] = true " .... when faced with the above definition , isabelle automatically proves termination ( in this case , by observing that the first argument always decreases in length with each recursive call ) . let us first prove that any list extends the empty list . we enter into isabelle : .... lemma extends1 : " extends a [ ] " .... it responds with the propositions we need to prove : .... goal ( 1 subgoal ) : 1 . extends a [ ] .... it is natural to prove this by induction on a , by entering the command ` apply(induct a ) ` . this results in two proof obligations , one for the base case and the other for the inductive step : .... goal ( 2 subgoals ) : 1 . extends [ ] [ ] 2 . ! ! a a. extends a [ ] = = > extends ( a # a ) [ ] .... the first proposition is one of the cases in our definition of extend . in the second ! ! denotes universal quantification and this similarly follows from one of our definition cases . we tell isabelle to simplify these expressions with the command ` apply(simp_all ) ` . isabelle manages to simplify all these expressions down to true , using rewrite rules for simplifying conjunctions , variable identity , and expanding the definition of extends . as a result we get : .... goal : no subgoals ! .... having completed the proof , we compactly store it in the following format : .... lemma extends1 : " extends a [ ] " apply(induct a ) apply(simp_all ) done .... we can now attempt our original goal : .... lemma extends2 : " extends ( a@b ) a " goal ( 1 subgoal ) : 1 . extends ( a @ b ) a .... the concatenation of lists a and b is denoted a @ b. we again induct with the command ` apply(induct a ) ` , then simplify with the command ` apply(simp_all ) ` .... goal ( 2 subgoals ) : 1 . extends ( [ ] @ b ) [ ] 2 . ! ! a a. extends ( a @ b ) a = = > extends ( ( a # a ) @ b ) ( a # a ) goal ( 1 subgoal ) : 1 . extends b [ ] .... since we proved this before , we use the command ` apply(simp only : extends1 ) ` to reuse our previous result , completing the proof . in sum : .... lemma extends2 : " extends ( a@b ) a " apply(induct a ) apply(simp_all ) apply(simp only : extends1 ) done .... 01 [ thm : kc2 ] suppose @xmath115 is a primitive recursive enumeration of requests such that @xmath389 then there exists a provably prefix - free machine @xmath23 and a primitive recursive enumeration @xmath117 of @xmath118 such that the following is provable in pa : 1 . @xmath119 , 2 . @xmath120 for all @xmath121 , 3 . @xmath122 for all @xmath121 . @xmath124 , @xmath126 , @xmath127 , @xmath128 . let @xmath129 be the longest element of @xmath103 of length at most @xmath125 . if no such string exists , terminate . @xmath131 . define @xmath132 . @xmath135 . the proof of the kraft - chaitin theorem is algorithmic : it describes a particular algorithm ( algorithm [ alg : kc ] of theorem [ thm : kc ] ) for selecting strings of the required lengths , and proves that the algorithm is correct . in what follows we will implement this algorithm in isabelle and will prove its correctness . the following isabelle code implements algorithm [ alg : kc ] . we give the definition of each function , then explain what it does . .... fun extend : : " nat list = > nat = > nat list list " where " extend l 0 = [ l ] " # tl ( extend l n ) " .... for l a binary list representing a binary string , and n a natural number , extend l n computes the list @xmath390.\ ] ] for example , in isabelle the expression extend [ 0,0,1 ] 5 evaluates to this corresponds to the set @xmath391 of binary strings . the set of unallocated prefixes @xmath123 and the set of allocated strings @xmath140 are represented by lists of strings . the free prefixes are ordered by decreasing length , the allocated strings by the order of allocation . consider one iteration of the main loop . let a be the list of previously allocated strings , f the list of free prefixes , and n the length of the string we want to allocate at this step . ( these are denoted @xmath140 , @xmath123 , and @xmath125 in the original algorithm . ) kcstep a f n returns the updated pair of allocated strings and free prefixes ( @xmath153 and @xmath392 ) . .... consts kcstep : : " nat list list = > nat list list = > nat = > ( nat list list * nat list list ) " primrec " kcstep a [ ] n = ( a , [ ] ) " ( * fail case * ) " kcstep a ( f # f ) n = ( if length f < = n then ( ( hd ( extend f ( n - length f ) ) ) # a , ( tl ( extend f ( n - length f ) ) ) @ f ) else ( fst ( kcstep a f n ) , f # snd ( kcstep a f n ) ) ) " .... kcstep searches through the list f of free prefixes for the longest string of length at most n. one it finds it , it calls extend , which returns a list of extended prefixes . it takes the first string in the list , guaranteed to have exactly length n , and adds it to the allocated strings list . the rest of the strings are placed on the free prefixes list . for example , kcstep [ ] [ [ ] ] 2 evaluates to ( [ [ 0,0 ] ] , [ [ 0,1 ] , [ 1 ] ] ) which corresponds to the list @xmath393 of allocated strings and the set @xmath394 of free prefixes . .... consts kcloop : : " nat list = > ( nat list list * nat list list ) = > ( nat list list * nat list list ) " primrec " kcloop [ ] x = x " " kcloop ( l#ls ) x = ( kcstep ( fst ( kcloop ls x ) ) ( snd ( kcloop ls x ) ) l ) " .... for a list of lengths l and a pair ( a , f ) of allocated strings and free prefixes , kcloop l ( a , f ) runs kcstep to allocate strings for every length in l. for example , kcloop [ 3,4,2 ] ( [ ] , [ [ ] ] ) allocates a string of length 2 , then one of length 4 , then one of length 3 , starting from the initial state where no strings are yet allocated ( ) and the empty string is our free prefix ( ) . for example , kcloop [ 3 , 2 ] ( [ ] , [ [ ] ] ) evaluates to ( [ [ 0,1,0 ] , [ 0,0 ] ] , [ [ 0,1,1 ] , [ 1 ] ] ) which corresponds to the list @xmath395 of allocated strings ( note that we reverse the list ) , and the set @xmath396 of free prefixes . .... fun kc : : " nat list = > nat list list " where " kc ls = ( fst ( kcloop ls ( [ ] , [ [ ] ] ) ) ) " .... for a list of lengths l , kc l returns the list of strings allocated by running kcloop on the list starting from the initial state where no strings have been allocated . for example , kc [ 4,3,2 ] evaluates to which corresponds to the sequence @xmath397 of allocated strings . this implements kraft - chaitin s algorithm , for we will prove that : 1 . if our list of lengths obeys kraft s inequality , @xmath398 , then kc ls is a list of strings , and the @xmath110th element of kc ls has length equal to the @xmath110th element of ls . kc ls is always a prefix - free list ( no two distinct elements of the list are prefixes of each other ) . if we add new lengths to the start of ls , then this adds new strings to the end of kc ls without changing the old ones . that is , once a string of a given length is allocated it is not changed . to prove the above we need to define what a prefix - free list is , a function to evaluate kraft s inequality , a function which checks whether the lengths of one list match the lengths in another , and a tool to check whether one list extends another . .... fun prefixes : : " nat list = > nat list = > bool " where " prefixes [ ] x = true " consts incomparable : : " nat list = > nat list list = > bool " primrec " incomparable x [ ] = true " " incomparable x ( y # ys ) = ( ~(prefixes x y ) & ( incomparable x ys ) ) " consts prefixfree : : " nat list list = > bool " primrec " prefixfree [ ] = true " " prefixfree ( x # xs ) = ( ( incomparable x xs ) & ( prefixfree xs ) ) " .... if x is a prefix of y , or vice versa , then prefixes x y. for example prefixes [ 0,0,1 ] [ 0,0 ] is true . incomparable x a holds if x is not a prefix of any string in a , for instance incomparable [ 0,0 ] [ [ 1,0 ] , [ 1,1,1 ] ] holds . prefixfree l holds if the list l is prefix - free , for instance prefixfree [ [ 0,0 ] , [ 1,0 ] , [ 1,1,1 ] ] holds . .... consts expn2 : : " nat = > rat " primrec " expn2 0 = 1 " " expn2 ( suc n ) = ( 1/2 ) * expn2 n " consts meas_nat : : " nat list = > rat " primrec " meas_nat [ ] = 0 " " meas_nat ( f#f ) = ( expn2 f + meas_nat f ) " .... we define expn2 n equal to @xmath399 . meas_nat f computes the `` measure '' of a sequence of natural numbers f , for example meas_nat [ 4,3,2 ] equals 7/16 . .... fun lengthsmatch : : " nat list list = > nat list = > bool " where " lengthsmatch [ ] [ ] = true " .... the expression lengthsmatch x y holds if the lengths of each string in x matches the corresponding number in y. for example , we have lengthsmatch [ [ 0,0 ] , [ 1,0 ] , [ 1,1,1 ] ] [ 2,2,3 ] is true . .... fun extends : : " ' a list = > ' a list = > bool " where " extends [ ] [ ] = true " .... finally , extends a b holds if the list a extends the list b , so extends [ 0,1 ] [ 0 ] holds . with the above definitions we can state the three results which establish correctness : .... theorem kc_correct1 : " meas_nat ls < = 1 = = > lengthsmatch ( kc ls ) ls " theorem kc_correct2 : " prefixfree ( kc ls ) " theorem kc_extend : " extends ( rev ( kc ( l2 @ l1 ) ) ) ( rev ( kc l1 ) ) " .... the first says that if ls is a list of natural numbers @xmath125 which satisfies kraft s inequality @xmath400 , then the strings kc ls allocated by running algorithm [ alg : kc ] on this list have exactly the lengths ls we asked for . the second says the strings allocated are prefix - free . the last says that when algorithm [ alg : kc ] allocates additional strings it does not change strings it has previously allocated . to see this , note that when we run kc l the algorithm allocates strings starting from the end of the list l. this means , the first element of kc l is the last string allocated . kc ( l2 @ l1 ) is the list of strings allocated if we allocate strings with lengths in l1 then strings with lengths in l2 . together , establishing these would show that the kc algorithm constructively establishes the kraft - chaitin theorem . all the above merely formalised the algorithm and stated the theorem we wish isabelle to prove . this gets the order mixed slightly , since formalising this theorem unearthed a mistake in the algorithm , so the process was mutual . in some sense this formalisation of the theorem is the major creative work , the rest is just technical detail . as one might guess , however , most of the work is in these details . to prove the above theorems we must guide isabelle to them by establishing numerous intermediate lemmas , and telling isabelle which proof techniques to use to establish each . often we just advise isabelle to induct on a variable then simplify , but sometimes we must give more detailed guidance . the isabelle proof follows the proof given for theorem [ thm : kc ] : we establish that the inner loop preserves some invariants , and use these invariants to establish correctness . recall the algorithm has two variables : the list of allocated strings and the list of free strings . each pass through the loop will ( potentially ) add one new allocated string , and modify the free strings . we then show that these two lists combined remain prefix - free , their joint measure never decreases , and that there are never two free strings of the same length . for reasons of space we give only the definitions required to state the above intermediate results and show how they are formalised in isabelle . the proof in its entirety is available online @xcite . .... fun strictlysorted : : " nat list list = > bool " where " strictlysorted [ ] = true " & ( strictlysorted ( x2 # xs ) ) ) " .... strictlysorted l holds if the strings in l are ordered by ( strictly ) decreasing length . in particular , this means there can be no two strings of the same length in l. .... fun inv1 : : " nat list list * nat list list = > bool " where " inv1 x = strictlysorted ( snd x ) " fun inv2 : : " nat list list * nat list list = > bool " where " inv2 x = prefixfree ( ( fst x ) @ ( snd x ) ) " .... the first invariant is that the list of free strings is strictly sorted . this is needed to show both that there is at most one string of any given length and to show that the algorithm will always select the longest string it is able to . .... fun inv : : " nat list list * nat list list = > bool " where " inv x = ( ( inv1 x ) & ( inv2 x ) ) " theorem kcstep_inv : " inv ( a , f ) = = > inv ( kcstep a f n ) " .... this says simply that if the invariants held of the variables before running through the loop once , then they hold afterwards . .... consts meas : : " nat list list = > rat " ( * the measure of a prefix free set * ) primrec " meas [ ] = 0 " " meas ( x # xs ) = expn2 ( length x ) + meas xs " .... this defines the measure of a list of strings : the usual @xmath401 . .... lemma kcstep_meas : " meas ( ( fst ( kcstep a f n ) ) @ ( snd ( kcstep a f n ) ) ) = meas ( a@f ) " .... this says that measure is preserved at each step of the loop . this measure will be 1 for all the intermediate states of the kc algorithm , but we need this more general result for the inductive proofs to work . a number of further intermediate results are required both to establish the above invariants and to apply them to the main theorems . below are three of the most important , which one may recall from the proof of theorem [ thm : kc ] ( in total , there are 102 theorems and lemmas proved ) . .... theorem meas_alloc : " [ \| expn2 n < = meas f ; strictlysorted f \| ] = = > length ( last f ) < = n " lemma kcstep_correct1 : " [ \|inv ( a , f ) ; expn2 n < = meas f\| ] = = > ( tl ( fst ( kcstep a f n ) ) = a ) & ( length ( hd ( fst ( kcstep a f n ) ) ) = n ) " lemma kcstep_correct2 : " [ \|inv ( a , f ) ; expn2 n < = meas f\| ] = = > meas ( fst ( kcstep a f n ) ) = meas a + expn2 n " .... theorem meas_alloc formalises the result that if @xmath399 is smaller than the measure of a set @xmath402 , and that set has no two strings of the same length , then there is a string of length at most @xmath31 in @xmath402 . lemma kcstep_correct1 says that if the invariants are satisfied by the current variables a and f , and the measure of f is at least @xmath399 , then kcstep succeeds . this means that we allocate one new string of length exactly @xmath31 , leaving the old strings untouched . lemma kcstep_correct2 expresses an implied result : if the algorithm succeeds , then the measure of the list of allocated strings increases by exactly @xmath399 ( isabelle will often not notice conclusions that seem obvious to the prover ; they must be spelt out ) . if pa receives an algorithm for a machine @xmath8 , a proof that @xmath8 is universal and prefix - free , an integer @xmath207 , and a computable increasing sequence of rationals converging to a real @xmath113 , then pa can prove that @xmath209 is c.e . and random . similarly , if pa receives an algorithm for a machine @xmath0 , a proof that @xmath0 is universal and prefix - free , then it can prove that @xmath1 is c.e . and random . this implies that every c.e . random real is provably c.e . and random as stated in solovay s email @xcite . we have offered two representations for c.e . and random reals from which pa can prove that the real is c.e . and random . in the first we fix a provably universal prefix - free machine @xmath8 and we vary the integer @xmath207 and the c.e . real @xmath113 to get via the formula @xmath403 all c.e . and random reals . in the second we vary all provably universal prefix - free machines @xmath0 to get via @xmath68 all c.e . and random reals . a key result was to show that the theorem that `` a real is c.e . and random iff it is the halting probability of a universal machine '' @xcite can be proved in pa . our proof , which is simpler than the standard one , can be used also for the original theorem . we proved two negative results showing the importance of the representation for provability of randomness : a ) there exists a universal machine whose universality can not be proved in pa , b ) there exists a universal machine @xmath0 such that , based on @xmath0 , pa can not prove the randomness of @xmath68 . chaitin @xcite explicitly computed a constant @xmath85 such that if @xmath404 is larger than the size in bits of the program for enumerating the theorems of pa plus @xmath85 , then pa can not prove that a specific string @xmath98 has complexity greater than @xmath404 , @xmath405 . consequently , pa can not prove randomness of almost all random ( finite ) strings . our positive result shows an interesting difference between the finite and the infinite cases of ( algorithmic ) randomness . does our positive result contradict chaitin and solovay s negative results discussed in the introduction ? the answer is negative because the digits of the binary expansion of a random c.e . real are not computable . our positive result would not be satisfactory without demonstrating our proofs with an automatic theorem prover . we have chosen isabelle @xcite to obtain an automatic proof of our version of the kraft - chaitin theorem , one of the key results of this paper . the paper contains a description of the formalisation ( for isabelle ) of the kraft - chaitin theorem and the description of the main steps of the automatic proof ; the full proof is available online @xcite . finally we speculate about the role of the automatic prover . how can an automatic theorem prover help understanding / proving a mathematical statement ? there are at least three possibilities . a ) use the prover to verify the theorem by discovering a proof , call it `` solovay mode '' ( because this corresponds to the result reported in this paper : bob solovay communicated to one of us the statement to be proved and we found a proof ) . it is worth observing that the kraft - chaitin theorem has two `` roles '' : one , as an algorithm , to be executed , the other , as a mathematical statement , to be proved . previous formalisation efforts focused only on the first part ; our present work was directed towards the second . one could imagine that mathematical journals might use such systems in the process of refereeing @xcite . b ) the second possibility is to use the prover to verify a human - made proof a full isabelle proof for all results in this paper is under construction . c ) the third possibility is to use the prover as some kind of `` assistant '' in an interactive process of discovery / proving . during the work to automate the proof of the kraft - chaitin theorem a mistake in our human - made argument was unearthed and corrected . we also used the experience with isabelle to test the adequacy of the representation of a c.e . random real in meeting the goal : to obtain the pa proof of randomness . we thank bob solovay for suggesting the result of this paper and useful comments , jeremy dawson for helpful advice on the isabelle proof , and greg chaitin , liam fearnley , bruno grenet , mathieu hoyrup , andr nies , cristobal rojas , frank stephan , garry tee and the anonymous referee for useful comments which improved our paper . 99 w. s. brainerd , l. h. landweber . _ theory of computation _ , wiley , new york , 1974 . c. s. calude . chaitin @xmath246 numbers , solovay machines and incompleteness , _ theoret . comput . _ 284 ( 2002 ) , 269277 . c. s. calude . a characterization of c.e random reals , _ theoret . _ 271 ( 2002 ) , 314 . c. s. calude _ information and randomness . an algorithmic perspective _ , 2nd edition , revised and extended , springer verlag , berlin , 2002 . c. s. calude , p. hertling , b. khoussainov , and y. wang . recursively enumerable reals and chaitin @xmath246 numbers , in : m. morvan , c. meinel , d. krob ( eds . ) , _ proceedings of the 15th symposium on theoretical aspects of computer science ( paris ) _ , springer verlag , berlin , 1998 , 596606 . full paper in _ theoret . _ 255 ( 2001 ) , 125149 . r. m. solovay . a version of @xmath246 for which zfc can not predict a single bit , in c.s . calude , g. pun ( eds . ) . _ finite versus infinite . contributions to an eternal dilemma _ , springer - verlag , london , 2000 , 323334 . r. m. solovay . _ draft of a paper ( or series of papers ) on chaitin s work done for the most part during the period of sept.dec . 1974 _ , unpublished manuscript , ibm thomas j. watson research center , yorktown heights , new york , may 1975 , 215 pp .
the near - infrared galaxy luminosity function is an important characteristic of the local galaxy population . it is a much better tracer of evolved stars , and hence of the total stellar content of galaxies , than optical luminosity functions which can be dominated by young stellar populations and are also strongly affected by dust extinction . hence , infrared luminosities can be much more directly related to the underlying stellar mass of galaxies and so knowledge of the present form and evolution of the infrared galaxy luminosity function places strong constraints on the history of star formation in the universe and on galaxy formation models ( cole and references therein ) . the local k - band luminosity function has been estimated from optically selected samples by mobasher , sharples and ellis ( ) , szokoly ( ) and loveday ( ) and from k - band surveys by glazebrook ( ) , and gardner ( ) . the existing k - band surveys are small . the largest , by gardner et al . , covers only 4 deg@xmath12 and contains only 510 galaxies . the recent survey of loveday covers a much larger solid angle . in this survey the redshifts were known in advance of measuring the k - band magnitudes and this was exploited by targetting bright and faint galaxies resulting in an effective sample size much larger than the 345 galaxies actually measured . however , like all optically selected samples , it suffers from the potential problem that galaxies with extremely red infrared to optical colours could be missed . in this paper we combine the 2-micron all sky survey ( ) with the 2df galaxy redshift survey ( ) to create an infrared selected redshift survey subtending deg@xmath12 . currently the sky coverage of both surveys is incomplete , but already the overlap has an effective area of deg@xmath12 . within this area the redshift survey is complete to the magnitude limit of the catalogue and so constitutes a complete survey which is 50 times larger than the previous largest published infrared selected redshift survey . a new catalogue of a similarly large area , also based on , has very recently been analysed by kochanek ( ) . they adopt isophotal rather than total magnitudes and concentrate on the dependence of the luminosity function on galaxy morphology . in section [ sec : selection ] we briefly describe the relevant properties of the and catalogues . section [ sec : matching ] is a detailed examination of the degree to which the matched galaxies are a complete and representative subset of the catalogue . section [ sec : mags ] examines the calibration of the total magnitudes and section [ sec : compl ] demonstrates that the catalogue and the inferred luminosity functions are not affected by surface brightness selection effects . in section [ sec : pop ] we present the method by which we compute k - corrections and evolutionary corrections and relate the observed luminosities to the underlying stellar mass . the estimation methods and normalization of the luminosity functions are described briefly in section [ sec : lf_est ] . our main results are presented and discussed in section [ sec : results ] . these include estimates of the j and ( k - short ) luminosity functions , the @xmath9 and j@xmath9 colour distributions as a function of absolute magnitude and the distribution of spectral type . we also estimate the stellar mass function of galaxies , which can be integrated to infer the fraction of baryons in the universe which are in the form of stars . we conclude in section [ sec : conc ] . the data that we analyze are the extended source catalogue from the second incremental release of the 2-micron all sky survey ( http://pegasus.phast.umass.edu ) and the galaxy catalogue of the 2df galaxy redshift survey ( http://www.mso.anu.edu.au/2dfgrs ) . here , we present the relevant properties of these two catalogues and investigate their selection characteristics and level of completeness . the is a ground - based , all - sky imaging survey in the j , h and bands . details of how extended sources are identified and their photometric properties measured are given by jarrett ( ) . the detection sensitivity ( 10@xmath13 ) for extended sources is quoted as 14.7 , 13.9 and 13.1 magnitudes in j , h and respectively . the complete survey is expected to contain 1 million galaxies of which approximately 580,000 are contained in the second incremental data release made public in march 2000 . the is selected in the photographic band from the apm galaxy survey ( @xcite , , ) and subsequent extensions to it , that include a region in the northern galactic cap ( maddox in preparation ) . the survey covers approximately deg@xmath12 consisting of two broad declination strips . the larger is centred on the sgp and approximately covers @xmath14@xmath15@xmath16@xmath15@xmath17 , @xmath18@xmath19@xmath20@xmath19@xmath21 ; the smaller strip is in the northern galactic cap and covers @xmath22@xmath15@xmath16@xmath15@xmath23 , @xmath24@xmath19@xmath20@xmath19@xmath25 . in addition , there are a number of randomly located circular 2-degree fields scattered across the full extent of the low extinction regions of the southern apm galaxy survey . there are some gaps in the sky coverage within these boundaries due to small regions that have been excluded around bright stars and satellite trails . the aims to measure the redshifts of all the galaxies within these boundaries with extinction - corrected magnitudes brighter than 19.45 . when complete , at the end of 2001 , 250,000 galaxy redshifts will have been measured . in this paper we use the redshifts obtained prior to . the overlap of the two surveys is very good . there are some gaps in the sky coverage due to strips of the sky that were not included in the second incremental release , but overall a substantial fraction of the deg@xmath12 of the is covered by . the homogeneity and extensive sky coverage of the combined dataset make it ideal for studies of the statistical properties of the galaxy population . = 8 truecm here we consider whether all the galaxies within the survey region have counterparts and assess the extent to which the fraction of galaxies with measured redshifts represents an unbiased sub - sample . the astrometry in both and is , in general , very good and it is an easy matter to match objects in the two catalogues . we choose to find the closest pairs within a search radius equal to three quarters of the semi - major axis of the j - band image ( denoted j_r_e in the database ) . scaling the search radius in this way helps with the matching of large extended objects . this procedure results in the identification of counterparts for of the objects , when at random one would only expect to find a handful of such close pairs . moreover , the distribution of separations shown in fig . [ fig : seps ] peaks at 0.5 arcsec , with only 3% having separations greater than 3 arcsec . a significant part of this tail comes from the most extended objects as is evident from the dotted histogram in fig . [ fig : seps ] which shows objects with semi - major axes larger than 12 arcsec . thus , we can be very confident in these identifications . the objects for which we have found secure counterparts amount to % of the extended sources that fall within the boundary of the . as discussed below , a more restrictive criterion that includes only sources fainter than j@xmath26@xmath27 that are confidently classified as galaxies by , increases the fraction with matches to % . the remaining % are missed for well understood reasons ( star - galaxy classification : % ; merged or close images : % ; miscellaneous : % ) , none of which ought to introduce a bias . this is confirmed explicitly , in the middle row of fig . [ fig : hists ] , by the close correspondence between the photometric properties of the missed % and those of the larger matched sample . hence , in estimating luminosity functions no significant bias will be introduced by assuming the matched sample to be representative of the full population . furthermore , the distribution shown in the bottom row of fig . [ fig : hists ] shows that the subset of galaxies for which we have measured redshifts is a random sample of the full matched catalogue . this summary is the result of a thorough investigation , which we describe in the remainder of this section , into the reasons why % of the sources are missed and whether their omission introduces a bias in the properties of the matched sample . = 15.0 truecm we first consider objects in the catalogue which based on their images and colours are not confidently classified as galaxies . in the database a high e_score or g_score indicates a high probability that the object is either not an extended source or not a galaxy . a cc_flag@xmath280 indicates an artifact or contaminated and/or confused source . for detailed definitions of these parameters we refer the reader to jarrett ( ) . rejecting all objects which have either e_score@xmath151.4 , g_score@xmath151.4 or cc_flag@xmath280 removes just % of the total . however , removing these reduces significantly the fraction of the sample that does not match with the catalogue , from % to % . thus , it is likely that about 30% of the objects which have e_score@xmath151.4 , g_score@xmath151.4 or cc_flag@xmath280 are not galaxies . the may contain a tail of very red objects that are too faint in the -band to be included in the @xmath1919.45 sample . [ fig : colours ] shows the distribution of -j colours for the matched objects with j@xmath1914.7 . ( here , the j - band magnitude we are using is the default magnitude denoted j_m in the database . in section [ sec : mags ] we will consider the issue of what magnitude definition is most appropriate for estimating the luminosity function . ) the vertical dashed line indicates the colour at which this sample starts to become incomplete due to the @xmath1919.45 magnitude limit of . the colour distribution cuts off sharply well before this limit , suggesting that any tail of missed very red objects is extremely small . in other words the is sufficiently deep that even the reddest objects detected at the faintest limits of ought to be detected in . in the top row of fig . [ fig : hists ] we compare the distributions of magnitude , colour and surface brightness for the matched and missed objects . in general , the properties of the missed subset overlap well with those of the much larger matched subset . however , we do see that the distributions for missed objects contain tails of bright and blue objects . it is quite likely that this is due to the extended source catalogue being contaminated by a small population of saturated or multiple stars . the dotted histograms in the top row of fig . [ fig : hists ] show the distributions of magnitude , colour and surface brightness for the bright subset of the missed objects with j@xmath1912 . here we clearly see bimodal colour and surface brightness distributions . the blue peak of the colour distribution is consistent with that expected for stars ( see jarrett ) . excluding these bright , j@xmath1912 , objects which are clearly contaminated by stars reduces the fraction of missed objects from % to % . the magnitude , colour and surface brightness distributions for this remaining % are shown in the middle panel of fig . [ fig : hists ] . we see that the missed objects are slightly under - represented at the faintest magnitudes and also slightly bluer on average than the matched sample , while the distribution of surface brightness is almost indistinguishable for the two sets of objects . these differences are small and so will introduce no significant bias in our luminosity function estimates . to elucidate the reasons for the remaining missed % of objects we downloaded 100 1@xmath291 arcmin images from the stsci digitized sky survey ( dss ) centred on the positions of a random sample of the missed objects . in each image we plotted a symbol to indicate the position of any galaxies within the 1@xmath291 arcmin field . we also plotted symbols to indicate the positions and classifications of all images identified in the apm scans from which the catalogue was drawn , down to a magnitude limit of @xmath30 . these images are classified as galaxies , stars , merged images ( galaxy+galaxy , galaxy+star or star+star ) or noise . this set of plots allows us to perform a census of the reasons why some objects are not present in the survey . = 8.5 truecm the main cause for the absence of objects in the is that the apm has classified these objects as stars . these amount to % of the missed sample ( % of the full sample ) . in some cases , the dss image shows clearly that these are stars and in others that they are galaxies . however , the majority of these objects can not easily be classified from the dss images . thus , they could be galaxies that the apm has falsely classified as stars or stars that has falsely classified as galaxies . the first possibility is not unexpected since the parameters used in the apm star - galaxy separation algorithm were chosen as a compromise between high completeness and low contamination such that the expected completeness is around 95% with 5% stellar contamination ( @xcite ) . it is hard to rule out the possibility that this class of object does not include a substantial fraction of stars , but if so , their presence appears not to distort the distribution of colours shown in fig [ fig : colours ] . another % of the random sample ( % of the full sample ) are classified by the apm as mergers or else consist of two close images in the dss but are classified by the apm as a single galaxy offset from the position . the remaining % of the random sample ( % of the full sample ) are missed for a variety of reasons including proximity to the diffraction spikes of very bright stars and poor astrometry caused by the presence of a neighbouring unclassified image . = 9.0 truecm = 18.0 truecm the extended source database provides a large selection of different magnitude measurements . in the previous section we used the default magnitudes ( denoted j_m and k_m in the database ) these are magnitudes defined within the same circular aperture in each waveband . for galaxies brighter than @xmath26@xmath31 , the aperture is the circular -band isophote of 20 mag arcsec@xmath32 and for galaxies fainter than @xmath2614 it is the circular j - band isophote of 21 mag arcsec@xmath32 . these are not the most useful definitions of magnitude for determining the galaxy luminosity function . since we are interested in measuring the total luminosity and ultimately the total stellar mass of each galaxy , we require a magnitude definition that better represents the total flux emitted by each galaxy . we consider kron magnitudes ( kron ) and extrapolated magnitudes . kron magnitudes ( denoted j_m_e and k_m_e in the database ) are measured within an aperture , the kron radius , defined as 2.5 times the intensity - weighted radius of the image . the extrapolated magnitudes ( denoted j_m_ext and k_m_ext in the database ) are defined by first fitting a modified exponential profile , @xmath33 $ ] , to the image from 10 arcsec to the 20 mag / arcsec@xmath12 isophotal radius , and extrapolating this from the kron radius to 4 times this radius or 80 arcsec if this is smaller ( jarrett private communication ) . note that improvements are being made to the extended source photometry algorithms developed and employed team and so in the final data release the definitions of the kron and extrapolated magnitudes may be slightly modified ( jarrett private communication ) . [ fig : mags ] compares the default , kron and extrapolated magnitudes in the j and bands for the matched catalogue . the upper panel shows that while the median offset between the j - band isophotal default magnitudes and the pseudo - total kron magnitudes is small there is a large spread with some galaxies having default magnitudes more than 0.5 magnitudes fainter than the kron magnitude . the kron magnitudes are systematically fainter than the extrapolated magnitudes by between approximately 0.1 and 0.3 magnitudes . this offset is rather larger than expected : if the kron radius is computed using a faint isophote to define the extent of the image from which the intensity weighted radius is measured , then the kron magnitudes should be very close to total . for an exponential light profile ( @xmath34 ) , the kron radius should capture 96% of the flux , while for an @xmath35 law ( @xmath36 ) , 90% of the flux should be enclosed . in other words , the kron magnitude should differ from the total magnitude by only 0.044 and 0.11 magnitudes in these two cases . however , the choice of isophote is a compromise between depth and statistical robustness . in the case of the second incremental release , an isophote of 21.7(20.0 ) mag arcsec@xmath32 in j ( ) was adopted ( jarrett private communication ) . these relatively bright isophotes , particularly the -band isophote , could lead to underestimates of the kron radii and fluxes for lower surface brightness objects and plausibly accounts for much of the median offset of 0.3 magnitudes seen in fig . [ fig : mags ] between the -band kron and extrapolated magnitudes . this line of reasoning favours adopting the extrapolated magnitudes as the best estimate of the total magnitudes , but , on the other hand , the extrapolated magnitudes are model - dependent and have larger measurement errors . to understand better the offset and scatter in the magnitudes we have compared a subset of the data with the independent k - band photometry of loveday ( 2000 ) . the pointed observations of loveday have better resolution than the 2mass images and good signal - to - noise to a much deeper isophote . this enables accurate , unbiased kron magnitudes to be measured . note that the offset between the -band and the standard k - band used by loveday is expected to be almost completely negligible ( see carpenter ) . the left hand panels of fig . [ fig : loveday ] compare these measurements with the corresponding kron and extrapolated magnitudes . the right hand panels show -band kron and extrapolated magnitudes computed by taking the j - band kron and extrapolated magnitudes and subtracting the j@xmath9 colour measured within the default aperture . these indirect estimates are interesting to consider as they combine the profile information from the deeper j - band image with the j@xmath9colour measured within the largest aperture in which there is good signal - to - noise . the straight lines plotted in fig . [ fig : loveday ] show simple least squares fits and the slope and zero - point offset of these fits are indicated on each panel along with bootstrap error estimates . also shown in the inset panels is the distribution of residual magnitude differences about each of the fits and a gaussian fit to this distribution . the rms of these residuals and a bootstrap error estimate is also given in each panel . from these comparisons we first see that all the fits have slopes entirely consistent with unity , but that their zeropoints and scatters vary . the zero - point offsets , @xmath37 , between both the kron magnitude measurements and those of loveday confirm that the kron magnitudes systematically underestimate the galaxy luminosities . in the case of the direct -band magnitudes the offset is @xmath38 magnitudes . in the case of the kron magnitudes inferred from the deeper j - band image profiles , the offset is reduced to @xmath39 magnitudes . conversely the extrapolated magnitudes are systematically brighter than the loveday kron magnitudes by @xmath40 and @xmath41 magnitudes , where one would expect an offset of only @xmath42 to @xmath43 due to the difference in definition between ideal kron and true total magnitudes . for both estimates of the extrapolated magnitude and for the directly estimated kron magnitude the scatter about the correlation is approximately 0.14 magnitudes and we note a slight tendency for the scatter to increase at faint magnitudes . the magnitude estimate that best correlates with the loveday measurements is the kron magnitude estimated from the j - band kron magnitude and the default aperture j@xmath9 colour . here the distribution of residuals has a much reduced scatter of only 0.1 magnitudes and has very few outliers . our conclusions from the comparison of kron magnitudes is that it is preferable to adopt the -band magnitude inferred from the j - band kron or extrapolated magnitude by converting to the -band using default aperture colour , rather than to use the noisier and more biased direct -band estimates . with this definition , we find that the kron magnitudes slightly underestimate the galaxy luminosities while the extrapolated magnitudes slightly overestimate the luminosities , particularly at faint fluxes . we will present results for both magnitude definitions , but we note that to convert to total magnitudes we estimate that the kron magnitudes should be brightened by @xmath44@xmath45 magnitudes and the extrapolated magnitudes dimmed by @xmath46@xmath43 magnitudes . here we define the magnitude limited samples which we will analyze in section [ sec : lf_est ] and test them for possible incompleteness in both magnitude and surface brightness . for the kron and extrapolated magnitudes , the catalogue has high completeness to the nominal limits of j@xmath1914.7 and @xmath1913.9 . however , to ensure very high completeness and avoid any bias in our luminosity function estimates , we made the following more conservative cuts . for the kron magnitudes , we limited our sample to either j@xmath1914.45 or @xmath1913.2 , and for the extrapolated magnitudes to either j@xmath1914.15 or @xmath1912.9 . these choices are motivated by plots such as the top panel of fig . [ fig : mags ] . here the isophotal default magnitude limit of j@xmath19@xmath47 is responsible for the right hand edge to the distribution of data points . one sees that this limit begins to remove objects from the distribution of kron magnitudes for j@xmath48@xmath49 . an indication that the survey is complete to our adopted limits is given by the number counts shown in fig [ fig : counts ] , which only begin to roll over at fainter magnitudes . = 8.5 truecm = 8.0 truecm more rigorously , we have verified that the samples are complete to these limits by examining their @xmath50 distributions . here , @xmath51 is the redshift of a galaxy in the sample , @xmath52 is the maximum redshift at which this galaxy would satisfy the sample selection criteria , and @xmath53 is the survey volume that lies at redshift less than @xmath54 . if the sample is complete and of uniform density , @xmath55 is uniformly distributed within the interval 0 to 1 . to evaluate @xmath56 we made use of the default k+e corrections described in the following section , but the results are not sensitive to reasonable variations in the assumed corrections or in the cosmology . the solid histograms in fig . [ fig : v_vmax ] show these distributions for each of our four magnitude limited samples . note that the -band magnitudes are those inferred from the j - band magnitudes and aperture colours . the dashed histograms in the lower panels show the corresponding distributions for the directly measured -band magnitudes . in all these cases we have computed @xmath57 simply from the imposed apparent magnitude limits and have ignored any possible dependence of the catalogue completeness on surface brightness . if the samples were incomplete the symptom one would expect to see is a deficit in the @xmath58 distributions at large @xmath58 and hence a mean @xmath59@xmath19@xmath60 . there is no evidence for such a deficit in these distributions . in fact each has a mean @xmath59 slightly greater than @xmath60 . the slight gradient in the @xmath58 distribution is directly related to the galaxy number counts shown in fig [ fig : counts ] , which are slightly steeper than expected for a homogeneous , non - evolving galaxy distribution . a similar result has been found in the bright -band counts ( @xcite ) . the -band result has variously been interpreted as evidence for rapid evolution , systematic errors in the magnitude calibration , or a local hole or underdensity in the galaxy distribution ( @xcite ) . here we note that the gradient in the @xmath58 distributions ( and also in the galaxy counts ) becomes steeper both as one switches from kron to the less reliable extrapolated magnitudes and as one switches from the j - band data to the lower signal - to - noise -band data . this gives strong support to our decision to adopt the -band magnitudes derived from the j - band kron and extrapolated magnitudes and aperture j@xmath9 colours . it also cautions that the mean @xmath59@xmath15@xmath60 can not necessarily be taken as a sign of evolution or a local underdensity , but may instead be related to the accuracy of the magnitude measurements . the comparison to the observations of loveday ( ) shows no evidence for systematic errors in the magnitudes , but does not constrain the possibility that the distribution of magnitude measurement errors may become broader or skewed at fainter magnitudes . such variations would affect the @xmath58 distributions and could produce the observed behaviour . we conclude by noting that while the shift in the mean @xmath59 is statistically significant , it is nevertheless quite small for the samples we analyze and has little effect on the resulting luminosity function estimates . = 8.0 truecm we now investigate explicitly the degree to which the completeness of the catalogue depends on surface brightness by estimating @xmath57 as a function of both absolute magnitude and surface brightness . this is an important issue : if the catalogue is missing low - surface brightness galaxies our estimates of the luminosity function will be biased . the approach we have taken follows that developed in cross ( ) for the . we estimate an effective central surface brightness , @xmath61 , for each observed galaxy assuming an exponential light distribution , that the kron magnitudes are total and that the kron radii are exactly five exponential scale - lengths . this is then corrected to redshift @xmath62 using @xmath63 to account for redshift dimming and k+e corrections ( c.f section [ sec : pop ] ) . note that in the catalogue , galaxies with estimated kron radii less than @xmath64 arcsec , have their kron radii set to 7 arcsec . this will lead us to underestimate the central surface brightnesses of these galaxies , but this will only affect high surface brightness objects and will not affect whether a galaxy can or can not be seen . the distribution in the @xmath65@xmath66 plane of our kron j - band selected sample is shown by the points in fig . [ fig : sb ] . cross ( ) use two different methods to estimate the value of @xmath57 associated with each position in this plane . the first method uses the visibility theory of phillipps , davies & disney ( ) . we model the selection characteristics of the extended source catalogue by a set of thresholds . the values appropriate in the j - band are a minimum isophotal diameter of @xmath67 arcsec at an isophote of @xmath68 mag arcsec@xmath32 , and an isophotal magnitude limit of j@xmath19@xmath47 at an isophote of @xmath69 mag arcsec@xmath32 ( jarrett et al . in addition , we impose the limits in the kron magnitude of @xmath70@xmath19j@xmath19@xmath71 that define the sample we analyze . we then calculate for each point on the @xmath65@xmath66 plane the redshift at which a such a galaxy will drop below one or other of these selection thresholds and hence compute a value of @xmath57 . the results of this procedure are shown by the contours of constant @xmath57 plotted in fig . [ fig : sb ] . note that these estimates of @xmath57 are only approximate since we have made the crude assumption that all the galaxies are circular exponential disks . in addition , the diameter and isophotal limits are only approximate and vary with observing conditions . the second method developed by cross ( ) consists of making an empirical estimate of @xmath57 in bins in the @xmath65@xmath66 plane . they look at the distribution of observed redshifts in a given bin and adopt the 90@xmath72 percentile of this distribution to define @xmath56 and hence @xmath57 . it is more robust to use the 90@xmath72 percentile rather than the 100@xmath72 percentile and the effect of this choice can easily be compensated for when estimating the luminosity function ( @xcite ) . note that in our application to the data we do not apply corrections for incompleteness or the effects of clustering . the result of this procedure is to confirm that for the populated bins , the @xmath57 values given by the visibility theory are a good description of the data . = 8.5 truecm in fig . [ fig : sb ] we see that the distribution of galaxies in the @xmath65@xmath66 plane is well separated from the low surface brightness limit of approximately @xmath68 mag arcsec@xmath32 where the @xmath57 contours indicate that the survey has very little sensitivity . thus , there is no evidence that low - surface brightness galaxies are missing from the catalogue . furthermore , in the region occupied by the observed data , the @xmath57 contours are close to vertical indicating that there is little dependence of @xmath57 on surface brightness . the way in which the @xmath58 distribution is modified by including this estimate of the surface brightness dependence is shown by the dotted histogram in the top - left panel of fig . [ fig : v_vmax ] . its effect is to increase the mean @xmath58 slightly , suggesting that this estimate perhaps overcorrects for the effect of surface brightness selection . even so , the change in the estimated luminosity function is negligible as confirmed by the three estimates of the kron j - band luminosity function shown in fig . [ fig : lf_sb ] . these are all simple @xmath73 estimates , but with @xmath57 computed either ignoring surface brightness effects or using one of the two methods described above . these luminosity functions differ negligibly , indicating that no bias is introduced by ignoring surface brightness selection effects . = 8.5 truecm = 8.5 truecm the primary aim of this paper is to determine the present - day j and -band luminosity functions and also the stellar mass function of galaxies . since the survey spans a range of redshift ( see fig . [ fig : dn_dz ] ) , we must correct for both the redshifting of the filter bandpass ( k - correction ) and for the effects of galaxy evolution ( e - correction ) . in practice , the k and e - corrections at these wavelengths are both small and uncertainties in them have little effect on the estimated luminosity functions . this is because these infrared bands are not dominated by young stars and also because the survey does not probe a large range of redshift . we have chosen to derive individual k and e - corrections for each galaxy using the stellar population synthesis models of bruzual & charlot ( ) . we have taken this approach not because such detailed modelling is necessary to derive robust luminosity functions , but because it enables us to explore the uncertainties in the derived galaxy stellar mass functions , which are , in fact , dominated by uncertainties in the properties of the stellar populations . the latest models of bruzual & charlot ( ) provide , for a variety of different stellar initial mass functions ( imfs ) , the spectral energy distribution ( sed ) , @xmath74 , of a single population of stars formed at the same time with a single metallicity , as a function of both age , @xmath75 , and metallicity , @xmath76 . we convolve these with an assumed star formation history , @xmath77 , to compute the time - evolving sed of the model galaxy , @xmath78 we take account of the effect of dust extinction on the seds using the ferrara ( ) extinction model normalized so that the v - band central face - on optical depth of the milk - way is 10 . this value corresponds to the mean optical depth of @xmath79 galaxies in the model of cole ( ) which employs the same model of dust extinction . we assume a typical inclination angle of 60 degrees which yields a net attenuation factor of 0.53 in the v - band and 0.78 in the j - band . by varying the assumed metallicity , @xmath76 , and star formation history , we build up a two - dimensional grid of models . then , for each of these models , we extract tracks of @xmath9 and j@xmath9 colours and stellar mass - to - light ratio as a function of redshift . our standard set of tracks assumes a cosmological model with @xmath7 , @xmath8 , hubble constant @xmath80 km s@xmath81 mpc@xmath81 , and star formation histories with an exponential form , @xmath82/\tau)$ ] . here , @xmath83 is the age of the universe at redshift @xmath54 and the galaxy is assumed to start forming stars at @xmath84 . for these tracks , we adopt the kennicutt imf ( @xcite ) and include the dust extinction model . the individual tracks are labelled by a metallicity , @xmath76 , which varies from @xmath85 to @xmath86 and a star formation timescale , @xmath87 , which varies from @xmath88 gyr to @xmath89 gyr . examples of these tracks are shown in fig . [ fig : tracks ] , along with the observed redshifts and colours of the galaxies . we can see that the infrared j@xmath9 colour depends mainly on metallicity while the @xmath9 colour depends both on metallicity and star formation timescale . thus , the use of both colours allows a unique track to be selected . note from the bottom panel that , for all the tracks , the k+e correction is always small for the range of redshift spanned by our data . we can gauge how robust our results are by varying the assumptions of our model . in particular , we vary the imf , the dust extinction and cosmological models , and include or exclude the evolutionary contribution to the k+e correction . also , we consider power - law star formation histories , @xmath90^{-\gamma}$ ] , as an alternative to the exponential model . the results are discussed at beginning of section [ sec : results ] . the procedure for computing the individual galaxy k+e corrections is straightforward . at the measured redshift of a galaxy , we find the model whose @xmath9 and j@xmath9 colours most closely match that of the observed galaxy . having selected the model we then follow it to @xmath62 to predict the galaxy s present - day j and -band luminosities and also its total stellar mass . we also use the model track to follow its k+e correction to higher redshift in order to compute @xmath56 , the maximum redshift at which this galaxy would have passed the selection criteria for inclusion into the analysis sample . we use both the simple @xmath73 method and standard maximum likelihood methods to estimate luminosity functions . we present schechter function fits computed using the sty method ( sandage , tammann & yahil ) and also non - parametric estimates using the stepwise maximum likelihood method ( swml ) of efstathiou , ellis & peterson ( ) . our implementation of each of these methods is described and tested in norberg ( ) . the advantage of the maximum likelihood methods is that they are not affected by galaxy clustering ( provided that the galaxy luminosity function is independent of galaxy density ) . by contrast the @xmath73 method , which makes no assumption about the dependence of the luminosity function with density , is subject to biases produced by density fluctuations . the two maximum likelihood methods determine the shape of the luminosity function , but not its overall normalization . we have chosen to normalize the luminosity functions by matching the galaxy number counts of jarrett ( in preparation ) . these were obtained from a 184 deg@xmath12 area selected to have low stellar density and in which all the galaxy classifications have all been confirmed by eye . the counts are reproduced in fig . [ fig : counts ] . by using the same 7 arcsec aperture magnitudes as jarrett ( in preparation ) and scaling the galaxy counts in our redshift survey , we deduce that the effective area of our redshift catalogue is @xmath91 deg@xmath12 . note that normalizing in this way by - passes the problem of whether or not some fraction of the missed objects are stars . [ fig : counts ] also shows the kron and extrapolated magnitude j and counts of the redshift survey . in the lower panel , these counts are seen to be in agreement with the published k - band counts of gardner ( ) and glazebrook ( ) . we also checked the normalization using the following independent estimate of the effective solid angle of the redshift survey . for galaxies in the parent catalogue brighter than @xmath92 , we computed the fraction that have both measured redshifts and match a galaxy . for a faint @xmath93 this fraction is small as the catalogue is much deeper than the catalogue , but as @xmath93 is made brighter , the fraction asymptotes to the fraction of the area of the parent catalogue covered by the joint redshift survey . by this method we estimate that the effective area of our redshift catalogue is @xmath94 deg@xmath12 , which is in good agreement with the estimate from the counts of jarrett . it should be noted that for neither of these estimates of the effective survey area do the quoted uncertainties take account of variations in the number counts due to large scale structure . to estimate the expected variation in the galaxy number counts within the combined - survey due to large scale structure we constructed an ensemble of mock catalogues from the @xmath95cdm hubble volume simulation of the virgo consortium ( evrard ; evrard in preparation ; http://www.physics.lsa.umich.edu/hubble-volume ) . mock catalogues constructed from the virgo hubble volume simulations ( baugh in preparation ) can be found at http://star-www.dur.ac.uk/@xmath96cole/mocks/hubble.html . we simply took these catalogues and sampled them to the depth of over a solid angle of @xmath97 deg@xmath12 . to this magnitude limit we found an rms variation in the number of galaxies of 15% . we took this to be a realistic estimate of the uncertainty in the number counts and propagated this error through when computing the error on the normalization of the luminosity function . [ cols= " < , > , > , > , > , > , > , > , > , > , > , > , > , > , > , > , > , > , > " , ] [ tab : madau ] it is interesting to compare our values with what is inferred by integrating the observational estimates of the mean star formation history of the universe . [ fig : madau ] shows observational estimates for one particular choice of cosmology and imf and illustrates how the rates are sensitive to the assumed dust extinction . by fitting a smooth curve through these estimates , we can calculate the mass of stars formed by the present day and how this depends on the imf and assumed dust extinction . the upper smooth curve shown in fig . [ fig : madau ] is of the form @xmath98yr@xmath81mpc@xmath3 , where @xmath99 . the data points uncorrected for dust extinction are fit with @xmath100 . as for our estimates above , we assume that no mass goes into forming brown dwarfs and multiply the star formation rate by @xmath101@xmath9@xmath102 , where @xmath102 is the recycled fraction , so as to form an estimate of the mass locked up in stars . values of @xmath103 estimated in this way are listed in table [ tab : madau ] . the values in this table are for an @xmath7 , @xmath8 cosmology , but they are insensitive to this choice . they depend slightly on the assumed metallicity of the stellar population and would be 10% lower if half solar , rather than solar metallicity were assumed . note that the @xmath104 values inferred from the star formation history of the universe scale differently with the assumed hubble constant than those inferred above from the ir luminosity functions . for @xmath105 our estimates from become @xmath106 for the kennicutt imf and @xmath107 for the salpeter imf . comparison with table [ tab : madau ] shows that these values are consistent with those inferred from the cosmic star formation history only if the dust correction assumed in the latter is modest , e(b - v)@xmath108 . this value is 50% smaller than the value preferred by steidel ( ) . the new generation of very large surveys currently underway make it possible to characterize the galaxy population with unprecedented accuracy . in this paper , we have combined two such large surveys , the infrared imaging 2mass and the 2df galaxy redshift survey and [ tab : smfswml ] are available at http://star-www.dur.ac.uk/@xmath96cole/2dfgrs-2mass . ] to obtain a complete dataset which is more than an order of magnitude larger than previous datasets used for statistical studies of the near - infrared properties of the local galaxy population . we have used this combined catalogue to derive the most precise estimates to date of the galaxy j and -band luminosity functions and of the galaxy stellar mass function . characterizing the near - infrared properties of galaxies offers several advantages . firstly , the near - infrared light is dominated by established , old stellar populations rather than by the recent star formation activity that dominates the blue light . thus , the j and k - band luminosity functions reflect the integrated star formation history of a galaxy and , as a result , provide particularly important diagnostics of the processes of galaxy formation . for the same reason , the distribution of stellar mass in galaxies the galaxy stellar mass function can be derived from the near - infrared luminosities in a relatively straightforward way , with only a weak model dependence . finally , corrections for dust extinction as well as k - corrections are much smaller in the near - infrared than in the optical . due to the size of our sample , our determination of the j- and -band galaxy luminosity functions have , for the most part , smaller statistical errors than previous estimates . furthermore , since our sample is infrared - selected , our estimates are free from any potential biases that might affect infrared luminosity functions derived from optically - selected samples . we find that the j- and -band galaxy luminosity functions are fairly well described by schechter functions , although there is some evidence for an excess of bright galaxies relative to the best - fit schechter functional form . in general , the swml estimates are a truer representation of the luminosity functions . our k - band estimates are in overall agreement with most previous determinations , but have smaller statistical errors . the exception is the k - band luminosity function inferred from the near infrared sdds photometry ( @xcite ) . the difference between the k - band luminosity function we infer from their data and our own estimate is too large to be explained by photometric differences . the difference between the two estimates is better described by a difference in overall number density of a factor of @xmath109 . a similar discrepancy is seen in the -band between the sdss and luminosity function estimates ( see @xcite and @xcite ) . the suspicion is that the uncertainty in the overall normalization of the luminosity functions induced by large - scale structure within the large , but finite , survey volumes could be to blame . however , the errors that we quote for the -luminosity functions already include an estimate of this sampling uncertainty as derived from realistic mock catalogues . a similar exercise for a catalogue with the same area and depth as that of blanton ( ) indicates that the required overdensity of a factor of @xmath109 is unlikely . so probably there is more than one contributory factor at work and the hope is that these will be identified as the surveys progress . using our j - band luminosity function , @xmath9 and j@xmath9 colours and simple galaxy evolutionary tracks , we have obtained the first estimate of the galactic stellar mass function derived directly from near - infrared data . we find that this mass function is also fairly well described by a schechter form . an integral over the stellar mass function gives @xmath104 , the universal mass density locked up in luminous stars and stellar remnants , expressed in terms of the critical density . @xmath104 is a key component of the overall inventory of baryons in the universe . an accurate determination of this quantity is essential for detailed comparisons with other quantities of cosmological interest such as the total baryonic mass , @xmath110 , inferred from big bang nucleosynthesis considerations ( e.g. burles & tytler 1998 ) , the cosmic star formation rate ( e.g. steidel ) , and the cosmic evolution of the gas content of the universe ( storrie - lombardi & wolfe 2000 ) . the statistical uncertainty in our estimate of @xmath104 is about 15% , several times smaller than the best previous determination by salucci & persic ( ) . in fact , the errors in @xmath104 are dominated by systematic uncertainties associated with the choice of stellar imf in the galaxy evolution model . a kennicutt imf gives @xmath111 while a salpeter imf gives @xmath11 . for @xmath105 , these values correspond to less than 11% of the baryonic mass inferred from big bang nucleosynthesis ( @xcite ) . our values of @xmath104 today are only consistent with recent determinations of the integrated cosmic star formation if the correction for dust extinction is modest . we thank john lucey for the initial suggestion to look at the database . we thank tom jarrett for advice and detailed explanations of the 2mass data reduction procedure . we also thank ian smail for help in manipulating dss images , jon loveday for supplying his luminosity function data in electronic form and simon white for useful suggestions . we are grateful to the virgo consortium for allowing us to use the hubble volume simulation data prior to publication.the redshift data used here were obtained with the 2-degree field facility on the 3.9 m angloaustralian telescope ( aat ) . we thank all those involved in the smooth running and continued success of the 2df and the aat . smc acknowledges a pparc advanced fellowship , iprn an snsf and ors studentship and csf a leverhulme research fellowship . the 2df galaxy redshift survey was made possible through the dedicated efforts of the staff at the anglo - australian observatory , both in creating the two - degree field facility and supporting it on the telescope .
within the last decade an increasing interest in molecular electronics has developed , with the expectation of realising molecular diodes and transistors . this is based on the progress in manipulation techniques , which now allow the controlled attachment of atomic scale structures like molecules to mesoscopic leads . with these new devices one is able to determine the conductance properties of molecular structures . explaining and predicting the electronic behaviour of such devices is an essential step towards their design and use as nano - scale electronic circuits . to this end a number of theoretical studies have been performed with the aim of reproducing measured @xmath0 characteristics . these studies differ in the way they take the electronic levels of the molecules , their modification by the coupling to the leads and the change of electrostatic potential due to bias into account . semi - empiric methods@xcite have been used , as well as first principles techniques,@xcite the latter being restricted to systems of moderate size . the wide range of experimentally observed behaviour ( see section [ sec : experiment ] ) suggests , that not only the structure of the molecule , but also the details of the device fabrication affect the conduction properties of molecular devices . the crucial step is the deposition of molecules onto the surface of the lead . as this is achieved by self - assembly the amount of adsorbed molecules and their individual positions can not be exactly controlled and therefore remains unknown . a satisfactory understanding of the interplay between geometrical alignment of the molecules and measured conductance properties has thus not yet been achieved ( for a recent review see e.g. nitzan et al.@xcite ) . in this paper we study the way in which changing the geometrical alignment of the mono - layer has an influence on the conduction properties of a molecular device . in so doing we can rule out a number of explanations which have previously been considered@xcite to explain the occurrence of additional structure in the conductance - voltage ( @xmath1 ) characteristics . the outline is the following : first we summarise some of the recent experimental findings . then the method we use ( based on esqc@xcite ) for calculating the conductance properties of molecular devices , is discussed . calculations for the conductance of a sam , being build of 1,4-phenylene diisocyanide ( pdi ) and sandwiched between gold leads are then presented . the results for three qualitatively different geometrical constitutions of the mono - layer are compared with experimental data . by this we can rule out all but one and conclude that the only geometrical alignment , which gives rise to several peaks in the conductance curve , is a mono - layer with additional molecules perturbing the periodicity . the devices built to study conductance properties of molecular structures differ not only in amount and chemical structure of the molecules in use but also in the way these are attached to metallic or semi - conducting leads . single or few molecules are accessible in mechanically controllable break junctions ( mcb ) and with the scanning tunneling microscope ( stm ) . many molecules are involved in sandwiched self - assembled mono - layer ( sam ) experiments . the observed properties depend on the exact geometry of the device . the conductance differs in orders of magnitude and the qualitative voltage dependency of the current ranges from simple ohmic behaviour to negative differential resistance ( ndr).@xcite in the past reed et al.@xcite have measured the electrical conductance of a self - assembled molecular mono - layer bridging an mcb at room temperature . molecules of 1,4-benzene dithiol ( i.e. having two thiol end - groups , which are known to couple strongly to au - atoms ) were used and the @xmath1 characteristic was found to be symmetric with one peak in the voltage range of @xmath2v . they measured a current of the order of @xmath3na at a bias voltage of @xmath4v , which they claim is produced by transport through one single active molecule . reichert et al.@xcite also used an mcb with molecules having two thiol end groups , but being considerably longer . the measured current amplitude was about @xmath5na at @xmath6v , i.e. although the molecule was more than twice as long , the current was ten times larger . with a different setup , where a sam is sandwiched between two metallic leads , chen et al.@xcite have found negative differential resistance ( ndr ) , namely one peak at @xmath4v in the @xmath0 curve . the molecule under investigation had one thiol end - group only and was attached to au - leads at both ends . the measurements were taken at room temperature and the measured current maximum was of the order of @xmath6na . only recently , sandwiched sam devices at @xmath7k were studied,@xcite where a benzene ring with two cyanide instead of thiol end - groups was used . the measurements exhibited currents of the order of @xmath8na . the @xmath1 characteristic for this molecule revealed more structure , in form of three to five peaks within a voltage range of @xmath6v . such a behaviour was not observed with previous devices containing other molecules . in the literature there has been presented quite a number of techniques to calculate non - equilibrium electronic transport through molecular systems attached to mesoscopic leads . usually the landauer formalism is applied , which describes current as elastic electron transmission and therefore requires the transmission function @xmath9 . to this end one needs a framework that allows for a description of the molecular device on the atomic level . this involves not only the molecules themselves , but also the surface and bulk region of the leads . quantum chemistry provides such a framework , and one can choose the level of theory according to the size of the system under consideration and the computational effort one is willing to spend . using a quantum chemistry method , the transmission function can be obtained from an effective one - particle hamiltonian , which is an appropriate description for strong coupling of the molecules to the leads ( as in the case of covalent binding ) . the methods differ in the generation of the one - particle hamiltonian , which might be based on semi - empirical grounds @xcite or on first - principles and self - consistent techniques . @xcite a different approach,@xcite taking many - particle effects explicitly into account , uses a master equation with transition rates calculated perturbatively using the golden rule . this approach is appropriate for weak coupling . we use the landauer formalism , as the molecules are assumed to be chemically bonded to the gold contacts ( i.e. strong coupling ) , together with the semi - empiric extended hckel quantum - chemistry method esqc.@xcite the molecular structure is optimised@xcite beforehand . this approach , though limited as compared to more sophisticated quantum chemistry methods , is yet justified , because we want to gain a qualitative understanding of a many molecule experiment , which can not be described by first - principle techniques , as the number of atoms involved is beyond the practical limitations of to - date computer resources . according to the landauer formula , current along a defect region is the result of electron transmission from the source to the drain lead . this is described by the transmission function @xmath9 and net electron transfer happens at all those energies , where there are more states occupied in the source lead than in the drain . the difference in occupation is a result of the applied bias @xmath10 , which raises the chemical potential @xmath11 of the source lead above the one of the drain lead @xmath12 and thus changes the distribution function @xmath13 . @xmath14 the landauer formula is valid under the condition that transport is coherent across the molecule , which is plausible as the typical mean - free path of electrons within metals is of the order of @xmath5 nm , while the molecular gap between source and drain lead is only about @xmath15 nm in length . furthermore it is assumed that the coupling of the leads to macroscopic reservoirs is reflection - less , and that the tunneling electrons equilibrate only deep inside the leads . this ensures that the distribution function for incoming electrons in a lead can be taken to be spatially constant , even close to the molecular region . ( colour online ) partitioning of the system into three parts : the two semi - infinite leads @xmath16 ( surrounded by boxes ) and the molecular region @xmath17 between them.,scaledwidth=50.0% ] the system is formally partitioned into three regions @xmath18 , two of them ( @xmath16 ) containing the semi - infinite leads , the third one ( @xmath19 ) being the finite region containing all molecules as well as a few surface layers of each lead ( see fig . [ fig : system ] ) . we use periodic boundary conditions in the directions perpendicular to the surface normal . by a tight binding approximation , the infinite - dimensional hamiltonian of the entire system can be composed of quantum chemical one - particle block hamiltonians of finite dimension : @xmath20 the first summation describes the isolated molecular region , by an on - site energy @xmath21 and a hopping term . the indices @xmath22 and @xmath23 run over the orbital basis set within that region . the next two summations describe the isolated leads , labelled by @xmath24 . layer by layer , starting with the surface layer @xmath25 , the first term accounts for intra - layer interactions , while the second one describes the interaction between layers . ( the size of each layer is chosen such that only adjacent layers have non - zero interaction . it therefore depends on the details of the tight - binding approximation . ) finally the last term describes the coupling between the molecular region and each lead . note that only the first layer @xmath25 contributes to that term and that there is no interaction between different leads . these are only formal restrictions , as parts of each lead can be included into the molecular region . the determination of the transmission function involves two steps . first the conduction properties of the isolated leads have to be calculated . thereby each lead will be decomposed into conducting and non - conducting in- and out - going channels . these correspond to propagating and evanescent solutions moving in one of two possible directions respectively . in a second step , the channels are connected to each other via the molecular region . this is described by the scattering matrix and the transmission function is finally obtained by summing up the contribution from each channel . the calculation can be performed either using green s function techniques@xcite or equivalently@xcite using esqc@xcite , which is a scattering - matrix approach . we present the details of the calculation in the second scheme , as individual contributions from each channel to the transmission function can then be easily studied . first we will restrict our attention to the semi - infinite lead hamiltonians , which do not have to be identical . the hamiltonians of equation for one lead @xmath26 are layer independent , if one assumes periodicity , i.e. @xmath27 and @xmath28 . using bloch s theorem one can reduce the infinite dimensional system of equations to an @xmath29-matrix equation ( @xmath30 being the number of orbital basis functions in one layer ) @xmath31 with @xmath32 , @xmath33 , and @xmath34 is the overlap matrix between orbitals in layer @xmath35 and layer @xmath36 for cases when one does not deal with an orthonormal basis set ( otherwise @xmath37 ) . with @xmath38 we denote the lattice spacing . defining @xmath39 one can easily see that equation is an @xmath29 quadratic eigenvalue equation . it can be transformed into a @xmath40 linear eigenvalue equation : @xmath41 \end{aligned}\ ] ] ( where we have dropped the energy dependency for notational ease ) . this layer - to - layer propagator @xmath42 reduces the problem of finding solutions for the entire isolated lead hamiltonian to specifying the wave function coefficients at two adjacent layers @xmath43 and @xmath44 only . this is due to the fact that given these coefficients any other pair of coefficients @xmath45 can now be obtained via : @xmath46 all possible solutions at energy @xmath47 can be decomposed into independent channels , by solving for the eigenvalues of equation . these eigenvalues come in pairs such that for each eigenvalue @xmath48 , there exists a corresponding eigenvalue @xmath49 satisfying the relation @xmath50 , as can be seen by transposing equation . eigenvalues with @xmath51 , i.e. complex @xmath52 , belong to exponentially diverging solutions ( see equation ) . these are of course non physical , as long as the lead is infinite . in semi - infinite leads however ( which we are dealing with ) , exponentially decaying coefficients at the boundary will contribute to the surface wave function and must not be neglected . the contribution from a single channel to the net current can not directly be seen from equation . it depends on the current density associated with a solution to the schrdinger equation @xmath53 and is obtained via the continuity equation . the probability amplitude @xmath54 for a stationary solution is constant in time @xmath55 because @xmath56 and @xmath57 are hermitian . for the probability amplitude at all layers between @xmath58 and @xmath59 one therefore has @xmath60 \langle l_2 , l_2 + 1\| \gamma\rangle \\\nonumber & - \langle\gamma \|l_1 - 1,l_1\rangle \frac{i}{\hbar } \left[\begin{array}{cc } 0 & -h\\ h^\dagger & 0 \end{array}\right ] \langle l_1 - 1 , l_1\| \gamma\rangle , \end{aligned}\ ] ] with the projectors @xmath61 . this gives rise to the definition of the current operator @xmath62 for layer @xmath35 as @xmath63 \langle l , l+1\|.\ ] ] now let both @xmath64 and @xmath65 be solutions at fixed energy @xmath47 with the eigenvalues @xmath66 and @xmath67 respectively . because the expectation value for @xmath62 is layer independent ( equation [ eq : currentoperator2 ] ) one has : @xmath68 this equation describes the connection between the current properties of a solution @xmath64 and its eigenvalue @xmath69 . we summarise the results of a detailed analysis of this equation , which is given in appendix [ ap : current ] . each channel @xmath70 can be assigned a current value @xmath71 , defined as @xmath72 where we have used the layer independence of @xmath62 in simply writing @xmath73 . channels with eigenvalue modulus @xmath74 , i.e. evanescent waves have zero current value . they therefore do not contribute to the current . ( yet they are important at the surface , as already mentioned above . ) only channels with an eigenvalue of modulus one ( @xmath75 ) contribute to the current . the sign of @xmath71 determines the direction of charge transport . therefore the eigenvalues can be sorted into incoming and outgoing solutions ( @xmath48 and @xmath49 respectively ) , according to the following scheme : if @xmath76 it represents an incoming evanescent solution , if @xmath77 then it is outgoing evanescent . only if @xmath78 it belongs to a propagating solution , which is incoming for @xmath79 and outgoing otherwise . we now define @xmath80 and @xmath81 as the two @xmath29 diagonal matrices composed of all incoming and outgoing eigenvalues @xmath82 . the @xmath83-matrix @xmath84 , which diagonalises @xmath42 : @xmath85,\ ] ] has the following quadratic block form : @xmath86.\ ] ] after this transformation into the diagonal basis of the propagator , we can easily obtain all physically relevant solutions of the infinite lead by specifying the amplitudes of all propagating waves at one lattice site . up to now , we have considered the isolated leads only . these are now assumed to be each coupled to the molecular defect region and thereby indirectly coupled to one another . we are interested in stationary solutions , which consist of an incoming propagating wave in one lead , being scattered among all the accessible outgoing channels ( propagating and evanescent ones ) . this information is contained in the scattering matrix @xmath57 @xmath87 which determines the wave amplitudes of all outgoing waves @xmath88 given the incoming ones @xmath89 . ( 0,0 ) ( 2499,1221)(439,-799 ) ( 631,254)@xmath90 ( 721,-736)@xmath91 ( 2476,-736)@xmath92 ( 2476,254)@xmath93 in appendix [ ap : smatrix ] it is shown , that @xmath57 is of the form @xmath94 it is important to notice , that the scattering matrix is always quadratic , because in each lead there are the same amount of incoming and outgoing channels . this is opposed to the transfer matrix @xmath95 , which determines the amplitudes of in- and outgoing waves @xmath96 in one lead given the in- and outgoing waves @xmath97 of a second lead : @xmath98 this matrix is quadratic only if both leads have the same number of channels . it then is of the form@xcite @xmath99 and the relation to the scattering matrix is@xcite @xmath100 methods calculating the scattering matrix via the transfer matrix@xcite fail , if two types of leads are used , because @xmath101 is then no longer quadratic and can not be inverted . therefore one commonly takes source and drain lead to be identically constituted . but even in such cases , the method becomes numerically unstable , with increasing distance between the molecular region and one lead . this is because the matrix elements of @xmath101 and @xmath102 ( in equation ) diverge exponentially , with increasing lead separation . taking the inverse of @xmath101 is therefore a numerically critical procedure . both these problems are avoided by the direct calculation of the scattering matrix , which we present in appendix [ ap : smatrix ] . this calculation is well defined without any restrictions to the number of leads and their composition . that means that it is not necessary to restrict to identical leads . furthermore it allows a numerically stable determination of the scattering matrix , even for weak coupling . the transmission function is the sum over the contributions from each combination of incoming channel in the source lead to outgoing channel in the drain lead : @xmath103 . the relation between scattering matrix @xmath57 and these transmission function elements is @xmath104 where @xmath105 is the lower left block of @xmath57 as defined in equation . the weighting with velocity factors comes about because the scattering matrix @xmath106 does not relate current densities , but wave amplitudes . the current densities are obtained from these wave amplitudes by multiplication with the corresponding velocity factor @xmath107 . the factor @xmath71 in the denominator normalises the transmission function to be exactly one for perfect transmission . if @xmath108 then @xmath109 , because incoming evanescent waves have zero amplitude at the surface . the total current is made up of the contribution from each @xmath52-state@xcite : @xmath110 here the summation over @xmath52 has been transformed into an integral over @xmath47 with a factor of 2 accounting for spin @xmath111 . low temperature experiments with pdi sam s sandwiched between two metallic leads show several peaks in the @xmath1-diagram.@xcite the typical voltage differences of these peaks are in the range of @xmath112 ( i.e. there are about 5 peaks within @xmath113 and @xmath114 ) . the commonly adopted explanation for the occurrence of such peaks is the following . each molecular orbital that enters the energy window , which is opened by the applied voltage , enables resonant tunneling . this increases the conductance and therefore results in a peak within the @xmath1-diagram . typically , the energy gap between molecular orbitals is in the range of @xmath115 . in other words , for applied voltages up to @xmath114 there should be only a single accessible orbital per molecule , giving rise to only a single peak in the @xmath1-diagram . therefore the following question arises : are there geometrical alignments of the molecules such that the additional peaks in the @xmath1 diagram can also be explained by resonant tunneling through molecular orbitals ? during the device fabrication , the step under least experimental control is the adsorption of the molecules onto the leads . therefore the exact geometrical alignment of the molecular sam and , at least in the sandwich geometry , also the atomic shape of the top metallic lead , is not exactly known . one therefore has to expect not only one specific but rather quite a variety of molecular alignments to be produced . as one is interested in the conduction properties of the resulting device , it is important to understand the influence of each type of geometrical alignment to the transmission function . to this end , we have investigated three such possible alignments , which will be discussed separately . 1 . by adding metal - atoms on top of the molecular mono - layer in a random way , there might occur metallic clusters on top of the mono - layer . these affect the electronic configuration of the molecules individually , and might therefore have an influence to the transmission function . 2 . in a sam experiment , there is not just one molecule , but rather a few hundred molecules involved . if the contribution to the transmission function was different for each molecule , then @xmath9 would change qualitatively , with a change in size of the mono - layer . we therefore analyse how the transmission functions depends on the number of molecules involved . 3 . if the molecular mono - layer is not strictly periodic , then there will be defects . for example can the distance between two molecules be reduced , such that inter - molecular bonds can be build . each of these defects will have a specific electronic structure and will therefore influence the transmission function . in the sandwich geometry , first the bottom metallic lead is created . then the molecular mono - layer is adsorbed on top of it by self - assembly . finally the top metallic lead is build upon the molecular mono - layer . the exact shape of neither metallic surface is known and may be anything but flat and regular . it is likely that the surface atoms of the top metallic lead build up clusters on top of the molecular layer ( as for example in fig . [ fig : singleandcluster ] b ) . which influence do they have on the electronic configuration of the molecule they are in contact with ? and do the clusters act as small molecules with new electronic levels ? the influence of an au cluster on the molecular electronic structure is twofold . first it introduces new electronic levels , and second the existing molecular electronic levels will be shifted , by an amount which depends on the strength of the coupling between cluster and molecule . the latter effect will be observed as a shifted peak in the transmission function , only if the coupling between cluster and molecule is different to the coupling between top electrode and molecule . for clusters similar to the one shown in fig . [ fig : singleandcluster ] ( b ) , this is however not the case . the energetic peak positions are identical , as can be seen in fig . [ fig : singleandcluster ] ( c ) . furthermore , there are no additional peaks , which one might have expected because of the additional electronic levels of the cluster . the explanation of their absence is the following : an electronic level gives rise to a peak in the transmission function only , if the corresponding orbital wave function overlaps with both the top and bottom electrode . the overlap with the electrode the cluster is attached to ( say top electrode ) is of course large . the overlap with the bottom electrode consists of two parts . the direct overlap and the indirect overlap via the molecule . the direct overlap is negligible due to the large spatial separation . the indirect overlap depends on the molecular orbital wave function . if the energy of the cluster level does not coincide with a molecular energy level , then there is no indirect overlap . only if two levels coincide , the indirect coupling is large , but in that case , there already exists a transmission peak due to the molecule itself . ( 0.98,0.72 ) ( 0,0 ) ( a ) ( colour online ) structure for a molecule without cluster . ( b ) ( colour online ) sturcture for a molecule with a gold cluster on top . ( c ) transmission function @xmath9 for both structures . energy scale is relative to the homo - lumo gap , such that @xmath116 corresponds to the middle of the gap . , title="fig : " ] ( 0,0.72 ) ( 0.98,0.72 ) ( 0,0 ) ( a ) ( colour online ) structure for a molecule without cluster . ( b ) ( colour online ) sturcture for a molecule with a gold cluster on top . ( c ) transmission function @xmath9 for both structures . energy scale is relative to the homo - lumo gap , such that @xmath116 corresponds to the middle of the gap . , title="fig : " ] ( 0,0.72 ) + ( 0.98,0.98 ) ( 0,0 ) ( a ) ( colour online ) structure for a molecule without cluster . ( b ) ( colour online ) sturcture for a molecule with a gold cluster on top . ( c ) transmission function @xmath9 for both structures . energy scale is relative to the homo - lumo gap , such that @xmath116 corresponds to the middle of the gap . , title="fig : " ] ( 0,0.98 ) therefore if transmission is already suppressed by the molecule ( at all off - resonant energies ) , it can either be further reduced by off - resonant tunneling through the cluster , or it can ( at best ) be left unchanged by resonant tunneling through the cluster . under no circumstances can transmission , once suppressed by the molecule , be afterwards increased by the cluster . this in turn means , that metallic clusters can not give rise to additional peaks in the transmission spectrum . ( colour online ) the structure for one , two , three , and four molecules adsorbed within an au-@xmath117 super - cell . this setup was used to test the sum rule.,title="fig : " ] ( colour online ) the structure for one , two , three , and four molecules adsorbed within an au-@xmath117 super - cell . this setup was used to test the sum rule.,title="fig : " ] ( colour online ) the structure for one , two , three , and four molecules adsorbed within an au-@xmath117 super - cell . this setup was used to test the sum rule.,title="fig : " ] ( colour online ) the structure for one , two , three , and four molecules adsorbed within an au-@xmath117 super - cell . this setup was used to test the sum rule.,title="fig : " ] how do we expect the transmission function @xmath118 for @xmath22 periodically arranged molecules to look like ? as long as the inter - molecular interactions are small ( compared to the intra - molecular ones ) the molecular levels of each molecule will not be significantly changed . furthermore as the mono - layer consists of only one kind of molecule , all of them will have the same electronic structure . therefore we expect each molecule to contribute the same amount to the transmission function : @xmath119 , where @xmath22 runs over all @xmath120 adsorbed molecules . ( 1.0,1.0 ) ( 0,0 ) ( a ) transmission function for one , two , three , and four pdi molecules . ( b ) when plotted against each other , the transmission functions reveal a linear relationship : @xmath121 ( bottom).,title="fig : " ] ( 0,1.0 ) ( 1,1 ) ( 0,0 ) ( a ) transmission function for one , two , three , and four pdi molecules . ( b ) when plotted against each other , the transmission functions reveal a linear relationship : @xmath121 ( bottom).,title="fig : " ] ( 0,1 ) we calculated the transmission function for @xmath120=1 to 4 molecules within an au super - cell of size @xmath117 ( the structures are shown in fig . [ fig : onetofourmolecules ] ) . the distance between the molecules is chosen to be a multiple of the closest au - au separation @xmath122 ( @xmath123 , with @xmath124 ) . to our knowledge , the parameters of the pdi - sam mono - layer have never been determined experimentally , which is why we have to assume the above values . however stm studies@xcite and also theoretical calculations@xcite have been performed for alkanethiol mono - layers , and these parameters motivated our choice . independent from the number of molecules present , the transmission functions have the same amount of peaks , at identical energetic positions ( see fig . [ fig : fourmoleculespectrum ] a ) . this result is also obtained for all larger distances of the molecules , where the inter - molecular interaction is even smaller . furthermore , the sum rule is indeed fulfilled , as shown in fig . [ fig : fourmoleculespectrum ] ( b ) . each @xmath118 is plotted against @xmath125 for @xmath126 . the calculated transmission values ( 300 discrete values each ) clearly show a linear correlation . the straight lines are linear fits to the data , and their slope does very well agree with the theoretically expected value of @xmath127 . the deviation is below @xmath128 , as can be seen in table [ tab : fitparameters ] , where we summarise all the fitted values for @xmath129 . .[tab : fitparameters ] the fitted values for @xmath130 together with their deviation from the theoretical value @xmath131 and a measure for the quality of the fit @xmath111 , where @xmath132 for @xmath133 discrete energy values @xmath134 . [ cols="<,<,^,^,^",options="header " , ] we conclude the following : a mono - layer , where the inter - molecular distance is large enough to not let inter - molecular interactions play a significant role , has the same number of distinct electronic levels as a single molecule . these levels are then highly degenerate . a @xmath1-diagram will therefore have the same number of peaks . only the net current will be increased by a factor @xmath130 compared to the single molecule case . the mere fact , that one deals with a mono - layer instead of a single molecule does not imply that the transmission function changes qualitatively . we have seen , that one does not observe additional peaks in the transmission function , as long as the inter - molecular influence is small . and this is the case for distances which occur in typical sam structures.@xcite we now investigate cases , where the molecular interactions are not negligible . this occurs for example , when the periodic structure of the mono - layer is perturbed by an additional molecule , such that a molecular cluster is formed . it is sufficient to study the transmission function of an isolated cluster only , because we have already seen that molecules in the periodic sam arrangement do not influence each other . the sum of the transmission function for the periodic sam and the transmission function for the molecular cluster is , due to the sum rule , the total transmission function for defect and sam . we study the influence of a shorter distance between two , three , and four molecules on the transmission spectrum and relate it to the discrete energies of the isolated molecules . the molecules are now separated by @xmath135 which corresponds to the au - au atom spacing . the atomic structure for this calculation is shown in fig . [ fig : molecularinteraction ] ( a ) , the resulting transmission functions in fig . [ fig : molecularinteraction ] ( b ) and ( c ) . ( 1.0,0.25 ) ( 0,0 ) ( a ) ( colour online ) two , three , and four molecules with a shorter inter - molecular distance . ( b ) the transmission functions for two and three molecules . ( c ) the transmission functions for three and four molecules . in contrast to all previous cases , the peaks are shifted with respect to each other and there are also additional peaks . these changes are due to the increase in inter - molecular interaction , which alters the electronic levels.,title="fig : " ] ( 0.33,0 ) ( a ) ( colour online ) two , three , and four molecules with a shorter inter - molecular distance . ( b ) the transmission functions for two and three molecules . ( c ) the transmission functions for three and four molecules . in contrast to all previous cases , the peaks are shifted with respect to each other and there are also additional peaks . these changes are due to the increase in inter - molecular interaction , which alters the electronic levels.,title="fig : " ] ( 0.66,0 ) ( a ) ( colour online ) two , three , and four molecules with a shorter inter - molecular distance . ( b ) the transmission functions for two and three molecules . ( c ) the transmission functions for three and four molecules . in contrast to all previous cases , the peaks are shifted with respect to each other and there are also additional peaks . these changes are due to the increase in inter - molecular interaction , which alters the electronic levels.,title="fig : " ] ( 0,0.25 ) ( 1.0,1.0 ) ( 0,0 ) ( a ) ( colour online ) two , three , and four molecules with a shorter inter - molecular distance . ( b ) the transmission functions for two and three molecules . ( c ) the transmission functions for three and four molecules . in contrast to all previous cases , the peaks are shifted with respect to each other and there are also additional peaks . these changes are due to the increase in inter - molecular interaction , which alters the electronic levels.,title="fig : " ] ( 0,1.0 ) ( 1.0,1.0 ) ( 0,0 ) ( a ) ( colour online ) two , three , and four molecules with a shorter inter - molecular distance . ( b ) the transmission functions for two and three molecules . ( c ) the transmission functions for three and four molecules . in contrast to all previous cases , the peaks are shifted with respect to each other and there are also additional peaks . these changes are due to the increase in inter - molecular interaction , which alters the electronic levels.,title="fig : " ] ( 0,1.0 ) by reducing the molecular separation from @xmath136 to @xmath137 , the transmission function qualitatively changes . the number of peaks roughly doubles . the new peak positions are different from the ones we have obtained in the previous calculations . and this time , the peak positions do depend on the number of molecules involved . this is an important point , because if there are several molecular clusters with different molecular distances , then they all give rise to peaks at different energy values . the resulting transmission function is the sum of the individual functions and will thus contain far more peaks , than the transmission function for the non perturbed periodic layer . we now show that the new peaks are a result of the increase in molecular interaction due to the decrease in spatial separation . for non - interacting molecules , the molecular energies are identical and therefore degenerate . an interaction between molecules breaks this degeneracy and therefore new energy levels occur . by performing a diagonalisation of the molecular hamiltonian ( without leads ) one can determine the levels of the molecular cluster . in fig . [ fig : moleculesandenergies ] we have again plotted the transmission function for three and four molecules , this time together with the discrete energy levels of the corresponding molecular cluster . the inset is identical to fig . [ fig : molecularinteraction ] ( c ) , while the plot itself is a magnification , to better resolve the discrete energy levels ( which are shown as points along the transmission function ) . each of the transmission peaks is related to at least one discrete energy value . but not each energy value can be related to a peak in the transmission function . why is that ? the discrete energies can only give rise to new peaks in the transmission function , if they are not suppressed by a weak coupling to one of the leads . all levels which are not related to any peak belong to this category . if the position of the peak is shifted away from a corresponding energy level , then this is due to the coupling between molecules and leads . this coupling is absent in the diagonalisation of the molecular hamiltonian , but present in the calculation of the transmission function . magnification of the transmission functions for three and four closely spaced molecules . additionally the discrete energy levels of the system without leads are plotted as points along the transmission function . to each peak there belongs at least one discrete energy level . a detailed discussion is given in the main text . inset : transmission function ( original scale ) for three and four closely spaced molecules ( identical to fig . [ fig : molecularinteraction ] c ) . ] finally we show , that the additional peak structure in the transmission function for a scenario with an increased inter - molecular interaction gives rise to a number of steps in the @xmath0-curve . figure [ fig : defectcurrent ] contains an @xmath0 calculation for a molecular structure containing all three molecular clusters shown in fig . [ fig : molecularinteraction ] ( a ) . in this calculation , the bias voltage @xmath138 enters as a shift of the fermi levels for source and drain lead : @xmath139 . the molecular energy has been set to @xmath140 , where @xmath141 is the zero bias displacement of the molecular levels and @xmath142 , because of the symmetric coupling to the leads . compared to the experiments@xcite the number of steps in the @xmath0curve is well reproduced by our calculation . the obtained current is at least one order of magnitude larger than the experimental values.@xcite this is a phenomenon common to all theoretical methods based on the landauer formula.@xcite a satisfactory explanation for this discrepancy as well as for the broad range of experimentally observed current values has not yet been found . @xmath0calculation for a molecular region containing all three molecular clusters shown in fig . [ fig : molecularinteraction ] ( a ) . there are three distinct steps within the voltage range of 1v . ] we have shown , that the peak structure of the transmission function is robust against changes in the number of adsorbed molecules , as long as the distance between molecules is considerably large ( @xmath143 ) . and also does the exact shape of the top metallic lead not influence the qualitative structure of the transmission function . only if the distance between molecules becomes so small , that inter - molecular interactions are no longer negligible ( which is below 6 in our case ) , does the transmission function undergo a qualitative change . namely an additional peak structure occurs . how does this finding compare to the experimental data ? as we have pointed out in section [ sec : experiment ] , only in devices using molecules with two cyanide end - groups a more or less random peak structure was observed in the @xmath1 characteristic@xcite . in other devices , molecules with at least one thiol end - group are typically used . these show significantly less peak structure . we therefore give the following interpretation : the thiol end - group is known to bind strongly to au atoms . it is therefore likely , that thiol - based mono - layers stably adsorb to gold leads . resulting periodic structures are then robust against distortions . the conductance of such structures is proportional to the corresponding single molecule conductance , i.e. the number of molecules involved changes the absolute value of the current only , not the peak structure . the random like peak structure in devices , made up of cyanide based molecules suggests , that there are a some molecular clusters present in the mono - layer . these clusters might occur , because the binding of a cyanide end - group to au is considerably weaker compared to that of a thiol end - group , and weaker binding results in a less robust periodic structure . we would like to thank xavier bouju , as well as udo beierlein for helpful discussions . the current properties of each channel can be related to the corresponding eigenvalue . we start from equation : @xmath144 let s first consider @xmath145 , i.e. @xmath146 , i.e. @xmath147 . for each channel with eigenvalue @xmath148 one then must have @xmath149 , i.e. this channel does not itself carry any current . this is consistent with our terminology of an evanescent wave . if , however , @xmath150 , then @xmath151 is purely imaginary , because @xmath152 is an anti - hermitian operator . we can therefore define the velocity of a propagating wave to be @xmath153 . now we consider the case of two different solutions @xmath154 and define @xmath155 . if their eigenvalues do not satisfy @xmath156 , then the current between these two solutions is zero @xmath157 . so let s assume @xmath158 . because if @xmath159 then @xmath160 , a current can flow between an evanescent left going wave and an evanescent right going wave . but if we restrict ourselves to solutions with finite amplitudes in a semi - infinite lead , then either the left or right going wave amplitude must be zero . therefore evanescent waves do neither carry a current themselves nor do they exchange current with other channels , that is they do not at all contribute to the net current . finally we are left with the case @xmath161 , with @xmath162 . this is equivalent to @xmath163 , i.e. the case of degenerate eigenvalues . therefore propagating waves to degenerate eigenvalues do exchange current . that in turn means , that the current of a superposition of two such waves does not necessarily equal the sum of the two individual currents , which is problematic as we want to express the total current as a sum of independent channels . however , the propagating and evanescent waves were obtained by diagonalising the propagator @xmath42 . this transformation is unique up to rotations in every degenerate eigenvalue subspace . because @xmath73 is anti - hermitian we can diagonalise these subspaces and the resulting diagonal elements will be purely imaginary . so the net current may be written as a summation over all the individual contributions of propagating channels , only if these subspace rotations are performed . summarising we have shown that the transformation @xmath84 diagonalising the propagator @xmath42 ( i.e @xmath164 ) can be chosen such that the transformation @xmath165 of the current operator is diagonal in the subspace of propagating waves with purely imaginary diagonal elements . all the other diagonal entries are zero and the only non - zero non - diagonal elements belong to evanescent waves in opposite directions . the part of the hamiltonian containing the molecular region and its coupling to the leads can be written as @xmath166 \|\psi\rangle = 0.\ ] ] ( using this order for the coefficients it is straight forward to extend all formulas to the general case of more than two leads . ) the indices 1 and 2 indicate source and drain lead surface layers , while the index 0 is used for the molecular region . @xmath167 are the coupling matrices from source / drain to the molecules . we now transform into the basis of incoming and outgoing channels , i.e. we apply @xmath168 , \quad\text{with } u^i=\left[\begin{array}{cc } u^i _ > & u^i _ < \\ u^i_>\lambda^i _ > & u^i_<\lambda^i_<\\ \end{array}\right]\ ] ] from the right to equation : @xmath169,\ ] ] with @xmath170 the first and third column act on the surface layer of the incoming channels , the second and fourth act on outgoing ones , while the fifth column , acting on the molecular region , remains unchanged . the scattering matrix expresses the outgoing channel amplitudes in terms of the incoming ones . therefore we split the matrix of equation into two parts , one containing the outgoing columns , the other one containing the incoming ones as well as the molecular column : @xmath171\quad\text{and}\\ m_\text{in } & : = \left [ \begin{array}{cc } a^1 _ > & 0 \\ 0 & a^2_>\\ b^1 _ > & b^2_>\\ \end{array } \right ] . \end{aligned}\ ] ] the first matrix @xmath172 is a square matrix and by inverting it , we obtain the scattering matrix @xmath173 m. j. frisch , g. w. trucks , h. b. schlegel , g. e. scuseria , m. a. robb , j. r. cheeseman , v. g. zakrzewski , j. a. montgomery , jr . , r. e. stratmann , j. c. burant , s. dapprich , j. m. millam , a. d. daniels , k. n. kudin , m. c. strain , o. farkas , j. tomasi , v. barone , m. cossi , r. cammi , b. mennucci , c. pomelli , c. adamo , s. clifford , j. ochterski , g. a. petersson , p. y. ayala , q. cui , k. morokuma , d. k. malick , a. d. rabuck , k. raghavachari , j. b. foresman , j. cioslowski , j. v. ortiz , a. g. baboul , b. b. stefanov , g. liu , a. liashenko , p. piskorz , i. komaromi , r. gomperts , r .l . martin , d .j . fox , t. keith , m .a . al - laham , c. y. peng , a. nanayakkara , c. gonzalez , m. challacombe , p. m .w . gill , b. g. johnson , w. chen , m. w. wong , j. l. andres , m. head - gordon , e. s. replogle and j. a. pople , gaussian 98 ( revision a.7 ) , gaussian , inc . , pittsburgh pa , 1998 .
during the past several years considerable interest has been shown in exploring localized nonlinear waves in the variable coefficient nonlinear schrdinger ( nls ) equation and its generalizations @xcite . the motivation comes from the fact that nls equation and its variants appear in several branches / topics of physics , including nonlinear optics @xcite and bose - einstein condensates ( becs ) @xcite , etc . focusing our attention on becs alone , it is well known that the gross - pitaevskii ( gp ) equation governs the evolution of macroscopic wave function at ultra low temperatures @xcite . in particular , for cigar shaped becs , it has been shown that the gp equation can be reduced to the 1d variable coefficient nls equation @xcite @xmath0 where @xmath1 is the condensate wave function , @xmath2 is the dimensionless normalized time , @xmath3 is the dimensionless normalized coordinate in the axial direction , @xmath4 represents the effective scattering length and @xmath5 is the axial trap frequency . a simple and straightforward way of exploring the localized / periodic structures of ( [ eq:1d - gp ] ) is by transforming it into a constant coefficient nls equation through a suitable transformation . from the known solutions of the latter equation the solutions of the former equation can be identified . using this procedure a class of solutions , in particular various localized solutions , have been identified for the model ( [ eq:1d - gp ] ) , including soliton , breather , and rogue wave ( rw ) solutions . in order to appreciate the relevance of the above type of localized structures for eq . ( [ eq:1d - gp ] ) , we may first consider their existence in the case of constant coefficient nls equation , @xmath6 it is well known that the standard nls equation ( [ nls ] ) admits the following basic localized profiles and their higher - order versions @xcite . [ [ i - envelope - soliton ] ] ( i ) envelope soliton : + + + + + + + + + + + + + + + + + + + + + it is a solitary wave ( localized envelope along with a carrier wave ) that retains its characteristics ( amplitude / shape and velocity ) under collision with another soliton , except for a change in phase . the intensity profile of the soliton is shown in fig . [ figa](a ) . the typical form of the envelope soliton is @xmath7 where @xmath8 is a complex constant , @xmath9/2+\eta_{1i}^{(0)}$ ] and @xmath10 . here @xmath11 and @xmath12 are constant parameters . [ [ ii - breather ] ] ( ii ) breather : + + + + + + + + + + + + + + it is a localized solution with temporally and/or spatially periodic structures having constant background exhibiting internal oscillations and bound states of nonlinear wave packets @xcite which is represented in fig . [ figa](b ) . its typical form reads @xcite @xmath13 \label{nls2}\end{aligned}\ ] ] where @xmath14 is the amplitude of the plane wave , @xmath15 , @xmath16 , @xmath17 , @xmath18 are constant parameters , @xmath19 , @xmath20 , @xmath21 , @xmath22 , @xmath23 , @xmath24 , @xmath25 , and @xmath26 is a complex constant . this solution is also called a general breather ( gb ) since it is periodic both in space and in time . two important special cases are the following : ( i ) when @xmath27 and @xmath28 , the gb solution corresponds to an akhmediev breather ( ab ) which is periodic in space and localized in time . ( ii ) if we take @xmath29 and @xmath30 , ( [ nls2 ] ) becomes a ma breather ( mb ) which is periodic in time only and localized in space . [ [ iii - rogue - wave ] ] ( iii ) rogue wave : + + + + + + + + + + + + + + + + + a further specialized structure , which is localized both in space and in time with a constant plane wave background , is the so called peregrine soliton or rw @xcite . it can be obtained by taking the limits @xmath31 and @xmath32 , and @xmath33 , where @xmath34 , @xmath35 and @xmath36 are constants . ( [ nls2 ] ) reduces to the following form , @xmath37 very often the above localized nonlinear wave is described as a wave that appears from nowhere and disappears without a trace " @xcite as shown in fig . [ figa](c ) . it was first observed in the area of oceanography @xcite and is traditionally defined as a wave whose wave - height ( the distance from trough to crest ) is more than twice the significant wave height ( swh ) . the latter is generally defined as the average wave height among one third of highest waves in a given time series @xcite . further , it has also been explained that the above structure ( [ nls3 ] ) arises due to a modulation instability ( mi ) @xcite of the plane wave solution of the constant coefficient nls equation . very recently higher - order rws ( hrws ) which correspond to the higher - order rational solutions of the nls equation ( [ nls ] ) have been deduced @xcite . the explicit expressions for the second and third - order rws of the nls equation are given in appendix a. these hrws have higher amplitudes than the first - order rw . the rws have also been observed experimentally in physical systems such as water wave tank @xcite , capillary waves @xcite and nonlinear optics @xcite . several theoretical studies on the dynamics of rws in nonlinear fiber optics @xcite , plasma physics @xcite , laser - plasma interactions @xcite , and even econophysics @xcite , described by scalar nls equation , have been made in recent times . now , as mentioned in the beginning , the dynamics of a cigar shaped bec at absolute zero temperature is usually described by the mean - field gp equation ( [ eq:1d - gp ] ) , which is a generalized form of the ubiquitous constant coefficient nls equation ( [ nls ] ) , for the wave function of the condensate . since the nls equation ( [ nls ] ) admits breather and rw solutions , it is natural to expect that rws and breathers may also be found in bec systems as well . in this context , the rws can correspond to a sudden increase of peaks in the condensate clouds similar to the nature of high peaks in the open sea , while breathers are generalizations of the rws . the formative mechanism for the matter rws in becs is the accumulation of energy and atoms towards its central part and their spreading out to a constant density background . the formation of matter breather is the periodic exchange of atoms between the profile and the plane wave background . from an experimental point of view , the existence of rw and breather structures in a bec system can be effectively controlled by tuning the nonlinear interaction between atoms by feshbach resonance technique @xcite and modulating the trapping frequency of the external potential . it will therefore be of great interest to study the characteristics and/or controlling of structures of rws and breathers due to their localization both in space and in time in bec experiments . past explorations of gp equation in becs have paved the way for important developments in manipulating coherent matter waves for application , including atom interferometry @xcite , coherent atom transport @xcite and quantum information processing or quantum computation @xcite . therefore , it is of high significance to study the dynamics of rws and breather profiles of the gp equation ( [ eq:1d - gp ] ) . however only few attempts have been made to identify and analyze the rws and breather solutions of ( [ eq:1d - gp ] ) @xcite . to the best of our knowledge neither higher - order rw solutions ( with certain free parameters ) nor higher order breather solutions of ( [ eq:1d - gp ] ) have been taken up for study . motivated by these observations , in this work , we construct the aforementioned localized and periodic solutions of ( [ eq:1d - gp ] ) . besides constructing these two families of solutions we also investigate how to manipulate the rws and breathers through the effective scattering length and the strength of trap parameter . having stated the motivation we now proceed to construct a transformation that transforms eq . ( [ eq:1d - gp ] ) to the standard nls equation ( [ nls ] ) . following the standard procedure @xcite we find that the required similarity transformation should be of the form @xmath38 \notag \\ & \times \exp \left[{i\left(c_1r_0 ^ 2 rx-\frac{r_t}{r}x^2-\frac{1}{2}c_1 ^ 2r_0 ^ 4\int{r^2(t)}dt\right)}\right],\end{aligned}\ ] ] where [ a4 ] @xmath39 and @xmath40 is the solution of the standard nls equation ( [ nls ] ) . in the above @xmath41 are arbitrary constants and the modulational functions @xmath4 and @xmath42 should satisfy the following condition @xmath43 which is a riccati type equation with dependent variable @xmath44 and independent variable @xmath2 . regardless of what @xmath4 is , as long as the condition ( [ a5 ] ) is satisfied , the gp equation is integrable @xcite . we also note here that the painlev singularity structure analysis performed on eq . ( [ eq:1d - gp ] ) confirms the same restriction ( [ a5 ] ) on the system coefficients @xcite . we further note that the solution ( [ a7 ] ) provides us some flexibility to generate new structures related to the rws which may be useful for the bec experiments . even though one can arbitrarily choose the functions @xmath4 and @xmath5 that satisfy the constraint ( [ a5 ] ) and generate the required solutions , in this paper we consider the trap frequency to be of the following three forms : ( i ) @xmath45 , ( ii ) @xmath46 $ ] and ( iii ) @xmath47 $ ] , where @xmath48 is a constant . as shown in reference @xcite the effective scattering length for these three cases turn out to be ( i ) @xmath49 , ( ii ) @xmath50 and ( iii ) @xmath51 . we consider the trap frequency to be in the above forms since it has been shown that they are valid forms in bec experiments @xcite . we consider each one of the cases separately and substitute them in ( [ a7 ] ) along with the rw and breather solutions of ( [ nls ] ) . we then analyze in detail how the nature of the rw and breather structures get modified by the above two functions @xmath5 and @xmath4 . our analysis shows that the amplitude parameter @xmath52 plays a vital role in the formation of rw , and the trap frequency and the effective scattering length modify the structure of the rw and breather profiles , allowing one the possibility of manipulating the rws and breathers in specific ways . the paper is organized as follows . in the following section , we construct rw solutions for time - independent and time - dependent traps and study their characteristics in detail . to the best of our knowledge , for the first time in the literature , we observe that the second- and third - order rws transform to first - order rw - like structures when we tune a parameter which appears in the harmonic traps ( time - independent and time - dependent traps ) . in sec . iii , we also construct rw solutions with free parameters which allow us to split the symmetric form solution into a multi - peaked solution for ( [ eq:1d - gp ] ) and investigate how these rw structures get modified in the plane wave background by increasing the strength of the trap . in sec . iv , we construct one - breather and two - breather solutions of ( [ eq:1d - gp ] ) and investigate their characteristics when we alter the trap parameter . finally , in sec . v , we present a summary of the results and our conclusions . to begin with , we consider the case in which the trap frequency is a constant , that is @xmath42 = constant = @xmath53 . a time - independent trap frequency implies that the frequency does not change with time and space . we then consider the trap frequency to be time - dependent and investigate the associated rw solutions . substituting @xmath45 in the integrability condition @xmath54 , we find that the time - dependent interaction term should be of the form @xmath49 , where @xmath35 is an integration constant . ) through split - step crank - nicolson method for the time - dependent nonlinearity coefficient @xmath49 and time - independent trap frequency @xmath45 . the initial condition chosen corresponds to the analytic solution of fig . [ fig8](d ) . ] plugging this expression in ( [ a7 ] ) , we find @xmath55 where @xmath56\right\ } , \notag\end{aligned}\ ] ] and @xmath57 s , @xmath58 , are the first , second and third - order rw solutions of the nls equation ( [ nls ] ) whose explicit expressions are given in the appendix ( vide eqs . ( [ a8 ] ) , ( [ a11 ] ) and ( [ a12 ] ) ) . also @xmath59 and @xmath60 have the forms as given in eqs . ( [ a4 ] ) . in fig . [ fig8 ] , the first , second , and third columns represent the density profiles of the first , second , and third - order rws obtained from ( [ a15 ] ) . in this figure we present the formation of rws in cigar - shaped becs . in the context of becs , it is the fluctuation in the density of atoms , localized in space and time , which is what we observe as rws . for instance consider the first - order rws depicted in figs . [ fig8](a)-(d ) . these may be interpreted as follows : atoms in the condensate , initially at a constant density background , suddenly accumulate to form a hump towards the center of the condensate at finite time while leaving voids in the density which appear as troughs in the rws , depending on the initial state . the crest and troughs reach their extrema as time progresses , and then the condensate atoms spread out so as to recover the constant density background in finite time , thus revealing the unstable nature of rws . further , by tuning the amplitude parameter @xmath61 in ( [ a15 ] ) , we can visualize the formation and manipulation of rws . for example , when we increase the amplitude parameter @xmath61 smoothly from @xmath62 to @xmath63 , we can observe in figs . [ fig8](a)-(d ) the formation of more and more localized first - order rws with increasing amplitude . at @xmath64 one can visualize a large amplitude wave which is sufficiently localized both in space and in time which in turn confirms the formation of the first - order rw in bec . the formation of second and third - order rws is also demonstrated in figs . [ fig8](c)-(h ) and [ fig8](i)-(l ) , respectively , for the same set of parameter values of @xmath61 . from these figures we infer that the time - dependent nonlinear interaction between the atoms induces density fluctuations over the condensate which gets more and more localized both in space and in time as we increase the order of the rw . in order to confirm the existence of the rws further , we have also performed a direct numerical simulation of ( [ eq:1d - gp ] ) with the aid of the split - step crank - nicolson method using an initial wave function which is the same as the function ( [ a15 ] ) and with space step @xmath65 and time step @xmath66 @xcite . in fig . [ num - rog ] , we have presented the computer generated density profile of the first - order rw and the corresponding contour plot with the parameters chosen as @xmath64 , @xmath67 , @xmath68 , and @xmath69 which are same as that of fig . [ fig8](d ) . one can easily observe a very good agreement between the numerical results and the analytical predictions for the emergence of the rws . we have also verified numerically the presence of second- and third - order rws of ( [ eq:1d - gp ] ) as well , replicating figs . [ fig8 ] . next we demonstrate how these localized structures vary with respect to the trap parameter @xmath48 . [ fig1 ] displays the first - order rw for the same nonlinearity management parameter @xmath49 and the external trap frequency @xmath45 . the nature of the first - order rw for @xmath68 is depicted in fig . [ fig1](a ) . when we increase the strength of the trap parameter the density profiles corresponding to the first - order rw become more and more localized in time as shown in figs . [ fig1](b ) and [ fig1](c ) , respectively . the corresponding contour plots are given in figs . [ fig1](d)-(f ) . [ fig1a ] displays the density profiles of the second - order rws for the same nonlinearity management parameter as a function of @xmath48 . in fig . [ fig1a](a ) we display the second - order rw for @xmath68 . when the strength of the parameter @xmath48 is increased to @xmath63 , the wave subcrests start to stretch as shown in fig . [ fig1a](b ) . the wave subcrests become more and more localized in time when we increase the value of @xmath48 and finally the second - order rw attains a new structure as given in fig . [ fig1a](c ) for @xmath70 . the resultant structure looks almost like a first - order rw , see fig . [ fig1](c ) . figs . [ fig1a](d)-(f ) are the corresponding contour plots . we also observe similar effects in the third - order rw case as well when we increase the value of @xmath48 . [ fig3a ] demonstrates the changes in the third - order rw when we vary the interaction strength . at @xmath71 , the third - order rw transforms into the second - order rw - like structure as shown in fig . [ fig3a](b ) . when we increase the value of the parameter @xmath48 to @xmath72 , one obtains a first - order like rw which is displayed in fig . [ fig3a](c ) . these facts are also confirmed by the corresponding contour plots , figs . [ fig3a](d)-(f ) . the aforementioned results reveal that when we increase the strength of the trap parameter @xmath48 , the second and third - order rws become more localized in time and delocalized in space , approaching the structure of a first - order rw . thus the robustness of the density profiles can be controlled by varying the strength of trap frequencies . next we consider the time - dependent trap frequency in the form @xmath46 $ ] . for this choice , the relation ( [ a5 ] ) fixes the interatomic interaction term to be of the form @xmath50 . the first , second , and third - order rw solutions for this trap frequency and strength of interatomic interaction are found to be @xmath73where @xmath74 and @xmath75}-c_1r_0 ^ 2\left[1+\tanh{\left(\frac{\beta_0 t}{2}\right)}\right]x + \frac{c_1 ^ 2r_0 ^ 4\left(\beta_0 t+2\log\left[\cosh(\frac{\beta_0 t}{2})-\tanh(\frac{\beta_0 t}{2})\right]\right)}{\beta_0}\right]\right\}. \notag\end{aligned}\ ] ] the qualitative nature of the first , second , and third - order rws for @xmath76 and @xmath77 $ ] turns out to be the same as in the previous case ( fig . [ fig8](d ) , ( h ) and ( l ) ) when the amplitude parameter @xmath61 is varied and so we do not display the outcome here separately . on the other hand , we identify interesting structures while varying the parameter @xmath48 which is discussed in the following . in fig . [ fig4 ] , we depict the first - order rw for these choices of @xmath4 and @xmath5 . when @xmath78 the first - order rw is as shown in fig . [ fig4](a ) ( see also fig . [ fig4](d ) for the corresponding contour plot ) . by altering the value of the trap parameter @xmath48 to @xmath79 , the structure of the first - order rw gets modified as shown in fig . [ fig4](b)/fig . [ fig4](e ) . one can also see from fig . [ fig4 ] that , as the trap parameter @xmath48 is increased , the rw gradually becomes more localized in time and the condensate atoms settle down to a slightly higher density background due to the attractive nature of the potential . when @xmath70 the modified structure of first - order rw is given in fig . [ fig4](c)/fig . [ fig4](f ) , where this feature is even more prominent . the density profiles of the corresponding second - order rws are presented in fig . when we tune the parameter @xmath48 from @xmath80 upwards the wave subcrests start to stretch . from the contour plots we can observe that the stretches occur on one side of the rw only . when we increase the value of @xmath48 further the second - order rw gets modified to a first - order rw like structure which is demonstrated in fig . [ fig5](c ) . a similar transition has also been observed in the third - order rw case as well which is illustrated in fig . the third - order rw acquires a new structure as shown in fig . [ fig6](b)/fig . [ fig6](e ) when we increase the value of the parameter @xmath48 from @xmath80 to @xmath79 . at @xmath81 , we observe that the third - order rw acquires a further modified structure which is displayed in fig . [ fig6](c ) and fig . [ fig6](f ) . note the similarity in the central part with that of the first - order rw as given in fig . [ fig4](c ) and fig . [ fig4](f ) . in the case of the third choice we consider the time - dependent periodic trap frequency to be of the form @xmath47 $ ] so that the strength of the time - dependent periodic interatomic interaction turns out to be @xmath51 . substituting these two expressions in ( [ a7 ] ) , we find @xmath82 where @xmath83\right\}.\notag\end{aligned}\ ] ] here @xmath57 , @xmath84 , are again the first , second and third - order rw solutions of the nls equation ( vide eqs . ( [ a8 ] ) , ( [ a11 ] ) , and ( [ a12 ] ) ) . in fig . [ fig7 ] , we present the density profiles of the first , second , and third - order rws ( first row ) and their corresponding contour plots ( second row ) for the strength of time - dependent interatomic interaction @xmath51 and time - dependent periodic trap frequency @xmath47 $ ] . fixing the value of @xmath48 at @xmath85 , from fig . [ fig7 ] we see that the rw exists on a periodic background for the above forms of @xmath4 and @xmath86 . further we have also verified that the behaviour of the rws as a function of @xmath48 follows the same qualitative picture discussed above for the other forms of @xmath4 and @xmath86 . very recently it has been shown that one can generalize the expressions for the higher order rw solutions of the scalar nls equation ( [ nls ] ) given in appendix a further by introducing certain free parameters which allow one to split a symmetric form rw solution into a multi - peaked solution and that by varying these free parameters one can extract certain novel patterns of rws @xcite . the introduction of free parameters decompose the higher - order rw solutions into @xmath87 first - order forms , where @xmath88 is the order of the rw , however maintaining the symmetry of the higher - order solutions even in their decomposed forms . further the free parameters are shown to determine the size and orientation of the first - order solutions @xcite . triplets are symmetry preserving first - order structures revealing the fact that the ( higher- ) second - order rw solution is a family of three first - order rational solutions . it is evident that the existence of triplets in an ocean corresponds to three big waves on the water surface in a row or three sister waves " @xcite . inspired by such a possibility , in the following , we consider the second- and third - order rw solutions of ( [ eq:1d - gp ] ) with suitable free parameters and analyze the symmetrical structures that arise due to these free parameters when we vary the strength of nonlinearity and the trap parameter . to begin with , we confine our attention to the second - order rw solution . in this case , we have the following modified expressions @xcite for @xmath89 , @xmath90 and @xmath91 in eq . ( [ a9 ] ) , that is @xmath92 , \notag \\ h_2 = & \ , 24 \big\ { t\big [ 15 - 16 x^4 + 24 x^2 - 48 l x-8(1 - 4 x^2)t^2 \notag \\ & \ , -16 t^4 \big]+6m(1 - 4 t^2 + 4x^2)\big\ } , \notag \\ d_2 = & \ , 64 x^6 + 48 x^4 ( 4t^2 + 1)+12x^2(3 - 4t^2)^2 + 64 t^6\notag \\ & \ , + 432 t^4 + 396 t^2 + 9 + 48m\big[18m+t(9 + 4 t^2 \notag \\ & \ , -12x^2)\big]+48l\big[(18l+x(3 + 12t^2 -4x^2)\big ] . \end{aligned}\ ] ] note that this rw solution now contains two free parameters , namely @xmath93 and @xmath94 . the parameters @xmath93 and @xmath94 describe the relative positions of the first - order rws in the triplet . substituting the above expressions in ( [ a15 ] ) , ( [ a22 ] ) and ( [ a24 ] ) , for @xmath95 , we obtain the corresponding second - order rw solutions to the gp equation ( [ eq:1d - gp ] ) with the free parameters @xmath93 and @xmath94 included . when @xmath96 , this solution coincides with the one given earlier ( vide eq . ( [ a11 ] ) ) which contains one largest crest and four subcrests with two deepest troughs ( fig . [ fig8](h ) ) . when @xmath93 and @xmath94 are not equal to @xmath97 , the second - order rw splits into three first - order rws . these waves emerge in a triangular fashion ( a triplet pattern ) . the three first - order rws form a triangular pattern with @xmath98 degrees of angular separation between them @xcite . in figs . [ fig8b](a)-(c ) we display the triplet pattern for @xmath49 and @xmath45 when @xmath99 and @xmath100 . the triplet rw pattern for @xmath68 is shown in fig . [ fig8b](a ) . when we increase the parameter @xmath48 to @xmath101 we observe that the triplet pattern has started to collapse as shown in fig . [ fig8b](b ) and a complete collapse is observed as in fig . [ fig8b](c ) when we increase the @xmath48 value further to @xmath79 . figs . [ fig8b](d)-(g ) represent the triplet pattern for @xmath76 and @xmath77 $ ] . the formation of triplet rws is shown in fig . [ fig8b](d ) when @xmath68 . when we increase the value @xmath48 to @xmath101 one of the single rws in the triplet pattern vanishes which is illustrated in fig . [ fig8b](e ) . by increasing the parameter @xmath48 further we observe that two first - order rws are more localized in time as shown in fig . [ fig8b](f ) . figs . [ fig8b](g)-(i ) represent the triplet pattern for @xmath51 and @xmath47 $ ] . the form of the triplet pattern for @xmath68 is displayed in fig . [ fig8b](g ) . here also when we increase the value of @xmath48 we observe the collapse of the triplet pattern in the periodic wave background ( vide fig . [ fig8b](i ) ) . we then move on to investigate the structure of the third - order rw solution with four free parameters , namely @xmath102 and @xmath103 . the third - order rw solution with four free parameters is much lengthier than the one without free parameters and so we do not give the explicit expression here and analyze the results only graphically . here also we analyze the solution with respect to the free parameters . when @xmath104 , we have the classical third - order rw solution which is shown in fig . [ fig8](l ) . it has one largest crest and six subcrests with two deepest troughs . for non - zero values of @xmath93 , @xmath94 , @xmath105 and @xmath103 , the third - order rw splits into six separated first - order rws . when we increase the value of the free parameters , the six first - order rws take new positions . the sextet pattern is displayed in figs . [ fig8h](a)-(c ) for @xmath49 and @xmath45 when @xmath106 , @xmath107 , @xmath108 and @xmath109 . for @xmath68 the set of six first - order rws is shown in fig . [ fig8h](a ) . when we increase the value of @xmath48 to @xmath110 , three peaks disappear and the remaining peaks start to bend as shown in fig . [ fig8h](b ) . when we increase the value of @xmath48 further the rws bend in the plane wave background as given in fig . [ fig8h](c ) . figs . [ fig8h](d)-(f ) represent six first - order rws for the time - dependent nonlinearity strength @xmath76 and the time - dependent external trap frequency @xmath77 $ ] . the formation of six first - order rws at @xmath68 is shown in fig . [ fig8h](d ) . when we increase the strength of the parameter @xmath48 to @xmath110 , one of six first - order rws vanishes as seen in fig . [ fig8h](e ) . if we increase the value @xmath48 further three out of six peaks bend in the plane wave background and eventually collapse which is shown in fig . [ fig8h](f ) . figs . [ fig8h](g)-(i ) represent the sextet pattern for @xmath51 and @xmath47 $ ] . the form of the six first - order rws for @xmath111 is displayed in fig . [ fig8h](g ) and for further increase in @xmath48 , the modified structures in the periodic wave background are as shown in figs . [ fig8h](h ) and [ fig8h](i ) . in the previous two sections we have analyzed how the rw profiles get modified by the variations of the distributed coefficients present in the variable coefficient nls eq . ( [ eq:1d - gp ] ) . in this section we analyze how the breather structures get modified in the condensates when we vary the strength of the external trap parameter . to begin with we consider the first - order breather solution of the nls equation ( [ nls ] ) , which is given in @xcite and is a special case of the gb solution ( [ nls2 ] ) , @xmath112 + 2 i \,f v \sinh[\alpha ( t - t_1)]}{2 \cosh[\alpha ( t - t_1 ) ] - 2v \cos [ f ( x - x_1 ) ] } \notag \\ & \ , -1\biggr ] \times \exp{(it ) } , \label{b1}\end{aligned}\]]where the parameters @xmath113 and @xmath114 are expressed in terms of a complex eigenvalue ( say @xmath115 ) , that is @xmath116 and @xmath117 , and @xmath118 and @xmath119 serve as coordinate shifts from the origin . the real part of the eigenvalue represents the angle that the one - dimensionally localized solutions form with the @xmath120 axis , and the imaginary part characterizes frequency of periodic modulation . the parameter @xmath121 @xmath122 in ( [ b1 ] ) is the growth rate of modulation instability . substituting this breather solution of the nls equation into ( [ a15 ] ) , ( [ a22 ] ) and ( [ a24 ] ) , we study the underlying dynamics of ( [ eq:1d - gp ] ) . , ( b ) mb for @xmath123 when @xmath49 and @xmath45 . ( c ) and ( d ) are their corresponding contour plots . the other parameters are @xmath124 , @xmath125 , @xmath67 , and @xmath69 . ] for illustration , we consider the case @xmath49 and @xmath45 , and plot the outcome in fig . [ fig9 ] the first row in fig . [ fig9 ] represents an akhmediev breather ( ab ) solution for the eigenvalue @xmath126 and ma breather ( mb ) solution for @xmath123 and the second row represents their contour plots . the ab and mb solutions are localized in time and space , respectively , as discussed in sec . i. this is clearly demonstrated in the two columns of fig . when we tune the parameter @xmath48 , a new structure against a breather background is obtained . to visualize this we fix the value of @xmath48 to be @xmath127 . for this value a stretching occurs in space in the case of ab ( fig . [ fig9a](a ) ) whereas in the case of mb ( fig . [ fig9a](b ) ) the stretching occurs in time [ fig9a](c)-(d ) illustrate the ab and mb profiles for the time - dependent nonlinearity coefficient @xmath76 and time - dependent trap frequency @xmath128 $ ] . here also we tune the strength of the trap parameter @xmath48 to 0.8 and observe that stretching occurs over space in ab which is depicted in fig . [ fig9a](c ) and the mb gets more localized and when @xmath129 the breather profile completely disappears as shown in fig . [ fig9a](d ) . our results reveal the fact that when we tune the parameter @xmath48 in the obtained breather solution , the breather gets a modified structure corresponding to a distortion of the breather profile . , ( b ) , ( d ) and ( f ) mb with @xmath123 for the three different forms of @xmath4 discussed in the text . ( a)-(d ) @xmath130 , ( e)-(f ) @xmath71 . the other parameters are @xmath124 , @xmath67 and @xmath69 . ] figs . [ fig9a](e ) and ( f ) respectively represent the ab and mb in the periodic background for @xmath51 and @xmath47 $ ] when @xmath71 . next , we proceed to construct the two - breather solutions of ( [ eq:1d - gp ] ) and analyze how these solutions are distorted by the variations of modulation parameters . the two - breather solution of the nls equation is given by @xcite @xmath131\exp{(it)},\end{aligned}\ ] ] where @xmath132 , @xmath133 , and @xmath134 are given by @xmath135 , \end{aligned}\ ] ] @xmath136 , \end{aligned}\ ] ] @xmath137 , \end{aligned}\ ] ] where the modulation frequencies , @xmath138 , @xmath139 , are described by the ( imaginary ) eigenvalues @xmath140 . in the above expressions , @xmath141 , @xmath142 , @xmath139 , represent the shifted point of origin , @xmath143 @xmath144 is the instability growth rate of each component and @xmath145 and @xmath146 are shifted variables . with two purely imaginary eigenvalues , @xmath140 , @xmath139 , the solution ( [ 2b1 ] ) is capable of describing a variety of possible second - order breather structures . the solution includes abs , mbs and the intersection of ab and mb solutions for certain combinations of eigenvalues . for example , when the imaginary parts of both the eigenvalues @xmath147 , @xmath139 , lie between @xmath97 and @xmath148 , we obtain the abs . on the other hand when both of them are greater than one ( @xmath149 ) we obtain the mbs and in the mixed possibility case , that is one of the eigenvalues is less than one ( @xmath150 ) and the other eigenvalue ( @xmath151 ) is greater than one , we obtain the intersection of ab and mb solutions . substituting the two - breather nls solution ( [ 2b1 ] ) in ( [ a15 ] ) , ( [ a22 ] ) and ( [ a24 ] ) we obtain the general two - breather solution of ( [ eq:1d - gp ] ) . [ br2 ] displays the evolution of the two - breather solution of ( [ eq:1d - gp ] ) for @xmath49 and @xmath152 with imaginary eigenvalues . to obtain the abs from ( [ 2b1 ] ) we consider the situation where the magnitudes of both the eigenvalues @xmath153 and @xmath154 are less than @xmath148 ( @xmath155 and @xmath156 ) . one ab developing with a time delay after another is shown in fig . [ br2](a ) , while in fig . [ br2](b ) we present the case where there is no such time delay . in fig . [ br2](c ) we depict the case when one ab along with a rw coexist for the choice @xmath155 and @xmath157 . when we take the eigenvalues @xmath153 and @xmath154 to be @xmath158 and @xmath159 , respectively , we obtain two mb solutions . similarly the evolution of the two mbs with and without spatial delay is shown in figs . [ br2](d ) and [ br2](e ) , respectively . we also observe that the distance between the mbs increases when we set both the eigenvalues to be nearly equal , say for example @xmath160 and @xmath161 which is not displayed here . when we take the eigenvalues as @xmath162 and @xmath163 , the ab intersects with the mb which is displayed in fig . [ br2](f ) . when we tune the strength of the trap parameter @xmath48 to @xmath164 , both the abs get stretched in the plane wave background which is demonstrated in fig . [ br2a](a ) . when @xmath165 , in the intersection of ab and mb solutions we observe that the ab gets a bending structure while the mb fully disappears in the plane wave background which is shown in fig . [ br2a](b ) . in fig . [ br2a](c ) we note that both the mbs develop a bending structure in the plane wave background when @xmath166 . figs . [ br2a](d)-(f ) display the evolution of the two - breather solution of ( [ eq:1d - gp ] ) for @xmath76 and @xmath77 $ ] with the imaginary eigenvalues . when @xmath167 one of the abs gets stretched which is not shown here and on further increase of the value of @xmath48 to @xmath168 , one of the abs gets annihilated and the other ab bends in the plane wave background which is demonstrated in fig . [ br2a](d ) . when @xmath169 the intersection of ab - mb structures is as shown in fig . [ br2a](e ) . in fig . [ br2a](f ) we observe that both the mbs get a bending structure in the plane wave background . figs . [ br2a](g)-(i ) show the evolution of the two - breather solution of ( [ eq:1d - gp ] ) for @xmath51 and @xmath47 $ ] with the imaginary eigenvalues . here also when we increase the value of @xmath48 we observe the collapse of the two - breather solution in the periodic wave background . in this work , we have constructed higher order rw solutions with and without free parameters for the quasi one - dimensional gp equation with time - dependent interatomic interation and external trap through the similarity transformation technique . by mapping the variable coefficient nls equation onto the constant coefficient nls equation we have derived these solutions . we have shown that the mapping can be done when the external trap and the nonlinearly interatomic interaction of atoms satisfy a constraint . from the known higher order rw and breather solutions of the constant coefficient nls equation , we have derived the solutions of ( [ eq:1d - gp ] ) . in our analysis , we have considered the harmonic trap frequency in three different forms , namely ( i ) time - independent expulsive trap , ( ii ) time - dependent monotonous trap , and ( iii ) time - dependent periodic trap , and correspondingly fixed the effective scattering length . we then studied the characteristics of the constructed rw solutions in detail . we have observed that the second- and third - order rws transform to first - order rw - like structures when a parameter appearing in the harmonic trap ( time - independent and time - dependent traps ) is varied . we have then analyzed the characteristics of triplet and sextet patterns of matter rws for ( [ eq:1d - gp ] ) . we have also constructed one - breather and two breather solutions of ( [ eq:1d - gp ] ) . we have investigated how these periodic localized waves change in the plane wave background when we tune the trap parameter in the obtained breather solutions . our results may provide possibilities to manipulate rws experimentally in a bec system . km thanks the university grants commission ( ugc - rfsms ) , government of india , for providing a research fellowship . the work of pm forms part of department of science and technology ( ref . sr / s2/hep-03/2009 ) and council of scientific and industrial research ( ref . 03(1186)/10/emr - ii ) , government of india funded research projects . the work of ms forms part of a research project sponsored by nbhm , government of india . the work forms part of an irhpa project and a ramanna fellowship project of ml , sponsored by the department of science and technology ( dst ) , government of india , who is also supported by a dae raja ramanna fellowship . in the absence of the trap , and the nonlinearity strength @xmath4 is equal to one , eq . ( [ eq:1d - gp ] ) reduces to the standard nls equation @xmath170 . several localized and periodic structures of standard nls are documented in the literature @xcite . equation @xmath170 admits @xmath171 order rw solution . we present the rw solution of the nls equation in the following form @xcite , @xmath172\exp{(it)},\end{aligned}\ ] ] where @xmath173 and @xmath174 are polynomials in the variables @xmath120 and @xmath175 . for the first - order @xmath176 rw solution @xmath177 , @xmath178 and @xmath179 . from ( [ a6 ] ) we get @xmath180 . for covenience we multiply this expression by @xmath181 $ ] and consider the solution in the form @xmath182}.\end{aligned}\ ] ] we use only this form of expression in our analysis . this solution is same as the one given in eq . ( [ nls3 ] ) when @xmath183 , @xmath184 and @xmath185 . for the second - order @xmath186 rw solution , the function @xmath89 , @xmath90 , and @xmath91 turn out to be @xcite @xmath187 and @xmath188 then we get @xmath189\exp{(it)}.\end{aligned}\ ] ] the profiles of second - order rw and third - order rw of the constant coefficient nls equation are shown in fig . [ nls - rog ] . for the third - order @xmath190 rw solution , we have @xmath191\exp{(it)},\end{aligned}\ ] ] where @xmath192 with @xmath193\notag \\ g_6 = & \frac{2}{45}\left[73 + 14 ( 2 x)^2 + \frac{7}{3}(2 x)^4\right],\notag \\ g_8 = & \frac{1}{15 } ( 11 + ( 2 x)^2 ) , \;\;\;\ ; g_{10 } = \frac{11}{675}\end{aligned}\ ] ] and @xmath194 with @xmath195 , \notag \\ h_2 = & \frac{2}{3 } \left[-11 - 28 ( 2 x)^2 - 2 ( 2 x)^4 - \frac{28}{45 } ( 2 x)^6 + \frac{(2 x)^8}{45}\right],\notag \\ h_4 = & \frac{4}{15 } \left[-107 + 19 ( 2 x)^2 - \frac{7}{3 } ( 2 x)^4 + \frac{(2 x)^6}{3 ^ 2}\right],\notag \\ h_6 = & \frac{4}{45 } \left[-29 - 2 ( 2 x)^2 + \frac{(2 x)^4}{3}\right],\notag \end{aligned}\ ] ] @xmath196 , \;\;\ ; h_{10 } = \frac{2}{675}.\end{aligned}\ ] ] the denominator is represented by @xmath197 where @xmath198,\notag \end{aligned}\ ] ] @xmath199,\notag \end{aligned}\ ] ] @xmath200,\notag \\ d_8 = & \frac{32}{15 } \left[\frac{83}{3 } + 2 ( 2 x)^2 + \frac{(2 x)^4}{3 ^ 2}\right),\notag \\ d_{10 } = & \frac{2 ^ 8}{225 } \left[7 + \frac{(2 x)^2}{3}\right ] , \;\;\ ; d_{12 } = \frac{2 ^ 9}{2025}.\end{aligned}\ ] ]
the dynamical response of black holes to any external disturbance is dominated , at intermediate times , by a discrete set of damped oscillations called quasinormal modes ( qnms ) . this result has been verified not only at linearized level , but also in fully numerical simulations for stellar gravitational collapse @xcite and collision of two black holes @xcite . the possibility of observation in the forthcoming years of the radiation due to the qnms opens a new perspective for the gravitational wave astronomy . from these observations it will be possible to obtain direct evidence of black holes existence , as well as to estimate their parameters ( mass , electric charge , and angular momentum ) @xcite . in addition to the astrophysical applications , an interpretation appeared some years ago of the qnms in terms of the ads / cft conjecture @xcite . according to this conjecture , there is a correspondence relation ( or duality ) between string theory in an anti - de sitter ( ads ) spacetime and a conformal field theory ( cft ) on the boundary of this space @xcite . it implies that a large black hole in the ads spacetime corresponds to an approximately thermal state in the cft . perturbing the black hole is equivalent to perturbing the thermal state , and the perturbation damping time gives the time scale for the return to the thermal equilibrium . therefore , by computing quasinormal ( qn ) frequencies in an ads space it is possible to obtain the thermalization time scale in the strongly coupled cft . the early works studying perturbations of ads black holes considered the evolution of a conformally invariant scalar field in such background spacetimes @xcite . after the advent of the ads / cft conjecture , qnms of spacetimes with a negative cosmological constant , such as kottler ( also called schwarzschild - ads ) @xcite , reissner - nordstrm - ads @xcite and kerr - ads @xcite , were computed for different fields , dimensions , and boundary conditions . within this program , other kinds of black hole solutions were also analyzed . for instance , in the baados - teitelboim - zanelli solution @xcite , analytical expressions for the qn frequencies presented exact agreement with the poles of the retarded correlation function in the dual cft @xcite . this result provided an important quantitative test of the ads / cft correspondence . as a complement to the aforementioned studies , analyses of the time evolution of some fields in asymptotically ads backgrounds have been performed @xcite . more recently , the conjecture relating the highly damped qnms to black hole area quantization has motivated a search for qn frequencies with large imaginary parts also in asymptotically ads spacetimes @xcite . the studies of perturbations in ads spacetimes were also extended to black holes with nonspherical topology . in ref . @xcite the evolution of a conformal scalar field in the family of topological backgrounds was investigated , which includes the plane - symmetric black hole as one of its members , while in ref . @xcite the case of a nonminimally coupled scalar field in the same kind of spacetimes was considered . lately , the lowest qn frequencies associated with the scalar , electromagnetic , and gravitational perturbations of the plane - symmetric black hole background were computed @xcite . the study of the scalar and electromagnetic fields was relatively complete . however , important questions were left without answers in connection with the gravitational perturbations . for instance , what are the wave equations governing general metric perturbations ? what is the physical interpretation of the modes with different wave numbers ? what are the qn frequencies in the intermediate and small black hole regimes ? what is the relation between these frequencies and the wave number ? why is the pure imaginary frequency not proportional to the horizon radius in the large black hole limit as it happens for ordinary modes ? is there some relation between the latter and the algebraically special frequency ? and what is the behavior of the qnm spectrum in the regime of highly damped overtones ? the objective of the present work is answering most of the previous questions as well as other ones that appeared during this work . some of the conclusions drawn here are extended to the schwarzschild - ads black hole . the plan of this paper is the following . in sec . [ solucao ] , we write the equations defining the background spacetime in a newman - penrose ( np ) null tetrad frame @xcite . in sec . [ perturb - np ] , we obtain schrdingerlike equations governing the axial and polar perturbations . in sec . [ qnm ] , we define the qnm boundary conditions to be used and prove the black holes stability against perturbations with a non - negative potential . a brief description of the horowitz - hubeny method ( sec . [ metodo ] ) follows , as well as subsequent exposition of the qn frequency values for different horizon radii and wave numbers ( sec . [ resultados ] ) . then , we study the large black hole limit of the axial qnms ( sec . [ limite ] ) and the non - qn character of the algebraically special frequency ( sec . [ algebrico ] ) . finally , we conclude by summarizing and commenting the main results ( sec . [ conclusao ] ) . throughout this work , we use the geometric system of units in which the speed of light and the newton s gravitational constant are set to unity . the background spacetime , moreover , has a natural lengthscale given by the ads radius @xmath5 , where @xmath6 is the negative cosmological constant . here we choose the length unit so that @xmath7 and measure everything in terms of the ads radius . in relation to notation , the asterisk ( @xmath8 ) denotes complex conjugation and the definition of the np quantities follows ref . with the aim of investigating the evolution of linear gravitational perturbations in the background of a plane - symmetric ads black hole @xcite , we summarize here the basic properties of such a spacetime . the metric may be written in the form @xmath9 where @xmath10 there is an event horizon at @xmath11 , and an essential singularity at @xmath12 . depending on the ranges of coordinates @xmath13 and @xmath14 , the metric may represent a cylindrical ( @xmath15 , @xmath16 ) , toroidal ( @xmath17 ) or planar ( @xmath18 ) black hole spacetime . for the torus , @xmath19 is related to the adm system mass @xcite ; for the cylinder , to the mass per length unit of @xmath14 line @xcite ; and for the plane , to the mass per area unit of the @xmath20 plane @xcite . according to the petrov classification , the above metric is algebraically type d. it is then convenient to define the np quantities in terms of a kinnersley - like null frame @xcite . in this case , the two real vectors of such a frame , @xmath21 and @xmath22 , are chosen to lie , respectively , on the ingoing and outgoing radial null geodesics and are defined by @xmath23 which are the double principal null directions of the weyl tensor . in order to complete the frame basis we have the vector @xmath24 and its complex conjugate @xmath25 . both of these vectors , @xmath26 and @xmath27 , are orthogonal to @xmath28 and @xmath29 , and satisfy the normalization condition @xmath30 . with the chosen tetrad basis , the only nonvanishing spin coefficients are @xmath31 and the only nonvanishing weyl and ricci scalars are @xmath32 here the equations for the linear gravitational perturbations are obtained by using the np formalism , which allows us to investigate general properties of such perturbations without any further simplifying assumption . in this approach , changes in the metric coefficients are directly related to changes in the tetrad null vectors . furthermore , perturbations in the vectors @xmath28 , @xmath29 and @xmath33 lead to first order alterations in weyl and ricci scalars , and also in the spin coefficients . the full system of linearized perturbation equations is obtained by expanding the complete np equations on the plane - symmetric black hole background , and retaining only the first order terms in the perturbation functions . generally , the linearized np equations form a large system of coupled equations whose solutions are difficult to be found . however , for petrov type d vacuum spacetimes , the problem reduces to solving a pair of decoupled equations for the gauge and tetrad invariants @xmath34 and @xmath35 @xcite . in the present case , these equations assume the simple forms @xmath36\psi_{0}=0 , \label{fund1}\\ [ ( \delta+3\gamma-\gamma^{\ast}+4\mu+\mu^{\ast})(d-\rho)- \delta^{\ast}\delta-3\psi_{2}]\psi_{4}=0 . \label{fund2}\end{gathered}\ ] ] the foregoing equations are already linearized since the weyl scalars @xmath34 and @xmath35 vanish in the equilibrium state , and then they are to be considered first order perturbations . the other quantities take their unperturbed values given in eqs . - . the background symmetries , together with the perturbation equations and , indicate that each one of the perturbation functions @xmath34 and @xmath35 can be written as a product of four different functions depending upon only one of the coordinates @xmath37 , @xmath38 , @xmath13 , and @xmath14 . moreover , due to the wavelike character of the equations , the dependence of these functions on the coordinates @xmath37 , @xmath13 and @xmath14 is of the form @xmath39 $ ] , where @xmath40 is the frequency , and @xmath41 and @xmath42 are wave numbers which may assume discrete or continuous values , depending on whether @xmath13 and @xmath14 are compact or noncompact coordinates , respectively . the radial functions , in turn , are conveniently written as @xmath43 where the subscripts @xmath44 are related to the spin weight and to the conformal weight of the weyl scalars @xmath34 and @xmath35 . the resulting equations for @xmath45 take the following standard form : @xmath46 where we introduced the differential operators @xmath47 here @xmath48 is the tortoise coordinate , defined in such a way that @xmath49 , and @xmath50 with @xmath51 . the chandrasekhar transformation theory @xcite can now be used to transform eqs . into one - dimensional schrdingerlike equations of the form @xmath52 this is done by means of the substitutions @xmath53 whose inverse transformation gets the form @xmath54 where @xmath55 in the foregoing equations , the potential functions @xmath56 are given by @xmath57 , \label{pot - polar}\ ] ] @xmath58 while the auxiliary functions @xmath59 are , respectively , @xmath60 the meaning of the signs @xmath61 comes from the metric approach to the perturbation problem , as it was performed in ref . they are related to the parity property of the perturbed metric functions under the exchange @xmath62 ( or @xmath63 ) . the variable @xmath64 represents the polar ( even ) perturbations , while @xmath65 represents the axial ( odd ) ones . in a certain sense , the wave equations can be thought of as a generalization to those found by cardoso and lemos @xcite . here the perturbations depend on both coordinates @xmath13 and @xmath14 , while in the aforementioned work the analysis was restricted to @xmath14-independent metric variations . this is equivalent to the vanishing of the wave number @xmath42 , resulting in @xmath66 . however , the plane symmetry of the background spacetime renders possible to find new coordinates ( @xmath67 , @xmath68 ) in such a way that one of the wave numbers , @xmath69 or @xmath70 , is zero . that is to say , it is always possible to find a coordinate system in which the perturbations depend only on two space coordinates ( @xmath38 and @xmath13 , or @xmath38 and @xmath14 ) . another result derived from the above analysis is related to the starobinsky constant @xcite , @xmath71 , where @xmath72 are given by eq . . the solutions of @xmath73 correspond to algebraically special frequencies in the sense of chandrasekhar @xcite . as we shall see in sec . [ resultados ] , one of such solutions , given by @xmath74 appears in the numerical results in connection to the purely damped qnms . the properties of the axial perturbations related to these algebraically special modes are analyzed in some detail in sec . [ algebrico ] . to solve the wave equations we must impose boundary conditions on the possible solutions . first , however , consider the case of the asymptotically flat schwarzschild spacetime . gravitational perturbations lead to schrdingerlike equations whose potentials vanish at infinity and at the event horizon . hence , close to these extremities , the solutions are plane waves , @xmath75 , where the @xmath48 coordinate ranges from @xmath76 to @xmath77 . qnms are then defined as solutions which are purely ingoing waves at the horizon , @xmath78 , and purely outgoing waves at infinity , @xmath79 . these requirements are satisfied only for a discrete set of complex @xmath40 called the quasinormal frequencies . in asymptotically ads spacetimes , we keep the requirement of purely ingoing waves at the horizon . in the meantime , since now the tortoise coordinate @xmath48 takes a finite value at spatial infinity , the boundary condition at infinity must be changed . this can be dirichlet , neumann , or robin boundary condition , depending on whether the field , its derivative , or a combination of both vanishes , respectively . here , as in most of the literature , we adopt dirichlet boundary conditions , since these lead to energy conservation in the field theory on ads space @xcite ( for alternative boundary conditions see ref . @xcite ) . having defined the qnms , we show below that the imaginary part of the qn frequencies is positive for non - negative potentials . the demonstration follows the one carried out in the schwarzschild - ads case by cardoso and lemos @xcite . initially , we exchange the variable @xmath80 by @xmath81 , so that eq . becomes @xmath82 where @xmath80 , @xmath83 and @xmath84 stand for both perturbation types , the polar @xmath85 and the axial ones @xmath86 . near the horizon and at infinity , the solutions @xmath83 satisfying the qnm boundary conditions are given , respectively , by @xmath87 } , \label{assint1}\ ] ] @xmath88 where @xmath89 and @xmath90 are arbitrary constants , and @xmath91 and @xmath92 are definite constants whose values depend on the perturbation type . now we multiply eq . by @xmath93 and integrate by parts using the asymptotic solutions and . after some algebra we obtain @xmath94dr= \frac{\|\omega\|^{2}\|\phi(r_{+})\|^{2}}{\mbox{im}(\omega)}. \label{int - final}\ ] ] since @xmath95 assumes only positive values outside of the black hole , the sign of the quantity on the left hand side of eq . depends crucially on the sign of the potential @xmath96 . for non - negative @xmath84 , it follows that the imaginary part of the qn frequencies is positive , @xmath97 . moreover , it is seen from eq . that @xmath98 is positive in the region @xmath99 for any @xmath1 and @xmath0 , and then we conclude that the plane - symmetric ads black holes are stable against polar perturbations . the same is true for axial perturbations as far as the wave numbers are restricted by @xmath100 . on the other hand , for @xmath101 , the potential @xmath102 takes negative values and the theorem can not be applied . another interesting property of frequencies associated to perturbations of a plane - symmetric black hole comes from the particular symmetry of such a background . this symmetry allows us to rescale @xmath40 , @xmath1 and @xmath0 by a pure scale transformation of coordinates @xmath103 , @xmath104 , @xmath105 , and @xmath106 , where @xmath107 is a constant @xcite . it implies that the frequency is a first degree homogeneous function of its parameters : @xmath108 . in particular , we can take @xmath109 so that @xmath110 then , after computing the qn frequencies for a given @xmath1 , e. g. , for @xmath111 , and for different values of @xmath0 , we can obtain @xmath112 by using the previous equation , and the values of @xmath40 for a fixed wave number @xmath1 are naturally divided into three different regimes : large ( @xmath113 ) , intermediate ( @xmath114 ) and small ( @xmath115 ) black holes . in the large ( small ) black hole limit , @xmath1 ( @xmath0 ) is negligibly small and it is expected that @xmath116 ( @xmath117 ) . we are now in a position to solve the perturbation equations and to find the values of the qn frequencies . this task , however , is more easily done through numerical procedures . therefore , it is appropriate summarizing here the method used for such a purpose . in the last decades a series of methods has been developed to numerically calculate the qnms of black holes . here we briefly describe one of these methods , used in ref . @xcite , which is of particular interest for the present work since it is suited for asymptotically ads spacetimes . in order to apply such a method we rewrite here the perturbation equations in the appropriate form . by introducing the independent variable @xmath118 , eq . becomes @xmath119 where @xmath120 and @xmath121 , with @xmath84 and @xmath122 now being considered as functions of @xmath123 . the foregoing equation has a regular singular point ( @xmath124 ) in the region of interest , @xmath125 . then , by fuchs theorem , any solution of can be written as a frbenius series @xcite . the behavior of the solutions near the horizon may be explored by writing @xmath126 , where @xmath127 is a constant to be determined . by substituting this expression into eq . and keeping terms only to the leading order , it follows that @xmath128 , which has the roots @xmath129 and @xmath130 . these roots correspond exactly to the ingoing and outgoing waves at the horizon , respectively @xcite . since we want only ingoing modes , we take @xmath129 and look for solutions of the form @xmath131 substituting eq . into , we find expressions for the @xmath132 s in terms of the frequency @xmath40 ( and of the parameters @xmath1 and @xmath133 ) . then , imposing the boundary condition at infinity , @xmath134 , we obtain @xmath135 thus , the problem of finding qn frequencies has been reduced to that of obtaining the roots of the infinite - order polynomial equation . of course , we can not carry out numerically the full sum in expression . in effect we truncate the series after a large number of terms and look for the zeros of this partial sum . the precision of the results is then verified by computing the relative variation between the zeros as we perform higher partial sums . we stop our search once we have a @xmath136 decimal digit precision . for each pair of @xmath1 and @xmath0 , we find a set of qn frequencies ( out of the infinity of possible frequencies ) . these frequencies are labeled as usual with the principal quantum number @xmath137 and ordered in a set beginning with the roots with the lowest imaginary parts . we use the horowitz - hubeny procedure described above to compute the gravitational qnms of plane - symmetric ads black holes . as it happens to other ads black holes , there are _ purely damped modes _ , whose frequencies are pure imaginary numbers , in addition to the _ ordinary modes _ , the frequencies of which also have nonzero real parts which are responsible for the oscillatory behavior of the perturbations . as mentioned earlier , the numerical analysis is divided into three different regimes according to the ratio @xmath138 values . that is to say , we have ( i ) large black holes when the ratio @xmath138 is large compared to unity ; ( ii ) intermediate black holes if @xmath138 is of the order of unity ; and ( iii ) small black holes when @xmath138 is small compared to unity . it is also worth mentioning here that some numerical analysis of the gravitational qnms was done in ref . however , such a study was restricted mainly to the large black hole regime for a specific value of @xmath1 , without generalizations for a generic wave number . in what follows , we perform a more detailed analysis , including small and intermediate black holes , and extend implicitly the numerical results to all values of @xmath1 . we also search the qn frequency values in the limit of highly damped overtones . to begin with , we discuss the case of regular ( ordinary ) qnms for which the frequencies have nonvanishing real and imaginary parts , @xmath139 . we list some numerical results of the lowest quasinormal frequencies for @xmath140 and some selected values of @xmath0 . the computed quantities include both the axial and polar perturbations , and the three regimes defined above . the data for large , intermediate , and small black holes are shown , respectively , in tables [ tlarge ] , [ tmedium ] , and [ tsmall ] . .the lowest ordinary qn frequencies for @xmath140 and several values of @xmath0 in the large black hole regime . [ cols="^,^,^,^,^,^ " , ] although numerical procedures need to be used to compute the qn frequencies , it is possible to study analytically the special case of axial perturbations of very large black holes . here we show in some detail the process of finding an asymptotic expression for the frequency of the purely damped modes . the analysis is performed in such a way that it holds for both the plane - symmetric and the schwarzschild - ads spacetimes . our starting point is the horowitz - hubeny approach described in sec . [ metodo ] . there , the qn frequencies of axial perturbations are associated to solutions of the equation @xmath141 hence , we need to obtain an analytical expression for the sum of this series in order to find exact frequency values . such a task is very difficult to be carried out for a general @xmath133 , but it becomes feasible for the large black hole limit of purely damped modes . the coefficients @xmath142 appearing in eq . are the frbenius coefficients of the power series solution to either eq . written for the axial functions , or the corresponding equation in the schwarzschild - ads case ( see ref . they are given in terms of the frequency , and of the other physical parameters of the problem , by the following recurrence formulas : @xmath143 . \label{rec2}\end{gathered}\ ] ] the last equation is valid for all @xmath144 and we have shortened the equations by introducing the quantities @xmath145x_{+}^{2 } , \label{def1}\\ b_{m}=(m^{2}-2m-3)\left[1+\kappa x_{+}^{2}\right ] , \label{def2}\\ f_{m}=\left[m(3+\kappa x_{+}^{2})+2i\omega x_{+}\right]^{-1 } , \label{def3}\end{gathered}\ ] ] where @xmath146 and @xmath147 depending on whether it is considered a plane or spherical hole , respectively , and now the foregoing expressions hold for any @xmath148 . in the large black hole regime , the wave number ( or the angular momentum ) of the perturbation is negligible and the frequency scales as @xmath149 . in terms of the variable @xmath133 , it implies that @xmath40 has a first order pole at @xmath150 . therefore , @xmath151 has a laurent series about @xmath150 , which is valid over the range @xmath152 , for some @xmath153 @xcite . such an expansion can be conveniently written as @xmath154 where @xmath155 are constants to be determined . however , not all of the elements @xmath155 are independent , since part of them can be eliminated on the basis of the parity property of the function @xmath151 . independently of what results from the sum on the left - hand side of eq . , the form of the quantities @xmath156 , @xmath157 , and @xmath158 ensures that the product @xmath159 is a function of @xmath160 . this means that @xmath40 changes its sign under the transformation @xmath161 , being an odd function of @xmath162 . accordingly , the series representing @xmath163 should be an even function of @xmath133 , which means that only the even powers of @xmath162 should survive , and then one should have @xmath164 for all @xmath165 . this result allows us to write @xmath166 where @xmath167 and @xmath168 are constants to be determined . the series representation of @xmath169 opens the possibility of taylor expanding the terms @xmath170 about @xmath171 . before doing this , we notice that the value of @xmath172 is arbitrary , since all the terms in the series of eq . are proportional to this element and the full sum is equal to zero . for simplicity , we choose here @xmath173 . then we substitute eq . into eq . and expand @xmath158 ( which depends on @xmath162 and also on @xmath40 ) about @xmath171 . when this expansion is substituted back into eqs . and , we obtain a set of taylor series for the elements @xmath170 ( one of such series for each @xmath174 ) , which may be represented by @xmath175 where the elements @xmath176 are functions of @xmath177 , @xmath178 , @xmath179 , @xmath180 , ... , @xmath181 . substituting the previous expansions into eq . and collecting terms of the same power in @xmath14 , we find @xmath182 + \left[\eta_{1}^{1}(c_{0},c_{1})+\eta_{2}^{1}(c_{0},c_{1 } ) + ... \right]z+ ... =0 . \label{colecao}\ ] ] yet the left - hand side of eq . is a power series in the variable @xmath14 . it implies that , in order for the equation being satisfied , all bracketed terms should vanish . therefore , the zeroth order coefficient provides an equation for @xmath179 , the first order coefficient , together with the earlier calculated value of @xmath179 , provides an equation for @xmath180 , and so on . here we begin with the computations that lead to @xmath179 . the first two relevant terms for this task are @xmath183 and @xmath184}$ ] . moreover , it can be seen from eq . that @xmath185 is identically zero , and as a consequence the term @xmath186 differs from @xmath187 only by a multiplicative factor . this implies , in particular , that the element @xmath188 is related to @xmath189 by @xmath190 such a relation , together with the formulas and , implies that all the subsequent elements are also proportional to @xmath189 . more specifically , the @xmath174th coefficient ( for @xmath191 ) can be written as @xmath192 , where @xmath193 is given by the recurrence formula @xmath194,\ ] ] with @xmath195 and @xmath196 . taking into account the last results , the @xmath14-independent term of eq . gives rise to the following equation for @xmath179 : @xmath197=0 . \label{puredampedzeros}\ ] ] clearly the above equation has the special solution @xmath198 , in addition to the roots that arise as zeros of the square - bracketed term . the existence of this special solution explains why the purely damped modes do not scale with the horizon size , as it happens to ordinary modes in the large black hole limit . in fact , the numerical computations show that purely damped modes are the only qnms which in first approximation are not proportional to @xmath0 . this fact suggests they should be associated to the root @xmath198 , and then it follows from eq . that for large horizon radii the corresponding frequencies go as @xmath199 . on the other hand , the numerical results also indicate that the ordinary axial qnms are proportional to @xmath0 ( in first approximation ) . from this one concludes that the remaining roots that are obtained from the zeros of the term among brackets in eq . correspond to the asymptotic values of the function @xmath200 for these ordinary qnms . it is seen from the above analysis that , in order to calculate the leading term of the large horizon limit for the purely damped mode frequency , we must take @xmath198 and calculate the next parameter @xmath180 . in such a case , the coefficients @xmath201 and @xmath202 reduce to @xmath203/3 $ ] and @xmath204 , while the other elements @xmath205 ( for @xmath206 ) assume the form @xmath207 . substituting these results into the first order coefficient of we obtain an equation that can be solved for @xmath180 , yielding @xmath208}{\left(4+\gamma \right)},\ ] ] where we have defined @xmath209 the problem now is summing the series that leads to @xmath210 . fortunately , we succeeded in summing up this series ( see appendix b ) and the exact result is @xmath211 . with this , the coefficient @xmath180 becomes @xmath212 where @xmath213 and @xmath214 are given by @xmath215 once we have obtained an expression for @xmath180 , the asymptotic formula of the pure imaginary qn frequency follows directly from eq . . in fact , considering a plane - symmetric ads black hole , for which @xmath216 and @xmath217 , we have @xmath218 , and the interesting result follows : @xmath219 alternatively , for a schwarzschild - ads black hole one has @xmath220 and @xmath221 , which implies in @xmath222 and @xmath223 , leading to @xmath224 both of the last two formulas for the purely damped frequencies are exact results at the asymptotic limit of large black holes , in perfect agreement with the approximate fits found through the numerical calculations . the numerical results of the qn frequencies revealed a close relation between the algebraically special modes and the purely damped qnms . for a fixed wave number , the imaginary frequency @xmath225 approaches the values of the special frequency @xmath226 as the horizon radius goes to zero ( see the comments at the end of section [ perturb - np ] ) . there also seems to exist a meeting point @xmath227 between the two frequencies for a special combination of parameters such that @xmath228 , as it can be seen from the numerical results shown in fig . [ fpuro ] . furthermore , the relation between the purely damped modes and the algebraically special frequency in the small horizon limit is not restricted to the plane - symmetric black hole case , but it also appears in axial perturbations in the schwarzschild - ads spacetime . all these facts render it convenient to analyze here the qn character of the algebraically special perturbation modes . the frequency function of these modes for the plane - symmetric black hole is given by eq . , while the corresponding expression in the spherical case is found in ref . @xcite . in a general form , the frequency @xmath226 can be written as @xmath229 } { 6(1+\kappa x_{+}^{2})}x_{+}^{3},\ ] ] where again @xmath146 and @xmath147 , as it is considered a plane or spherical black hole , respectively . to verify if the algebraically special modes satisfy the qnm boundary conditions we need to obtain the solution @xmath230 associated with the algebraically special frequency @xmath226 . since this function is a particular case of the general power series solution , we may use eqs . and to determine the coefficients @xmath132 for @xmath231 . once again we may take @xmath173 , so that the coefficients @xmath232 and @xmath233 are given , respectively , by @xmath234 @xmath235 + 6i\omega ( 1+\kappa x_{+}^{2})}{4x_{+}(3+\kappa x_{+}^{2}+i\omega x_{+ } ) ( 3+\kappa x_{+}^{2}+2i\omega x_{+})}.\ ] ] it is seen from the last equation that @xmath233 vanishes for the frequency @xmath231 . moreover , since all the other elements @xmath132 ( for @xmath206 ) differ from @xmath233 only by a multiplicative factor , we have @xmath236 . the solution for @xmath230 then reduces to @xmath237x_{+}^{2}}\left(1-\frac{x}{x_{+}}\right ) . \label{afunc}\ ] ] we know from the horowitz - hubeny method that the function represents an ingoing wave at the event horizon . here the problem is verifying if @xmath230 satisfies the dirichlet boundary condition at the spatial infinity . imposing the condition @xmath238 , we find an equation that is satisfied only if @xmath239 or @xmath240 , where @xmath214 and @xmath213 are given by eq . . for a plane - symmetric ads black hole , this means we need a vanishing wave number in order to have a qn frequency equal to @xmath226 . in this case , however , the resulting perturbation functions would be independent of time , which could not represent gravitational waves . similarly , for a schwarzschild - ads spacetime we need to have @xmath241 or @xmath242 . the first root is a physically forbidden value , while the second one leads to a static perturbation , which only adds a small rotation to the black hole @xcite . on the basis of these results , we conclude that @xmath226 can not be a qn frequency , because the associated wave function @xmath230 does not satisfy the dirichlet boundary condition at infinity . although the numerical calculations show that in the small horizon limit @xmath225 is very close to @xmath226 , the above result proves that the frequency function associated with purely damped modes is not exactly equal to the algebraically special frequency . the solid line in fig . [ fpuro ] does not coincide with the dots , which represent qnms , not even in the small black hole regime . there is , however , a special point @xmath227 where @xmath226 seems to coincide with @xmath243 . the answer to the question whether @xmath226 corresponds to a qn frequency at that point or not is given in the following . in spite of the generality of the analysis performed above , it fails to test the qn character of @xmath226 in the case where @xmath244 . this situation happens for a special family of parameters satisfying the condition @xmath245 in this case the roots of the frbenius index equation , which are @xmath129 and @xmath246 ( @xmath247 ) , differ by an integer number and the horowitz - hubeny method is not valid . we can circumvent this problem by writing explicitly the axial version of eq . for @xmath231 and parameters satisfying eq . . by doing this in such a way that the analysis holds for both the plane and the spherical ads black holes , it is found that @xmath248 \frac{d^{\,2}\phi_{a}^ { ( -)}}{dx^{2}}+ \left[3(1+\kappa x_{+}^{2})x + ( 3+\kappa x_{+}^{2})x_{+}\right ] \frac{d\phi_{a}^ { ( -)}}{dx}-3(1+\kappa x_{+}^{2 } ) \phi_{a}^ { ( -)}=0 . \label{fund - new}\ ] ] here @xmath249 is a regular point of the above differential equation , and the solution for @xmath250 can be expressed as a taylor series , @xmath251 substituting this expression into eq . , we find as usual that all coefficients @xmath252 ( for @xmath253 ) can be expressed in terms of the first two , @xmath254 and @xmath255 . the general solution of eq . then assumes the form @xmath256\\ & + d_{1}\left[(x - x_{+})+ \frac{(3 + 2\kappa x_{+}^{2})}{(3+\kappa x_{+}^{2 } ) } \sum_{m=2}^{\infty}\frac{\beta_{m}}{(-x_{+})^{m-1 } } ( x - x_{+})^{m}\right ] , \label{sol - geral } \end{split}\ ] ] where the parameters @xmath257 ( for all @xmath258 ) are given by the recurrence formula @xmath259 with @xmath260 and @xmath261 . the first term among brackets in eq . represents a perturbation function which is constant at the event horizon and corresponds to an ingoing wave , while the second term represents an outgoing wave , since it goes as @xmath262 in the limit @xmath263 . to satisfy the qnm boundary condition at the horizon , we must take @xmath264 and keep only the first term of the general solution . in addition , without loss of generality we may choose @xmath265 . finally , we must verify if the function @xmath230 satisfies the dirichlet boundary condition at the spatial infinity @xmath266 . by substituting @xmath267 , @xmath268 , and @xmath266 into eq . it follows that @xmath269 for plane - symmetric ads black holes , our numerical calculations show that @xmath270 , independently of the horizon radius . on the other hand , we obtain different values for @xmath271 as we change the schwarzschild black hole size . another way to see this variation is in terms of the angular momentum , since @xmath133 and @xmath4 are coupled through eq . . [ falgeb ] , we plot the numerical values of @xmath272 ( dashed line ) and of @xmath271 for @xmath273 ( dots ) . as the angular momentum goes to infinity , which means @xmath274 , the results found for the spherical case tend to the value of the plane black hole . since @xmath275 is always different from zero , we conclude that the algebraically special modes are not qnms , not even for the special parameters that lead to @xmath228 . for the plane - symmetric black hole and for @xmath140 , this corresponds to the point @xmath227 in fig . [ fpuro ] , where the lines representing the purely damped qn and the algebraically special frequencies tend to intercept each other . among other things , this result implies that the qnm spectrum of some black holes does not present a particular pure imaginary frequency , for those specific modes whose wave numbers ( or angular momentum ) satisfy eq . . for instance , an axial quadrupole ( @xmath276 ) perturbation is not able to excite a purely damped qnm in a schwarzschild - ads black hole with @xmath277 . whether this conclusion has some physical implication to the ads / cft conjecture or not is still an open question . we have studied the gravitational quasinormal modes of plane - symmetric anti - de sitter black holes . these modes are responsible for the late time behavior of small gravitational disturbances . the wave equations governing the axial and polar perturbations were derived by a procedure based on the newman - penrose formalism and on the chandrasekhar transformation theory . this approach allowed us to treat equally the two spacelike coordinates @xmath13 and @xmath14 , as well as provided the algebraically special frequencies associated to the gravitational waves radiated by the black hole . using the horowitz - hubeny method , we have computed the quasinormal frequencies for @xmath140 and several values of @xmath0 in the large , intermediate , and small black hole regimes . the wave - number dependence of the modes was obtained on the basis of the homogeneity property of the function @xmath112 . for large black holes , the ordinary frequencies scale linearly with @xmath0 , and the pure imaginary frequency goes as @xmath199 . for small black holes , the fundamental modes present @xmath278 and @xmath279 , while @xmath225 approaches the algebraically special frequency @xmath280 . in the limit of highly damped overtones , we found the qn frequencies are evenly spaced , with the spacing per unit of horizon radius being a constant which is independent of the wave - number value , perturbation parity , and black hole size . in relation to stability , we have shown analytically that for a non - negative potential the imaginary part of the frequency is positive definite for every mode . from the numerical results it also follows that all qn frequencies have a positive imaginary part , including the cases where the potential takes negative values . these results conjointly imply that plane - symmetric ads black holes are stable against gravitational perturbations . to study the large black hole limit we have developed an analytical method based on the expansion of the frequency in powers of @xmath199 . application of this method to axial perturbations showed that there is a special frequency for which the ratio @xmath281 goes to zero as @xmath282 . the first nonvanishing term is @xmath283 for the plane hole and @xmath284 for the schwarzschild hole . it is important to observe that these purely damped modes are particularly long living for large black holes , which is exactly the relevant regime for the ads / cft correspondence . if the above asymptotic formulas could be derived on the cft side , it could provide an important quantitative test to the maldacena s conjecture . as a last remark , it should be noted that the discussion of the physical interpretation of the modes with different wave - number values remains . preliminary results show that the zero wave - number axial perturbations yield only small rotations on the system , while the polar ones lead to an increase in the mass and the production of gravitational waves @xcite . this issue , however , must be investigated in more detail and shall be considered in our future work . we are indebted to v. cardoso for suggestions and for a critical reading of the manuscript . we also thank j. p. s. lemos for stimulating conversations . one of us ( a. s. m. ) thanks the conselho nacional de desenvolvimento cientfico e tecnolgico - cnpq , brazil , for a grant . in this appendix we discuss briefly how to find the modes of gravitational perturbations in pure plane - symmetric ads spacetimes . this is to be compared to the small black hole limit qnms , as calculated numerically in sec . [ resultados ] . the relevant wave equations for axial and polar perturbations may be obtained from eq . by assuming @xmath286 wherever it appears . for instance , the potentials @xmath287 reduce to the same constant value @xmath288 the resulting equations are then @xmath289 whose solutions in terms of the coordinate @xmath290 are of the form @xmath291 where @xmath292 , @xmath293 , @xmath294 , and @xmath295 are integration constants . the solution functions for @xmath296 are analytic in the whole range of the radial coordinate which is of interest for comparison to the small black hole qnms , @xmath297 . hence , the only condition @xmath40 must satisfy is @xmath298 in order to avoid the wave growing exponentially in the limit @xmath299 . however , additional restrictions to these modes arise by imposing the usual boundary conditions for qnms in ads spacetimes : no wave can come out from the black hole horizon , which is located at @xmath300 ; and no wave can propagate to spatial infinity , @xmath301 . by substituting these conditions into eqs . it follows that @xmath302 and @xmath303 , which means waves whose frequencies satisfy the relation @xmath296 do vanish everywhere . we are then left with the solutions for which @xmath304 only . in this case there is in fact no wave in the radial direction , so that the solutions @xmath305 , for @xmath306 , represent waves traveling along the spatial directions orthogonal to the @xmath38 direction . in other words , there are no waves coming out from the horizon nor propagating to spacelike infinity . considering these as the plane - symmetric pure ads modes , their frequencies , given by @xmath307 agree with the result found numerically for small black holes . here we show how the series presented in eq . can be summed up exactly . since it is an alternating series , it is important not only to prove its convergence , but also to test it for absolute convergence . by mathematical induction , we find that the series can be written as a combination of seven terms @xmath308 where the elements @xmath309 ( @xmath310 , @xmath311 , ... , @xmath136 ) are given by @xmath312 then we are left with a sum of six series , and by proving the absolute convergence of each one of them , the absolute convergence of the whole series @xmath313 is also proved . in order to do that , we calculate the limit @xmath314 and verify that it is smaller than unity for any @xmath315 , and then by applying the dalembert ratio test we conclude that every individual series is absolutely convergent . this result implies that the final sum for @xmath210 is independent of the element ordering in the alternating series . in particular , we can add the terms @xmath316 over all @xmath315 , for @xmath310 , @xmath311 , ... , @xmath136 , before performing the sum over @xmath174 . by doing that , we find the simple result @xmath317 for any value of @xmath174 , and hence the sum over @xmath174 gives zero . therefore , the sum of the series ( [ comb1 ] ) is @xmath211 .
diffuse emission in the infrared seems like a perfect laboratory to study the dynamics of the interstellar medium . recent large - scale surveys by the spitzer space telescope , specifically the glimpse project @xcite have provided us with unprecedented high resolution data of the diffuse emission in the mid - infrared ( mir ) . at first glance the wealth of structure exhibited in the flux density maps seems a striking argument by itself for structure analysis . however , the conspicuous structures themselves namely shells , bubbles , filaments and dark clouds ( see e.g. @xcite ) raise the question of how much of the observed structure actually corresponds to physical structure . flux density maps contain information about volume density , column density and excitation , but to extract one of them is only possible under assumptions . for the structure analysis , ideally , we are interested in volume density , which is accessible only indirectly , leaving us with column density as a second best at most . in fact , over a broad range of wavelengths from ultraviolet to infrared , the opportunities seem to be rather rare where we can interpret observed intensity maps of diffuse emission as information about the underlying column density structure . more often than not the medium is optically thick for the emitted radiation , or denser components of the ism act as absorbers . optical depth effects become weaker with increasing wavelength , which is why the mid - ir ( @xmath2 m ) takes a somewhat special position ( for a study of correlation between emission and column density in the far - ir see e.g. @xcite and @xcite ) . at longer wavelengths from the far - ir through millimeter , current observatories have rather poor spatial resolution , precluding study of small - scale interstellar structure . with a dust / pah extinction cross section of @xmath3@xmath4 per h - atom @xcite , the optical depth for mir emission in molecular clouds should range below or around @xmath5 , which would encourage a direct interpretation of flux density as column density . for the near - infrared ( nir ) this possibility has been discussed and supported by @xcite to interpret so - called `` cloudshine '' observations by @xcite , although there , the column densities would have to be substantially smaller than in the mir . although applicable to a wider range of surveys , this paper focuses on the mir diffuse emission as seen by the irac camera of the spitzer space telescope . to a large extent , the emission in the [ 5.8 ] and [ 8.0 ] bands comes from polycyclic aromatic hydrocarbons ( pahs ) ( to a lesser extent they also contribute to the [ 3.6 ] band , @xcite ) , mostly excited in the environment of nearby uv sources . two ( not unrelated ) issues are raised in the mir : ( 1 ) the medium is generally optically thick for the soft uv - radiation longward of @xmath6 that excites the pah emission . thus , only the outer layers of any structures unless really diffuse are excited and thereby traced out in the mir , giving the object a filamentary appearance . ( 2 ) observationally , there seems to be a strong morphological bias to shells in the pah - emission . these could be dynamical shells ( see e.g. @xcite ) , like e.g. windblown bubbles or hii regions . however , the destruction of pahs around the uv source could also lead to a shell - like structure . and finally , these shells could be irradiation effects as noted under ( 1 ) . and , of course , a combination of mechanisms is possible also . moreover , the usual projection problem introduces a bias to interpreting objects as being two - dimensional , while they could be very `` 3d '' : more diffuse material gets ionized or blown away first , leaving the ubiquitous elephant trunk remnant structures ( for an impressive example of this problem see the glimpse study of rcw49 , @xcite ) . we investigate the appearance of a model molecular cloud in the radiation field of a nearby uv - source , in order to quantify the correspondence of various measures and tracers between the original column density maps and the derived flux density maps . rescaling the density range in the models allows us to mimic different physical environments , from diffuse clouds to dense molecular clouds and cores . the appearance of the original cloud can be completely altered by irradiation . only for densities of up to @xmath7@xmath1 can the flux density actually be interpreted as column density . above that , mir self - extinction and strong shadowing effects in the uv will let the maps diverge . however , even for the highest density range ( up to @xmath8@xmath1 ) , where the flux density maps bear no resemblance to the column density , the structural properties of the original column density distribution still can be retrieved from the flux density . our results demonstrate that the diffuse emission data as made available by glimpse can serve as a powerful means to analyze the ( dynamical ) structure of the interstellar medium . this study aims at pointing out possible pitfalls and at giving a rule - of - thumb estimate where flux density structure could be trusted to represent column density . in the next section ( [ s : glimpse ] ) , we will give some observational motivation in the form of diffuse emission maps from the glimpse data . we are deferring the full structure analysis of the glimpse data to a future paper . the models and the details of the radiative transfer treatment are described in [ s : irradmodels ] . the results ( [ s : results ] ) are summarized in [ s : summary ] . the following maps are extracted from images of the glimpse project . the data were processed through the glimpse pipeline reduction system @xcite . point sources were extracted from each frame using a modified version of daophot @xcite , and the individual residual ( i.e without point sources ) frames produced by daophot were mosaicked . only sources fit well by a stellar point - spread - function ( psf ) were extracted . thus saturated and near - saturated stars remain in the images . in addition , there are some artifacts resulting from point - source subtractions due to the undersampling of the stellar psf at irac wavelengths . figures 1 - 5 show [ 8.0 ] residual mosaic images displayed in galactic coordinates . we will split the maps into three categories : volume illumination ( [ ss : volumeillum ] ) , absorption and shadowing ( [ ss : darkclouds ] ) , and high - density environments ( [ ss : highdens ] ) . figures [ f : g51.8 + 0.6 ] and [ f : g343.5 - 0.3 ] show two examples of `` volume - filling '' diffuse emission , i.e. the optical - depth effect described in [ s : motivation ] does not seem to be the dominant structure generation mechanism . rather , it seems , the diffuse emission traces the volume density . while figure [ f : g51.8 + 0.6 ] shows a clear shell - like structure , figure [ f : g343.5 - 0.3 ] exhibits dark filaments , probably extinction in the mir against a bright diffuse background . the volume illumination could either stem from material that is optically thin for the exciting uv provided by a nearby star or by the interstellar radiation field ( fig . [ f : g343.5 - 0.3 ] ) , or arise in an environment which although at higher densities contains several uv - sources , so that we still see large parts of the volume in pah - emission ( fig . [ f : g51.8 + 0.6 ] ) . residual image of diffuse emission at [ 8.0 ] micron , centered on @xmath9 . ] residual image of diffuse emission at [ 8.0 ] micron , centered on @xmath10 . ] figures [ f : g26.9 - 0.3 ] and [ f : g309.1 - 0.4 ] give two examples of structures that are typical of the glimpse data : the combination of bright and dark features in one and the same object . in figure [ f : g26.9 - 0.3 ] , the ( galactic ) north side of the filament is seen in emission , while the south side shows dark structures . figure [ f : g309.1 - 0.4 ] displays a structure composed of bright vertical filaments , which may be an example of volume illumination ( see previous section ) , while the object in question is located at smaller longitudes @xmath11 than the filament , at @xmath12 : the left side is emitting , while the right side is seen as dark structure . in contrast , the top of the same figure shows a `` pure '' dark cloud , without any irradiation features . the reader will find more examples in the other frames . this raises the question of whether we see two physically distinct classes of objects here , or whether the `` bright - dark '' clouds are just an illuminated version of the classic dark clouds . the `` bright - dark '' clouds can often be associated with irradiating sources ( see @xcite ) , so that we will identify the two morphological classes as one and the same type of object , namely dark clouds . the region around @xmath13 ( fig . [ f : g30.7 - 0.0 ] ) shows one of the many bubbles in the glimpse data @xcite . bright dust / pah emission and dark clouds form an intricately structured network . the northern part of the bubble exhibits several shell - like structures , whose brightness tapers off to the diffuse background more or less immediately . the region is compact and contains a lot of dark extinction regions in the bright emission regions , indicating that this is a high - density environment . note , however , that darker regions in emission could also just mean that the pahs have been destroyed around the uv source . volume illumination seems to play less of a role here , instead , the rim effects mentioned in the introduction seem to have come into full play . we employed two model series : series a is meant to resemble star formation regions in the galactic disk . it is derived from cloud dispersal models in a turbulent flow @xcite , and corresponds to a box length of @xmath14 pc , a background density of @xmath15 @xmath1 , and a cloud density of @xmath16 @xmath1 . model a is in approximate pressure equilibrium , with a temperature of @xmath17 k in the ambient medium . the model is not self - gravitating . series b stems from self - gravitating mhd - turbulence models in a periodic box @xcite . it is intended to be typical for conditions within a molecular cloud , with a box size of @xmath18 pc and a mean density of @xmath19 @xmath1 , an ( isothermal ) temperature of @xmath20 k and a turbulent _ rms _ velocity of @xmath5 km s@xmath21 . as we will see below , self - gravitation has already lead to core formation . both model series a and b have a resolution of @xmath22 grid cells . the density contrast can be adapted to test irradiation effects in various environments . the uv - source is assumed to be a o5v star that can be placed anywhere in the simulation domain . note that we are just investigating the radiation paths of the stellar uv - photons . the models do not account for ionization or stellar winds , and thus can not provide us with estimates how they would affect the emission structure . we further assume that only uv - photons with @xmath23 are available for pah - excitation , and that photons with shorter wavelength have already been absorbed within the hii region around the star . we follow the photons on rays through the simulation domain . the discretization in longitude and latitude follows the healpix prescription @xcite in order to guarantee equal areas for each ray . rays are traced within a tree - structure such that optical depths are known locally at each node of the tree . the number of rays is determined by the number of cells on a sphere with radius @xmath24 ( in grid cells ) as @xmath25 . thus , each cell on a given sphere gets at least one ray . densities are interpolated bi - linearly . for each location @xmath26 , the energy loss is determined by @xmath27 with the central ( stellar ) luminosity @xmath28 and the local optical depth ( measured from the uv source ) @xmath29 where @xmath30 is the atomic hydrogen particle density , and @xmath31@xmath32 is the pah absorption cross section per h - atom . the factor @xmath33 accounts for the ray splitting : the base level consists of @xmath34 rays , each of which spawns @xmath35 children at the next refinement level @xmath11 . the radial step size is @xmath36 , thus the volume at each location is known , and @xmath37 can be multiplied by the corresponding volume expression , resulting in the actual uv - energy loss . we integrated over the whole uv - spectrum of the incident radiation field between @xmath6 and @xmath38 ( see e.g. @xcite , fig . 1 ) with the constant absorption cross section @xmath39 given above . furthermore , we assume that all absorbed uv - energy is converted into pah emission , which we distribute according to the pah - spectrum by @xcite . as shown by @xcite the spectrum is independent of the radiation field over a large range in intensity . the emitted spectrum , sampled at @xmath40 wavelengths , is convolved with the irac response function @xcite for band 4 at @xmath41 $ ] micron , which contains the bulk of the pah emission @xcite . the resulting `` gray '' emission is integrated along the line of sight ( i.e. along a chosen grid axis ) following the equation of radiative transfer , including emission and extinction , with an extinction cross section @xmath42 @xmath4 per h - atom . the chosen absorption and extinction cross sections @xmath43 and @xmath44 are a factor of @xmath45 smaller relative to the values given by @xcite and @xcite . they give values of @xmath46 @xmath4 and @xmath47 @xmath4 . the slightly smaller cross sections can be accommodated by rescaling the density . note , however , that in both cases the cross sections for uv and mir differ by a factor of @xmath48 . thus , material which is optically thick at uv can be optically thin at the re - emitted mir . in fact , it is to a large extent this difference that leads to the wealth of structure in the observed mir diffuse emission . the irradiation effects will alter the appearance of the cloud , depending on gas density and geometry ( [ ss : morphologies ] ) . the ( point ) source needs to be subtracted ( [ ss : intensdist ] ) to permit a meaningful analysis of the spatial structure ( [ ss : spatstruct ] ) and to facilitate a test of the traditional structure analysis tools , namely power spectra ( [ sss : powspec ] ) and structure functions ( [ sss : structfunc ] ) . the overall correlation between column density and flux density deteriorates with increasing volume density ( [ ss : correl ] ) . under certain conditions , the flux density maps can even be used to trace the three - dimensional structure of the cloud ( [ ss : pahemissedge ] ) . we begin by discussing the morphology of the irradiated objects depending on the source s position and the density contrast . the appearance of the irradiated cloud depends strongly on the absolute gas density and on the location of the source . figures [ f : irradclouda0 ] [ f : irradclouda3 ] give a first impression of the appearance of an irradiated cloud . the top row of each figure shows the column density along each of the three cartesian axes . each column stands for one of the cartesian axes along which the pictures have been taken . the first digit of the labels in the panels gives the model series name ( a or b ) . the second digit stands for the position of the star along the @xmath49-axis in units of @xmath50 , where @xmath51 pc is the box length . a value of @xmath5 in the first column of images means that @xmath52 of the box length is between the observer and the star , e.g. panel a1x0 represents a situation where the source is located between observer and most of the cloud , and thus the cloud `` face '' in the plane of sky as seen by the observer is irradiated . thus , in the left column , the star is moved along the line of sight , and in the center and right columns , it is moved from left to right . the cloud itself is contained in approximately the central half of the box . the third digit ( @xmath53 ) gives the cartesian coordinate axis along which the model has been projected . the last digit denotes the logarithm of the density contrast enhancement above the contrast already existing in the simulation , i.e. a value of @xmath54 denotes a density contrast enhancement of @xmath55 . thus , columns are a sequence in position , rows are a sequence in line - of - sight orientation , and each of the four figures [ f : irradclouda0 ] [ f : irradclouda3 ] contains frames of a fixed density contrast . the densities range between @xmath5@xmath1 at the edge of the box ( in the `` inter - cloud medium '' ) and @xmath55@xmath1 within the cloud for a contrast enhancement of @xmath5 , and between @xmath5 and @xmath56@xmath1 for a contrast enhancement of @xmath57 . question marks ( `` ? '' ) in the model names act as group specifications , e.g. a??0 refers to all panels in figure [ f : irradclouda0 ] , whereas a?x0 refers to the panels at varying star position , projected along the @xmath49-axis ( i.e. left column of figure [ f : irradclouda0 ] ) . for a density contrast of @xmath5 ( fig . [ f : irradclouda0 ] ) , the cloud s column density structure is well traced by the pah - emission . the object is still optically thin enough for the uv photons to excite its whole volume . panel a3x0 and the row a?y0 show slight self - extinction of the pah - emission ( slightly darkened regions within the cloud ) . the model sequence a??0 could be compared to figures [ f : g51.8 + 0.6 ] and [ f : g343.5 - 0.3 ] . the extinction structures get more pronounced at the next higher density contrast of @xmath20 or a density range @xmath58 @xmath1 ( fig . [ f : irradclouda1 ] ) . panels a??1 tell us more about these regions : the dark structures are a combination of uv - absorption _ and _ mir extinction . since the star is moving away from the observer from panel a1x1 to panel a3x1 , the dark fuzzy region in a3x1 must be located between the star and the observer . if it were solely due to uv - absorption , it should show up as a `` bright '' ( light gray ) region in the pah - emission . the fact that it is dark in the pah - emission maps thus shows that it is not only a uv - shadowing effect , but that it is also dense enough to extinct the background mir radiation . a similar effect can be seen in panel a3z1 , at the lower left of the cloud . the region is uv - shadowed by the bulk of the cloud , however , in contrast to a3x1 , the optical depth in the mir is smaller , so that while the background mir is extincted , its brightness is close to the background brightness . as one can see in the corresponding column density map above a3z1 , the region is still within the cloud , i.e. there remains gas available for extinction of the mir . a wisp of denser gas in front of the shadowed region is irradiated by the source , showing up in lighter gray . the effect of the background intensity will be discussed for the model series b. increasing the density contrast further to @xmath48 , corresponding to a density range of @xmath59 @xmath1 , ( fig . [ f : irradclouda2 ] , panels a??2 ) leads to stronger absorption , and in some cases to complete shadowing . simultaneously , the diffuse gas still mostly optically thin for the uv photons gets brighter . overall , the maps seem to be more `` structured '' than their lower density counterparts , which is partly an effect of the combination of emission and extinction , and partly because at higher densities , smaller changes in density are traced out . that shadowing effects play a large role can be seen in figures [ f : irradclouda1 ] and [ f : irradclouda2 ] ( models a??1 and a??2 ) . moreover , the spatial structure of the cloud continues from the emission regions to the extinction regions ( panel a1z2 ) , an effect often seen in the diffuse emission maps ( see the examples in [ ss : darkclouds ] as well as @xcite ) . figure [ f : irradclouda3 ] , panels a??3 stands for the extreme case : it corresponds to a molecular cloud with densities ranging up to @xmath60 @xmath1 . except for the close vicinity around the star , the object stays dark : all the uv is absorbed directly ( see row a2[x , y , z]3 ) . in panel a3?3 , we actually catch a glimpse of the far side of the cloud : the irradiating source is on the far side of the cloud , and while the cloud is optically thick to the exciting uv , it is still ( marginally ) optically thin for the pah - emission . the `` dark fuzzy '' region discussed in a3x1 is still visible and has acquired a `` halo '' of excited pahs . the irradiation by the central star introduces shell - like patterns in an object that does not have any shell structure whatsoever ( see column density panels on top of figure ) : especially , panel a1z3 gives the impression of a star sitting in a 3d - shell . in other words , it is relatively simple to confuse such structures with the signposts of interaction between a star and its surrounding medium ( although obviously it does not rule out such an interaction ) . this situation reminds one of the shell - like structures north of @xmath13 , figure [ f : g30.7 - 0.0 ] . model a1z3 in a zoomed - in view . the central source is now located in the upper left corner . over - plotted are two rays through the column density and flux density map respectively , whose profiles are shown in the top and bottom panel . vertical lines denote points of local maxima in the flux density along the ray . in order to help identification in the maps , these points are marked with small vertical lines on the ray . not all flux density maxima have corresponding column density maxima . to identify the shell - like structures more easily , compare to panel a1z3 in figure [ f : irradclouda3 ] . ] figure [ f : pahcuts ] presents a closer look at a sub - region of panel a1z3 in order to learn more about the mechanisms that lead to the shell - like structures in the flux density maps . the two center panels show the column density and the flux density , where the central source is now located in the upper left corner of the frames . for an identification of the shell , it helps to compare this figure to panel a1z3 of figure [ f : irradclouda3 ] . the top and bottom panels show column and flux density profiles for two rays indicated on the gray - scale maps . vertical lines denote local maxima in the flux density , marked as well on the gray - scale maps by short vertical lines . generally , we would expect emission peaks at locations where the radial volume density gradient is positive , and its curvature negative . however , of the three locations , one does not follow this expectation : we observe an emission peak for a negative _ column _ density gradient . since enhanced emission can only occur at positive radial density gradients , this might seem surprising , but it only tells us that the flux density maps in fact trace out the 3d structure of the cloud : while the overall column density drops with increasing distance from the source , a local density enhancement ( most likely in the foreground because otherwise it would be masked ) leads to the peak in the flux density . thus , the shell - like structure observed here has nothing to do with a physical shell , but is a result of a lucky combination of smaller density enhancements predominantly located in the foreground . observationally , such objects could be distinguished from physical shells by comparing them at other wavelengths , e.g. in the dust continuum at @xmath61 micron . we have to mention a caveat here , though . our models do not account for pah destruction in the vicinity of the star . this would introduce a further bias towards shell - like structures , independent of any dynamical effects such as stellar winds or hii regions . while the overall resemblance between observed flux density maps and the modeled ones ( e.g. a??2 ) is suggestive , there is nevertheless one striking difference : the modeled maps exhibit dark radial streaks caused by the total absorption of uv and thus the lack of pah emission along those rays . such structures are generally not observed , for the following reasons : ( 1 ) in the observed maps , the lines of sight are substantially longer than in the models , and thus diffuse material in the foreground and/or background will lead to a much higher uniform flux , which easily would cover any faint low - brightness structures . ( 2 ) the restriction to one source and neglecting the interstellar radiation field introduces a directional bias in our models which then leads to the shadows seen in the models . only if the absorbing material is directly between the observer and the source can shadowing effects be observed , namely as dark clouds . clearly , the overall correspondence between column density maps and the pah - emission maps deteriorates for increasing densities . figure [ f : irradcloudb ] is a variation on the previous theme , namely flux density maps of a self - gravitating turbulent medium , irradiated by a star . as before , a sequence in increasing density contrast is shown . one difference between model a and b is obvious : model b does not exhibit a well - defined irradiated cloud structure ( or at least the rims of that ) . one could see model b as a `` close - up '' of model a , taken of a small region around the star inside the cloud of model a. in model b , the uv - radiation is mainly stopped by the numerous filaments arising from shock compressions . in addition to those filamentary structures , the emission has a `` fog - like '' appearance : in contrast to model a , the density within the cloud is already high enough to absorb sufficient uv to trace even low - density regions , and the scales of the frame are smaller by a factor of nearly @xmath62 , so that we are seeing the close environment around the o - star . the small black objects visible at the higher density contrasts are self - gravitating cores that have already collapsed . while it is possible to identify irradiated filaments with structure in the column density map ( top ) , the overall agreement between flux density and column density ( even for the lowest contrasts ) is rather poor . the densities range between @xmath20 and @xmath63@xmath1 in the original density distribution ( density contrast enhancement of @xmath64 , model b??0 ) . an enhancement of @xmath5 then corresponds to a maximum density of @xmath65@xmath1 , @xmath54 to @xmath60@xmath1 and finally @xmath57 to @xmath66@xmath1 . the dark regions in the lower row of figure [ f : irradcloudb ] are reminiscent of the dark clouds in figure [ f : g30.7 - 0.0 ] , however , for the wrong reason : the observed dark regions arise from back - illumination by the diffuse @xmath67 micron emission along the whole line of sight , while the dark ray - like regions in model b are a result of the complete uv - absorption of the single source available in the model . model sequence b does not seem representative of observed diffuse mir emission . the column density maps prominently exhibit filamentary structure over the whole domain . these filaments are shock fronts caused by the supersonic turbulence in the box . however , they are not reproduced in the flux density map . obviously , because of their highly transient nature , the filaments contain only little mass , so that they do not represent a major obstacle to the star s uv - radiation . stars form in these models predominantly in the ( abundant ) intersections of filaments . the high shock compression leads to a close to instantaneous destabilization and fragmentation of these intersections , resulting in the formation of collapsed cores after less than one dynamical time in the simulations ( see e.g. @xcite , @xcite ) . in other words , cores form too soon in these simulations for the filaments to gather sufficient mass to be seen in mir diffuse emission . before we can analyze the intensity distributions ( this section ) and the spatial structure ( [ ss : spatstruct ] ) , we need to `` remove '' the central source , which would otherwise dominate the analysis . to that purpose , we average the intensities azimuthally , centered on the flux density peak , and subtract the resulting profile from the frame . an alternative approach would be to just cut off the peak of the intensity corresponding to the central source . however , this leaves the imprint of the @xmath68-dilution of the radiation field , which is a direct effect of the central source and would confuse the structure analysis . the subtraction of the @xmath68-profile could be seen as a modified photometry algorithm for point sources irradiating a surrounding , optically thin diffuse medium . the residual frame ( fig . [ f : cleanedfluxa ] ) still shows artifacts due to the central source , but the cloud structure is more pronounced than in the `` directly observed '' maps . comparing the residual maps to the column density maps ( top row of fig . [ f : irradclouda0][f : irradclouda3 ] ) demonstrates that the _ irradiation _ besides introducing additional structure via absorption ( dark regions ) _ generates more small - scale structure than observable in the column density map_. this is mostly an optical depth effect in the sense that already small increases in the optical depth ( shocks , filaments , smooth gradients in the density ) leads to additional absorption which then will be traced out by the re - emitted mir . this effect gets amplified if the `` obstacle '' uses up all the remaining uv photons in that ray , i.e. if the optical depth changes from @xmath69 to @xmath70 within that structure . the result is a bright region facing the central source , and a dark , shadowed region facing away from the source . subtracting the central source in model b ( fig . [ f : cleanedfluxb ] ) is not as prone to produce artifacts in the residual flux density maps because the model is pretty much isotropic by construction . in contrast , since the cloud center of model a does not coincide with the uv - source , artifacts are unavoidable and easily recognized as darker circular regions . residual flux density maps for the three lower density contrasts of model a. artifacts unfortunately are unavoidable . ] residual flux density maps for the three lower density contrasts of model b. since the model tends to be more isotropic by construction , artifacts of the subtraction carry less weight . ] in the previous section we saw that a `` by eye''-comparison of the column density maps and the flux density maps reveals additional structure in the flux density . the next two sections will discuss under which conditions , and how strongly , the irradiation affects structure measures . we test the reliability of two traditional structure analysis tools , namely power spectra ( [ sss : powspec ] ) and structure functions ( [ sss : structfunc ] ) . since the flux density maps at higher densities are dominated by global irradiation effects , power spectra turn out to retrieve structure information less reliably than structure functions . the latter can only be trusted if applied to small enough regions not to be influenced by global irradiation effects . how reliable is the structure information extracted from diffuse emission ? figure [ f : speca ] ( left column ) gives an impression for our irradiated cloud ( model a ) . the top panel shows the power spectrum for the original column density map , without any source or irradiation . the three lower panels in turn give the spectra of the irradiated cloud including the source ( solid lines on top of panels ) , and the spectra of the residual flux density , i.e. with the central source subtracted as described in [ ss : intensdist ] . obviously , the maps including the source are completely useless for structure analysis : the source dominates the whole spectrum . the slope ranges around @xmath71 ( not shown in the plot ) . a completely different picture is revealed when analyzing the residual maps : the resulting slopes approach the values of the column density maps . _ the slightly flatter slopes indicate that irradiation introduces additional small - scale structure . _ the artifacts introduced by the subtraction of the central source in model a2?0 are mirrored in the slight hump of the corresponding spectrum around @xmath72 . note that the power law exponents refer to power spectra , i.e. a kolmogorov law would be represented by @xmath73 . to conclude for model sequence a , the power spectra reproduce the original scale distribution within the errors , which however are substantial , due to the circular averaging . the situation changes drastically for model series b ( fig . [ f : speca ] , right column ) , because the high - density cores lead to a strong power increase at small scales ( basically , they act as noise ) in the column density spectrum , which is why this spectrum is much flatter than expected for a turbulent power spectrum ( see also @xcite for a discussion of this effect ) . note that the column densities in the cores range approximately 2 orders of magnitude above the mean column density ( fig . [ f : irradcloudb ] ) . since only a tiny fraction of a dense core can emit in the mir , these objects generally do not show up in the emission maps , and then only in extinction . again , the central source dominates the `` uncleaned '' spectrum . thus , in a sense , the mir maps of model b are more suitable for structure analysis than the original column density maps granted that the major part of the map is unaffected by shadowing or self - extinction . structure functions are a more appropriate tool for analyzing the spatial distribution of an incomplete data set , since masking introduces artificial signals in the power spectra . only pixels with a measurable signal would contribute to structure functions of observational data . however , due to the irradiation and absorption effects the flux density map generally will not cover the whole cloud , as demonstrated in [ ss : morphologies ] . thus , in order to compare the column density and flux density maps , we selected the former for column densities @xmath74 @xmath75 , which traces out the bulk of the cloud . applying the resulting mask to the flux density maps and determining the structure functions @xmath76 with lag @xmath11 and spatial coordinate @xmath49 , results in figure [ f : structcheck ] . structure functions of column density @xmath77 ( thin lines ) and residual flux density maps ( thick lines ) for the three spatial directions . the center block shows models a2 ? ? , i.e. with the source dead - center , while the upper and lower block correspond to a1 ? ? and respectively , i.e. the source offset to smaller / larger @xmath49 . since model a2?3 would not be very revealing ( see fig . [ f : irradclouda3 ] ) , we dropped it . ] the four top rows belong to model a1 ? ? , with the source being offset to smaller @xmath49 . the three center rows refer to model a2 ? ? , with the source dead - center , and a3 ? ? is represented by the four bottom rows , where the source is offset to larger @xmath49 , so that a3x ? refers to a configuration where the cloud is located between observer and source . each panel shows the structure function of the original column density at @xmath74 ( thin line ) . the structure function for the corresponding flux density is plotted in thick line style . the numbers @xmath78 in each panel refer to the logarithmic slope of the structure function , @xmath79 taken over the first decade in @xmath11 . although a column density threshold of @xmath74 leads to most of the high - density gas in one contiguous object , there are a few isolated cloud fragments . since these cloud fragments are separated from the main body of the cloud , they tend to have the largest lags with respect to the main body , however , it turns out that their column densities are close to the mean column density within the main body , so that the structure functions show a drop at large lags . since figure [ f : structcheck ] is somewhat messy to interpret , figure [ f : structchecksum ] summarizes the degree to which the original column density structure information is retrieved from the flux - density maps normalized difference between logarithmic slope of structure functions ( eq . [ e : sfpower ] ) for column density ( @xmath80 ) and flux density ( @xmath81 ) , against the density contrast . for reasons of clarity , the three coordinate directions are plotted in open symbols centered around their respective density contrast , i.e. a?x1 can be found left of @xmath82 , a?y1 at @xmath5 , and a?z1 right of @xmath5 . filled symbols denote the mean over the axis directions @xmath49 , @xmath83 and @xmath84 . symbol types refer to the position of the source . ] just looking at the filled symbols ( averages over the three coordinate directions ) one might get the impression that _ larger _ density contrasts lead to more accurately retrieved slopes . this is somewhat surprising . however , the scatter between the three coordinate directions is substantial and suggests to take a closer look at figures [ f : irradclouda0][f : irradclouda3 ] again . as will be shown in [ ss : pahemissedge ] , the lowest density contrast allows a `` volume - rendering '' of the cloud : the medium is optically thin enough so that the exciting uv can light up the whole cloud , nearly acting as a uniform background field . if the source is within the cloud , though , it tends to over - represent the volume directly around it . thus , the best fit is given by a3y0 , where the source is located seemingly at the tip of a filament , farthest away from the bulk of the cloud . the effect of over - emphasizing close - by structures can be well observed even in a3y0 : above the source , a filament is traced out which does not show up clearly in the column density map . comparing this to a3x0 we see that it suffers from `` neighborhood - illumination '' by the source : internal structure of the cloud thus is just smeared out : the flux - density slope is flatter than the column density slope ( fig . [ f : structcheck ] ) . a3z0 fares slightly better , although here the source over - emphasizes the ring - like structure close by , leading to a steeper slope . the center - position maps ( a2?0 ) all suffer from the source being located inside the cloud , while models a1?0 , where the source ( along the @xmath49-axis is located in the front , again tell a mixed story . for a1z0 , the source is located at the upper left end of the cloud , in a fairly diffuse environment and sufficiently far away from the bulk of the cloud in order not to affect the structure function ( third cross symbol for @xmath85 in fig . [ f : structchecksum ] ) . for a1x0 and a1y0 , the source is already too embedded . moving on to the next density contrast , @xmath86 , again , a3y1 shows the lowest deviation , however , because of the shadowing effects in the cloud s left ( `` eastern '' ) part , the deviation is larger than for a3y0 . that a3x1 provides better fits than a2x1 and a1x1 is understandable again because of shadowing , although it is not directly clear why the match should be better than that for a3x0 . checking the corresponding panels a3x0 and a3x1 in figure [ f : structcheck ] gives the answer : on the small scales , the structure functions agree , however , on large scales they differ . the better matches for a1z1 and a2z1 ( compared to a3z1 ) might be a consequence of the shadowed regions , which still show substructure : _ this substructure would enter the structure function on the small scales and provide the same information as structure in emission . _ thus granted that we chose a small enough field for analysis to some extent structure in absorption and emission can be used simultaneously for analysis . at density contrasts @xmath87 and @xmath57 , matching between the overall shapes of structure functions for column and flux density deteriorates on the large scales ( fig . [ f : structcheck ] ) , as one would expect by looking at figures [ f : irradclouda2 ] and [ f : irradclouda3 ] . however , there is still enough signal from irradiating small density variations for the structure functions to match on the small scales . again , given that we take a small enough region , the structure function can be reproduced . another limitation on the size of the region is given by the available dynamic range in the observed flux density maps . in the diffuse emission , it often turns out that the dynamic range is spanning @xmath5 , at most @xmath54 , decades , which limits the spatial scales available for analysis . to summarize , structure functions seem to be a much more viable tool to investigate localized diffuse emission than power spectra . despite the fact that the irradiation sometimes modifies the underlying density information beyond the point of recognizability , the resulting structure functions still retrieve the salient scale information given that the field investigated is small enough . because of the global irradiation effects , there is only little hope to retrieve the large - scale information accurately . the most promising way would be to identify small , `` uncontaminated '' regions and reconstruct the structure information from those . the deviations ( fig . [ f : structchecksum ] ) are within the errors on the mean of the structure function ( see the analysis of power spectra ) . again , these stem from the circular averaging . the large deviations in fact point to a major problem when applying two - point correlators to data exhibiting anisotropic structures . for example , filaments will lead to different structure information along and perpendicular to their long axis , in turn leading to a mixing of the spectral information when averaging the directions . since the observed structure in the interstellar medium seems to be anisotropic to a large extent , the application of unmodified two - point correlators might raise questions . we will discuss this problem and possible remedies in a future paper . how reliably are structures in column density reproduced in the flux density maps ? the correlation coefficient @xmath88 serves as a first crude estimator for the similarity of column density maps @xmath89 and flux density maps @xmath90 . the means over the map are denoted by @xmath91 and @xmath92 respectively . for all density contrasts larger than @xmath5 , @xmath77 and @xmath93 are basically uncorrelated ( fig . [ f : correlcoeff ] ) . correlation coefficient ( eq . [ e : correlcoeffdef ] ) for model series a and b. for larger density contrasts , the maps are uncorrelated . ] the correlation coefficient provides an overall estimate for how `` similar '' the maps are . unfortunately , equation ( [ e : correlcoeffdef ] ) is only of limited use for a point - to - point similarity measure . although one could plot a map of the summands , the normalization is unclear . one possible similarity measure is given by @xmath94 at locations where both maps show either strong or weak signals , @xmath95 , while positions with anti - correlated signals will exhibit @xmath96 . correlation maps ( eq . [ e : correlmapdef ] ) for model series a. ] figure [ f : correlfluxa ] shows @xmath97 for model a. it can be directly compared to figures [ f : irradclouda0][f : irradclouda2 ] and [ f : cleanedfluxa ] for the lower density contrasts , the column density map correlates well with the emission ( light gray to white shades ) , while for higher density contrasts , the absorption regions lead to a strong anti - correlation . the overall positive correlation deteriorates for larger density contrasts . if the uv from the central source excites pah - emission in the rims of the surrounding molecular cloud , the emitted flux density could be used as an edge sensor or gradient indicator . to test this , for each position in the simulation domain and the column density map we calculate the 3d radial density gradients @xmath98 and the two - dimensional column density gradients respectively @xmath99 the hatted quantities are bi - linear interpolations of the ( column ) density at the positions @xmath100 with @xmath101 and @xmath102 . the radial increments @xmath103 and @xmath104 measure one grid cell , and @xmath105 as well as @xmath106 in equations ( [ e : ddr ] ) and ( [ e : ddr ] ) refer to the coordinates @xmath107 , @xmath108 respectively . the gradients are computed with respect to the central source ( @xmath109 and @xmath110 at position of source ) and selected for positive values ( figs [ f : gradienta ] , [ f : gradienta2 ] ) . positive gradients @xmath111 indicate locations of enhanced uv - absorption and thus enhanced pah - emission , while @xmath112 would denote the same if the physical cloud structure were truly 2d . we compute @xmath113 and @xmath112 for each cell in the simulation domain and column density map respectively , i.e. a slight subtlety enters regarding the 3d - case : we still need to project the gradients on a plane . in order to keep the 3d - information at least partially , we color - coded the distance along the line of sight , where blue colors denote short distances ( towards the observer ) and red colors denote long distances ( at the far end of the volume ) . the intensity gives the strength of the gradient . the distance information is weighted with the gradient strength , i.e. two gradients of same strength located at one quarter and at three quarters of the box length will show up in purple . gradient maps ( see text ) for model a , source centered ( a2 ? ? ) . column density ( top row ) and flux density ( two bottom rows ) are shown for comparison . the color bar for the 3d gradient gives the distance along the line of sight in box units , with @xmath114 ( blue ) closest to the observer , and @xmath5 ( red ) located at the far end of the box . ] gradient maps ( see text ) for model a , source at position 1 ( a1 ? ? ) . ] at first sight , the correspondence between gradient maps and flux density is at most poor . a close look however reveals some interesting details . we focus on figure [ f : gradienta ] ( models a2?1 and a2?2 ) first . the top row shows the original column density , followed by the 2d gradient @xmath112 and the projected 3d gradient @xmath113 . the two bottom rows give the flux density maps for models a2?1 and a2?2 , i.e. for increasing density contrast . the color scale is a measure of the location of the gradient , with blue closest to the observer , and red at the far end of the box . clearly , the gradients making up the map come from all possible locations along the line of sight . taking a2y1 , the l - shaped structure visible in extinction in the flux - density maps is traced out by a hazy bluish gradient - indicator : the corresponding density jump is close to the observer , which makes sense , since dense material between the source and the observer leads to the extinction . further right of the l - shape , the gradient map of a2y ? shows a red ( albeit dark ) region : weak gradients at the far side of the source . correspondingly , we see a bright irradiated region in the flux density map of a2y1 : the ( far away ) structures are irradiated by the central source . this is not seen anymore in a2y2 : the density has increased and leads to extinction of the far - away emission . similarly , the dark region to the lower left of the center in a2x1 corresponds to gradients close to the observer . in this case however , there is no clear blue signal in the 3d gradient map , which indicates that the dark region is caused by a combination of uv - shadowing and mir extinction of diffuse background . moving northwards of that direction , the source of emission is receding from the observer : again , this information is lost at higher density contrast ( a2x2 ) , and thus the cloud structure is only incompletely `` sampled '' . interestingly , for a2z2 , the situation is nearly converse : the dark gradient - regions link up with strong emission signals ( to upper left of the center ) , while the extincted regions to the south are farther away from the observer . thus , those dark regions might again be caused by a combination of missing irradiation and extinction of diffuse background . the `` handle '' to the right is located at the far side of the central source . compared with this richness of structure , the 2d gradients come as sort of a big disappointment : they resemble the cloud structure only in the widest sense . beginning with a2y1 again , the l - shape is recovered , but this is already all we can see . , the `` handle '' to the right of the central source is showing up slightly in @xmath112 . however , moving to the larger density contrast ( a2?2 ) , the situation changes slightly : the flux density maps look more `` rim''-like , more easily identified with the 2d gradient map . carrying this idea to the extreme , figure [ f : gradienta2 ] shows the gradient maps and flux densities for a1?3 ( as fig . [ f : irradclouda3 ] shows , model a2?3 would be not exactly informative in this context ) . the 3d gradients now are hardly reproduced ( with the noticeable exception of the southern absorption rim in a1x3 ) . the 2d gradients however tend to trace now the absorption rims ( see e.g. a1y3 and a1z3 ) . _ thus , for lower density contrasts , or a more diffuse environment , the emission structures are intrinsically 3d , while for higher density regions , they tend to get more and more 2d ( although 3d effects are still not to be neglected ) . _ the abundance of structure in mir diffuse emission as observed in e.g. the glimpse data seems to offer a perfect laboratory to study the dynamics of the dense ism . however , a large part of the observed structure could be irradiation effects due to pah emission . pahs are excited by uv photons from nearby stellar sources or the interstellar radiation field , and re - emit in the mir . because the respective cross sections differ by a factor of approximately @xmath48 , gas , which is optically thick in the uv , can still be optically thin in the ( emitted ) mir . thus , pah emission is often seen in filamentary structures , probably the `` rims '' of denser clouds . these irradiation effects might spoil the opportunity to study the ism gas structure , since this requires interpretation of the observed flux density in terms of volume or column density . motivated by a few examples taken from glimpse , we identified possible limitations of this interpretation . we quantified the reliability of flux - density maps of diffuse emission in the mir to reproduce the underlying ( column ) density information . we used two model sets , one corresponding to a ( more or less ) isolated `` molecular '' cloud , and the other imitating a region deep inside a molecular cloud , both irradiated by an o5 star . pahs absorb the uv and re - emit the energy in the mir , which then is integrated along the line - of - sight , including mir extinction . in the following , we will summarize how reliable flux density maps reproduce column density in our models , and how this affects the structure analysis , with possible applications to mir observations . if the medium is optically thin for the irradiating uv , then the mir emission maps could be used for a high - resolution study of the column density structure of the medium . @xcite supported this possibility for the nir , based on the observations of `` cloudshine '' by @xcite . since they were interested in the appearance of the diffuse emission , they used an isotropic radiation field mimicking a uv background , in contrast to our models that employ a point source for irradiating the surrounding medium . for higher - density environments such as molecular cloud in the vicinity of strong uv sources , interpreting mir emission as column density requires some caution : the transition from @xmath69 to @xmath115 can lead to strong signals in the mir flux density , but not necessarily paired with a corresponding signal in the column density : the maps bear no resemblance to the column density ( figs [ f : irradclouda0][f : irradclouda3 ] and [ f : irradcloudb ] ) . in the extreme case , pah - emission is only excited at the rims of clouds ( e.g. @xcite ) , causing the impression of a highly filigree structure in the gas : irradiation introduces more small - scale structure than observable in the underlying column density maps . as soon as shadowing is obvious , the structure seen in emission will generally not represent column density . some of the observed shell - like structures could be just irradiation effects , and by themselves indicate that shadowing ( or `` rimming '' ) has set in ( figs . [ f : g30.7 - 0.0 ] , [ f : irradclouda1 ] and [ f : pahcuts ] ) . since our models do not include pah destruction around the star , we expect them to exhibit fewer shell - like structures than observed . pah cavities would lead to `` shells '' even if the cavities were not associated with e.g. wind - blown bubbles . power spectra are only partially useful for an analysis of diffuse emission structure . their well - known main drawback is that they tend to confuse the information about extent of a region and the separation of regions . furthermore , masking is always an issue in power spectra , since it tends to introduce a signal by itself . power spectra containing the central source are completely dominated by that ( fig . [ f : speca ] , left column ) , while spectra of residual maps are slightly flatter than the underlying column density distribution . this could be a projection effect and/or the result of additional small - scale structure traced out by irradiation . the spectral slope is pretty much insensitive to the density contrast within the error bars , which are significant . for collapsed regions ( fig . [ f : speca ] , right column ) , the column density spectrum flattens considerably because of the strong point source contribution . compared to that , the flux density spectra steepen because the point sources are not fully irradiated and thus do not show up ( except in extinction ) . structure functions seem to be a more viable tool to investigate localized diffuse emission . despite the fact that the irradiation may modify the underlying density information beyond the point of recognizability , the resulting structure functions still retrieve the salient scale information given that the field investigated is small enough not to be contaminated by global irradiation effects . there is little hope to retrieve the _ large - scale _ information accurately by applying a global structure measure such as power spectra . structure in extinction can be used as a continuation of structure seen in emission , although this raises the issue of an appropriate choice of background for the extincted region ( figs . [ f : g26.9 - 0.3 ] and [ f : irradclouda1 ] ) . the deviations ( fig . [ f : structchecksum ] ) are within the errors on the mean of the structure function . since the overall structure in the ism tends to be anisotropic , applying two - point correlators seems at least questionable . averaging in @xmath116 or @xmath11 space leads to substantial errors on the mean , which themselves indicate that the underlying structure is anisotropic to a large extent . since the conversion of uv to mir will occur predominantly at regions of large positive radial density gradients ( as long as there are photons left ) , the flux density maps might offer the opportunity to gather information about the ( 3d ) density structure of the cloud . to test this , we compared the radial volume and column density gradients to the flux density maps . for lower density contrasts , the flux density maps tend to trace out the 3d structure of the cloud , and in fact they can be used as 3d gradient indicators . for higher density contrasts , they revert to an indicator of the column density gradients : the structures seemingly become two - dimensional . thus , more diffuse regions are intrinsically `` more 3d '' , while higher - density environments tend to be 2d . \(1 ) if the volume density is low enough that the exciting uv can irradiate the whole cloud , one might question how long the line - of - sight actually is , and whether angular effects leading to scale - mixing in a structure analysis would play a role . \(2 ) on the other hand , the volume density might be large so that the irradiating uv will be absorbed more or less directly `` at the rim '' ( if such a thing exists ) of the cloud . then , depending on the geometry of observer , irradiated medium and irradiation source , the observer might see predominantly 1d structures or 2d structures , implying projection effects in the spectral information . depending on the diagnostics , mir flux density maps of diffuse emission from pahs excited by a nearby uv source can be used to extract information about the density structure of the underlying ( molecular ) cloud , though this statement needs some qualification . \(3 ) as long as structure studies are restricted to areas small enough not to be contaminated by any large - scale effects , flux density and column density show similar structural properties . however , the application of unmodified two - point correlation functions introduces substantial errors on the mean due to the underlying anisotropy in the ism structure . employing mir diffuse emission to extract structure information about the underlying interstellar medium requires close attention to the environment . this study attempts to provide some guidelines to chose appropriate locations . bearing the limitations in mind , analyzing the ism structure with the help of the glimpse data will be a promising task . we thank e. bergin for a critical reading of the manuscript . computations were performed at the ncsa ( ast040026 ) ( fh ) and on the sgi - altix at the usm , built and maintained by m. wetzstein and r. gabler . support for this work was provided by the university of michigan ( fh ) , nasa s astrophysics theory program ( nng05gh35 g ) ( bw ) , the spitzer space telescope legacy science program through contracts 1224988 ( bw ) and 1224653 ( ec , bb , mm ) and nasa s spitzer space telescope fellowship program ( ri ) . benjamin , r. a. , et al . 2003 , , 115 , 953 bethell , t. , zweibel , e. g. , heitsch , f. , & mathis , j. s. 2004 , , 610 , 801 churchwell , e. , et al . 2004 , , 154 , 322 churchwell et al . 2006 , , _ submitted _ draine , b. t. 2003 , , 41 , 241 foster , j. b. , & goodman , a. a. 2006 , , 636 , l105 grski , k. m. , hivon , e. , banday , a. j. , wandelt , b. d. , hansen , f. k. , reinecke , m. , bartelmann , m. 2005 , , 622 , 759 heitsch , f. , mac low , m .- , & klessen , r. s. 2001 , , 547 , 280 heitsch , f. , zweibel , e. g. , mac low , m .- m . , li , p. , & norman , m. l. 2001 , , 561 , 800 heitsch , f. slyz , a. d. , devriendt , j. e. g. , burkert , a. 2006 , _ submitted _ heitsch , f. , whitney , b. a. , indebetouw , r. , meade , m. r. , babler , b. l. , churchwell , e. 2006 , _ in preparation _ jackson , j. et al . 2006 , _ in preparation _ kim , s .- h . , martin , p. g. , & hendry , p. d. 1994 , , 422 , 164 klessen , r. s. , heitsch , f. , & mac low , m .- m . 2000 , , 535 , 887 lger , a. , dhendecourt , l. , dfourneau , d. 1989 , , 216 , 148 li , a. , & draine , b. t. 2001 , , 550 , l213 li , a. , & draine , b. t. 2001 , , 554 , 778 mercer , e. p. et al . 2006 , , _ submitted _ padoan , p. , juvela , m. , & pelkonen , v .- m . 2006 , , 636 , l101 peeters , e. , van diedenhoven , b. , van kerckhoven , c. , hony , s. , tielens , a. g. g. m. , allamandola , l. j. , hudgins , d. m. , & bauschlicher , c. w. 2003 , astrophysics of dust , 41 reach et al . 2006 , the irac handbook , http://ssc.spitzer.caltech.edu/irac/dh/ schnee , s. , bethell , t. , & goodman , a. 2006 , , 640 , l47 slyz , a. d. , devriendt , j. e. g. , bryan , g. , & silk , j. 2005 , , 356 , 737 slyz , a. d. , devriendt , j. e. g. , bryan , g. , heitsch , f. & silk , j. 2005 , , _ submitted _ stetson , p. b. 1987 , , 99 , 191 whitney , b. a. et al . 2004 , , 154 , 315
nonlinearities prevail in fluid dynamics when the reynolds number , @xmath4 , is large @xcite . for geophysical flows , the reynolds number is often larger than @xmath5 and for some astrophysical flows values of @xmath6 is not unreasonable . the number of degrees of freedom ( _ dof _ ) in the flow increases as @xmath7 for @xmath8 in the kolmogorov framework @xcite ( hereafter k41 ) . such a huge number of _ dof _ makes direct numerical simulations ( dns ) of turbulence at high @xmath4 infeasible on any existing or projected computer for decades to come . because of this intractability , simulations of turbulence are always carried out in regions of parameter space far from the observed values , either with : ( a ) an unphysical lack of scale separation between the energy - containing , inertial , and dissipative ranges while parameterizing the missing physics , or ( b ) a study of the processes at much smaller length scales , often with periodic boundaries ( unphysical at large scales but used under the hypothesis of homogeneity of turbulent flows ) . clearly , modeling of unresolved small scales is necessary . two main approaches have been developed over the years to model the effects of the unresolvable small scales in turbulence on the scales resolved in the simulations . the first approach is large eddy simulations ( les , see @xcite ) . les is widely used in engineering , in atmospheric sciences , and to a lesser extent in astrophysics . however , in the les approach , the reynolds number is not known . instead , one attempts modeling the behavior of the flow in the limit of very large @xmath4 . another approach models the effects of turbulence at higher reynolds numbers than are possible with a dns on a given grid , by using a variety of techniques that can be viewed as filtering of the small scales a rather novel approach to modeling of turbulent flows employs _ regularization modeling _ as a sgs model @xcite . unlike closures which employ eddy - viscosity concepts ( modifying the dissipative processes ) , the approach of regularization modeling modifies the spectral distribution of energy . existence and uniqueness of smooth solutions can be rigorously proven , unlike many les models ( e.g. , eddy - viscosity ) , as well as the fact that the subgrid model recovers the navier - stokes equations in the limit of the filter width going to zero . their robust analytical properties ensure computability of solutions . proved the existence ( but not smoothness , or uniqueness ) of solutions to the navier - stokes equations in @xmath9 , n=2,3 using the leray model @xcite . geurts and holm @xcite began using the leray model with ( a three - point invertible approximation of ) an inverse - helmholtz - operator filter of width @xmath2 . later it was dubbed leray@xmath0 and an upper bound for the dimension of the global attractor was established @xcite . the global existence and uniqueness of strong solutions for the leray model is a classical result @xcite . leray@xmath0 has been compared to dns simulations on a grid of @xmath10 in a doubly - periodic compressible channel flow domain @xcite . its performance was found to be superior to a dynamic mixed ( similarity plus eddy - viscosity ) model ( with an even greater reduction in computational cost ) . however , it possessed a systematic error of a slight over - prediction of the large scales accompanied by a slight under - prediction of the small scales . it possessed both forward- and back - scatter , but exhibited too little dissipation . the leonard tensor - diffusivity model @xcite ( sometimes known as the clark model @xcite ) is the first term of the reconstruction series for the turbulent sub - filter stress for all symmetric filters that possess a finite nonzero second moment . this leading order approximation of the subgrid stress is thus generic @xcite . in _ a priori _ testing , it reconstructs a significant fraction ( @xmath11 , but not all ) of the subgrid stress , provides for local backscatter along the stretching directions while remaining globally dissipative , and possesses a better reconstruction of the subgrid stress than the scale - similarity model . used as a les , in _ a posteriori _ testing the leonard tensor - diffusivity model required additional dissipation ( a dynamic smagorinsky term ) to achieve reasonable gains in computation speed for 3d periodic flows and for channel flows @xcite . the leonard tensor - diffusivity model does not conserve energy in the non - viscous limit . cao , holm , and titi @xcite developed a related ( conservative ) subgrid model which they dubbed clark@xmath0 . the clark@xmath0 model applies an additional inverse - helmholtz filter operation to the reynolds stress tensor of the clark model . the global well - posedness of the clark@xmath0 model and the existence and uniqueness of its solutions were demonstrated , and upper bounds for the hausdorff ( @xmath12 ) and fractal ( @xmath13 ) dimensions of the global attractor were found @xcite . the third regularization model we will consider is the incompressible lagrangian - averaged navier - stokes ( lans@xmath0 , @xmath14model , also known as the viscous camassa - holm equation @xcite ) . it can be derived , for instance , by applying temporal averaging to hamilton s principle , where taylor s frozen - in turbulence hypothesis ( the only approximation in the derivation ) is applied as the closure for the eulerian fluctuation velocity in the reynold decomposition , at linear order in the generalized lagrangian mean description @xcite . in this derivation , the momentum - conservation structure of the equations is retained . for scales smaller than the filter width , lans@xmath0 reduces the steepness in gradients of the lagrangian mean velocity and thereby limits how thin the vortex tubes may become as they are transported , while the effect on larger length scales is negligible @xcite . lans@xmath0 may also be derived by smoothing the transport velocity of a material loop in kelvin s circulation theorem @xcite . consequently , there is no attenuation of resolved circulation , which is important for many engineering and geophysical flows where accurate prediction of circulation is highly desirable . an alternative interpretation of the @xmath14model is that it neglects fluctuations in the smoothed velocity field , while preserving them in the source term , the vorticity @xcite . lans@xmath0 has previously been compared to direct numerical simulations ( dns ) of the navier - stokes equations at modest taylor reynolds numbers ( @xmath15 @xcite , @xmath16 @xcite , and @xmath17 @xcite ) . lans@xmath0 was compared to a dynamic eddy - viscosity les in 3d isotropic turbulence under two different forcing functions ( for @xmath18 and @xmath19 ) and for decaying turbulence with initial conditions peaked at a low wavenumber ( with @xmath20 ) as well as at a moderate wavenumber ( with @xmath21 ) @xcite . lans@xmath0 was preferable in these comparisons because it demonstrated correct alignment between eigenvectors of the subgrid stress tensor and the eigenvectors of the resolved stress tensor and vorticity vector . the lans@xmath0 and the leray@xmath0 dynamic mixed model ( similarity plus eddy - viscosity ) in a turbulent mixing shear layer ( with @xmath22 ) @xcite . lans@xmath0 was found to be the most accurate of these three les candidates at proper subgrid resolution , but the effects of numerical contamination can be strong enough to lose most of this potential . while lans@xmath0 has the greatest grid - independent accuracy of the three models , it also requires the greatest resolution . from the les perspective , this could pose some limitations on the practical use and application of lans@xmath0 for high @xmath4 cases . indeed , recent high - resolution simulations of lans@xmath0 showed that energy artificially accumulates in the sub - filter - scales , giving as a result only a modest computational gain at very high reynolds number @xcite . we propose to pursue these previous studies of leray@xmath0 and lans@xmath0 further at high reynolds number , one goal is to as the three regularizations are related via truncation of sub - filter stresses , for lans@xmath0 , the predicted -scale spectra is @xmath23 @xcite . this scaling has been observed to be subdominant to an energy spectrum @xmath3 which corresponds to `` enslaved rigid body '' or `` polymerized '' portions of the fluid @xcite . the sub - filter scaling observed in the third - order structure function corresponded to the predicted @xmath24 scaling of the energy spectrum . however , regions were observed in the flow where no stretching was acting in the sub - filter scales . these regions , which give no contribution to the energy cascade , and hence do not affect the third order structure functions , are responsible for the @xmath3 scaling in the lans@xmath0 energy spectrum . for clark@xmath0 , the correct time scale for vortex stretching is difficult to determine and its spectrum is found to range between @xmath23 and @xmath25 @xcite . leray@xmath0 has the same difficulty and the spectrum can range between @xmath26 and @xmath27 @xcite . the determination of these scaling laws is needed to quantify the computational gain if each model is to be used as a sgs model . as a result , to determine empirically the sub - filter scale spectra . we present the three models and describe how they are related , derive the krmn - howarth equation for clark@xmath0 and leray@xmath0 ( from which exact scaling laws for third order quantities follow ) , and review theoretical predictions of inertial range scaling in section [ sec : models ] . in section [ sec : sfsstudy ] . we first compute a fully resolved dns of the navier - stokes equations at a resolution of @xmath28 ( @xmath29 , @xmath30 , and @xmath31 ) . we then perform model runs with the exact same conditions at a resolution of @xmath32 . we take @xmath2 to be @xmath33th the box size , which was found in an earlier study to be large enough to exhibit both navier - stokes and sub - filter - scale lans@xmath0 dynamics @xcite . this large filter case is important because it gives insight into the behavior of the models at scales much smaller than the filter width without requiring higher resolution than is feasible . we compare the three regularizations as subgrid models guided by a previous study of lans@xmath0 @xcite , we take @xmath2 to be @xmath34th the box size . this choice was found to produce an optimal @xmath14les ( in the sense of being optimal for the class of lans@xmath0 models , with respect to the value of @xmath2 ) . finally , we review bounds on the size of the attractors and use these bounds to comment on the computational savings of the three regularizations viewed as sgs models . the incompressible navier - stokes equations are given in cartesian coordinates by @xmath35 filtering these equations with a convolution filter , @xmath36 in which @xmath37 ( resp . @xmath38 ) denotes the filtered ( resp . unfiltered ) field , yields @xmath39 in which by convention we denote @xmath40 and the reynolds turbulence stress tensor , @xmath41 , represents the closure problem . it has been shown @xcite that for all symmetric filters possessing a finite nonzero second moment , the first term of the reconstruction series for the turbulent sub - filter stress is @xmath42 where @xmath43 is the fourier transform of the convolution kernel ( @xmath44 is the convolution kernel where @xmath45({\mathbf}{r})=\int g({\mathbf}{r}-{\mathbf}{r { ' } } ) z({\mathbf}{r { ' } } ) d^3{\mathbf}{r{'}}$ ] ) . this approximation of the subgrid stress is then generic and is known as the leonard tensor - diffusivity model @xcite ( or , often , the clark model @xcite ) . related to this model , the clark@xmath0 model is @xcite @xmath46 and its subgrid stress for @xmath47 is given by @xmath48 here the filter is the inverse of a helmholtz operator , @xmath49 . the clark@xmath0 model conserves energy in the @xmath50 norm instead of the @xmath51 norm , @xmath52 in which the clark@xmath0 energy @xmath53 is expressed as @xmath54 and the clark@xmath0 energy dissipation rate is given by @xmath55 where @xmath56 and @xmath57 . for @xmath58 , we note that @xmath59 which implies that the turbulent sub - filter stress tensor for the tensor - diffusivity model given by eq . ( [ eq : carati ] ) is @xmath60 which is proportional to the clark@xmath0 stress tensor to second order in @xmath2 . hence , the _ a priori _ tests of @xcite should apply to clark@xmath0 , at least at this order . in 1938 , krmn and howarth @xcite introduced the invariant theory of isotropic hydrodynamic turbulence , and derived from the navier - stokes equations the exact law relating the time derivative of the two - point velocity correlation to the divergence of the third - order correlation function . the corresponding krmn - howarth theorem for lans@xmath0 in the fluid case was derived in @xcite . the relevance of the krmn - howarth theorem for the study of turbulence can not be underestimated . as a corollary , rigorous scaling laws in the inertial range can be deduced . in this section , we derive these results for the clark@xmath0 case . for the sake of simplicity , we consider the case @xmath61 , since the dissipative terms may be added at any point in the derivation . we denote @xmath62 and begin our investigation of the correlation dynamics by computing the ingredients of the partial derivative @xmath63 . the clark@xmath0 motion equation ( [ eq : clark ] ) may be rewritten as @xmath64 combining eqs . ( [ eq : subgrid ] ) and ( [ eq : sgs_clark ] ) , we arrive at the fluctuation - velocity equation , @xmath65 where @xmath66 ( @xmath67 ) . multiplying eq . ( [ eq : clarkv2 ] ) by @xmath68 and eq . ( [ eq : clarku2 ] ) by @xmath69 , then adding the result yields @xmath70 where we have used statistical homogeneity @xmath71 we symmetrize eq . ( [ fluct - eqn ] ) in the indices @xmath72 by adding the corresponding equation for @xmath73 . we then use homogeneity again as @xmath74 and define the tensors @xmath75 we can drop @xmath76 because the terms with the pressures @xmath77 and @xmath78 vanish everywhere , as follows from the arguments of isotropy @xcite . finally , we obtain @xmath79 this is the krmn - howarth equation for clark@xmath0 ( compare to eq . ( 3.8 ) in @xcite for lans@xmath0 ) . by dimensional analysis , the energy dissipation rate in clark@xmath0 is @xmath80 and eq . ( [ eq : kh1 ] ) implies @xmath81 for large scales ( @xmath82 ) , we recover the navier - stokes scaling known as the four - fifths law , @xmath83 @xcite . here , @xmath84\cdot{\mathbf}{l}/l$ ] is the longitudinal increment of @xmath85 . strictly speaking , the four - fifths law expresses that the third - order longitudinal structure function of @xmath85 , @xmath86 , is given in the inertial range in terms of the mean energy dissipation per unit mass @xmath87 by @xmath88 or , equivalently , that the flux of energy across scales in the inertial range is constant . we also recover the kolmogorov 1941 @xcite ( hereafter , k41 ) energy spectrum , @xmath89 or , equivalently , @xmath90 for sub - filter scales ( @xmath91 ) , we have @xmath92 and the first and third right - hand terms in eq . ( [ eq : full_clark ] ) are equivalent . in this case , we are left with two different possible scalings depending on the prefactors in eq . ( [ eq : full_clark ] ) . if the first ( or third ) right - hand term is dominant , our scaling law becomes @xmath93 for our sub - filter scale energy spectrum we would then have @xmath94 , or , equivalently , @xmath95 this result is the same as for the @xmath14model @xcite . if , however , the second right - hand term in eq . ( [ eq : full_clark ] ) is dominant , then the k41 results are recovered , with @xmath96 substituted for @xmath85 . in that case , one finds the alternative clark@xmath0 sub - filter scale spectral energy scaling , @xmath97 we review here the derivation by dimensional analysis of the spectrum which follows the scaling ideas originally due to kraichnan @xcite and which is developed more fully in ref . @xcite . in examining the nonlinear terms in eq . ( [ eq : clark ] ) , it is not entirely clear which of three possible scales for the average velocity for an eddy of size @xmath1 , @xmath98 @xmath99 or @xmath100 should result . therefore , three corresponding `` turnover times '' , @xmath101 , for such an eddy may be proposed @xmath102 the term `` turnover time '' is used advisedly here , since only the velocity @xmath103 is composed of the fluid transport velocity . we define the ( omnidirectional ) spectral energy density , @xmath104 , from the relation @xmath105 since , @xmath106 , we have @xmath107 then , the total energy dissipation rate , @xmath108 is related to the spectral energy density by @xmath109 which yields , finally , the predicted energy spectra for clark@xmath0 , @xmath110 , @xmath111 for scales much larger than @xmath2 ( @xmath112 ) the kolmogorov scaling for navier - stokes is recovered , @xmath113 whereas for scales much smaller ( @xmath114 ) , the spectrum becomes @xmath115 these arguments constrain the clark@xmath0 sub - filter scale spectrum to lie between @xmath1 and @xmath116 . the leray model in cartesian coordinates is @xmath117 where the flow is advected by a smoothed velocity , @xmath96 . by comparison with eq . ( [ eq : subgrid ] ) we see that the leray model approximates the subgrid stress as @xmath118 , or , with @xmath119 , @xmath120 as has been noted previously @xcite , the subgrid stress of clark@xmath0 in eq . ( [ eq : sgs_clark ] ) is a truncation of the subgrid stress of leray@xmath0 , in eq . ( [ eq : sgs_leray ] ) . for leray@xmath0 , the @xmath121 norm is the quadratic invariant that is identified with energy , @xmath122 where @xmath123 and @xmath124 as was pointed out in @xcite , the incompressibility of the velocity field @xmath85 only implies a divergenceless filtered velocity @xmath96 under certain boundary conditions for leray@xmath0 . when @xmath125 , the energy @xmath126 is no longer conserved ( helicity and kelvin s theorem are not conserved for leray@xmath0 ) . in our numerical study , we employ periodic boundary conditions , for which @xmath127 implies @xmath128 and leray@xmath0 conserves energy in the usual sense of @xmath51 . in this section we derive the krmn - howarth equation for the leray@xmath0 case . following section [ sec : kh - clark ] , we begin our investigation of the correlation dynamics by computing the ingredients of the partial derivative @xmath129 . ( [ eq : leray ] ) may be rewritten as : @xmath130 multiplying eq . ( [ eq : lerayv2 ] ) by @xmath131 yields @xmath132 we can make this equation symmetric in the indices @xmath72 , by adding the equation for @xmath133 . we define the tensors @xmath134 again , we may drop @xmath135 because the terms with the pressure @xmath136 vanish everywhere and thereby obtain @xmath137 this is the krmn - howarth equation for leray@xmath0 . the energy dissipation rate for leray@xmath0 is denoted by @xmath138 and it satisfies @xmath139 . by dimensional analysis , eq . ( [ eq : kh1leray ] ) implies @xmath140 for large scales ( @xmath82 ) , we recover the navier - stokes scaling , eqs . ( [ eq : fourfifths ] ) and ( [ eq : k41 ] ) . for sub - filter scales ( @xmath91 ) our scaling law becomes @xmath141 for our small scale energy spectrum we would then have @xmath142 ( where we employed @xmath92 ) , or , equivalently , cf . ( [ eq : clarka_spectrum2 ] ) , @xmath143 we review here the derivation by dimensional analysis of the spectrum for leray@xmath0 as we did for clark@xmath0 in section [ sec : pheno_clark ] . this analysis is developed more fully in ref . we argue again that there are three possible scales for the average velocity for an eddy of size @xmath1 , eqs . ( [ eq : uk0 ] ) , ( [ eq : uk1 ] ) , and ( [ eq : uk2 ] ) , with the turn - over time , @xmath144 given by eq . ( [ eq : turnover ] ) . since , @xmath145 , we have @xmath146 then , the total energy dissipation rate , @xmath138 is related to the spectral energy density by @xmath147 which yields , finally , the predicted energy spectra for leray@xmath0 , @xmath148 , @xmath149 for scales much larger than @xmath2 ( @xmath112 ) the k41 spectrum is recovered , eq . ( [ eq : k41 ] ) , and for scales much smaller ( @xmath114 ) the spectrum is @xmath150 these arguments constrain the leray@xmath0 sub - filter scale spectrum to lie between @xmath151 and @xmath152 . lans@xmath0 is given by @xmath153 for lans@xmath0 , the usual choice of filter is again @xmath154 with this filter , the subgrid stress tensor is given by @xmath155 as has been previously noted @xcite , the subgrid stress of leray@xmath0 , eq . ( [ eq : sgs_leray ] ) , is a truncation of the subgrid stress of lans@xmath0 eq . ( [ eq : sgs_lans ] ) . like clark@xmath0 , energy is conserved in the @xmath156 norm instead of the @xmath51 norm . additionally , lans@xmath0 is the only model of the three examined here that conserves the helicity ( and kelvin s circulation theorem ) . for lans@xmath0 in the fluid case the krmn - howarth theorem was derived in @xcite . we summarize here the dimensional analysis argument for the lans@xmath0 inertial range scaling that follows from this theorem , beginning from equation ( 3.8 ) in ref . @xcite . in the statistically isotropic and homogeneous case , without external forces and with @xmath157 , taking the dot product of eq . ( [ eq : lans ] ) with @xmath158 yields the equation @xmath159 the trace of this equation is the fourier transform of the detailed energy balance for lans@xmath0 ; @xmath160 is the second - order correlation tensor while @xmath161 and @xmath162 are the third - order correlation tensors for lans@xmath0 and @xmath163 is the sub - filter scale stress tensor . for @xmath164 this reduces to the well - known relation derived by krmn and howarth . the energy dissipation rate for lans@xmath0 , @xmath165 , satisfies @xmath166 . by dimensional analysis in eq . ( [ eq : holm38 ] ) we arrive at @xmath167 for large scales ( @xmath82 ) , we recover the navier - stokes scaling eqs . ( [ eq : fourfifths ] ) and ( [ eq : k41 ] ) . for sub - filter scales ( @xmath91 ) our scaling law becomes eq . ( [ eq : lcube ] ) and our sub - filter scale spectra is given by @xmath168 in this case , by the phenomenological arguments , we know that eddies of size @xmath1 are advected by the smoothed velocity , eq . ( [ eq : uk2 ] ) . this scaling is confirmed in ref . but it coexists with a @xmath169 energy spectrum corresponding to `` enslaved rigid bodies '' or `` polymerized '' portions of fluid which do not contribute to the turbulent energy cascade . to this end , we compute numerical solutions to eqs . ( [ eq : ns ] ) , ( [ eq : clark ] ) , ( [ eq : leray ] ) , and ( [ eq : lans ] ) in a three - dimensional ( 3d ) cube with periodic boundary conditions using a parallel pseudospectral code @xcite . we employ a taylor - green forcing @xcite , @xmath170 \label{eq : tg}\ ] ] ( with @xmath171 ) , and employ dynamic control @xcite to maintain a nearly constant energy with time . the taylor - green forcing , eq . ( [ eq : tg ] ) , is not a solution of the euler s equations , and as a result small scales are generated rapidly . the resulting flow models the fluid between counter - rotating cylinders @xcite and it has been widely used to study turbulence , including studies in the context of the generation of magnetic fields through dynamo instability @xcite . we define the taylor microscale as @xmath172 and the mean velocity fluctuation as @xmath173 . the taylor microscale reynolds number is defined by @xmath174 and the reynolds number based on a unit length is @xmath175 . to compare the three regularizations ( clark@xmath0 , leray@xmath0 , and lans@xmath0 ) we compute a fully resolved dns of the navier - stokes equations at a resolution of @xmath28 ( @xmath29 , @xmath176 ) and model runs with the exact same conditions at a resolution of @xmath32 . the details of the flow dynamics of the dns have already been given @xcite . in particular , the reynolds number based on the integral scale @xmath177 ( where @xmath178 is the total energy ) is @xmath179 , where @xmath180 is the _ r.m.s . _ velocity and the reynolds number based on the taylor scale is @xmath31 . the dns was run for nine turnover times @xmath181 ( in the following results , time @xmath182 is in units of the turnover time ) . we employ a filter width of @xmath183 for which lans@xmath0 exhibits both navier - stokes and lans@xmath0 inertial ranges in the third - order structure function @xcite . from these we hope to obtain the behavior of the models for scales much smaller than @xmath2 for leray@xmath0 and in the dns , @xmath184 for lans@xmath0 and clark@xmath0 ) . dns ( @xmath176 ) is shown as solid black lines , lans@xmath0 as dotted red , clark@xmath0 as dashed green , and leray@xmath0 as blue dash - dotted . the cyan long - dashed line represents a @xmath185 dns ( @xmath186 , @xmath187 ) . here each run is calculated only until it reaches a statistical steady state . leray@xmath0 reduces the dissipation , @xmath188 , and increases the time scale to reach a statistical turbulent steady - state . both effects are greater as @xmath2 is increased . by comparison with the @xmath189 run , we see that these two effects are consistent with a reduced effective reynolds number . a smaller reduction in flux ( but not an increase in time to steady - state ) is also observed for lans@xmath0 and is likely related to its rigid bodies.,width=338 ] in fig . [ fig : enstrophy ] we present the time evolution of the enstrophy ( @xmath190 for leray@xmath0 and dns , @xmath184 for lans@xmath0 and clark@xmath0 ) , which is proportional to the dissipation ( @xmath191 or @xmath192 depending on the case ) . also shown is a well - resolved @xmath185 dns of a less turbulent flow ( @xmath193 , @xmath194 , @xmath187 , ( cyan online ) long - dashed line ) . here each run is calculated only until it reaches a statistically steady state . we see that the dissipation is reduced and the time - scale is increased for leray@xmath0 . we see by comparison with the @xmath194 dns that a reduced effective reynolds number in the leray@xmath0 run . for lans@xmath0 , the dissipation is decreased although the time to reach steady state is not increased . this is probably related to the enslavement of its rigid body regions which would have no internal dissipation . of the three models , clark@xmath0 the total dissipation for a large range of @xmath2 . indeed , as it is the order @xmath195 approximation of navier - stokes , this dissipation behavior for clark@xmath0 may continue to hold until @xmath2 becomes quite large . figure [ fig : eavg ] shows a comparison of the energy spectrum at the turbulent steady state , for all the runs in fig . [ fig : enstrophy ] . the isotropic energy spectra are calculated as follows , @xmath196 where @xmath197 ( the @xmath198 norm is employed for clark@xmath0 and lans@xmath0 ) . the length scale @xmath2 is indicated by a vertical dashed line and the plotted energy spectra are compensated by @xmath199 ( i.e. , leading to a flat k41 @xmath152 spectrum ) . from fig . [ fig : eavg ] it is clear that the clark@xmath0 spectral behavior is close to the predicted @xmath1 spectrum , rather than the @xmath200 from eq . ( [ eq : clarka_spectrum2 ] ) , or the other possible spectrum from eq . ( [ eq : clarka_spectrumn ] ) . likewise , the @xcite . leray@xmath0 , on the other hand , possesses a very steep sub - filter scale spectrum as well as enhanced large - scale energy as has been previously observed @xcite . the spectrum of leray@xmath0 in fig . [ fig : eavg ] gives a good approximation to the @xmath189 dns in the range @xmath201 $ ] ( i.e. , to @xmath193 rather than to @xmath202 which was employed ) . this result , leray@xmath0 s increased characteristic time scales and its reduced dissipation imply that the leray@xmath0 model is operating at a much lower effective reynolds number . this is also clear from the rapid drop in the spectrum at small scales , shown in fig . [ fig : eavg ] . indeed , we can build an effective reynolds number in the large scales as @xmath203 . since @xmath204 is controlled in this simulation by the forcing scale , the drop in the dissipation rate implies a reduced nonlinearity in leray@xmath0 . this is also consistent with a direct comparison of the nonlinear terms in leray@xmath0 with , for instance , lans@xmath0 . the nonlinear terms in lans@xmath0 , eq . ( [ eq : lans ] ) , may be written as @xmath205 ( where the suffix @xmath206 denotes a transposition ) , while the nonlinear term in leray@xmath0 , eq . ( [ eq : leray ] ) , is only @xmath207 . both nonlinear terms in lans@xmath0 are of order @xmath208 ; so the absence of one of the nonlinear terms in leray@xmath0 be understood as a reduction in the nonlinearity . validation of the krmn - howarth equation scalings , eqs . ( [ eq : lcube ] ) and ( [ eq : lcubeleray ] ) , enables us to measure scaling laws in the inertial range and , thus , compare the intermittency properties of the models . the third - order correlations involved in the theorems , namely @xmath209 for clark@xmath0 and lans@xmath0 , and @xmath210 for leray@xmath0 , and @xmath211 for navier - stokes , are plotted versus @xmath212 in fig . [ fig : theorem ] . in fig . [ fig : theorem ] we can see validation of the krmn - howarth scaling for scales smaller than @xmath2 for both lans@xmath0 and clark@xmath0 . in particular , we note the observed scaling for clark@xmath0 verifies the @xmath213 scaling and not the ( theoretically possible ) @xmath214 ( @xmath215 ) scaling . the predicted scaling is not observed in leray due to its reduced effective reynolds number . with these scalings in hand , we may proceed to observe the scaling of the longitudinal structure functions , @xmath216 where we again replace the @xmath217 norm , @xmath218 , for the @xmath219 norm , @xmath220 , in the case of clark@xmath0 and lans@xmath0 . we utilize the extended self - similarity ( ess ) hypothesis @xcite which proposes the scaling @xmath221 and normalize the results by @xmath222 to better visualize the deviation from linearity ( which serves as a measure of intermittency ) . the results are displayed in fig . [ fig : exponents ] . in fig . [ fig : exponents ] , we may observe the intermittency properties of the models at sub - filter scales . we note a reduced intermittency for both leray@xmath0 and the @xmath194 dns . this is consistent with the smoother , more laminar fields ( due to the reduction of the effective @xmath4 ) possessed by both . interestingly , though lans@xmath0 and clark@xmath0 both possess the same cascade scaling ( eq . ( [ eq : lcube ] ) , as confirmed in fig . [ fig : theorem ] ) , the clark@xmath0 model is markedly more intermittent than lans@xmath0 . if artificially truncated local interactions ( in spectral space ) is taken as a cause of enhanced intermittency @xcite , then the increased intermittency observed in clark@xmath0 is the expected result of truncation of the higher - order terms in the sub - filter stress tensor . moreover , if the lans@xmath0 s @xmath3 spectrum is indeed associated with rigid bodies , these would serve to decrease the intermittency ( no internal degrees of freedom being available in a rigid body ) which is consistent with the results shown here . due to this effect , lans@xmath0 of the three regularization models the high - order intermittency of @xmath223 corresponding to an optimal @xmath14les @xcite . note that the value of @xmath2 has been optimized for neither clark@xmath0 nor leray@xmath0 ; as a consequence , these models might perform better in other parameter regimes than the results indicate in this study . for leray@xmath0 and in the dns , @xmath184 for lans@xmath0 and clark@xmath0 ) . dns ( @xmath176 ) is shown as solid black lines , lans@xmath0 as dotted red , clark@xmath0 as dashed green , and leray@xmath0 as blue dash - dotted . an under - resolved ( @xmath32 ) navier - stokes run is shown as a pink dash - triple - dotted line.,width=338 ] fig . [ fig : enstrophy2 ] along with that of an under - resolved navier - stokes solution at a resolution of @xmath32 ( @xmath202 , ( pink online ) dash - triple - dotted line ) . we see that both lans@xmath0 and clark@xmath0 reproduce the proper amount of dissipation and are within @xmath224 of the time required by the dns to reach a statistical turbulent steady state . as has been observed before , leray@xmath0 is under - dissipative @xcite . we also note that it takes longer than the compensated spectra averaged over several eddy turn - over times are shown for the sgs case ( i.e. , @xmath225 ) in fig . [ fig : eavg2 ] . note that as the subgrid models are averaged over a different time interval , no meaningful comparison to the dns is possible for @xmath226 . even without an optimal choice for the value of @xmath2 , clark@xmath0 best reproduces the dns spectrum for scales larger than @xmath2 . we compute root - mean - square spectral errors where @xmath227 is the wavenumber for the forcing scale , @xmath228 is the dns spectrum ( in the @xmath121 norm ) , and @xmath229 is the subgrid model spectrum ( in the appropriate norm ) . we see that only clark@xmath0 improves the estimate of the power spectrum at this resolution for leray@xmath0 performs the poorest of the three regularization models , but it is also not optimized . as previously argued , its effective reynolds number is too low to accurately model the dns flow . either a decrease in the viscosity @xmath230 , or a decrease in the filter size @xmath2 ( and , hence , an increase in the nonlinearity ) , or both would likely improve the accuracy of leray@xmath0 as an sgs model . due to its frozen - in ( or enslaved ) rigid - body regions , the lans@xmath0 model can not reproduce the dns spectrum unless @xmath2 is only a few times larger than the dissipation scale @xcite . [ cols="^,^,^ " , ] another measure of the success of a subgrid model is the reproduction of structures in the flow . in figure [ fig : vapor ] we have 3d volume rendering of the enstrophy density @xmath231 ( @xmath232 for lans@xmath0 and clark@xmath0 ) for the dns , the three sgs - model simulations ( @xmath233 ) , the @xmath32 under - resolved navier - stokes solution , all at a reynolds number of @xmath234 , and the @xmath194 dns . due to the late times depicted ( longer than a lyapunov time ) there can be no point - by - point comparison between the simulations . instead , we note that there are four horizontal bands where the forcing causes a maximum shear . this large - scale feature of the flow is missing only from leray@xmath0 and the @xmath194 run . the three other runs reproduce this feature well ( note that the apparently thicker tubes present in clark@xmath0 are vortex tube mergers ) . the results lead again to the conclusion that the under - resolved navier - stokes , the clark@xmath0 , and the lans@xmath0 models are better subgrid models than leray@xmath0 due to its reduced effective @xmath4 . for the sgs models , the predicted @xmath235 from the krmn - howarth theorem for navier - stokes is well - reproduced by all models at super - filter scales . in fig . [ fig : exponents2 ] the sgs model intermittency results . all models reproduce the intermittency up to the tenth - order moment within the error bars ( although there is a small decrease in intermittency for leray@xmath0 ) . thus , we conclude that with adequately chosen values of @xmath2 ( and of @xmath230 for leray@xmath0 ) , all three models can reproduce the intermittency of the dns ( to within the error bars ) . the rationale behind using is that it leads to adequate solutions at a reduced computational cost , because it computes fewer _ dof _ ; indeed , for an , the ratio of navier - stokes s _ dof _ to the model s _ dof _ , a prediction for memory savings and hence computation time savings for numerical simulation , is a crucial factor . consequently , analytical bounds on the sizes of the attractors for the three regularization subgrid models may be useful indicators of their computational savings . the _ dof _ for lans@xmath0 is derived in @xcite and confirmed in @xcite , @xmath236 where @xmath237 is the integral scale ( or domain size ) . we may compare this to the _ dof _ for navier stokes , @xmath238 which immediately yields @xmath239 it was found , however , that to reproduce the energy spectrum of an equivalent dns , the filter - width @xmath2 must be no larger than a few times the dissipation scale , @xmath240 @xcite . this is the result of the `` polymerization '' of the flow in lans@xmath0 , and the associated @xmath241 scaling at sub - filter scales . with this added caveat , it follows that the reduction in _ dof _ is independent of @xmath4 ( and a net factor of about @xmath242 ) . our study here illustrates that the high - order structure functions may be reproduced for much larger values of @xmath2 . therefore , in applications where the spectrum is not of great concern , much greater reduction in numerical resolution would be feasible . for clark@xmath0 there is an upper bound on the hausdorff , @xmath12 , and fractal , @xmath13 , dimensions of the attractor , @xmath243 where @xmath244 is the kolmogorov dissipation length scale corresponding to the clark@xmath0 model @xcite . from its observed @xmath1 spectrum , we may estimate @xmath244 or , equivalently , @xmath245 . for dissipation the large wavenumbers dominate and , therefore , combining the clark@xmath0 energy balance eq . ( [ eq : lansa_balance ] ) with its sub - filter scale energy spectrum eq . ( [ eq : clarka_spectrum ] ) allows us to implicitly specify its dissipation wavenumber , @xmath246 , by @xmath247 then we have , @xmath248 it follows that @xmath249 this is similar to the prediction for lans@xmath0 , but as energy spectra are more easily reproduced for larger values of @xmath2 than with lans@xmath0 ( but not the intermittency properties ) , it may be the case that @xmath2 is not tied to the kolmogorov dissipation scale @xmath240 . if so , then the computational saving might increase as @xmath250 which is promising for use of clark@xmath0as an les model . this conclusion is bolstered to the extent that the results in section [ sec : super ] for @xmath233 ( @xmath251 ) are acceptable . if , there is still a greater reduction in _ dof _ ( a factor of 20 ) for clark@xmath0 than for lans@xmath0 . for leray@xmath0 , we have the following upper bounds on the hausdorff dimension , @xmath12 , and fractal dimension , @xmath13 , of the global attractor , @xmath252 where @xmath253 is the dissipation length scale for leray@xmath0 @xcite . again , we estimate the dissipation wavenumber for leray@xmath0 @xmath254 . then , from eqs . ( [ eq : leray_balance ] ) and ( [ eq : leray_spectrum ] ) , that is , assuming the @xmath151 spectrum resulting from the krmn - howarth equation , we find @xmath255 consequently we have , @xmath256 it follows that @xmath257 our results suggest that for an effective les the viscosity @xmath258 must be chosen to be _ smaller _ than @xmath230 . this leads to an upper bound on the computational savings for leray@xmath0 , @xmath259 if we further assume that @xmath2 is directly proportional to the dissipative scale @xmath240 , we arrive at @xmath260 which is not exceedingly promising for use as a les . all such estimates are , however , purely conjectural until the proper choices of @xmath2 and @xmath258 are determined . we derived the krmn - howarth equations for the leray@xmath0 and clark@xmath0 models . these two models may be viewed as successive truncations of the sub - filter scale stress of the lagrangian - averaged navier - stokes @xmath14model ( lans@xmath0 ) . in the case of clark@xmath0 two different inertial range scalings follow from the dimensional analysis of this equation . the case of leray@xmath0 is simpler as a single scaling is predicted . this is the case for navier - stokes and lans@xmath0 as well . solutions for a @xmath28 dns under periodic boundary conditions ( @xmath261 ) using a @xmath32 resolution under the same exact conditions for lans@xmath0 , leray@xmath0 , clark@xmath0 , and an under - resolved @xmath32 solution of the navier - stokes equations . we employed two different filter widths @xmath2 . the first choice @xmath262 was used to and the second choice @xmath263 was employed to test the sgs potential of the models . in comparing these two choices , we found for leray@xmath0 that an increase in @xmath2 substantially decreases the nonlinearity ( and hence decreases the effective reynolds number @xmath4 ) . for this reason , we were unable to confirm either the inertial range scaling from its krmn - howarth equation or its sub - filter scale energy spectrum . for clark@xmath0 we were able to determine the dominant krmn - howarth inertial range scaling to be @xmath264 associated @xmath1 energy spectrum , the performance of the three regularizations as sgs models ( for a resolution of @xmath32 and @xmath225 ) was comparable to that of the under - resolved navier - stokes solution in reproducing the dns energy spectrum . only clark@xmath0 showed a clear improvement in approximating the spectrum . from 3d volume rendering of enstrophy density we found that clark@xmath0 and lans@xmath0 were comparable to the under - resolved solution . even at @xmath265 , leray@xmath0 s 3d spatial structures are consistent with a significantly reduced @xmath4 flow ( e.g. , comparable to a @xmath266 dns ) . we note that the value of @xmath2 was chosen optimally for lans@xmath0 at the resolution of @xmath32 , and that for clark@xmath0 ( and especially for leray@xmath0 ) smaller resolutions may have comparable results for this value . such a comparison is beyond the scope of the present work . although lans@xmath0 and clark@xmath0 exhibit the same inertial range scaling arising from similarities in their krmn - howarth equations , clark@xmath0 is decidedly more intermittent than at the same time , lans@xmath0 is only slightly more intermittent than navier - stokes . these results are consistent with the artificial truncation of local nonlinear interactions ( in spectral space ) in the sgs stress tensor of each model . this effect is reduced for lans@xmath0 by the `` rigid - body regions '' enslaved in its larger scale flow which possess no internal degrees of freedom . the reduced intermittency observed for leray@xmath0 is related to its smoother , more laminar fields as a result of its reduced effective @xmath4 . finally , we analyzed the reduction in the number of _ dof _ in the models , as compared to navier - stokes ( and , hence , their les potential based on their computational savings ) . we noted that as lans@xmath0 reproduces the intermittency properties of a dns quite well even for larger values of @xmath2 , some further reduction in numerical saving might be achieved provided the contamination due to its @xmath169 rigid - body energy spectrum were not important in a given application . as clark@xmath0 possesses a similar reduction in _ dof _ to lans@xmath0 , its les potential is tied to the optimal value of @xmath2 for les . our study indicates that clark@xmath0 may be applicable ( especially with regards to the energy spectrum ) for larger values of @xmath2 than lans@xmath0 . in fact , if its optimal value is not a function of @xmath4 , the computational resolution savings increases as @xmath250 for clark@xmath0 . for the case of leray@xmath0 , the prediction is complicated by the effective reduction in @xmath4 as @xmath2 increases . prediction of optimized values of @xmath2 and of effective dissipation @xmath258 are required to assess its les potential . future work should include such a study for both leray@xmath0 and clark@xmath0 . all three regularizations were shown to be successful , in that their control of the flow gradient reduces the degrees of freedom and saves computation while preserving a properly defined reynolds number ( albeit for leray@xmath0 that definition is not yet ) . clark@xmath0 accurately reproduces the total dissipation , the time scale to obtain a turbulent statistical steady - state and the large - scale energy spectrum of a dns . these results seem to result from clark@xmath0 being an order @xmath195 approximation of navier - stokes . we have shown that leray@xmath0 reduces the effective reynolds number of the flow . the last of the three models , lans@xmath0 restores kelvin s circulation theorem ( advected by a smoothed velocity ) and the conservation of a form of helicity . using spectra as a measure of the success of a subgrid model , lans@xmath0 is less than optimal , due to its however , other measures of the success of a subgrid model are possible : for example , in regard to intermittency , lans@xmath0 may be considered a superior model . for clark@xmath0 , intermittency may be a function of filter width while for lans@xmath0 , intermittency does not vary much with @xmath2 . we remark that the computational gain thus far achieved by any of these regularizations is insufficient for applications at very high reynolds numbers , and the three subgrid stress tensors discussed here may need to be supplemented with an enhanced effective viscosity . this is a common practice when implementing the clark model ( see e.g. , ref . @xcite for a study of this model with an extra smagorinsky term ) . in this light , the present study may be useful as an analysis of the properties of the sgs tensors of the regularizations , and to pick best candidates , before the addition of enhanced dissipation . studies similar to that in ref . @xcite will also need to be done for the cases of clark@xmath0 and leray@xmath0 to quantify their computational savings before the addition of such dissipative terms . computer time provided at ncar is gratefully acknowledged , as well as nsf - cmg grant 0327888 . the work of ddh was partially supported by the us department of energy , office of science , advanced scientific computing research , and by the royal society of london wolfson award for meritorious research . a. leonard . . in _ turbulent diffusion in environmental pollution ; proceedings of the second symposium , charlottesville , va . , april 8 - 14 , 1973 . volume a. ( a75 - 30951 13 - 47 ) new york , academic press , inc . , 1974 , p. 237 - 248 . nasa - supported research . _ , pages 237248 , 1974 . j. pietarila graham , d. holm , p. mininni , and a. pouquet . highly turbulent solutions of the lagrangian - averaged navier - stokes alpha model and their large - eddy - simulation potential . , 76:056310+ , 2007 .
the low energy processes with a hadron participation are permanently in the center of theoretical and experimental activity . new measurements and more precise experimental data force us to look for new theoretical approaches . there are two known and widely used methods of effective lagrangian s approach in hadron physics : the effective lagrangian deriving from the qcd principles immediately @xcite-@xcite , and the use of various dynamical symmetries to build the basic lagrangian in a phenomenological way @xcite-@xcite . the linear sigma - model @xmath1 is the most popular and examined part of the second way . it is the main approach to the effective analysis of nucleon - nucleon @xcite-@xcite and quark - meson interaction @xcite together with the vector dominance model @xcite-@xcite . an exact theory of the meson interactions should be nonperturbative . that is why a lot of hadron characteristics ( effective couplings , formfactors etc . ) are calculated , in particular , in the qcd sum rules method operating with some nonperturbative vacuum parameters . and that is precisely why we need in effective meson interaction theory giving an information on the low - energy effects and hadron structure . an effective model of @xmath2 chiral interactions can be deduced from the fundamental qcd @xcite , using the bosonization procedure . in such a way , the gauge structure of meson interactions follows from the quark level qft @xcite . thus , the initial quarks with current masses transform to the constituent ones with effective masses @xmath3 , including some contribution of the internal gluon substructure . therewith , quark loops simulate an uprising of nonperturbative model parameter . so , the proportionality of pion mass to the `` quark condensate '' is restored in the framework of such effective models @xcite . further , to incorporate electromagnetic and strong ( vectorlike ) interactions into the effective theory an appropriate vector fields are introduced as the gauge fields . the vector meson treatment as the gauge fields realizing some dynamical symmetry , is useful to diminish some theoretical uncertainty of phenomenological description of the hadron interactions . in refs . @xcite-@xcite the gauge models were successfully used for the consideration of some low - energy aspects of baryon - meson interaction . it means that the fundamental quantum field principles can be applied to the describing of interaction at the different hierarchical levels . in the case under consideration , it provides a transition from the quark - level sigma model ( @xmath4 ) to the nucleon - level sigma model ( @xmath5 ) . to construct an effective meson interaction model , we will use the gauge scheme based on @xmath0 group . this group is the simplest one to consider the light unflavored vector mesons , @xmath6 and @xmath7 , together with the photon , as the gauge fields . moreover , it has the necessary symmetry to describe the electromagnetic and strong meson - meson and quark - meson interactions , which are insensitive to the chiral structure . so , to analyze the electromagnetic and strong effects only , it is sufficient to localize the diagonal sum of the global chiral @xmath8 . this sum corresponds to @xmath9 subgroup of the total gauge group @xmath0 . the extra @xmath10 groups are introduced for the gauge realization of the vector meson dominance ( see the next section ) . note that in the model we deal with the constituent quarks not with the fundamental current ones . obviously , the electromagnetic interaction violates the isospin symmetry . after the spontaneous breaking of the symmetry , higgs fields can be associated with the scalar mesons @xmath11 and @xmath12 ( see the next section ) . as it was noted above , this model reproduces the relation @xmath13 , if the quark condensate is taken into account in the equation for the vacuum shift . further , quarks arise in the model as the gauge group representations , and these degrees of freedom account for the internal meson structure . in fact , quark - meson models ( see also @xcite ) are necessary hybrid approaches to reproduce both nonperturbative and structure effects of the low - energy meson interactions . in this paper , the gauge scheme interactions of @xmath7 , @xmath6 and @xmath14 with quarks and mesons are analyzed . in particular , this scheme describes some important features of radiative decays of vector mesons . namely , the electromagnetic quanta mixing with @xmath7- and @xmath6- fields leads to the gauge variant of vector dominance , i.e. all these interactions are considered in the same way . so , we deal with the @xmath15 which incorporates gauge interactions of vector mesons with the constituent quarks . then , the model accounts for hadronic internal degrees of freedom , which are described in the quantum field approach . really , two - level ( quark - meson ) structure of the model realizes some analog of the `` bootstrap '' idea . the model contains a number of free parameters in the gauge sector , which can be fixed from the well measured two - particle decays of vector mesons . as it will be shown later , the gauge model predictions for the radiative vector meson decays are strict ( without any free parameters ) . as a rule , the radiative decays @xmath16 and @xmath17 were described in a phenomenological way . within the framework of the gauge model , these processes arise at the tree level . there is a good agreement between the theory and the experimental data on the vector meson radiative decays . so , these decays can be successfully studied in the gauge field approach . internal meson structure is described by the quark - meson sector of the model . these aspects of the model have been tested in @xmath18 , @xmath19 and @xmath20 decays and results of calculations are in a good agreement with the experimental data . analysis of the decay @xmath21 @xcite , which arises at the loop level due to the quark - meson interaction only , is more cumbersome . calculations done show that the three - pion decays of @xmath22 and @xmath23 mesons are well descripted in this model . the obtained results are in preparation now . the structure of the paper is as follows . in section 2 the gauge field model is described . numerical results for tree decays @xmath24 and @xmath25 are presented in section 3 . we consider loop decays @xmath26 , @xmath27 and @xmath20 in section 4 . the gauge model of the low - energy meson interactions is based on the conceptual assumption concerning the transfer of gauge principles from the fundamental quark level to the effective hadron one . just the same transfer was deduced in ref . @xcite for the case of quark - meson hierarchy levels . the applicability of the gauge approach to the baryon - meson interaction was demonstrated in refs . @xcite-@xcite . in this work , we show the validity of the gauge scheme for the case of meson - meson and quark - meson effective interactions . we describe @xmath7- and @xmath6- mesons together with @xmath14- field within the framework of the gauge approach . thus , the strong and electromagnetic interactions are united in the gauge scheme . the simplest variant of corresponding dynamical symmetry is based on @xmath28 group . as it must be , the global symmetry of the chiral model takes place here , and , in a general case , the total chiral group of symmetry should be localized . however , for the application of the gauge approach to the @xmath23 and @xmath22 strong and electromagnetic decays , it is reasonably to localize the diagonal sum of @xmath29 subgroups together with the additional @xmath10 groups . the traditional @xmath30- model is a part of the model involving nonlinear terms of self - action . the @xmath30-model symmetry can be realized in various representations of the global groups : @xmath31 , where @xmath9 means the diagonal sum of @xmath29 subgroups . for all these cases we have six independent parameters of transformations . transformation of the quark doublet is : @xmath32 where @xmath33 are the parameters of the @xmath34 group and @xmath35 are the parameters of the @xmath36 group . thus , here we use @xmath37 representation for @xmath38 and @xmath30 fields . then , the corresponding transformation properties fo @xmath39 are : @xmath40 having two sets of independent parameters @xmath41 and @xmath35 , to describe strong and electromagnetic interactions it is sufficient to localize the @xmath41 set together with the parameters of @xmath10 groups . ( to consider weak interactions the localization of the @xmath35 parameters should be added . ) as it is seen , the triplet of pions is the adjoined representation of the gauge group . at the same time , the quark doublet @xmath42 is the fundamental representation of this group . thus , we consider the triplet @xmath43 , the singlet @xmath30-meson and @xmath44-quarks as fields of matter . the physical scalar fields , emerging from the initial higgs multiplets , can be associated with the scalar mesons @xmath11 and @xmath12 . from these considerations the model lagrangian is : @xmath45 here @xmath46 - is the first generation quark doublet ; @xmath47 - two scalar fields doublets with hypercharges @xmath48 , @xmath49 and @xmath50 . the gauge derivatives and field strengthes are : @xmath51 in principle , the gauge invariant term @xmath52 can be added to the lagrangian . however , this term can be diagonalized from the very beginning by the orthogonal transformation of the initial @xmath53 and @xmath54 fields . so , we get the same lagrangian with the redefined @xmath53 and @xmath54 fields . to describe the realistic processes of decays and scattering we introduce also the interactions of leptons with @xmath55 and @xmath10 fields in the gauge form : @xmath56 where interaction of @xmath54 field with leptons is driven by the phenomenological coefficient @xmath57 . note that the origin of the last term in the above formula is specific . namely , this term is caused by the nondiagonal @xmath52 part of the initial lagrangian . thus , the interaction of leptons with the @xmath54 vector field occurs allowing to describe the decays @xmath58 ( see the section 4 ) . the introducing of the vector fields to the theory in a gauge way provides universality of coupling with vector fields , i.e. it strictly bounds the number of free parameters . so , this approach raises the predictability of the model . the analogous universality of vector and pseudoscalar meson interactions was analyzed in @xcite in a phenomenological way . physical states are formed by the primary fields mixing when mass quadratic forms of scalar and vector fields are diagonalized . at the tree level the mass forms arise as a result of vacuum shifts : @xmath59 note , the loop contributions to the mass matrix can play an essential role in the mixing of mesons . so , in a general case this matrix has the form : @xmath60 where @xmath61 is formed by the shifts and @xmath62 by the self - energy insertions . in our case these insertions are significant for the description of the @xmath63 mixing . as the result of the diagonalization of vector and scalar mass forms , we get the model spectrum of the vector and scalar multiplets . the gauge status of vector mesons @xmath64 and @xmath65 is confirmed by the calculation of their model mass spectrum and decay properties . these model parameters are in agreement with the experimental data for dominant decay channels of vector mesons @xcite . moreover , due to the presence of free parameters in the scalar sector , it is possible to describe the mass spectrum and decay channels of the scalar mesons @xmath30 ( or @xmath66 ) , @xmath11 and @xmath12 , which have the status of higgs scalars in the model ( see also eqs.([e:14 ] ) and comment ) . an emergence of primary hadron component in the photon is an important consequence of the mixing in the gauge sector . the structure of the vector boson physical states can be illustrated in a tree approximation ( when @xmath67 ) by the following expressions : @xmath68 where the mixing angle @xmath69 is determined under the diagonalization of the vector fields quadratic form . due to the mixing ( [ e:11 ] ) the processes with initial `` unphysical '' photon , @xmath70 , give contributions to the processes with the intermediate hadron states : @xmath71 some parameters of the mixing can be fixed from the experimental data on the vector meson decays . it will allows to describe some decays properties of vector mesons considered in this paper . the tree form of the mixing ( [ e:11 ] ) is caused by the diagonalization of the real mass matrix ( this matrix is generated by the vacuum shifts only , without an account of any self - energy insertions ) with the help of the real orthogonal matrix . together with the tree real gauge couplings , it leads to the absence of the relative phase in the @xmath72 and @xmath73 channels at the tree level ( see ( [ e:12a ] ) ) . experiments , however , indicate that there is a nonzero relative phase shift between amplitudes of @xmath74 and @xmath75 processes ( it is interpreted as @xmath76 mixing effect in the pion formfactor ) . in a phenomenological approach , the complexity is introduced into the superposition of pure isospin @xmath77 and @xmath78 states . so , the needed relative phase is a free parameter which is determined from the fit to the experiment @xcite . note , such important feature of vector mesons physics as the @xmath79 mixing , can be reproduced in the model at the loop level . as a rule , physical vector states of @xmath23 and @xmath22 mesons are described as linear combinations of pure isospin states with complex coefficients . it results to the relative @xmath80 mixing phase . in our model the complexity can not be directly introduced into the superposition of operator fields ( [ e:11 ] ) , because it describes the physical neutral fields . to reproduce the relative phase , we would treat the complexity in a following way . namely , instead of tree @xmath81 and @xmath82 we should have absolute values of the renormalized parameters of mixing in the superposition ( [ e:11 ] ) . then , the corresponding phase should be introduced into the renormalized couplings as the vertex factors ( see ( [ e:12a ] ) and ( [ e:12b ] ) ) . this treatment is equivalent to the known approach when physical meson states are the complex superpositions of pure isospin vector states @xcite . the necessary phase value can be provided in the framework of this model due to presence of free parameters , in particular , in the scalar sector ( @xmath83 couplings ) . certainly , these free parameters should be fixed simultaneously from the analysis of the @xmath84- decays . for the processes , which are under consideration in this paper , the relative @xmath85 phase is inessential , so we do not use it here . an analysis of the vector mesons mass matrix with an account of loop contributions and the study of the @xmath86 mixing in the model will be the subject of a forthcoming paper . now , we give the main part of the physical lagrangian which will be used for calculations : @xmath87 in eqs.([e:12a ] ) : @xmath88 where @xmath89 and @xmath90 are the complex ( renormalized ) parameters of the @xmath91 mixing with account of the self - energy insertion to the mass matrix . there are also another parts of the interaction lagrangian , which have the structures @xmath92 and @xmath93 . here @xmath94 and @xmath95 are the sets of scalar and vector physical fields . the used identification of the gauge and higgs fields components provides the invariance of the physical lagrangian under the @xmath96 transformations . some relations arise as a direct consequences of the model : @xmath97 the value of @xmath98 can be found from the second relation @xmath99 . values of @xmath100 and @xmath101 ( or @xmath102 ) are fixed from the experimental values of @xmath103 and @xmath104 . it should be noted , the last decay takes place due to the mixing ( [ e:11 ] ) only . thus , we extract the values of main model parameters which we use to describe another decay properties of mesons : @xmath105 these values were applied for calculation of the vector meson radiative decay widths to verify the gauge vector dominance approach . in our strategy of calculations the strong couplings are extracted from the above mentioned processes as the effective final values . so , we do not need in loop corrections to these couplings . at the same time , electromagnetic vertices should be renormalized by the strong interactions ( for details , see the decay @xmath20 which is considered in the fourth section ) . in addition to ( [ e:12a ] ) we give also a part of the lagrangian , describing an interaction of the scalar mesons with @xmath38-mesons : @xmath106 this part of the lagrangian contains free coupling constants and makes it possible to describe the dominant decay channels of scalar mesons - @xmath107 and @xmath108 @xcite . the decay channel @xmath109 is not observed and from the model coupling @xmath110 it follows that @xmath111 . so , in the scalar sector the residual global @xmath9- symmetry approximately takes place after the shift . exact symmetry , @xmath112 , is not possible because it forbids the decay channel @xmath113 , which is observable experimentally . evidently , the inequality @xmath114 , violating the global isotopic @xmath9 symmetry , is related with an account of the electromagnetic interaction in the model . there are no two - photon decays of scalar mesons at the tree level , i.e. these channels are suppressed in the frame of the model . thus , the status of scalars as the higgs mesons does not contradict to their observed mass spectra and decay properties . because of the model two - level structure corresponding to the meson - meson and quark - meson interactions , both of them are tested independently in two - particle and three - particle channels . it is seen from ( [ e:12a ] ) that the meson - meson lagrangian describes both tree and loop level processes , while the quark - meson interactions occur in the model at the loop level only . radiative decays of neutral vector mesons of the type @xmath115 are an object of steady attention during some decades ( see @xcite and also @xcite-@xcite ) . experimental investigation of these reactions and accompanied theoretical speculations contribute to the understanding of hadron intermediate states and low energy dynamic of meson interactions . in the channels where the charged pair @xmath116 is formed , dominant contribution comes from tree diagrams corresponding to the vector dominance approximation @xcite . radiative decays @xmath117 and @xmath118 are described by meson - meson sector of the gauge model . at the tree level , the former decay is represented by the diagrams in fig.1 . the total amplitude for the process is : @xmath119.\ ] ] here @xmath120 and @xmath121 are polarization vectors for @xmath122 - meson and photon , @xmath123 are 4-momenta for all particles in the process . in ( [ e:1 ] ) we omit all terms which are equal to zero on the mass shell in the transversal gauge . for comparison with the experimental spectrum of photons ( see the work of dolinsky , @xcite ) the differential width is presented in the form : @xmath124 where : @xmath125.\end{aligned}\ ] ] our numerical results , which follow from ( [ e:2 ] ) and ( [ e:3 ] ) , agree with the results of @xcite given by the vector dominance approach . in fig.2 the theoretical spectrum of photons in comparison with the experimental data ( from @xcite ) are represented . the curve ( 2 ) in fig.2 describes the spectrum @xmath126 , normalized by the total width . here @xmath127 , and it does not depend on the model coupling constants . the curve ( 1 ) represents the model fit of the experimental data with the help of a single free multiplicative parameter . as it is seen , this fitted curve ( 1 ) improves the prediction at the low energy range . this effect can be caused by the renormalization of @xmath128 coupling . to improve the theoretical spectrum near @xmath129 it was suggested to consider the loop corrections ( see @xcite and @xcite ) . however , this consideration is reasonable only if the more detailed and reliable experimental data on the photon spectrum are available . integration of ( [ e:2 ] ) from @xmath130 up to @xmath131 gives the value of partial @xmath6- meson branching @xmath132 which slightly exceeds the experimental value @xmath133 from @xcite . an account of the loop contributions ( with the phenomenological couplings ) leads to the result : @xmath134 @xcite . it should be noted that there is some discrepancy between the experimental branching and the one following from the integration of spectrum ( see fig.2 ) . namely , an excess of @xmath135 over @xmath136 can be caused by the deviation of the theoretical spectrum from the experimental one at the energy range @xmath137 ( to the left of the first point of the experimental data in fig.2 ) . decay characteristics of the process @xmath138 at the tree level are computed analogously with the following replacement in ( [ e:1 ] ) - ( [ e:3 ] ) : @xmath139 in @xmath140 , and @xmath141 in @xmath142 and @xmath143 . the partial width for the decay is damped by the small mixing parameter , @xmath144 , so we have @xmath145 and @xmath146 for @xmath147 and @xmath148 , respectively . these estimations do not contradict to the experimental restriction @xmath149 @xcite and agree with the theoretical results of @xcite . thus , the same set of the fixed parameters ( the gauge constant and two mixing tree angles ) describes two different decays at the tree level . certainly , loop corrections can be important for the case due to the smallness of the tree contribution , see , for example @xcite . in analogy with the results of these papers , loop corrections can increase @xmath150 up to @xmath151 , which does not contradict again to the upper limit of the experiment . as it was noted above , consideration of the loop corrections to the decay @xmath152 in @xcite-@xcite was intended to describe the photon spectrum fine structure . however , it increases the discrepancy between the model and experimental values of the total width . quantitatively , this effect depends on the underlying model . moreover , in effective theories an account of loop diagrams ( for some process arising at the tree level ) has some subtleties connected with the compensation of divergencies and renormalizability . for the processes which occur at the loop level only , these problems are absent all divergencies are summed to zero when all external lines are on the mass shell @xcite . an examples of such loop processes will be given in the next section . the usage of parameters from ( [ e:13 ] ) together with the constants @xmath153 and @xmath154 extracted from the widths of @xmath155 and @xmath156 decays , leads to the correct estimation of the cross section in the peaks vicinity , @xmath157 . as it was noted above , the tree approximation ( without an account of the @xmath76 mixing at the loop level ) does not allow us to describe the resonance curve in this region in details . a relative @xmath85 phase consideration is a very essential for the resonance curve description . in the most papers these decays are defined by the phenomenological vertices which are introduced at the tree level @xcite . in the model considered , radiative decays @xmath161 and three - particle decays @xmath162 occur via quark loops with the gauge vertices . one - loop diagrams for the decays @xmath163 are shown in fig.3 , where @xmath164 denote quark fields . total amplitude for the process @xmath165 has the following form @xmath166 here @xmath167 is the color factor , vertex constant @xmath168 ( @xmath169 is the constant of @xmath170-interaction , see ( [ e:12a ] ) ) and @xmath171 is the three - point passarino - veltman function @xcite . for the constituent quark mass we suppose @xmath172 . substituting the function @xmath173 into ( [ e:4 ] ) , we have : @xmath174 where @xmath175 and the decay constant @xmath176 . in ( [ e:5 ] ) it was used the goldberger - treiman relation @xmath177 ( see , for example , @xcite ) . it defines the coupling @xmath169 through the constituent quark mass . then the constituent quark mass value , which is taken as a free parameter , can be found from the widths fit . the decay @xmath178 is described by the diagrams in fig.3 with the corresponding replacement of the coupling constant . the expression for the width is @xmath179 follows from the isotopic structure of @xmath180 , @xmath181 and @xmath182 vertices ( [ e:12a])-([e:12b ] ) ( here we omit small terms proportional to @xmath183 and @xmath89 ) . both the widths are in a good agreement with the experimental data @xcite when the effective quark mass value is @xmath184 : @xmath185 thus , the fitting of two widths by one model parameter - mass of the constituent quark @xmath186 , gives @xmath187 . this quark mass value can be understood in a connection with the schematic representation of the quark - gluon content of hadrons . specifically , for the nucleon and @xmath6- meson masses we can write approximately @xmath188 here we use an effective quark mass , @xmath186 , and suppose the same gluon component of meson and baryon structure , @xmath189 . from this system of equations it follows @xmath190 and @xmath191 . the latest value nearly coincides with the quark mass assessment resulting from the decay channels analysis . note that a sufficiently low quark mass value @xmath192 together with the @xmath193 are followed from the composite - meson model with the four - quark interaction @xcite . so , the model discussed introduces the constituent quark which is differed from the traditional effective quark component of a hadron with @xmath194 ( see , however , @xcite ) . moreover , the value of @xmath189 makes it possible to interpret the meson @xmath195 as the ( mostly ) glueball state @xcite . it does not contradict to some known ideas on the gluon nature of @xmath196 meson ( see @xcite and references therein for the review of scalar meson properties ) . if it is the case , the low value of the constituent quark mass can be treated as the consequence of a model separation of the quark and gluon degrees of freedom in a hadron . hence , the value of @xmath186 , which is essentially lower than @xmath197 , provides the agreement of the model results with the experiment for the processes under consideration . of course , if we explanate the @xmath30-meson as the effective glueball state , we need in an detailed analysis of its properties and manifestations in the model . this studying will be presented in a forthcoming paper . note also that the @xmath196 meson was carefully analyzed , supposing the meson is a @xmath198 state , in @xcite , for instance . in an analogy with the process @xmath199 , the decays under consideration can contain an anomaly contribution caused by the @xmath200 vertex . this contribution is small here due to the smallness of the ratio @xmath201 , that is approximation @xmath202 takes place with a good accuracy . it is known that the same approximation is not admissible for the decay @xmath203 , where the anomaly contribution dominates . an analysis of this decay in our model is coincide with the one in @xcite , for example . namely , due to the dominant anomaly contribution , the decay amplitude is again @xmath204 i.e. it depends on the @xmath205 and does not contain the quark mass . the constant @xmath205 arises in the amplitude from pion - quark coupling which is @xmath206 , as it dictates by the goldberger - treiman relation . so , our value of the quark mass corresponds to the ( lower ) pion - quark coupling , @xmath169 , providing , however , the reasonable values for @xmath207 widths . note also that in the consideration of loop three - particle decays @xmath208 we will have in this model a non - negligible ratio @xmath209 as a consequence of this relatively low quark mass ( cf . with @xcite ) . nevertheless , an accurate calculations give for @xmath210 the values which are in agreement with the data ( these results will be discussed in the forthcoming work too ) . one of the most important features of the model is the implantation of electromagnetic ( em ) interaction into the gauge scheme of strong interactions . the mixing of the gauge fields leads to the effect of mutual renormalization of the interaction couplings . now we discuss the decay @xmath20 which is caused mainly by the em interaction . at the tree level this process is described by the vertex following from ( [ e:10 ] ) and the mixing ( [ e:11 ] ) . corresponding decay coupling is @xmath211 , so it gives @xmath212 , while the experimental decay rate is @xmath213 @xcite . here we show that this discrepancy can be eliminated due to renormalization of the decay coupling @xmath153 by the strong interactions . in the one - loop approximation there are two types of diagrams which give the main contribution to the renormalization ( see fig.4 ) . in this figure , solid lines correspond to @xmath214 and @xmath215 ( or all set of scalar particles , @xmath216 ) , and wavy lines correspond to @xmath14 and @xmath122 . all extra diagrams are damped by the small mixing parameters . as a result , we get full amplitude of the process : @xmath217,\ ] ] where @xmath218 and @xmath219 is the sum of the loop contributions . finally , the decay rate has the form : @xmath220 where @xmath221 and @xmath143 is the renormalization parameter . the function @xmath219 in the @xmath222 scheme is : @xmath223 here we use normalized quantities : @xmath224 in the ms scheme an additional term arises in ( [ e:18 ] ) : @xmath225 where @xmath226 . with the help of the expressions ( [ e:16])-([e:20 ] ) we have found that @xmath227 when the renormalization parameter @xmath228 in the @xmath222 scheme and @xmath229 in the ms scheme . so , strong interactions lead to the large renormalization of the em coupling and the physical result @xmath230 is obtained within the framework of @xmath222 scheme when the renormalization scale @xmath143 is in the range of typical meson masses . the value of @xmath231 depends on the masses of quarks and scalar mesons very weakly . the quark mass changing from @xmath232 to @xmath233 leads to the variation of @xmath143 lower than @xmath234 . variation of @xmath143 due to the increasing of scalar mass from @xmath235 up to @xmath236 is approximately @xmath237 . an analogous effect takes place for the decay @xmath238 , but in this case we have use free parameter ( @xmath57 from ( [ e:10 ] ) ) . the presence of this parameter makes it possible to get the accordance @xmath239 for the same values of @xmath143 . the linear sigma - model is the most popular approach in the description of low - energy hadron interaction . we have considered the gauge generalization of this model and include quark degrees of freedom explicitly . in the gauge scheme the vector meson dominance occurs in the tree - level processes . the quark - meson sector describes some loop - level processes , when quark structure of mesons plays a noticeable role . the model is applied to some radiative decays of mesons which are intensively discussed in the literature . decays @xmath24 and @xmath25 have been considered within the framework of the vector dominance gauge variant . the results are in good accordance with the experimental data . the radiative decays @xmath158 and @xmath240 take place at the loop level only , when the quark - meson interaction is important . for these cases we get results in agreement with the experiment too . decay channel @xmath20 can be described by the model when we take into account the electromagnetic decay constant renormalization by strong interactions . from the analysis we fulfilled , it follows that the radiative decay rates of vector mesons can be calculated with a good accuracy within the framework of the gauge model . quantum field approach to the vector dominance in meson interaction can be also supplemented by the quark - meson interaction which expands the linear sigma - model . 99 y. nambu and g. jona - lasinio , _ dynamical model of elementary particles based on an analogy with superconductivity . * 122 * ( 1961 ) 345 . y. nambu and g. jona - lasinio , _ dynamical model of elementary particles based on an analogy with superconductivity . ii , phys . rev . _ * 124 * , ( 1961 ) 246 . m. k. volkov , _ low - energy meson physics in the quark model of superconductivity type , sov . and nuclei . _ * 17 * ( 1986 ) 186 ; _ effective chiral lagrangians and the nambu - jona - lasinio model , sov . j. part . and nuclei . _ * 24 * ( 1993 ) 81 . m. k. volkov , a. e. radzhabov , _ the nambu - jona - lasinio model and its development , sov . j. phys - usp . _ * 49(6 ) * ( 2006 ) 551 . g. v. efimov , m. a. ivanov and v. e. lyubovitskiy , _ predictions for semileptonic decay rates of charmed baryons in the qurk confinement model , z.phys . c - particles and fields _ * 52 * ( 1991 ) 149 [ jinr preprint e2 - 90 - 24 , dubna , 1990 ] . n. i. karchev and a. a. slavnov , _ effective chiral lagrangians from @xmath241 quantum chromodynamics , teor . mat . fiz . ( ussr ) _ * 65(1985 ) 192 . t. goldman and r. w. haymaker , _ dynamically broken chiral symmetry with bag confinement phys d24 * ( 1981 ) 724 . d. espriu , e. de rafael and j. taron , _ the qcd effective action at long distances , nucl . * b345 * ( 1990 ) 22 . d. ebert and h. reinhardt , _ effective chiral hadron lagrangian with anomalies and skyrme terms from qurk flavour dynamics , nucl . phys . _ * b271 * ( 1986 ) 188 . hsieh , _ effective meson field theory from qcd _ [ hep - ph/0306143 ] . s. s. olean , _ effective lagrangian for baryons and baryon - meson interaction _ [ hep - ph/0310005 ] . s. l. adler , _ axial vector vertex in spinor electrodynamics , phys . * 177 * ( 1969 ) 2426 . j. s. bell and r. jackiw , _ a pcac puzzle : @xmath242 gamma gamma in the sigma model , nuovo cim . _ * 60 * ( 1969 ) 47 . j. d. walecka , _ a theory of highly condensed matter , ann . phys . _ * 83 * ( 1974 ) 491 . b. d. serot , _ a relativistic nuclear field theory with pi and rho mesons , phys . b _ * 86 * ( 1979 ) 146 . k. sailor , i. lovas , _ gauge - invariant qhd lagrangian , acta phys . hung . _ * 64 * ( 1988 ) 123 . r. raczka , _ a spontaneously broken quintet baryon model for strong interactions , czech . b _ * 37 * ( 1987 ) 421 . g. m. vereshkov , v. i. kuksa , _ u(1)su(3 ) gauge model of barion - meson interactions , yad . fiz . _ * 54 * ( 1991 ) 1700 . j. c. polkinghorne , _ renormalization of axial vector coupling , il nuovo cim . _ * 8 * ( 1958 ) 179 . j. schwinger , _ field theory commutators , phys . lett . _ * 3 * ( 1959 ) 296 . m. gell - mann , m. levy , _ the axial vector current in beta decay , il nuovo cim . _ * 16 * ( 1960 ) 705 . r. delbourgo and m. d. scadron , _ dynamical generation of the gauged @xmath9 linear sigma model , mod . lett . _ * a10 * ( 1995 ) 251 [ hep - ph/9910242 ] . a. bramon , riazuddin and m. d. scadron , _ double counting ambiguities in the linear sigma model , j. phys . _ * c24 * ( 1998 ) 1 [ hep - ph/9709274 ] . r. delbourgo , d. liu and m. d. scadron , _ meson pvv interactions are determined by quark loops , int . * a14 * ( 1999 ) 4331 ( 1999 ) [ hep - ph/9905501 ] . r. p. feynman , _ photon - hadron interactions , w. a. benjamin , inc . _ ( 1972 ) . a. bramon , a. grau and g. pancheri , _ intermediate vector meson contributions to @xmath243 decays , phys . _ * b283 * ( 1992 ) 416 ; _ chiral perturbation theory and radiative @xmath243 decays , phys.lett._ * b289 * ( 1992 ) 97 ; _ radiative vector - meson decays in @xmath244 broken effective chiral lagrangians , phys.lett._ * b344 * ( 1995 ) 240 . d. black , m. harada and j. schechter , _ vector meson dominance model for radiative decays involving light scalar mesons , phys . * 88 * ( 2002 ) 181603 [ hep - ph/0202069 ] . baboukhadia , v. elias , m.d . scadron , _ linear sigma model linkage with nonperturbative qcd , j. phys . _ * g23 * ( 1997 ) 1065 [ hep - ph/9708431 ] . m.d . scadron , _ quark level linear @xmath30 model ( l@xmath30 m ) via loop graphs _ [ hep - ph/0007184 ] . j. l. lucio m. and m. napsuciale , m. d. scadron , v. m. villanueva , _ the @xmath245 transition and @xmath246 decay , phys . * d61 * ( 2000 ) 034013 [ hep - ph/9902349 ] . p. singer , _ decay mode @xmath247 , phys . * 128 * ( 1962 ) 2789 ; _ radiative @xmath6-meson decay , phys . rev . _ * 130 * ( 1963 ) 2441 . m. gell - mann , d. sharp and w. g. wagner , _ decay rates of neutral mesons , phys . * 8 * ( 1962 ) 261 . e. a. kuraev , z. k. silagadze , _ once more about the @xmath248 contact term , phys . * 58 * ( 1995 ) 1589 [ hep - ph/9502406 ] . yao et al . , _ review of particle physics , particle data group j. phys . _ * g33 * ( 2006 ) 1 . f. klingl , n. kaiser , w. weise , _ effective lagrangian approach to vector mesons , their structure and decays@xmath249 , z.phys._ * a356 * ( 1996 ) 193 [ hep - ph/9607431 ] . r. urech , _ @xmath85 mixing in chiral perturbation theory , phys.lett._ * b355 * ( 1995 ) 308 [ hep - ph/9504238 ] . oconnell , b.c . pearce , a.w . thomas , a.g . williams , _ @xmath85 mixing , vector meson dominance and the pion form - factor , prog.nucl.part.phys._ * 39 * ( 1997 ) 201 [ hep - ph/9501251 ] . yan , j .- h . jiang , x .- j . wang , _ on the mixed propagator approach to @xmath76 mixing , commun.theor.phys._ * 38 * ( 2002 ) 195 [ hep - ph/0111281 ] . m. benayoun et.al . _ new results in @xmath122 meson physics , eur.phys.j._ * c2 * ( 1998 ) 269 [ hep - ph/9707509 ] . m. benayoun et.al . _ the pion form factor within the hidden local symmetry model , eur.phys.j._ * c29 * ( 2003 ) 397 [ nucl - th/0301037 ] . a. bramon , r. escribano , _ a consistent description of @xmath250 decays including @xmath251- meson effects , jhep _ * 0402 * ( 2004 ) 048 [ hep - ph/0305043 ] . a. gokalp , s. solmaz and o. yilmaz , _ scalar sigma meson effects in radiative @xmath122-meson decays , phys.rev._ * d67 * ( 2003 ) 073007 [ hep - ph/0302129 ] . y. oh and h. kim , _ higher meson resonances in @xmath252 and @xmath253 , phys * d68 * ( 2003 ) 094003 [ hep - ph/0307286 ] . s. i. dolinsky et al . , _ summary of experiments with the neutral detector at the @xmath254 storage ring vepp-2 m , phys . rep . _ * 202 * ( 1991 ) 99 . a. gokalp , a. kucukarslan and o. yilmaz , _ vmd , chiral loops , @xmath30-meson , and @xmath255 mixing in @xmath253 decay , phys . * d67 * ( 2003 ) 073008 [ hep - ph/0302240 ] . a. gokalp , a. kucukarslan , s. solmaz and o. yilmaz , _ sigma - meson and @xmath255 mixing effects in @xmath256 decay , acta . polon . _ * b34 * ( 2003 ) 4095 [ hep - ph/0306044 ] . v. kuksa , _ compensation of divergences in rare processes , yad . fiz . _ * 66 * ( 2003 ) 1785 . d. bardin and g. passarino , _ standard model in the making , oxford university press _ ( 1999 ) . d. ebert , m.k . volkov , _ composite - meson model with vector dominance based on @xmath257 invariant four - quark interactions , z.phys.c-particles and fields _ * 16 * ( 1983 ) 205 . k. hagiwara et al . , _ review of particle physics , particle data group , phys . * d66 * ( 2002 ) 1 . m.d . scadron , f. kleefeld , g. rupp , _ constituent and current quark masses at low chiral energies , eurphys . * 80 * ( 2007 ) 51001 [ hep - ph/0710.2273 ] . h. jin , x. zhang , _ scalar glueball decay into pions in effective theory , phys . _ * d66 * ( 2002 ) 057505 [ hep - ph/0208120 ] . anisovich , _ systematics of @xmath258- states , scalar mesons and glueball , aip . * 619 * ( 2002 ) 197 [ hep - ph/0110326 ] . scadron , _ scalar sigma meson via chiral and crossing dynamics , eur . phys . j. _ * c6 * ( 1999 ) 141 [ hep - ph/9710317 ] . e. van beveren , f. kleefeld , g. rupp , m.d . scadron , _ remarks on the @xmath259 scalar meson as the dynamically generated chiral partner of the pion , mod . phys.lett._ * a17 * ( 2002 ) 1673 [ hep - ph/0204139 ] .
many quantum phenomena , often radically different from our intuitive predictions , are attributed to the fundamental principle of quantum superposition that a physical system can be in a linear superposition of two distinct states . when the principle of quantum superposition is applied to multipartite physical systems , it manifests another interesting feature of quantum mechanics , namely quantum entanglement . quantum entanglement is nonclassical correlation between local systems and is now widely referred to as a key resource for quantum information processing . quantum phenomena are observed typically on microscopic scales . however , as illustrated in schrdinger s famous cat paradox @xcite , quantum mechanics does not preclude , in principle , the possibility of a macroscopic object being in a quantum superposition or being a part of quantum entanglement . natural questions then follow . if `` macroscopic _ and _ quantum '' is somehow a possible combination , how can we define , characterize and quantify `` macroscopic quantumness '' or `` quantum macroscopicity '' ? further , how and to what extent can we implement such macroscopic quantumness ? of course , scientists have tried to answer these questions . in the early days of quantum mechanics , such attempts were primarily in the area of interpretations or philosophical discussions , as they are not within the reach of experimental tests or implementations . since then , however , a remarkable development of quantum and atom optics has been brought about , which has paved a way to control and detect individual quantum systems at the level of single photons and atoms . based on this progress , further efforts are being made to collectively control larger quantum systems , that is also closely related to the ability to perform quantum information processing @xcite . now , we may say that a significant amount of efforts made in physics and optics research in the last few decades are more or less related to explorations of macroscopic quantumness . in this paper , we review two major research topics on macroscopic quantumness its quantifications and physical implementations using optical fields and attempt to make remarks on them . we want to know whether and/or how much a physical system is _ both _ macroscopic _ and _ quantum . needless to say , it is not sufficient for a state to be either macroscopic or quantum . it should be _ quantum - mechanically macroscopic _ or _ macroscopically quantum_. this point may sound too obvious to make , but it is nontrivial to technically define and quantify macroscopic quantumness . leggett posed a question along this line as `` what is the correct measure of ` schrdingers - cattiness ' ? '' and commented , `` ideally , one would like a quantitative measure which corresponds to our intuitive sense '' @xcite . a number of proposals have been made for quantification of macroscopic quantum superpositions @xcite based on the effective number of particles that involve the superposition @xcite , distinguishability between the constituent states @xcite and operational interpretations @xcite . in the context of this paper , we would like to pose three requirements for a desirable measure of macroscopic quantumness . first , it should be applicable to a wide range of states , not limited to a specific type of states . second , we prefer to have a measure that quantifies the degree of a genuine superposition against a classical mixture , together with its effective size factor . in other words , it should be applicable not only to pure states but also to mixed states . third , if a state is given , the degree should be unambiguously determined . these points are important because our motivation is to compare different types of states being considered as candidates for macroscopic quantum superpositions using a conclusive measure . in view of these points , we shall review in chronological order such measures proposed to quantify macroscopic quantumness . in 1980 , leggett in his pioneering work defined a measure called disconnectivity " @xmath0 @xcite that quantifies genuine multipartite quantum correlations . suppose that we are interested in characterizing an @xmath1-mode bosonic system @xmath2 . we first obtain a reduced density operator @xmath3 ( @xmath4 ) from @xmath5 by tracing out every mode except for @xmath6 arbitrarily chosen modes . quantity @xmath7 is introduced as @xmath8 where @xmath9 $ ] is von - neumann entropy of @xmath3 . by definition , @xmath7 is set to be @xmath10 when both the numerator and the denominator are zero , and @xmath11 . the disconnectivity @xmath0 is defined as the largest integer @xmath6 that makes @xmath7 the smallest . for an ideal greenberger - horne - zeilinger ( ghz ) state @xmath12 with an orthogonal basis @xmath13 , it is clear that @xmath14 and @xmath15 for @xmath16 . its disconnectivity is then @xmath17 that is the largest @xmath6 minimizing @xmath7 to be @xmath18 . however , @xmath0 is always @xmath10 for mixed states @xmath19 regardless of the values of @xmath1 and @xmath20 as far as @xmath21 . this means that macroscopic quantumness for partially mixed states can not be identified by @xmath0 . leggett pointed out that so called macroscopic quantum phenomena " such as superconductivity or superfluidity do not require the existence of a high-@xmath0 state . superfluidity can be explained by a product of identical bosonic states of which disconnectivity is obviously @xmath10 . a superconducting system described by @xmath1 cooper pairs , @xmath22 , also shows a small value of disconnectivity @xmath23 regardless of @xmath1 . on the other hand , multi - mode quantum correlations , in the form of pure states , always give large values of @xmath0 . even though disconnectivity sensibly quantifies multipartite correlations for pure states , it is not always sensitive to distinguishability between the constituent states . this point shall be clearer in the following example discussed by dr _ et al . _ @xcite . dr _ et al . _ @xcite considered the effective size of a generalized form of the greenberger - horne - zeilinger ( ghz ) state : @xmath24 where @xmath25 . the disconnectivity of this state is found to be @xmath17 regardless of @xmath26 even though @xmath26 obviously contributes to the distinguishability between @xmath27 and @xmath28 . it would be of particular interest to know whether @xmath29 can still be a macroscopic superposition in comparison to an ideal ghz state when @xmath26 takes a small value . to put it concretely , they attempt to find out what size of the ideal ghz state ( i.e. @xmath30 ) state @xmath29 is equivalent to . two different methods lead to the same result that the effective size of state @xmath29 as a macroscopic superposition approaches @xmath31 when @xmath32 . the first method is based on the rate of decoherence and the second is related to the ghz entanglement distillation by local operations and classical communication ( locc ) . consider a dephasing process that is described by a completely positive map @xmath33 where @xmath34 , @xmath35 is time , @xmath36 is the dephasing rate , and @xmath37 is the pauli-@xmath38 operator . they show that the trace - norm of the off - diagonal elements for state @xmath29 is @xmath39 when @xmath32 and that it is @xmath40 for state @xmath30 . based on this comparison , they conclude that the effective size of the generalized ghz state @xmath29 is equivalent to that of an ideal ghz state @xmath30 of size @xmath31 for small values of @xmath26 . this conclusion is also derived from depolarizing decoherence where it gives exactly the same decay rate . of course , this analysis is limited only to a very specific type of states in the form of eq . ( [ eq : gen - ghz ] ) . bjrk and mana s suggestion @xcite is based on their observation that a quantum superposition is more sensitive than its constituent states for interferometric applications . let us consider a pure state @xmath41 with a measurement outcome distribution of observable @xmath42 as @xmath43 where @xmath44 is a function of a reasonably smooth form centered at @xmath45 with width @xmath46 . the operator @xmath47 is applied to state @xmath48 where @xmath49 corresponds to the degree of the interaction time and strength . generally , the overlap between the original and evolved states is @xmath50 which becomes zero when @xmath51 . in other words , the original state @xmath41 evolves to a state that is orthogonal to the original one at a certain interaction time @xmath52 . meanwhile , if there was only a single peak @xmath44 ( @xmath53 or @xmath54 ) for the distribution of @xmath48 , the overlap simply would be @xmath55 regardless of @xmath56 and its first local minimum ( or a half of the initial value ) would be found at @xmath57 . the measure of a macroscopic superposition in terms of interferometric sensitivity is defined as the ratio of the two interaction times : @xmath58 bjrk and mana s approach has a distinguishing feature as an operational measure based on a physical application even though it is devised only for pure states . there is an ambiguity applying this measure to mixed states about how close the evolved state should be to the orthogonal state in determining @xmath59 . this may lead to ambiguity in comparing different types of states . shimizu and miyadera proposed index @xmath60 @xcite that determines whether a given form of @xmath1-mode state @xmath62 becomes macroscopically quantum as @xmath1 increases . the index @xmath60 is obtained as @xmath63 where @xmath64 is the variance for observable @xmath65 with state @xmath48 , @xmath1 is the number of modes , and @xmath66 if @xmath67 is a nonzero constant . the maximum in eq . ( [ eq : pindex ] ) is taken over all possible additive observables represented by @xmath68 where @xmath69 is a local observable for mode @xmath6 . the value of index @xmath60 is found to be @xmath70 for ghz - type entanglement while @xmath71 for a simple product form of state . a state is considered to be macroscopically quantum if its index is @xmath70 . the index @xmath60 , however , can not be applied to mixed states . for example , a mixed state @xmath72 have the same value of index @xmath70 as an ideal ghz state . shimizu and morimae generalized the measure for arbitrary mixed states @xcite . the index @xmath61 for state @xmath73 is @xmath74 where @xmath75 is an arbitrary projection operator satisfying @xmath76 and @xmath77 $ ] . the correlation can be represented as @xmath78 where @xmath79 is an eigenstate of @xmath65 with eigenvalue @xmath80 . it becomes @xmath81 so that @xmath82 when there is non - negligible @xmath83 coherence @xmath84 between the macroscopically distinct states of @xmath85 . the indices @xmath60 and @xmath61 are equivalent for pure state , i.e. , index @xmath60 is a special case of index @xmath61 . the index @xmath61 , as well as index @xmath60 , does not give a value for a given _ state _ ; rather it identifies what _ kinds _ ( or forms ) of states can scale to be macroscopically quantum when they become large multipartite states of @xmath86 . therefore , it can not be directly applied to a single - mode state such as @xmath87 where its macroscopicity depends on @xmath88 , nor is it a quantifier of macroscopic quantumness for a given state . cavalcanti and reid proposed an inequality of which violations verify macroscopic superpositions of continuous - variable states @xcite . suppose a generalized macroscopic superposition @xmath89 and a pointer - measurement @xmath90 giving macroscopically ranged outcomes @xmath91 . the domain for @xmath91 can be partitioned into three regions for @xmath92 , @xmath18 , and @xmath93 which correspond to @xmath94 , @xmath95 and @xmath96 , respectively . if the reference value @xmath97 is sufficiently large , two region @xmath92 and @xmath93 are called macroscopically distinct . in contrast to eq . ( [ eq : gs ] ) , a mixed state @xmath98 is not a macroscopic superposition in the sense that the outcomes of @xmath99 only spread for @xmath100 @xmath101 and those of @xmath102 for @xmath103 @xmath104 . in other words , state ( [ eq : rholr ] ) does not incorporate a macroscopic superposition since the coherence element is @xmath105 . the authors derived an inequality that should be satisfied by state ( [ eq : rholr ] ) : @xmath106 with @xmath107 where @xmath108 are the normalized probability distributions for regions @xmath109 , and @xmath110 and @xmath111 are their means and variances , respectively . cavalcanti and reid s inequality can be applied to a wide range of states compared to previous measures , i.e. , applicable to arbitrary single - mode continuous - variable states and may be extended to multi - mode continuous - variable states if appropriate measurements are defined . however , it does not provide a degree of macroscopic quantumness , but it works as a criterion of macroscopic quantumness for a given state with respect to an arbitrarily chosen scale @xmath97 . its calculations involve nontrivial numerical integrations and optimization processes . et al . _ experimentally demonstrated violations of this inequality using gaussian states and showed that it critically depends on purity of the states @xcite . korsbakken _ et al_. @xcite suggested an effective size of an @xmath1-particle superposition state in the form of @xmath112 based on how many measurements are required to distinguish @xmath113 and @xmath114 . the size measure for state @xmath62 is defined by @xmath115 for given probability @xmath116 @xmath117 of distinguishing the two constituent states by measuring @xmath118 number of particles . the discrimination probability is calculated as @xmath119+\tr[\rho_b^{(n)}e_b^{(n)}]\right)\ ] ] where @xmath120 are reduced density matrices for @xmath6 particles ( @xmath121 ) and @xmath122 are positive - operator valued measurements ( povms ) acting nontrivially only on @xmath6 particles . the maximum value of the discrimination probability is obtained when the povm is a projective measurement in the eigenbasis of @xmath123 @xmath124 obviously , the two constituent states of an ideal ghz state @xmath125 can be distinguished by a single particle @xmath37 measurement so that @xmath126 regardless of @xmath127 . korsbakken _ et al . _ applied this measure to superconducting flux qubits @xcite where a superposition of macroscopically distinct values of currents and magnetic moments is observed . they found the effective size of the flux qubits is surprisingly ( but not trivially ) small despite the apparent large difference in macroscopic observables . the reason is that only a small fraction of all electrons contribute to the superposition , while their speeds are high enough to yield large currents or magnetic moments . there is an ambiguity in using this measure due to the arbitrariness of the choice of @xmath127 . for example , the size of the superposition in the form of @xmath128 is found to be @xmath129 @xcite . therefore , for example , which of the two states , @xmath130 and @xmath131 , is more `` macroscopically quantum '' depends on the choice of @xmath127 . marquardt _ et al . _ define the size of quantum superposition by a number of particles that effectively involve the superposition @xcite . more precisely , for given constituent states @xmath113 and @xmath114 they count how many single - particle operations are required to convert @xmath113 into @xmath114 or vice versa . as a simple example , an @xmath1-product horizontal - polarization state @xmath132 can be converted into and a vertical - polarization state @xmath133 by acting @xmath134 , where @xmath135 and @xmath136 are creation and annihilation operators for corresponding modes of the subscripts . this operation is an @xmath1 number of single - particle operations , and the size of superposition @xmath137 is @xmath1 . in general , @xmath114 is obtained by superposing different states as @xmath138 , where @xmath139 is a state converted from @xmath113 by at least @xmath140 single particle operations . the average effective particle number is then @xmath141 . in ref . @xcite , the authors considered @xmath142 and @xmath143 for modes @xmath144 and @xmath145 . the constituent state @xmath114 is expanded as @xmath146 with coefficients @xmath147 @xcite . the state @xmath148 can be obtained by applying @xmath140 times of the single particle operation @xmath149 to @xmath150 implying that its effective number is @xmath140 . the average effective number is then @xmath151 and it becomes @xmath1 for an ideal ghz state of @xmath152 . it is obvious that this measure can not be applied to mixed states . in the phase space of mutually conjugate variables , the wigner function of a macroscopic quantum superposition typically shows an interference pattern with a high frequency . taking note of this point , lee and jeong @xcite defined a measure of macroscopic quantumness for an arbitrary harmonic - oscillator state . it simultaneously quantifies two different kinds of essential information : the degree of quantum coherence and the effective size of the physical system that involves the superposition . the basic idea is to take an integral @xmath154 , where @xmath155 $ ] is the characteristic function for state @xmath73 , in order to measure `` frequency '' and `` magnitude '' of the interference fringes at the same time in the wigner representation . in a continuous - variable phase space , it effectively quantifies _ both _ `` how widespread '' the constituent states of a given state are _ and _ `` how quantum mechanically pure '' coherences between those constituent states are , at the same time . the formal definition of the measure @xmath156 for an @xmath157-mode harmonic oscillator system @xcite is only slightly different : @xmath158 \left\| \chi \left ( \xi_1,\ , \xi_2,\ , \cdots,\ , \xi_m \right ) \right\|^2\ ] ] where @xmath159 is the @xmath157-mode characteristic function , @xmath160 . the value @xmath161 has been inserted to make any coherent states or their product states ( regardless of their amplitudes ) a reference with @xmath162 , but it may be removed to guarantee positivity of the measure @xcite . the maximum possible value of @xmath153 for an optical state is shown to be its average photon number @xcite . assuming the photon loss condition , @xmath163 with @xmath164=(decay rate @xmath165 time ) , it turns out that the measure is equivalent to the purity decay rate of the state @xmath166 where @xmath167 $ ] is the purity of state @xmath73 . this is consistent with the rapid decay of macroscopic superpositions , and the purity decay rate itself and eq . ( [ eq : ad ] ) may be an alternative definition of @xmath153 . it is possible to detect @xmath153 for optical states using overlap measurements without full tomography of quantum states @xcite . the measure @xmath153 is applicable to arbitrary harmonic oscillator systems including mixed and multi - mode states . it is decomposition - independent and straightforward to calculate for any states represented in the phase space , giving definite values for direct comparison between different types of states . frwis and dr suggested using the fisher information as a measure of macroscopic quantumness for spin systems @xcite . when estimating an unknown parameter , @xmath168 , caused by an unitary evolution @xmath169 with hamiltonian @xmath42 , the classical limit of the estimation uncertainty is determined by @xmath170 where @xmath1 is the system size . it is well known that this limit can be lowered down to @xmath171 using a quantum mechanically correlated probe state . more generally , the limit of the estimation uncertainty using probe state @xmath73 is given by the cramr - rao bound @xcite , @xmath172 , where @xmath173 is the quantum fisher information where @xmath174 and @xmath175 are eigenvalues and eigenvectors of @xmath73 . ] for a given additive hamiltonian @xmath42 . therefore , @xmath176 implies that state @xmath73 is capable of the parameter estimation over the classical limit . noting that this improvement is attributed to long - range quantum correlations in the state @xmath73 , fr@xmath177wis and d@xmath178r call it a genuine macroscopic quantum effect , not an accumulation of microscopic effects @xcite . they call a quantum state @xmath73 macroscopic if @xmath179 , where the maximization is taken over the entire set @xmath180 of additive operators . however , there are some cases where individual particles do not show long - range quantum correlation while local groups of them do @xcite . therefore , the aforementioned measure needs to be generalized to @xmath181 where @xmath182 is a set of extended additive operator @xmath183 with each @xmath184 locally acts on @xmath185 distinct groups of particles sized @xmath83 . as indices @xmath60 and @xmath61 , frwis and dr s approach in the forms of eq . ( [ eq : d2 ] ) can not be directly applied to a single - mode state . such a measure is defined as @xmath186 giving a definite number for a given state . this proposal is applicable to arbitrary spin systems . it was applied to cloned quantum states to find that they are macroscopically quantum @xcite . frwis and dr also compared @xcite several measures for spin systems @xcite and concluded that index @xmath60 and their fisher - information - base approach detect the most broad set of macroscopic quantum states among those . nimmrichter and hornberger @xcite call a quantum superposition of a mechanical system to be macroscopic , if its experimental demonstration allows one to rule out even a minimal modification of quantum mechanics . in order to specify such minimal modification , they consider an additional generator @xmath187 to the von neumann equation for @xmath1 particle density matrix @xmath5 as @xmath188 + \mathcal{l}_n ( \rho_n ) , \label{eq : vonneumann}\ ] ] where @xmath189\end{aligned}\ ] ] and @xmath190 is the weighted sum of the single particle operators @xmath191 $ ] giving position translation @xmath192 and momentum boost @xmath193 of the @xmath6-th particle of mass @xmath194 . here , @xmath195 , @xmath196 , and @xmath197 are the mass , the coherence time parameter , and the normalized distribution function for the reference particle , respectively . the reference particle is chosen to be an electron , and the distribution @xmath198 is taken to be a gaussian distribution with standard deviations @xmath199 and @xmath200 for position and momentum respectively . the role of the generator @xmath201 in eq . ( [ eq : vonneumann ] ) is to wipe out the coherence from the original distribution of @xmath5 in the position - momentum phase space . for an experimental demonstration of a quantum superposition in a mechanical system , it rules out a certain region of the modification parameter so that there is a lower bound of the time parameter @xmath196 . if such a lower bound is larger , the superposition becomes more macroscopically quantum . the lower bounds of @xmath196 for interference experiments with neutrons , electrons , bose - einstein condensates , and molecules are obtained @xcite , and then the measure of macroscopicity is defined as @xmath202 where @xmath203 is the greatest lower bound of the time parameter @xmath196 . state - of - the - art interferometers achieve macroscopicities of up to @xmath204 @xcite . it can be applied to any mechanical phenomena and successfully addressed various experiments @xcite . however , it may require more investigations to find out whether macroscopicity witnessed by this measure is in line with the idea of genuine macroscopic superpositions , for example , in the context of refs . @xcite . sekatski _ et al . _ s measure for optical states is determined by how fuzzy ( or how classical ) a single - shot photon number measurement can be to distinguish two constituent states of a superposition @xcite . a pointer system is initially assumed to be in a gaussian position distribution @xmath205 with variance @xmath206 . when a quantum state @xmath207 interacts with the pointer , the final distribution of the pointer becomes @xcite @xmath208.\ ] ] for instance , if @xmath207 is an @xmath6 photon number state @xmath209 , the resultant distribution is exactly shifted by @xmath210 from the original one . when @xmath211 become larger , the resolution of the detector degrades and the detector is considered to be more `` classical . '' the probability of a correct discrimination between two constituent states , @xmath113 and @xmath114 , is @xmath212=\frac{1}{2}\left(1+d\left[p_a^\sigma(x),p_b^\sigma(x)\right]\right),\ ] ] where @xmath213=2^{-1}\int dx \vert p_a^\sigma(x)-p_b^\sigma(x)\vert$ ] is the trace distance between the outcome distributions @xmath214 and @xmath215 . the size of a superposition @xmath216 is then defined by the maximum tolerable @xmath211 for @xmath217 $ ] to reach a certain reference value @xmath218 . a straightforward example discussed in ref . @xcite is a superposition of the vacuum and a coherent state @xmath219 ( unnormalized ) . the measure is obtained as @xmath220 for large @xmath221 , which is proportional to the average photon number @xmath222 for a fixed @xmath218 . a more controversial example is an entangled state generated by applying the displacement operation @xmath223 on single - photon entanglement @xcite @xmath224 this state can be rewritten as @xmath225 with @xmath226 and @xmath227 can be considered as the constituent states of entanglement . the measure is then found to be @xmath228 which is proportional to @xmath221 indicating that such a state is macroscopically quantum . there arises a nontrivial question . the coherent states are considered to be the most classical states among all pure states @xcite . a coherent state @xmath229 can be rewritten as @xmath230 , and @xmath231 and @xmath232 are macroscopically distinguishable according to this approach . we then have to conclude that a coherent state , as far as its amplitude is large enough , is a macroscopic superposition , which would probably be unacceptable to many researchers in the community . it could also be controversial whether merely a local displacement operation can generate genuine macroscopic quantumness . various attempts have been made for generating and detecting macroscopic / mesoscopic quantum superpositions using atomic and molecular systems @xcite , optical setups @xcite , superconducting circuits @xcite , and other mechanical systems @xcite . quantum optics has provided a testbed for such experimental implementations in free - traveling light fields . for example , there has been remarkable progress in generating superpositions of coherent states ( scss ) and entangled coherent states ( ecss ) . another example is entanglement between macroscopic and microscopic systems in several different forms @xcite in line with schrdinger s original paradox @xcite . there are yet other examples for generating macroscopic quantum states such as squeezed macro - macro entanglement @xcite , ghz - type entanglement @xcite and noon states @xcite that are beyond the scope of this paper . in this section , we review several of such states and their generation schemes that have been suggested and experimentally performed . we also discuss their degrees as macroscopic quantum superpositions or entanglement based on some of the measures discussed in the previous section . let us first consider an scs in the form of @xmath233 where @xmath234 are coherent states of amplitudes @xmath235 , @xmath236 is a real relative phase factor , and @xmath237 is the normalization factor . the scss are often referred to as `` schrdinger cat states , '' provided that @xmath221 is reasonably large , probably due to the following two reasons . first , the coherent states are known as classical states due to several reasons @xcite . a classic criterion of nonclassicality for a quantum state is whether its @xmath238-function @xcite is well defined @xcite . the coherent states are the only kind of pure states that have well - defined @xmath238-functions @xcite . they are also robust against decoherence as `` pointer states '' @xcite and the closest analogy of classical point particles in the quantum phase space @xcite . second , the two coherent states , @xmath234 , are macroscopically distinguishable when @xmath221 is sufficiently large ; they can be efficiently discriminated using homodyne detection with a limited efficiency . therefore , an scs generates quantum interferences between these two `` classical '' but `` macroscopically distinct '' states as an analogy of schrdinger s paradox . _ studied nonclassical properties of scss such as sub - poissonian and oscillatory photon statistics @xcite . one may also consider entangled coherent states ( ecss ) and one of the simplest forms of such states is @xmath239 with normalization factor @xmath240 . several measures of macroscopic quantum superpositions confirm that scss and ecss are clearly macroscopically quantum . the values of measure @xmath153 for an scs and an ecs are found to be @xmath241 and @xmath242 for @xmath243 . in the case of @xmath244 , they are @xmath245 and @xmath246 . as another evidence , we numerically obtain the maximum value of @xmath97 for an scs to violate cavalcanti and reid s inequality ( [ eq : cavalcanti ] ) @xcite using homodyne detection is @xmath247 which corresponds to the distance between the two peaks of the superposition in the phase space . yurke and stoler found that a coherent state in a kerr nonlinear medium evolves to an scs in the form of @xmath248 after a certain interaction time @xcite , based on the calculations for an anharmonic oscillator coupled to a zero - temperature heat bath by milburn and holmes @xcite . they explicitly referred it to as a `` quantum mechanical superposition of macroscopically distinguishable states '' @xcite , quoting einstein and schdinger @xcite . mecozzi and tombesi showed that a type of two - mode ecss can be generated using a two - mode nonlinear interaction @xcite . sanders proposed a scheme to generate ecss using a mach - zehnder interferometer with a single - mode kerr nonlinearity @xcite and later provided a comprehensive review on this kind of states @xcite . gerry suggested using a cross - kerr nonlinearity in a mach - zehnder inteferometer setup with two photodetectors to generate scss and ecss @xcite . similar schemes using cross - kerr effects and beam - splitter interferences were independently developed by dariano _ _ @xcite and howell and yeazell @xcite . paris devised a two - step scheme that generates a small scs using a kerr nonlinear effect and then amplifies it using @xmath249 nonlinearity @xcite . studied a method to generate scss and ecss via cross phase modulation in a double electromagnetically induced transparency ( eit ) regime @xcite . however , all these approaches have remained as theoretical proposals yet . the required levels of nonlinearities are extremely demanding even though there are developing techniques such as eit to obtain giant kerr effects @xcite . there have been trials to reduce the required levels of nonlinearities . _ found that it is possible to use a relatively weak single - mode kerr nonlinearity together with a conditioning homodyne measurement to generate scss @xcite . the idea of using a weak cross - kerr interaction has been discussed in the context of the development of the bell - state detection @xcite and quantum computation @xcite . it was explicitly shown @xcite that a cross - kerr nonlinearity can be used to generate scss and ecss and this approach can overcome decoherence due to photon losses . there exist nontrivial problems to overcome even at a theoretical level in order to utilize the approach based on a weak cross - kerr nonlinearity @xcite . recently , he _ _ showed that approximately ideal cross - kerr effects can be obtained using relatively weak interactions between photon pulses in atomic ensembles @xcite . non - deterministic methods based on linear optics and conditioning measurements have been investigated @xcite since initial attempts by several authors @xcite . _ showed that an scs with a small amplitude as @xmath250 can be well approximated ( @xmath251 ) by applying the squeezing operation on a single photon @xcite . it can be achieved by subtracting a single photon from a squeezed state and a number of experiments have been performed along this line @xcite . however , this type of experiment can not generate two separate peaks in the phase space with interferences between them because the resulting state is simply a squeezed form of the single photon . et al . _ attempted a different method : a number state is divided by a 50:50 beam splitter and conditioned by a homodyne detection to generate an scs @xcite . the experiment with a number state of @xmath252 has resulted in an approximate scs with @xmath253 with interferences between two clearly separate peaks in the phase space @xcite . it can be used to generate arbitrarily large scss if large number states are available @xcite . glancy and de vasconcelos provided a comprehensive review on various methods for producing scss in optical systems in 2008 @xcite . _ showed that multiple photon subtractions result in large squeezed scss with extremely high fidelities @xcite . in this way , demonstrations of subtracting two or more photons on a squeezed vacuum state have been performed to generate scss @xcite . -photon subtraction on a squeezed vacuum @xmath254 using a beam splitter ( bs ) with low reflectivity ( @xmath255 ) and a photodetector . ( b ) scs generation scheme using homodyne detection @xcite . homodyne conditioning for measurements outcomes @xmath256 using a 50:50 beam splitter ( bs ) and homodyne detection @xmath257 on a photon number state @xmath258 . the additional ( optional ) squeezed vacuum @xmath259 may be used to control the degree of squeezing @xmath260 of the the output scs . ( c ) the scs amplification scheme @xcite . two small scss with amplitudes @xmath221 may be used to conditionally generate an scs of a larger amplitude @xmath261 using two 50:50 beam splitters ( bs1 and bs2 ) , an auxiliary coherent state @xmath262 and two _ imperfect _ on / off photodetectors . ] many of the non - deterministic yet feasible schemes are categorized into two major approaches . one is based on the squeezing operation and photon subtractions @xcite and the other is based on the number state generation and homodyne detection @xcite . figures [ fig - scs](a ) and 1(b ) present schematics of the two approaches . in fact , either the multiple photon subtraction from a squeezed vacuum @xcite or homodyning on the one part of the number state divided by a beam splitter @xcite yields a _ squeezed _ scs , i.e. @xmath263 where @xmath264 is the squeezing operator with the squeezing parameter @xmath260 as depicted in fig . [ fig - scs ] . in principle , as far as one could perform the photon number subtraction of many photons @xcite or generate a large number state @xcite , a squeezed scs of a large amplitude and high fidelity can be obtained . in order to generate a normal scs ( without squeezing ) with high fidelity @xmath265 , in addition to the scheme in ref . @xcite , one may use an additional squeezed vacuum as shown in fig . [ fig - scs](b ) @xcite . a generation scheme for arbitrary scss with unbalanced ratios was suggested @xcite and experimentally demonstrated @xcite . it is worth noting that large scss may be obtained in a non - deterministic way out of small scss using realistic on / off detectors and an auxiliary coherent state as shown in fig . [ fig - scs](c ) @xcite , and there exists an alternative amplification scheme using homodyne detection @xcite . in general , the scss and ecss are sensitive to a lossy environment and detection inefficiency as their amplitudes become large @xcite . theoretical attempts were made using the squeezing operation to make the scss more robust against losses @xcite or to amplify them @xcite . schemes for entanglement purification for mixed ecss @xcite and concentration for pure ecss @xcite were investigated using bell - state measurements @xcite . et al_. devised and experimentally demonstrated a method to generate an ecs in a remote way using two scss and two photodetectors @xcite . et al_. suggested another scheme for the same purpose that is made to be more robust against detection inefficiency using an auxiliary coherent state @xcite . proposals to distribute ecss using the quantum repeater protocol were also suggested @xcite . superpositions and entanglement of multiple numbers of coherent states have been theoretically studied @xcite . experimentally demonstrated quantum teleportation of a single - photon - subtracted squeezed vacuum ( i.e. , an approximate scs ) using the continuous variable teleportation protocol @xcite . generalization of scss and ecss to highly mixed forms has been studied for testing quantum theory in an even more `` classical '' limit @xcite . the wide scope of applications using scss and ecss includes quantum teleportation @xcite , quantum computation @xcite , precision measurements @xcite , bell - type inequality tests @xcite , leggett - type inequality tests @xcite , quantum contextuality tests @xcite and quantum steering @xcite . in particular , quantum teleportation and computation schemes using scss and ecss are considered as a strong candidate for the optical implementation of quantum computation in terms of the resource requirement and loss tolerance @xcite . a proof - of - principle experiment of this type of teleportation scheme was performed by neergaard - nielsen _ et al_. @xcite . there have been attempts to generate entanglement between microscopic and macroscopic ( or quantum and classical ) states of light by using several different methods @xcite . implementations of such micro - macro entanglement are of special interest in relation to schrdinger s cat paradox where a cat as a macroscopic classical object and an atom as a microscopic quantum system are entangled @xcite . we limit our discussions to several examples using free - traveling optical systems mainly from a theoretical point of view , while there are other examples such as atom - field entanglement @xcite . de martini _ et al_. suggested @xcite and experimentally demonstrated @xcite a scheme depicted in fig . [ fig - micro - macro]a to amplify a local part of microscopic entanglement using quantum injected optical parametric amplification ( qiopa ) @xcite @xcite . the microscopic entanglement is supposed to be a polarization entangled state , @xmath266 , where @xmath267 ( @xmath268 ) is the left ( right ) circularly polarization state for mode @xmath269 ( @xmath270 ) . one of the modes , say @xmath270 , is amplified to be `` macroscopic '' using qiopa @xcite . the states @xmath271 and @xmath272 then evolve to two orthogonal states with different photon number distributions as @xmath273 where @xmath274 ( @xmath275 ) is the photon number state with @xmath276 photons and the right ( left ) circularly polarization , and @xmath277 are real coefficients ] with the amplification constant @xmath278 . the amplification results in entanglement between microscopic and macroscopic states : @xmath279 it is straightforward to verify that state @xmath280 has the maximum value of measure @xmath153 @xcite as @xmath281 that corresponds to its average photon number . this result is implied in the fact that the qiopa causes @xmath282 and @xmath283 to be macroscopically quantum as seen in eq . ( [ eq : process ] ) . in fact , @xmath284 and @xmath285 themselves have the maximum values of @xmath153 . however , it is a separate question whether @xmath284 and @xmath285 are truly macroscopically distinct in this type of entanglement . in ref . @xcite , the two states @xmath284 and @xmath285 are considered to be macroscopically distinct in the sense that their average photon numbers for the @xmath286-polarization are @xmath287 and @xmath288 , respectively . however , as we discussed in sec . [ dcn ] , this type of approach leads to a problem that becomes more noticeable when it is applied to a single - mode superposition . once again , the example of a strong coherent state , @xmath289 with @xmath244 , may be considered for a comparison . it can be simply shown that the difference between the average photon numbers of constituent states @xmath231 and @xmath232 is @xmath290 , i.e. , it monotonically increases with @xmath221 although a coherent state would not be regarded as a macroscopic superposition . in spite of this ambiguity , a strong nonclassical feature of state @xmath280 seems evident based on the high value of @xmath153 . a perfect discrimination of @xmath284 and @xmath285 by a single - shot measurement requires a photon - number resolving detector for a parity measurement . in the real experiment @xcite , an alternative procedure called the `` orthogonality filter '' was used where @xmath284 and @xmath285 are discriminated by field intensity measures for each polarization in a non - deterministic and approximate way . there have been considerable debates and discussions @xcite over whether the state generated in ref . @xcite possesses genuine entanglement and how to verify this type of entanglement more clearly . sekatski _ et al . _ proposed a scheme that uses the displacement operation for an amplification process to generate macroscopic entanglement @xcite . the state is generated by locally displacing each mode of single - photon entanglement as @xmath291 . they claim @xmath292 and @xmath293 are macroscopically distinguishable since they exhibit 3 times different photon number variances @xcite . _ @xcite and lvovsky _ et al . _ @xcite experimentally realized a simpler variant @xmath294 in eq . ( [ eq : gisinstate ] ) , where only one of the modes is displaced for generating micro - macro entanglement ( see fig . [ fig - micro - macro]b ) . in refs . @xcite , it was shown that a single shot photon number measurement can distinguish between @xmath292 and @xmath293 in the macroscopic part ( @xmath270 ) of state ( [ eq : gisinstate ] ) with 74% probability for @xmath244 . it can be improved up to 90% by changing the basis of the initial microscopic state into @xmath295 . however , we already discussed in secs . [ dcn ] and [ sec : amp - mm ] that this type of approach does not seem very convincing . the value of measure @xmath153 for @xmath296 , as well as @xmath297 , is exactly the same to that of microscopic single - photon entanglement @xmath298 . in fact , measure @xmath153 is invariant under the displacement operation that does not change the structure of phase space distributions of any states @xcite . the same principle is applied to cavalcanti and reid s inequality @xcite so that neither @xmath296 nor @xmath297 is found macroscopically quantum . in the experiments @xcite , entanglement was detected after applying the `` reverse '' displacement operation , @xmath299 , to state ( [ eq : gisinstate ] ) . _ observed entanglement with more than 500 photons using photodetectors @xcite . _ achieved the displacement operations with @xmath300 and entanglement was observed using homodyne tomography @xcite . ghobadi _ et al . _ @xcite investigated an alternative proposal to create micro - macro entanglement by applying the squeezing operation @xmath264 to one mode of single - photon entanglement as @xmath301 . it was shown that two states @xmath302 and @xmath303 can be discriminated by the mean photon number contained in each state which are 3 times different for large enough squeezing . the squeezing operation changes the structure of the phase space distribution and the degree of macroscopic quantumness @xmath304 increases as @xmath260 becomes larger while the tricky issue explained above on macroscopic distinctness between @xmath302 and @xmath303 remains . this state can be categorized into the same type of state with the one in eq . ( [ martinistate ] ) . recently , hybrid entanglement between particle - like ( or quantum ) and wave - like ( or classical ) states was experimentally demonstrated by jeong _ _ @xcite and morin _ a form of such state is @xmath305 it takes the maximum value of @xmath153 for @xmath244 as @xmath306 . s scheme uses an idea of superposing two distinct quantum operations . initially , a coherent state @xmath307 and the vacuum are prepared with two photon addition devices @xcite and a beam splitter with appropriate ratio as shown in fig . [ fig - micro - macro](c ) . when only of the two detectors d1 and d2 clicks , one does not know which mode the single photon was added to . the resulting state is then superposition as @xmath308 ( unnormalized ) that can be made to be an approximate state of @xmath309 by applying an appropriate displacement operation . however , hybrid entanglement with only small values of @xmath221 can be obtained in this way . experimentally , a fidelity of @xmath310 and entanglement of negativity @xmath311 were shown with @xmath312 @xcite . more sophisticated methods such as the tele - amplification method @xcite need be used to obtain large values of @xmath221 for such hybrid states @xcite . et al . _ s experiment @xcite is based on a similar type of idea but using the photon subtraction with a beam splitter and a photodetector , a two - mode squeezed state with a very low gain , and a single - mode squeezed state in order to generate a slightly different type of target state such as @xmath313 . entanglement of negativity @xmath314 and fidelity @xmath315 were observed with amplitude @xmath316 @xcite . this type of idea is also found in andersen and neergaard - nielsen s previous proposal where a single photon , a single - mode squeezed state and a detector with a beam splitter for the photon subtraction are required @xcite . the target states of all these schemes @xcite show strong properties as macroscopic quantum states , i.e. , the maximum values of @xmath153 for @xmath244 . kreis and van loock investigated how to classify and quantify various types of hybrid entanglement between discrete and continuous variable states @xcite . entanglement between microscopic and macroscopic states is closely related to schrdinger s gedankenexperiment @xcite . micro - macro ( or hybrid - type ) entangled states are useful for loophole - free bell - type inequality tests @xcite , quantum information processing @xcite , and exploring quantum gravity @xcite . in particular , optical hybrid states benefit from both their discrete- and continuous - variable features in such a way that they are useful resources for quantum teleportation @xcite , quantum computation @xcite , and quantum key distribution @xcite . sheng _ et al . _ proposed an entanglement purification scheme for hybrid entanglement using linear optics elements and photon number measurements @xcite . we have reviewed and discussed proposed measures of macroscopic quantumness together with several attempts for implementing macroscopic quantum states using optical fields . in line with our discussions , three @xcite among the discussed measures seem to be a little more general than the others in the sense that they are applicable to various kinds of quantum states and serve as unambiguous quantifiers . those measures are particularly suitable for light fields @xcite , spin systems @xcite , and mechanical systems @xcite , respectively . it seems that the former two @xcite stick to the idea of genuine macroscopic quantum superpositions @xcite , i.e. , superpositions of macroscopically distinct states or genuine macroscopic quantum effects over mere accumulation of microscopic quantum effects . on the other hand , the latter @xcite seems to approach macroscopic quantumness in a broader way . it would be interesting to find out relations between measures for different physical systems from an inclusive point of view . very recently , frwis _ et al . _ investigated macroscopic quantumness of several optical states using measures for spin systems @xcite . they mapped a photonic state into a spin state using an ideal interaction model and found that some different measures show strong mathematical connections and give a similar classification of macroscopic quantum states . there have been various attempts to generate macroscopic quantum states using light fields . the arbitrariness of the decomposition for a quantum state leads to a question of whether certain quantum states @xcite are truly macroscopically quantum , at a theoretical level , even though they seem so with certain criteria @xcite . we exemplified the case where even a coherent state , @xmath229 , is categorized into a macroscopic superposition under a certain decomposition using this type of approach @xcite . perhaps , this suggests that more stringent considerations are necessary for classifications of macroscopic quantum states . this work was supported by the national research foundation of korea ( nrf ) grant funded by the korea government ( msip ) ( no . 2010 - 0018295 ) and the center for theoretical physics at seoul national university . iskhakov , i. n. agafonov , m. v. chekhova , g. o. rytikov , g. leuchs , phys . rev . a 84 ( 2011 ) 045804 . iskhakov , m. v. chekhova , g. o. rytikov , g. leuchs , phys . 106 ( 2011 ) 113602 . iskhakov , i. n. agafonov , m. v. chekhova , g. leuchs , phys . ( 2012 ) 150502 . b. kanseri , t. sh . iskhakov , g. rytikov , m. v. chekhova , g. leuchs , phys . a 87 ( 2013 ) 032110 . b. sanders , phys . a 40 ( 1989 ) 2417 . a. n. boto , p. kok , d. s. abrams , s. l. braunstein , c. p. williams , j. p. dowling , phys . rev . lett . 85 ( 2000 ) 2733 . p. kok , h. lee , j. p. dowling , phys . a 65 ( 2002 ) 052104 . i. afek , o. ambar , y. silberberg , science 328 ( 2010 ) 879 - 881 . s. olivares , m. g. a. paris , j. opt . b : quantum semiclass . opt . 7 ( 2005 ) s616 . m. s. kim , e. park , p. l. knight , h. jeong , phys . a 71 ( 2005 ) 043805 . s. olivares , m. g. a. paris , laser physics 16 ( 2006 ) 1533 . k. s. lee , m. s. kim , s .- d . lee , v. buzk , j. kor . ( 1993 ) 197 . x. wang , phys . a 64 ( 2001 ) 022303 . s. j. van enk , phys . 91 ( 2003 ) 017902 . s. j. van enk , phys . a 72 ( 2005 ) 022308 . n. lee , h. benichi , y. takeno , s. takeda , j. webb , e. huntington , a. furusawa , science 332 ( 2011 ) 330 . h. jeong , t. c. ralph , phys . 97 ( 2006 ) 100401 . h. jeong , t. c. ralph , phys . a 76 ( 2007 ) 042103 . s. j. van enk , o. hirota , phys . a 64 ( 2001 ) 022313 . h. jeong , m. s. kim , j. lee , phys . a 64 ( 2001 ) 052308 . h. jeong , m. s. kim , phys . a 65 ( 2002 ) 042305 . t. c. ralph , a. gilchrist , g. j. milburn , w. j. munro , s. glancy , phys . a 68 ( 2003 ) 042319 . a. p. lund , t. c. ralph , h. l. haselgrove , phys . ( 2008 ) 030503 . p. marek , j. fiurek , phys . a. 82 ( 2010 ) 014304 . a. p. lund , t. c. ralph , h. l. haselgrove , new j. phys . 13 ( 2011 ) 115015 . j. kim , j. lee , s .- w . ji , h. nha , p. m. anisimov , j. p. dowling , arxiv:1012.5872 ( 2010 ) . c. c. gerry , r. a. campos , phys . a 64 ( 2001 ) 063814 . c. c. gerry , a. benmoussa , r. a. campos , phys . a 66 ( 2002 ) 013804 . r. a. campos , c. c. gerry , a. benmoussa , phys . a 68 ( 2003 ) 023810 . c. r. myers , t. c. ralph , phys . a 66 ( 2002 ) 023819 . j. joo , w. j. munro , t. p. spiller , phys . ( 2011 ) 083601 . o. hirota , k. kato , d. murakami , arxiv:1108.1517 ( 2011 ) . j. joo , k. park , h. jeong , w. j. munro , k. nemoto , t. p. spiller , phys . rev . a 86 ( 2012 ) 043828 . a. mann , b. c. sanders , w. j. munro , phys . rev . a 51 ( 1995 ) 989 . r. filip , j. ehck , m. duek , j. opt . b : quantum semiclass . opt . 3 ( 2001 ) 341 . d. wilson , h. jeong , m. s. kim , j. mod . 49 ( 2002 ) 851 . h. jeong , w. son , m. s. kim , d. ahn , c. brukner , phys . rev . a 67 ( 2003 ) 012106 . j. wenger , m. hafezi , f. grosshans , r. tualle - brouri , p. grangier , phys . a 67 ( 2003 ) 012105 . m. stobiska , h. jeong , t. c. ralph , phys . rev . a 75 ( 2007 ) 052105 . h. jeong , phys . a 78 ( 2008 ) 042101 . lee , h. jeong phys . a * 80 * , 052105 ( 2009 ) . c. c. gerry , j. mimih , a. benmoussa , phys . rev . a 80 ( 2009 ) 022111 . h. jeong , m. paternostro , t. c. ralph , phys . 102 ( 2009 ) 060403 . y. lim , m. paternostro , j. lee , m. kang , h. jeong , phys . rev . a 85 ( 2012 ) 062112 . b. t. kirby , j. d. franson , phys . rev . a 87 ( 2013 ) 053822 . b. t. kirby , j. d. franson , phys . rev . a 89 ( 2014 ) 033861 . m. paternostro , h. jeong , phys . a 81 ( 2010 ) 032115 . lee , m. paternostro , h. jeong , phys . a 83 ( 2011 ) 022102 . g. mckeown , m. g. a. paris , m. paternostro , phys . a 83 ( 2011 ) 062105 . lee , s .- w . ji , h. nha , j. opt . b 30 ( 2013 ) 2483 . t. c. ralph , g. j. pryde , prog . opt . 54 ( 2009 ) 209 . j. s. neergaard - nielsen , y. eto , c .- w . lee , h. jeong , m. sasaki , nat . photon . 7 ( 2013 ) 439 . f. de martini , phys . 81 ( 1998 ) 2842 . f. de martini , phys . lett . a 250 ( 1998 ) 15 . p. sekatski , b. sanguinetti , e. pomarico , n. gisin , c. simon , phys . a 82 ( 2010 ) 053814 . r. ghobadi , a. lvovsky , c. simon , phys . ( 2013 ) 170406 . n. bruno , a. martin , p. sekatski , n. sangouard , r. t. thew , n. gisin , nat . ( 2013 ) 545 . a. i. lvovsky , r. ghobadi , a. chandra , a. s. prasad , c. simon , nat . ( 2013 ) 541 . s. raeisi , p. sekatski , c. simon , phys . ( 2011 ) 250401 . s. raeisi , w. tittel , c. simon , phys . ( 2012 ) 120404 . p. sekatski , n. brunner , c. branciard , n. gisin , c. simon , phys . 103 ( 2009 ) 113601 . n. spagnolo , c. vitelli , m. paternostro , f. de martini , f. sciarrino , phys . rev . a 84 ( 2011 ) 032102 . m. stobi ' nska , p. sekatski , a. buraczewski , n. gisin , g. leuchs , phys . a 84 ( 2011 ) 034104 . h. kwon , h. jeong , phys . a 88 ( 2013 ) 052127 . m. stobi ' nska , f. tppel , p. sekatski , a. buraczewski , phys . rev . a 89 ( 2014 ) 022119 . f. de martini , f. sciarrino , c. vitelli , arxiv:0804.0341 ( 2008 ) . a. furusawa , p. van loock , quantum teleportation and entanglement : a hybrid approach to optical quantum information processing , john wiley and sons , new york , 2011 . p. van loock , laser & photon . rev . 5 ( 2011 ) 167 . k. park , s .- w . lee , h. jeong , phys . a 86 ( 2012 ) 062301 . lee , h. jeong , phys . a 87 ( 2013 ) 022326 . j. rigas , o. ghne , n. ltkenhaus , phys . a 73 ( 2006 ) 012341 . c. wittmann , j. frst , c. wiechers , d. elser , h. hseler , n. ltkenhaus , g. leuchs , opt . express 18 ( 2010 ) 4499 .
quantum correlation is an intrinsic aspect of quantum theory that enables the manifestation of several interesting phenomena beyond the realms of the classical world . the quintessential form of quantum correlation is entanglement which is an important resource in various quantum information and computational protocols @xcite . the fundamental need for scalability of quantum information processing and computation protocols require the generation of controlled quantum correlations distributed over large number of subsystems @xcite . in particular , multipartite quantum correlation is an indispensable resource in one - way quantum computing @xcite , secret - sharing protocols @xcite and quantum communication @xcite . recent experimental developments in quantum mechanics have enabled the generation of small multipartite entangled states to simulate one - way quantum computers @xcite , graph states @xcite and open - destination quantum teleportation @xcite , using trapped ions @xcite and photons @xcite . the practical realization of quantum information and computation protocols using multiparty quantum systems is severely challenged due to decoherence caused by the interaction of the system with the environment . such interactions create superfluous quantum correlations between the system and the environment leading to information being scattered in the intractable hilbert space of the environment . therefore , the fragility of multipartite quantum correlations makes the generation and preservation of quantum correlations in any quantum system a daunting task for experimentalists . from a theoretical point of view , the lack of a unique characterization of multipartite quantum correlations in quantum systems obstructs definitive study of decoherence - induced loss of correlations @xcite . the decoherence models used to study the robustness of multipartite states , are mostly local ; i.e. , each subsystem of the state interacts independently with the environment @xcite . therefore , in order to successfully implement a quantum protocol , the degradation of the quantum correlations , through local decoherence channels , must be suppressed during the application of the protocol . various schemes such as distillation protocols @xcite and quantum error correction @xcite have traditionally been used to protect entanglement under local decoherence ( cf . the efficiency of these methods depend on the robustness of the entanglement in the initial state . a more recent approach to tackling local decoherence is using the quantum zeno effect @xcite and the reversibility of quantum weak measurement @xcite . the experimental viability of implementing quantum weak measurements makes the latter method an elegant approach to counter decoherence . the suppression of decoherence on a single qubit using the reversal of weak measurement has been experimentally exhibited @xcite . further , the protection of bipartite entanglement in two - qubit systems using a reversal weak measurement scheme has been studied @xcite and experimentally demonstrated @xcite . the method has also been extended to include two - qutrit quantum correlations @xcite and few - body quantum communication tasks @xcite . a natural progression of the weak measurement approach is to consider the suppression of the degradation of multipartite quantum correlations under local decoherence as they are desirable for applications in scalable quantum information and computation protocols . an obvious difficulty in designing a weak reversal multipartite decoherence - protection scheme is the simultaneous characterization of the quantum correlation measure and the weak measurement technique in the multipartite setting . we use the local amplitude damping channel ( ladc ) as our decoherence model , and study its effect on the generalized greenberger - horne - zeilinger ( gghz ) state @xcite . such a channel produces a mixed state of the system from an initial multipartite pure entangled state , rendering procedural characterization of pure state multipartite quantum correlations irrelevant . hence , we characterize the mixed state multipartite quantum correlations using a multipartite extension of logarithmic negativity @xcite and a measure of global entanglement called the mayer - wallach measure @xcite . we then apply a multipartite generalization of the decoherence - protection protocol based on the reversibility of weak measurement . we observe that the protocol makes the multipartite quantum correlations robust against ladc . this is also evident from the enhanced critical damping value that corresponds to entanglement sudden death . to further elucidate the efficiency of the protocol , we investigate quantum teleportation @xcite in a one sender , many receivers setting and multiparty quantum information splitting @xcite using the amplitude - damped gghz state and observe an increase in the average fidelity of teleportation and information splitting under the weak measurement reversal protocol . the paper is organized in the following way . in sec.[wk ] , we discuss the weak measurement reversal protocol to be applied to the initial multipartite state under consideration . the characterization of the multipartite quantum correlation measure is done in sec.[meas ] . the results for the suppression of decoherence and enhancement of multipartite quantum correlation is shown in sec.[res ] . section [ telep ] discusses the results of the quantum teleportation and quantum information splitting tasks . we conclude in sec.[con ] with a brief discussion of the results and its ramifications . weak measurements were initially developed for pre - selected and post - selected ensembles of quantum systems by aharonov _ _ @xcite and later generalized to cases without post - selection @xcite . an alternative definition of a weak measurement is obtained from partial collapse measurement @xcite . in such instances , weak measurements are positive - operator valued measure with limited information access as compared to projective measurement thus allowing the operation to be non - unitarily reversed . the concept of reversing a partial collapse measurement was initially introduced in @xcite and subsequently applied to quantum error - correcting codes @xcite . in recent times , a weak measurement reversal scheme has been developed to suppress the decoherence in a single qubit state @xcite , and subsequently expanded to protect quantum correlations in two qubit states @xcite . the underlying principle for the weak measurement reversal scheme is the fact that any partial collapse measurement can be reversed @xcite . the weak measurement reversal for protection of multipartite quantum correlations can be described by the following protocol . first , a weak measurement of a given strength , say @xmath0 , is made on the initial multiparty entangled state before the state is subjected to local decoherence . for an @xmath1qubit initial entangled state , the measurement consists of @xmath1 single qubit weak measurements , each of strength @xmath0 , acting locally and independently on each qubit . secondly , the post weak measured state undergoes decoherence via a local amplitude damping channel , with damping parameter @xmath2 , acting independently on each qubit . finally , the decohered state is subjected to a reversal weak measurement , with an optimal reversal strength , @xmath3 , that is attained by maximizing the multipartite quantum correlations . again , the reversal weak measurement consists of @xmath1 single qubit reversal weak measurements , each of strength @xmath3 , acting locally and independently on each qubit . the protocol is graphically illustrated in fig . ( [ fig : reverse ] ) . in the following parts , we discuss the amplitude damping channel , the weak measurement , and the weak measurement reversal . amplitude damping channel is a very useful model of decoherence for studying various quantum phenomena such as spontaneous emission in quantum optics , energy dissipation in quantum open - systems and capacities of quantum channels . the decoherence due to amplitude damping results in the coupling of a qubit , locally , to its environment leading to irreversible transfer of a basis to the other . the effect of a ladc on a density matrix of a single qubit in computational basis is described by the following map : + @xmath4,\\\nonumber\end{aligned}\ ] ] where @xmath2 is the amplitude damping parameter and @xmath5 ( @xmath6 ) are the elements of @xmath7 in the computational basis . thus , the amplitude damping channel keeps the computational basis state @xmath8 unchanged but transfers the state from @xmath9 to @xmath8 with probability @xmath2 . the effect of a ladc of strength @xmath2 on an @xmath1qubit state , say @xmath10 , is given by the map @xmath11 . the weak measurement , used in our protocol is a positive operator valued measure , consisting of a set of positive operators @xmath12 , such that @xmath13 if we discard the outcome of the measurement @xmath14 , then the measurement @xmath15 is a null - result weak measurement of strength @xmath0 , that partially collapses the system to one of the basis states . the action of the weak measurement operator @xmath15 on a qubit density matrix in computational basis can be mapped as @xmath16,\end{aligned}\ ] ] where @xmath5 ( @xmath6 ) are the elements of @xmath7 in the computational basis . the quantum map of weak measurement is different from the ladc as it uses post - selection to selectively map the states . the discarded detections , via some ideal detector , ensures that the operator @xmath15 keeps the computational basis state @xmath9 unchanged with a probability @xmath17 and a norm less than unity . such post - selection in weak measurements can increase or decrease quantum correlations in bipartite systems @xcite . the action of the local quantum weak measurement , of strength @xmath0 , on an @xmath1qubit state , say @xmath10 , is given by the map @xmath18 . qubit gghz state . it then applies local weak measurements of strength @xmath0 on all individual qubits of the @xmath1party state . the weak measured qubits are then individually passed through a ladc with amplitude damping parameter @xmath2 . after local amplitude damping , each qubit is subjected to a local reversal weak measurement of strength @xmath19 . the multipartite quantum correlation of the final state is more robust than a @xmath1qubit state passing through ladc without weak measurement reversal protocol . , width=80 ] the weak measurement can be reversed using nonunitary operations . for every null - result weak measurement given in sec . [ subweak ] , we can obtain a nonunitary reverse operation @xmath20 with strength @xmath19 , that will produce the initial state with some probability . such an operation is termed as a reversal weak measurement . let us consider a measurement using the set of positive operators @xmath21 , such that @xmath22 @xmath23 again if we selectively discard the outcome of measurement @xmath24 , then @xmath20 constitutes a reversal weak measurement with strength @xmath19 . for some optimized reversal weak measurement strength @xmath25 , corresponding to the original weak measurement strength @xmath0 , one can generate the pre - weak measurement state with some probability . the action of the weak measurement operator @xmath20 on a qubit density matrix in computational basis can be mapped as @xmath26,\end{aligned}\ ] ] where @xmath5 ( @xmath6 ) are the elements of @xmath7 in the computational basis . similar to the weak measurement , in the reversal measurement , the discarded detections ensures that the operator @xmath20 keeps the computational basis state @xmath8 unchanged with a probability @xmath27 and a norm less than unity . the action of the local reversal quantum weak measurement of strength @xmath19 on an @xmath1qubit state , say @xmath10 , is given by the map @xmath28 . in our study , the amplitude damping channel , the weak and the reversal measurements act locally on each subsystem ( qubit ) of the gghz state . the ladc , in effect , renders the initial pure gghz state to a mixed quantum state . the study of multipartite quantum correlations in this decohered quantum state will require the characterization of a multipartite measure . though there are known measures of multipartite entanglement @xcite which can be computed in large pure quantum states @xcite , these measures do not uniquely generalize to the mixed state regime . in the following parts we attempt to discuss multipartite quantum correlations that shall prove useful for our study . logarithmic negativity ( ln ) is a useful and computable measure of entanglement introduced by vidal and werner @xcite based on the peres - horodecki criteria @xcite . the criteria notes that the negativity of the partial transpose of any bipartite state is a sufficient condition for entanglement along that particular cut . for an arbitrary bipartite state , @xmath29 , ln is defined as @xmath30,\end{aligned}\ ] ] where @xmath31 is called the negativity " , and @xmath32 is the trace norm of @xmath33 , which is the partially transposed state of the bipartite state @xmath29 with respect to the subsystem @xmath34 . the negativity , @xmath35 , is thus the sum of the absolute values of the negative eigenvalues of @xmath36 . for the gghz state , the negative partial transpose criteria of entanglement along all possible bipartite cuts is a sufficient condition . hence , for gghz states , ln is a measure of multipartite entanglement @xcite . a scalable and functional entanglement monotone for multiqubit pure states , called the global entanglement , was introduced by meyer and wallach @xcite . the meyer and wallach measure of global entanglement ( mw ) is closely related to the per qubit total nonlocal information @xcite in the system and hence can be easily related to the loss of nonlocal information per site under decoherence @xcite . the distribution of nonlocal information in a system is known to be related to quantum correlations in a multipartite quantum system @xcite . mw measure is , thus a measure of pure state multipartite quantum correlations . interestingly , the connection between the total nonlocal information and tangles can be used to characterize global entanglement in the decohered gghz state . the mw measure of global entanglement , for an @xmath1qubit state in terms of tangles , can be written as @xcite @xmath37,\end{aligned}\ ] ] where @xmath38 is the 2tangle , @xmath39 is the 3tangle and @xmath40 is the @xmath1tangle . the term on the right side of eq . ( [ mw ] ) , with a multiplicative factor ( @xmath1 ) , is the expression for total nonlocal information of a quantum system in terms of the tangles . the @xmath41tangle for an @xmath1qubit state can be obtained by generalizing the idea of @xmath42tangle @xcite . for the states belonging to gghz class , the @xmath1tangle ( @xmath43 ) is the only nonzero tangle and , therefore , the sole contributor to eq . ( [ mw ] ) @xcite . for multiparty states with even @xmath1 , the @xmath1tangle can be expressed in terms of the square of a quantity called the @xmath1concurrence @xcite , which is a suitable generalization of concurrence @xcite for a two qubit state . hence , for certain mixed multipartite states the mw measure of global entanglement can be exactly computed in terms of its @xmath1concurrence , where @xmath1 is even . a mathematical disposition of the mw measure for the weak measurement reversal protocol is given in sec . in this section , we apply the weak measurement reversal protocol to a local amplitude damped @xmath1qubit gghz state and study its effect on the decay of multipartite quantum correlations . to investigate the effectiveness of the protection protocol , we study the decay of these measures for different values of amplitude damping parameter @xmath2 and system size @xmath1 . we consider the @xmath1qubit generalized ghz ( gghz ) state @xcite as our initial multipartite entangled state , which can be written as @xmath44 where @xmath45 and @xmath46 with @xmath47 . to apply the local decoherence in the initial state , we pass each of the @xmath1 qubits of the gghz state , separately , through the amplitude damping channel , given in sec.[amp ] . the action of the amplitude damping channel on the initial gghz state yields the following decohered @xmath1qubit state @xmath48\nonumber\\ & + \|\beta\|^2\sum_{k=0}^{n}p^k\bar{p}^{(n - k)}[({\|{0}\rangle}{\langle{0}\|})^{\otimes k}\otimes({\|{1}\rangle}{\langle{1}\|})^{\otimes n - k}+ \cal{r}],\end{aligned}\ ] ] where @xmath49 . @xmath50 denotes all other diagonal terms that can be obtained by permutations of positions of 0 and 1 in the computational basis . we first consider , the calculation of the multipartite logarithmic negativity ( sec.[mln ] ) . to obtain the multipartite ln , we need to obtain the partial transpose of the state @xmath51 across all possible bipartite cuts . let us consider the @xmath52 cut for partial transposition . the state after partial transposition along this cut is given by @xmath53\nonumber\\ & + \|\beta\|^2\sum_{k=0}^{n}p^k\bar{p}^{(n - k)}[({\|{0}\rangle}{\langle{0}\|})^{\otimes k}\otimes({\|{1}\rangle}{\langle{1}\|})^{\otimes n - k}+ \mathcal{r}],\end{aligned}\ ] ] where the only transformation occurs in the non - diagonal terms of @xmath51 . the partial transposed matrix @xmath54 contains a diagonal block of dimension @xmath55 that is positive semi - definite . the negative eigenvalue(s ) can be obtained from the remaining @xmath56 block , which corresponds to basis vectors @xmath57 . the @xmath58 non - positive semi - definite block can be written as @xmath59 . \end{aligned}\ ] ] the eigenvalue of the above block which contributes to the entanglement , is given by @xmath60 , \end{aligned}\ ] ] where @xmath61 . using the negative eigenvalue obtained from eq.([ev ] ) , we calculate the multipartite ln ( defined by eq.([ln ] ) ) . the condition for nonzero entanglement , i.e. , @xmath62 to be negative is given by @xmath63 . thus , the entanglement of the decohered gghz state in the bipartition @xmath52 vanishes for @xmath64 , where @xmath65 is the critical damping value . @xmath66 is found to be independent of the bipartition used to calculate the ln . for @xmath67 case , @xmath68 , which is the case of maximum decoherence . however , numerically the entanglement approaches values close to zero , before the critical value @xmath68 . in our analysis , we define a new critical value of @xmath2 , denoted by @xmath69 , as the least value of @xmath2 above which the entanglement is less than or equal to @xmath70 . interestingly , for the case of @xmath71 , the multipartite ln becomes zero before the maximal value of decoherence parameter , @xmath72 , is attained . this phenomenon is called entanglement sudden death ( esd ) @xcite . now let us consider the operations under the weak measurement reversal protocol , as mentioned in sec.[wk ] . as discussed earlier , we first apply a weak measurement ( @xmath73 ) , of strength @xmath0 , on the initial @xmath1qubit gghz state . we then pass the weak measured state through an amplitude damping channel ( @xmath74 ) , with damping parameter @xmath2 . finally , we perform a reversal weak measurement ( @xmath75 ) , of strength @xmath19 , to obtain the final state . and @xmath88 . @xmath89 is the success probability of the reversal weak measurement . the transmissivity or the success probability of the final weak measurement reversal protocol , is given by @xmath90.\end{aligned}\ ] ] let us again consider the @xmath52 bipartition for the partial transposition of the final post reversal weak measurement state @xmath91 . following the steps for partial transpose of the non - protected amplitude damped state we again obtain a @xmath58 matrix block that contains the possible non - positive eigenvalue . the matrix form of such a block can be found to be @xmath92.\nonumber\\ \end{aligned}\ ] ] the eigenvalue , which can be negative is @xmath93 , \end{aligned}\ ] ] where @xmath94 . the condition for nonzero entanglement , i.e. , @xmath95 to be negative is given by @xmath96 . this shows that the weak measurement reversal protocol enhances the critical damping value of @xmath2 as compared to the unprotected case . thus , a higher damping parameter @xmath2 is required to remove entanglement . the optimal value of the reversal weak measurement strength @xmath19 is obtained by maximizing the multipartite ln for a fixed value of the damping strength @xmath2 , the number of qubits @xmath1 , the initial weak measurement strength @xmath0 and all possible bipartitions @xmath97 . we will denote this optimal value of ln by @xmath98 . for @xmath71 and specific values of the weak measurement strength @xmath0 , the observed esd under the amplitude damping can be suitably circumvented . figure ( [ fig1 ] ) represents the optimal value of the multipartite ln ( @xmath98 ) of the amplitude - damped gghz state ( with @xmath99 ) , under the weak measurement reversal protocol , as a function of the damping strength @xmath2 , for various initial weak measurement strength @xmath0 , and for @xmath1 qubits , at the bipartition @xmath97 = @xmath100 . we observe that the weak measurement reversal protocol suppresses the decay of multipartite quantum correlations , for different @xmath1 . from fig . ( [ fig1 ] ) , we observe that @xmath98 decreases with increase of the amplitude damping parameter @xmath2 and vanishes after a critical damping value @xmath101 , that depends on @xmath1 . for smaller values of @xmath1 ( @xmath102 and @xmath103 ) , the weak measurement reversal technique leads to a substantial increase in the multipartite entanglement for larger values of @xmath0 , making the state more robust against the noise . however , we observe later that the arbitrary high values of @xmath104 $ ] are not allowed . at higher values of @xmath1 , say @xmath105 , it is apparent from fig . [ fig1 ] ( also see fig . [ cri1 ] ) that @xmath101 becomes smaller and the weak measurement reversal protocol does not sufficiently enhance entanglement ( due to limitation on @xmath0 ) , rendering the state less robust against decoherence for very large values of @xmath1 . figure [ cri1 ] shows the variation of @xmath101 with @xmath1 , for fixed values of the weak measurement strength @xmath0 . we observe that the values of @xmath101 for all @xmath0 , get closer at higher values of @xmath1 , which implies that enhancement of the critical value for protected states is lower , rendering the protocol less efficient at very large @xmath1 . , for @xmath106 plotted against the number of qubits @xmath1 . the plot shows that @xmath101 of the protected state remains above the unprotected state for @xmath1 = 100 , though the difference diminishes with increasing @xmath1 . the plot label up is for the unprotected state ( @xmath0 = @xmath19 = 0).,width=188 ] the effective value of the weak measurement strength @xmath0 also depends on the success probability of the reversal protocol . to obtain the success probability of the protocol we calculate its transmissivity , defined in relation ( [ trans1 ] ) . for an initial gghz state with @xmath107 , the transmissivity is given by the expression @xmath108 . $ ] the transmissivity @xmath109 is calculated for the optimal value of the reversal weak measurement strength @xmath19 for which @xmath110 is maximal , i.e. @xmath111 . we observe from fig . [ fig2 ] , that the success probability reduces with increase in the weak measurement strength @xmath0 and hence , the protocol becomes less robust . thus , for larger values of the initial weak measurement strength , the success probability of the quantum state reversal using weak measurement becomes lower and hence the number of initial copies of the system needed to successfully implement the protection protocol increases . we again consider the @xmath1qubit gghz state given by eq . ( [ ghz ] ) . for the mw measure of global entanglement ( @xmath112 ) in eq . ( [ mw ] ) , it is known that the multipartite quantum correlation for the @xmath1qubit gghz state is contained only in the @xmath1tangle part of the nonlocal information , since all the other @xmath41tangles ( @xmath113 ) are zero @xcite . for an even number of qubits ( even @xmath1 ) , the @xmath1tangle can be calculated using the @xmath1concurrence measure @xmath114 , as shown in @xcite . @xmath114 is a generalization of two qubit measure of concurrence @xcite . therefore for an initial , even @xmath1qubit gghz state the mw global entanglement is given by the relation @xmath115 where @xmath114 is defined as @xmath116,\end{aligned}\ ] ] where @xmath117 s are the eigenvalues of the matrix @xmath118 ( for any density matrix @xmath7 ) in decreasing order , and @xmath119 is the pauli matrix . the gghz state , after passing through local amplitude damping channel is given by eq . ( [ gdamp ] ) . the eigenvalues of @xmath120 are give by @xmath121,\nonumber\\ & \lambda_2=\|\beta\|^2\bar{p}^n\big[(2\|\alpha\|^2+\|\beta\|^2p^n)-\sqrt{4\|\alpha\|^2(\|\alpha\|^2+\|\beta\|^2p^n)}\big],\nonumber\\ & \lambda_j=\|\beta\|^4(p\bar{p})^n , \{j=3,4, .. ,2^n\},\end{aligned}\ ] ] where @xmath122 is the spin conjugated state obtained after applying @xmath123 on @xmath124 . therefore , the @xmath1concurrence is given by @xmath125\nonumber\\ & = \max\big[0,\big\{2\|\alpha\|\|\beta\|\bar{p}^{n/2}\big[1-(2^{n-1}-1)\frac{\|\beta\|}{\|\alpha\|}p^{n/2}\big]\big\}\big].\end{aligned}\ ] ] the above expression shows that the @xmath1concurrence is zero for @xmath126 , where @xmath127 . in the limit of very large @xmath1 , the critical value @xmath128 becomes @xmath129 , independent of the parameters of the initial gghz state . the @xmath1concurrence indicates that the gghz can sustain decoherence upto @xmath2 = @xmath130 for very large @xmath1 , unlike the logarithmic negativity which , in principle can sustain decoherence upto @xmath2 = @xmath84 . the global entanglement , @xmath112 of the state @xmath51 is given by @xmath131 . we define the numerical critical value of @xmath2 in the case of @xmath112 , @xmath132 , as the value of @xmath2 below which the @xmath112 is less than @xmath70 . @xmath138\big ] . \ ] ] the expression for @xmath139 shows that the @xmath1concurrence is zero for @xmath126 , where latexmath:[$p^{mw}_c = \min\{1 , \frac{1}{\bar{s}}\left(\frac{\|\alpha\| } { in the limit of very large @xmath1 , the critical value of @xmath128 becomes @xmath141 , which is a function of initial weak measurement strength @xmath0 but independent of the parameters of the initial gghz state . the @xmath1concurrence of the protected gghz can sustain decoherence upto @xmath142 for very large @xmath1 . this indicates that the protected state is more robust against decoherence and esd occurs at higher amplitude damping values , as compared to the unprotected state . the global entanglement , @xmath112 is given by @xmath143 . the optimal value of the global entanglement , @xmath144 , after the weak measurement reversal protocol can be obtained by numerically maximizing @xmath112 , with respect to the weak reversal strength @xmath19 , at fixed values of @xmath1 , @xmath0 , and @xmath2 . [ fig3 ] shows the optimal value of global entanglement , @xmath144 , against the amplitude damping parameter @xmath2 , for different values of @xmath1 and @xmath0 . the plots show that @xmath144 increases with weak measurement strength @xmath0 . again , arbitrary large values of @xmath0 in @xmath145 $ ] are limited , as it will lead to low success probability of the weak measurement reversal protocol . it is also clear from the plots that the critical value @xmath132 is always greater for the weak measurement applied state compared to the unprotected state . we observe that the value of @xmath132 is not suitably enhanced for very large @xmath1 . , for @xmath144 is plotted against the number of qubits @xmath1 . the plot shows that the critical value , @xmath132 , of the protected state remains above the unprotected state for @xmath1 = 100 , though the difference diminishes with increasing @xmath1 . the plot label up is for the unprotected state ( @xmath0 = @xmath19 = 0).,width=50 ] in section ( [ res ] ) , we have seen that multipartite entanglement can be protected against ladc , using weak measurement reversal protocol . the immediate question that arises , is whether such a protection scheme can prove beneficial in some quantum communication task ? here we consider , two proximately related quantum communication tasks , namely , quantum teleportation of an unknown qubit state in a one sender , many receivers setting @xcite and multiparty quantum information splitting @xcite , that , under decoherence , are known to be associated with different aspects of multiparty quantum correlations @xcite . we consider the decohered gghz state as the shared resource in these quantum communication tasks and calculate the average fidelities under the weak measurement protection protocol . according to the one sender , many receivers quantum teleportation protocol , alice ( @xmath34 ) wants to send an unknown qubit in a state ( @xmath146 ) to ( @xmath147 ) bobs ( @xmath148 , @xmath149 to @xmath147 ) using a shared @xmath1qubit ghz state , given as @xmath150 . \label{sghz}\end{aligned}\ ] ] the success of a quantum teleportation protocol can be evaluated by the average fidelity between the unknown initial qubit to be teleported and the final qubit received during the process . the fidelity of teleportation is defined as @xmath151 , where @xmath152 is the unknown initial state and @xmath7 is the final teleported state . for a perfect quantum teleportation protocol , with unit fidelity , @xmath34 and @xmath148 s must share a maximally quantum correlated state @xcite . the maximum fidelity achievable using only classical communication , with no quantum correlations , is @xmath153 = @xmath154 @xcite . for the considered @xmath1qubit ghz state , given by ( [ sghz ] ) , the pre - shared amount of quantum correlation is sufficient to teleport the unknown state from alice to ( @xmath147 ) bobs with maximum fidelity . to elaborate , alice makes a bell measurement on the unknown qubit , @xmath152 , and the qubit ( @xmath34 ) of the @xmath1qubit pre - shared ghz state , in her possession , and classically communicates her results to ( @xmath147 ) bobs . the information to be teleported is encoded in the reduced ( @xmath147)-qubit state in bobs possession . on receiving alice s result , the ( @xmath147 ) bobs together apply suitable local unitaries to obtain the state @xmath155 applying local measurements , the ( @xmath147 ) bobs together can distill the unknown state , @xmath146 , with unit fidelity , from the encoded state , @xmath156 . however , for practical applications , the pre - shared resource is generally not optimal for perfect teleportation . let us consider a slightly non - idealistic situation where some arbitrary quantum state preparator prepares the @xmath1qubit gghz state , given by eq . ( [ sghz ] ) , and sends the qubits to alice and bobs . since alice and bobs are at distant locations , the state sent by the preparator suffers from decoherence due to environmental interaction . such a decohered state can be modelled by sending the prepared @xmath1qubit gghz state through @xmath1 single qubit ladcs ( sec.[amp ] ) to alice and bobs . thus , the state shared by alice and @xmath157 bobs is given in eq . ( [ gdamp ] ) ( with @xmath107 ) . the average fidelity of teleportation using such a shared state is known to decrease with increasing damping strength @xmath2 @xcite . in the weak measurement reversal protocol , the preparator first applies a weak measurement of strength @xmath0 on each shared qubit of the state . these weak measured states are then sent to alice and ( @xmath147 ) bobs through the local amplitude damping channel , with strength @xmath2 . now , alice and ( @xmath147 ) bobs locally apply reversal weak measurements of equal strength @xmath19 , to all the @xmath1 qubits . the final measured state , using eq . ( [ proghz ] ) , is now the pre - shared quantum state , given by @xmath158 + \bar{s}^{n}\sum_{k=0}^{n}(p\bar{r})^k\bar{p}^{(n - k)}[({\|{0}\rangle}{\langle{0}\|})^{\otimes k } \nonumber\\ & \otimes({\|{1}\rangle}{\langle{1}\|})^{\otimes n - k}+ \cal{r}_1]\big ] , \label{tel - res}\end{aligned}\ ] ] alice now wants to teleport the unknown qubit @xmath152 to ( @xmath147 ) bobs using the @xmath1qubit pre - shared state @xmath159 . the joint state is given by , @xmath160 alice now possesses a qubit ( @xmath34 ) of the @xmath1qubit pre - shared state @xmath159 and the unknown qubit @xmath152 and makes a bell basis measurement on the two qubits in her possession . the unknown state to be teleported is now encoded in some reduced ( @xmath147)qubit state in ( @xmath147 ) bobs possession . let , @xmath161 and @xmath162 , be the bell basis states . on measurement , alice gets @xmath163 as outcomes with probability @xmath164 and the normalized state of ( @xmath147 ) bobs is @xmath165 , where @xmath166_{b_1 .. b_{n-1 } } , \label{rho1}\end{aligned}\ ] ] and @xmath167 similarly , alice gets @xmath168 as outcome with probability @xmath169 and the normalized state of ( @xmath147 ) bobs is @xmath170 , where @xmath171_{b_1 .. b_{n-1 } } , \label{rho2}\end{aligned}\ ] ] and @xmath172 now depending on the measurement outcomes of alice , the ( @xmath147 ) bobs apply suitable local unitaries on the encoded state , @xmath173 ( @xmath174 ) , to obtain the optimal ( @xmath147)qubit state , say , @xmath175 ( @xmath176 ) , with corresponding probabilities @xmath177 = \{@xmath164 , @xmath169 } , that maximizes the fidelity of the teleportation . from relation ( [ dis ] ) , it is known that in the absence of ladc , the encoded state @xmath178 , allows ( @xmath147 ) bobs to distill the unknown state , @xmath179 , with unit fidelity . hence , the success of the protocol , depends on the fidelity of ( @xmath147 ) bobs optimal encoded state , @xmath180 , with the state @xmath178 . thus , the average fidelity of teleportation can be calculated using the relation , @xmath181 . using the weak measurement reversal protocol , the average fidelity ( @xmath182 ) of quantum teleportation is given by the expression @xmath183.\end{aligned}\ ] ] the average fidelity of teleportation is maximized with respect to the reversal weak measurement strength @xmath19 , keeping the parameters @xmath2 , @xmath0 , and @xmath1 fixed . without the protection protocol ( @xmath184 and @xmath185 ) , the average fidelity is given by the expression @xmath186 , as obtained in @xcite . figure ( [ fig5 ] ) is the plot for the numerically optimized average fidelity of quantum teleportation . it can be observed from the figure that the critical value @xmath187 ( beyond which the optimized average fidelity is less than or equal to the classical teleportation fidelity , @xmath154 ) has increased with the initial weak measurement strength @xmath0 , as compared to the average fidelity without any weak measurement protection . it is also observed from fig . ( [ fig5 ] ) that the optimal average fidelity always remains above or equal to classical fidelity that is @xmath188 , for finite weak measurement reversal application . in the one sender , many receivers quantum teleportation protocol , the average fidelity of teleportation is obtained by maximizing the fidelity of the encoded ( @xmath147)qubit states in possession of bobs with the state @xmath178 . the ( @xmath147 ) bobs perform as a single entity to optimize the outcome . an alternate approach is to consider the multiparty quantum information splitting protocol @xcite to decode the unknown state , @xmath179 . in this protocol , the ( @xmath147 ) bobs are independent , but co - operative , entities that can classically communicate with each other . a specific bob ( say ( @xmath147)@xmath189 bob ) is chosen as the pre - determined receiver with the other ( @xmath190 ) bobs assisting it to receive the unknown state , @xmath179 . the information is thus being split among many receivers and can be retrieved only by necessary co - operation . this forms the basis of many secret sharing protocols @xcite . in the quantum information splitting protocol , the encoding of the state to be teleported by alice remains the same as the quantum teleportation protocol . alice makes a bell basis measurement on the two qubits in her possession , the qubit ( @xmath34 ) of the @xmath1qubit pre - shared state @xmath159 and the unknown qubit @xmath152 . after alice s measurement , the encoded states , @xmath191 ( @xmath192 to @xmath193 ) , given by eq.([rho1 ] ) and ( [ rho2 ] ) , with corresponding probabilities @xmath194 , are split among the independent ( @xmath147 ) bobs . the ( @xmath190 ) non - receiver bobs measure their respective qubits in the eigenbasis of the pauli spin matrix @xmath195 and classically communicate their outcomes , which occur with probabilities , @xmath196 ( @xmath197 ) , to the ( @xmath147)@xmath189 bob . depending on the measurement outcomes of alice and the ( @xmath190 ) bobs , ( @xmath147)@xmath189 bob applies suitable unitary operations to obtain the single qubit state , say , @xmath198 ( @xmath197 ) , with probability @xmath199 . the average fidelity of the protocol can then be calculated using the relation , @xmath200 . for a maximally entangled pre - shared state , such as the @xmath1qubit ghz state , the ( @xmath147)@xmath189 bob can obtain the unknown state , @xmath152 , with unit fidelity . under the action of ladc and weak measurement reversal protocol , the average fidelity of teleportation using multiparty quantum information splitting ( @xmath201 ) , is given by the expression @xmath202.\end{aligned}\ ] ] the average fidelity using information splitting can be maximized with respect to the reversal weak measurement strength @xmath19 , keeping the parameters @xmath2 , @xmath0 , and @xmath1 fixed . without the protection protocol ( @xmath184 and @xmath185 ) , @xmath201 is given by @xmath203 , as obtained in @xcite . figure ( [ fig6 ] ) is the plot for the numerically optimized average fidelity of teleportation using quantum information splitting . it can be observed from the figure that the critical value @xmath187 ( beyond which the optimized average fidelity is less than or equal to the classical teleportation fidelity , @xmath154 ) has increased with the initial weak measurement strength @xmath0 , as compared to the average fidelity without any protection from the weak measurement reversal protocol . it is also observed from fig . ( [ fig6 ] ) that the optimal average fidelity always remains above or equal to classical fidelity that is @xmath204 , for finite weak measurement reversal application . we observe that both the communication tasks , viz . , one sender , many receivers quantum teleportation and multiparty quantum information splitting , have greater average fidelity under the weak measurement reversal protocol , when the pre - shared resources suffer from decoherence . the fundamental difference between the two communication tasks is in the optimization to obtain the desired outcome . while the first protocol requires joint optimization of the encoded states by ( @xmath147 ) bobs to allow maximal distillation of the outcome , the second protocol uses multiparty information splitting to decode the state , with the pre - determined bob performing the optimization . it is argued in @xcite , that the two protocols utilize different aspects of multipartite quantum correlations to get the average fidelity above the classical upper bound , using w - like state as a resource . however , in our investigation , we observe that the two quantum communication tasks , using a decohered multi qubit gghz state , do not exhibit any relation with a specific kind of multiparty entanglement measure . for example , in the absence of the weak measurement protection protocol , starting with a four qubit decohered gghz state , we observe that the mw measure goes to zero for @xmath205 but fidelity of information splitting is greater than 2/3 for @xmath205 . this clearly shows that the mw measure can not be the only unique resource for the information splitting protocol ( c.f . quantum correlations are the basic resources for various quantum information protocols . from the perspective of futuristic designs of quantum devices , the scalability of quantum resources is an important aspect of contemporary research . as such , quantum correlations in multipartite quantum systems need to be harnessed and generated . hence , the study of decoherence protected quantum systems are of fundamental and practical importance from the perspective of multipartite quantum correlations . in our study , we have considered a multipartite , @xmath1qubit gghz state undergoing local amplitude damping and quantified its mixed state multipartite quantum correlations using logarithmic negativity and global entanglement . we have then formulated a deocherence suppression and quantum correlation protection scheme based on weak measurement and reversal technique . using analytical characterization and numerical optimization we have evaluated the multipartite quantum correlations under the weak measurement reversal protocol . we observe that under such a protocol , the multipartite correlations are more robust and do not vanish for low damping . in particular we observe an enhanced value of the damping parameter corresponding to the esd . to investigate the efficacy of the weak measurement reversal protocol , we study two quantum communication tasks , namely , quantum teleportation in a one sender , many receivers setting and quantum information splitting through local amplitude damping channel . we observe that the protocol enhances both the average fidelity of teleportation and information splitting and prevents the channel from performing below the classical upper - limit . the weak measurement reversal protocol thus strengthens the robustness of global quantum correlations in multipartite systems under the effect of local amplitude damping . given the fact that such weak measurements can be experimentally performed , the protection protocol may prove practically useful in enhancing performance in quantum information tasks , in presence of noise . h.s.d . acknowledges university grants commission for financial support under the senior research fellowship scheme . thanks the harish - chandra research institute ( hri ) for hospitality and support during visits . the authors thank aditi sen(de ) , arun kumar pati and ujjwal sen for various helpful discussions . u.s . thanks rafael chaves and yong - su kim for helpful communications . d. bouwmeester , j .- w . pan , m. daniell , h. weinfurter , and a. zeilinger , phys . lett . * 82 * , 1345 ( 1999 ) ; a. raschenbeutel , g. nogues , s. osnaghi , p. bertet , m. brune , j. m. raimond , and s. haroche , science * 288 * , 2024 ( 2000 ) ; c. a. sackett , d. kielpinski , b. e. king , c. langer , v. meyer , c. j. myatt , m. rowe , q. a. turchette , w. m. itano , d. j. wineland , and c. monroe , nature ( london ) 404 , * 256 * ( 2000 ) ; j .- w . pan , m. daniell , s. gasparoni , g. weihs , and a. zeilinger , phys . 86 * , 4435 ( 2001 ) ; c. f. roos , m. riebe , h. hffner , w. hnsel , j. benhelm , g. p. t. lancaster , c. becher , f. schmidt - kaler , and r. blatt , science * 304 * , 1478 ( 2004 ) . m. zukowski , a. zeilinger , m. a. horne , and h. weinfurter , acta phys . pol . * 93 * , 187 ( 1998 ) ; r. demkowicz - dobrzanski , a. sen(de ) , u. sen , and m. lewenstein , phys . a * 80 * , 012311 ( 2009 ) ; n. gisin , g. ribordy , w. tittel , and h. zbinden , rev . mod . phys . * 74 * , 145 ( 2002 ) . d. leibfried , e. knill , s. seidelin , j. britton , r. b. blakestad , j. chiaverini , d. b. hume , w. m. itano , j. d. jost , c. langer , r. ozeri , r. reichle , and d. j. wineland , nature ( london ) * 438 * , 639 ( 2005 ) ; h. hffner , w. hnsel , c. f. roos , j. benhelm , d. chek - al - kar , m. chwalla , t. krber , u. d. rapol , m. riebe , p. o. schmidt , c. becher , o. ghne , w. dr , and r. blatt , _ ibid . _ * 438 * , 643 ( 2005 ) . c. h. bennett , g. brassard , s. popescu , b. schumacher , j. a. smolin , and w. k. wootters , phys . * 76 * , 722 ( 1996 ) ; c. h. bennett , h. j. bernstein , s. popescu , and b. schumacher , phys . rev . a * 53 * , 2046 ( 1996 ) ; d. deutsch , a. ekert , r. jozsa , c. macchiavello , s. popescu , and a. sanpera , phys . lett . * 77 * , 2818 ( 1996 ) ; n. gisin , phys . lett . a * 210 * , 151 ( 1996 ) . p. zanardi and m. rasetti , phys . lett . * 79 * , 3306 ( 1997 ) ; d. a. lidar , i. l. chuang , and k. b. whaley , _ ibid . _ * 81 * , 2594 ( 1998 ) ; l. viola and s. lloyd , phys . rev . a * 58 * , 2733 ( 1998 ) ; l. viola , e. knill , and s. lloyd , phys . lett . * 82 * , 2417 ( 1999 ) ; l. viola , s. lloyd , and e. knill , _ ibid . _ * 83 * , 4888 ( 1999 ) . a. n. korotkov and a. n. jordan , phys lett . * 97 * , 166805 ( 2006 ) ; y .- s . kim , y .- w . cho , y .- s . ra , y .- h . kim , opt . exp . * 17 * , 11978 ( 2009 ) ; q. sun , m. al - amri , and m. suhail zubairy , phys . a * 80 * , 033838 ( 2009 ) . n. katz , m. neeley , m. ansmann , r.c . bialczak , m. hofheinz , e. lucero , a. oconnell , h. wang , a. n. cleland , j. m. martinis , and a. n. korotkov , phys . * 101 * , 200401 ( 2008 ) ; j .- c . lee , y .- c . jeong , y .- s . kim , y .- h . kim , opt . * 19 * , 16309 ( 2011 ) . f. verstraete , j. dehaene , and b. de moor , phys . a * 68 * , 012103 ( 2003 ) ; t .- c . wei and p. m. goldbart , _ ibid . _ * 68 * , 042307 ( 2003 ) ; f. verstraete , m. popp , and j. i. cirac , phys . rev . lett . * 92 * , 027901 ( 2004 ) ; a. sen(de ) and u. sen , phys . rev . a * 81 * , 012308 ( 2010 ) ; q .- q . shi , r. orus , j. o. fjrestad , and h .- q . zhou , new j. phys . * 12 * , 025008 ( 2010 ) . r. prabhu , s. pradhan , a. sen(de ) , and u. sen , phys . a * 84 * , 042334 ( 2011 ) ; h. s. dhar , a. sen(de ) , and u. sen , phys . 111 * , 070501 ( 2013 ) ; h. s. dhar , a. sen(de ) , and u. sen , new j. phys . * 15 * , 013043 ( 2013 ) . m. horodecki , k. horodecki , p. horodecki , r. horodecki , j. oppenheim , a. sen(de ) , and u. sen , phys . lett . * 90 * , 100402 ( 2003 ) ; m. horodecki , p. horodecki , r. horodecki , j. oppenheim , a. sen(de ) , u. sen , and b. synak - radtke , phys . a * 71 * , 062307 ( 2005 ) .
the discovery of the @xmath0 gev scalar particle @xcite with the properties of the standard model higgs boson together with the so far negative results for searches of supersymmetry are pointing to a possibility of a very different model building paradigm based on a minimally extended standard model with manifest classical scale invariance . in this paper we will present the first , to the best of our knowledge , realisation of a slow - roll cosmological inflation mechanism in a classically scale - invariant bsm theory . the rational for advocating this ` heretical ' approach to model building based on classical scale invariance is as follows . first point in favour of ( approximate ) scale symmetry is a well - known ` experimental ' fact that there is just a single occurrence of a non - dynamical scale in the standard model ( sm ) the higgs mass - squared input parameter @xmath1 appearing in the sm higgs potential , @xmath2 by selecting a particular negative value for it , an expectation value @xmath3 for the higgs field , and the higgs mass @xmath4 are obtained , @xmath5 . classical scale invariance , broken by @xmath1 , is recovered by re - interpreting this scale in terms of a vacuum expectation value of a new scalar @xmath6 , coupled to the sm via the higgs portal interaction , @xmath7 @xmath8 the classical theory described by is scale - invariant , and insofar as the appropriate natural value for @xmath9 can be generated quantum mechanically , it will trigger the electroweak symmetry breaking . already 40 years ago in ref . @xcite coleman and weinberg discovered that a massless scalar field @xmath6 coupled to a gauge field dynamically generates a vacuum expectation value via dimensional transmutation of the logarithmically running coupling constants . the @xmath10 is non - vanishing , calculable in a weakly - coupled theory , and its value is naturally small ( i.e. exponentially suppressed ) relative to the uv cutoff of the theory , @xmath11 \ , \ll m_{\sst uv}\,,\ ] ] where @xmath12 is the gauge coupling of @xmath6 . in the appendix we review the coleman - weinberg mechanism to recall some of the key expressions for the convenience of the reader . in the sm@xmath13u(1)@xmath14 theory with the classical scalar potential , the @xmath6-vev in then generates the higgs vev @xmath15 and the higgs mass @xmath16 , as dictated by , @xmath17 and one an set @xmath18 , @xmath19 and @xmath20 . the u(1)@xmath14 theory coupled to the sm via the higgs portal with the scalar potential was first considered in @xcite . note that no input mass scales are allowed in the coleman - weinberg theory and , in the course of uv renormalisation , the subtraction scheme is chosen to set the _ renormalised masses _ at the origin of the field space to zero , @xmath21 where @xmath22 is the quantum - corrected effective potential . does the masslessness condition amount to a fine - tuning ? a theory with no input mass - scales is classically scale invariant @xcite . the scale invariance is not exact , but neither it is broken by an arbitrary amount . the violation of scale invariance is controlled in quantum theory by the anomaly in the trace of the energy - momentum tensor , or in other words , by the logarithmic running of dimensionless coupling constants and their dimensional transmutation scales which , in the weakly coupled perturbation theory , are much smaller than the uv cutoff . bardeen has argued in @xcite that in order to maintain anomalously broken scale invariance , one should select a regularisation scheme which does not explicitly break classical scale invariance ( so that it does not introduce positive powers of the uv cutoff scale into renormalised quantities ) . in dimensional regularisation , the equation is satisfied automatically . indeed , no power - like dependences on the cutoff scale appear in dimensional regularisation , and in theories like ours , which contain no explicit mass scales at the outset , no finite corrections to mass terms at the origin of the field space can appear either . in this regularisation which preserves classical scale invariance , the theory as it stands is not fine - tuned , at least , not in the technical sense . if the higher theory above the planck scale breaks the classical scale invariance of its ir effective theory , through heavy degrees of freedom coupled to the standard model , then from the perspective of this microscopic theory , the masslessness condition does amount to a fine - tuning . but from the perspective of our sub - planckian theory used for particle physics computations , there is no fine - tuning , and no practical problems arise associated e.g. with cancellations between large scales , etc . furthermore , once the masslessness condition is enforced at one scale , it holds at all rg scales @xcite . to summarise , within dimensional regularisation , the renormalisation condition and , more generally , equations , @xmath23 are self - consistent , contain no fine - tuning in the theory at hand , and are independent of the rg - scale . the phenomenology of the sm@xmath13u(1)@xmath14 model in the context of lhc , future colliders and low energy measurements was analysed recently in @xcite . the minimal model has only two remaining free parameters , the portal coupling , @xmath24 and the mass of the second scalar eigenstate @xmath25 ( with @xmath26 being the observed higgs mass ) . ] , and it was shown that the model is perfectly viable . in particular , the presently available higgs data constrains the portal coupling to be @xmath27 on the part of the parameter space where the second scalar is in the region between @xmath28 gev and @xmath29 . for heavier @xmath25 , the portal coupling is much less constrained experimentally , and has a theoretical upper limit of @xmath30 , see fig . 2 of ref . @xcite for more detail . other related studies of classically scale - invariant models can be found in @xcite . as the resulting theory has to be valid up to a very large uv scale , the classical scale invariance provides a powerful principle for the bsm model building . no vastly different scales which is treated separately . ] can co - exist in the theory : first , it is difficult to generate a large hierarchy of scales from the same @xmath31 . secondly , the large scales , if they do appear , would ultimately couple to the higgs and would destabilise it mass . the bsm theory is a minimal extension of the sm which should address all the sub - planckian shortcomings of the sm , such as the generation of matter - anti - matter asymmetry , inflation , dark matter , stabilisation of the sm higgs potential , without introducing scales higher than @xmath31 which itself is not much higher the electroweak scale . can one generate the matter - anti - matter asymmetry of the universe in these settings ? it was shown in ref . @xcite that the classically - invariant sm@xmath13u(1)@xmath14 theory where the coleman - weinberg u(1)@xmath14 sector is interpreted as the gauged @xmath32 symmetry of the sm , can indeed generate the observed value of matter - anti - matter asymmetry . the @xmath32 model @xcite is an appealing extension of the sm as it automatically contains three generations of right - handed majorana neutrinos necessary ( needed to cancel the gauge anomaly of u(1)@xmath33 ) . the standard see - saw mechanism generates masses of visible neutrinos and facilitate neutrino oscillations . the cw field @xmath6 carries the @xmath32 charge 2 and its vev generates the majorana neutrino masses and the mass of the u(1)@xmath33 @xmath34 boson . in the classically conformal settings , these models were first considered in @xcite . as well - known , however , the standard thermal leptogenesis formalism @xcite requires very heavy majorana neutrino masses of the order of @xmath35 gev , to generate the observed baryon asymmetry of the universe . such heavy neutrino masses go against the grain of our classically scale - invariant approach . ref . @xcite used a different realisation of the leptogenesis mechanism proposed in @xcite and further developed in @xcite the leptogenesis due to flavour oscillations of the right - handed majorana neutrinos . it was shown in @xcite that the the classically - invariant sm@xmath13u(1)@xmath33 model can generate the observed matter - anti - matter asymmetry on a large portion of the model parameter space without introducing any additional mass scales and without requiring any resonant fine - tuning . the majorana masses are in the range between 200 mev and 500 gev ; they , together with and a heavier mass for the z boson , are generated by the cw vev @xmath31 . ( 4 . ) _ vacuum stability : _ it is well - known that in the standard model the higgs quartic coupling becomes negative rendering the higgs effective potential unstable at above the scale @xmath36 gev . the next - to - next - to leading order analysis of the sm higgs effective potential carried out in @xcite found the best - fit value of the higgs self - coupling at the planck scale to be small but negative , @xmath37 with further refinements carried out in @xcite . it is remarkable that @xmath38 and its beta function nearly vanish at the planck scale , but without new physics , the absolute stability of the standard model higgs potential is ruled out @xcite . a minimal and robust way to repair the electroweak vacuum stability is provided by the higgs portal extension of the sm and this is provided automatically in the classically scale - invariant theory . there are two effects helping to stabilise the vacuum : first is the fact that the portal coupling gives a positive contribution to the beta function of the higgs quartic coupling , @xmath39 . the second effect occurs due to a vev of the second scalar , @xmath40 , leading to mixing between @xmath6 and the higgs and resulting in a threshold correction lifting the sm higgs self - coupling @xcite . in the present context of classically scale - invariant sm @xmath13 u(1)@xmath14 theory , the vacuum stabilisation was addressed in recent papers refs . the values and ranges of the couplings needed to stabilise the higgs potential depend on the model specific rg - running . gauge group . ref . @xcite showed that for the values of the portal coupling @xmath41 and the non - abelian cw coupling @xmath42 , the vacuum stability is recovered ( here the second scalar is heavy , @xmath43 gev ) . more generally , in ref.@xcite the allowed regions on the entire parameter space @xmath44 consistent with the vacuum stability and perturbativity were determined for @xmath45 and for @xmath46 , we refer to their fig . 1 for more detail . vacuum stability in the sm @xmath13 u(1)@xmath47 theory was addressed in @xcite and @xcite following a more constrained than our approach , where the vanishing ( rather than just positivity ) of the higgs potential was requested at @xmath48 . ] the conclusion we draw from this work is that the vacuum stability up to the planck scale is effectively restored on large portions of the parameter space for classically conformally invariant extensions of the standard model . we also mention that adding an additional singlet to the theory coupled to the higgs via a portal will further enhance the stability of the potential . _ inflation _ and _ dark matter _ within the classically scale - invariant bsm formulation is the subject of the present paper . in the following section we will extend the higgs portal model by introducing a single real scalar field @xmath49 with a non - minimal coupling to gravity . an exponentially flat potential for the singlet will be generated , as required for a successful implementation of a slow - roll cosmological inflation , preserving the classical scale invariance of the model and without the need of perturbative unitarisation at sub - planckian scales . quite pleasingly , the singlet @xmath49 is protected by a @xmath50 symmetry and gives a viable candidate for dark matter . thus the roles of the inflation and of the scalar dark matter particle are unified and are straightforwardly incorporated in the context of the classically scale - invariant higgs - portal bsm model . in the standard big bang cosmology the cosmological inflation was proposed in @xcite to solve the flatness , isotropy , homogeneity , horizon and relic problems . inflation is the leading theory of the early universe , confirmed by observations including the recent data from planck satellite @xcite which favour a simple inflationary scenario with only one slow rolling scalar field . slow - roll inflation occurs in many models constructed in the past , although , in general , the underlying microscopic particle physics of many inflation models is still unknown , as it involves energy scales far higher than can be probed at colliders . we will focus on the approach based on renormalisable qft lagrangians , which are then coupled to gravity , and in addition to the usual einstein - hilbert term also involve a non - minimal coupling of a scalar field to gravity . by taking the non - minimal coupling @xmath51 to be large , a flat slow - roll potential is generated and inflation takes place , as originally discussed in @xcite . an interesting extremely minimal proposal which bridges cosmology with the standard model particle physics is that the higgs itself could play the role of the inflaton @xcite . the coupling @xmath52 however introduces two sub - planckian scales in the theory , @xmath53 and @xmath54 , where the theory appears to violate unitarity , thus spoiling the self - consistency of the inflationary approach , which has lead to a considerable debate in the literature as to whether such models make sense quantum mechanically @xcite . from now and for the rest of this section we will be using the reduced planck mass @xmath55 gev and will refer to it simply as @xmath56 . in @xcite it was shown that the higgs inflation model can be unitarised by introducing ( or integrating in ) a new scalar with the mass @xmath57 participating in the inflation . this allowed the authors of @xcite to raise the unitarity bound up to the planck scale , well above the inflationary scale @xmath58 . this approach however is incompatible with the requirement of classical scale invariance of the theory which we want to maintain . the mass @xmath59 of the new scalar field , which interacts with the higgs , is a non - dynamical input scale . upon integrating out this massive degree of freedom the higgs mass is destabilised and requires fine - tuning . we take an alternative approach to inflation which is based on a classically _ massless _ and vev - less real scalar field @xmath60 . our inflaton is neither the higgs itself as in ref . @xcite , nor the heavy scalar as in ref . our approach is also different from the singlet - assisted higgs portal inflation of refs . @xcite where the inflaton was a combination of the higgs and the singlet with a non - vanishing vev ( for earlier and related work see also @xcite and references therein ) . with the inflaton ( or more precisely the non - minimally coupled to gravity scalar ) being a single - component field , we will be able to avoid problems with violations of unitarily at intermediate scales in pure gravity and kinetic sectors , along the same lines as in @xcite . we start from the classical scalar potential of the sm coupled to the cw sector which we rewrite as @xmath61 here the new @xmath6-self - coupling @xmath62 is defined in terms of the original parameters of via a shift @xmath63 . we now extend this model by adding a single degree of freedom , a one - component real scalar field @xmath49 . it is a gauge singlet which is coupled to the @xmath64 theory only via the scalar portal interactions with the higgs and @xmath6 , @xmath65 equations - describe the general renormalisable gauge - invariant scalar potential for the three massless scalars , @xmath66 as required by the classical scale invariance of the theory . the coupling constants in the potential are taken to be all positive ( or at least non - negative ) , thus the potential is stable and the positivity of @xmath67 and @xmath68 ensure that no vev is generated for the singlet @xmath49 . instead the cw vev @xmath31 generates the mass terms for the singlet , @xmath69 in the vacuum @xmath70 , @xmath71 , @xmath72 . we now proceed to couple our theory to gravity . to achieve the slow - roll inflation we need to introduce large non - minimal couplings to gravity of at least some of the scalars@xcite . as noted earlier , the central point of our approach is to non - minimally couple _ only _ the singlet @xmath60 , such that the lagrangian takes the form , @xmath73 the term @xmath74 is the non - minimal coupling of the singlet @xmath60 to gravity , with @xmath75 being the scalar curvature . it will follow that for successful inflation the value of the non - minimal coupling constant @xmath76 should be relatively large , @xmath77 . for this reason , we will treat @xmath76 ( and @xmath78 ) as large parameters @xmath79 . in this sense , @xmath60 is distinguished from the two other scalars , @xmath80 and @xmath6 , which in our case have either vanishing or small loop - induced non - minimal gravitational couplings ( @xmath81 and thus are neglected in without loss of generality ) . @xmath82 gev denotes the reduced planck mass ; in the lagrangian @xmath56 appears only in the einstein - hilbert term and does not couple directly to non - gravitational degrees of freedom . for the time being we will treat it as an explicit scale characterising the classical gravitational background . can be recast in terms of the vev of the dilation of the spontaneously broken local scale invariance . in dimensional regularisation , the scale - invariance can be kept manifest in quantum theory . we will return to this point in the conclusions section . ] finally , the subscript @xmath83 in defines the so - called jordan frame , where the scalar - to - gravity coupling is non - minimal and kinetic terms are canonically normalised . the second line on the _ r.h.s . _ of contains the scalar potential and , plus radiative corrections , plus kinetic terms for gauge and fermion fields and yukawa interactions . an immediate question arises concerning the non - minimal coupling of the singlet @xmath60 to gravity , @xmath84 in : will the appearance of the relatively large coupling @xmath85 render the theory ill - defined at an intermediate scale @xmath86 which is below the high scale @xmath56 ? if one expands around the flat space , @xmath87 , the non - minimal coupling generates the dimension-5 @xmath88 interaction . this interaction generates contributions to e.g. the @xmath89 scattering processes mediated by the graviton exchange . however , it was shown in ref . @xcite that when summed over @xmath60 , @xmath90 and @xmath91 channels , the leading order in @xmath76 contributions to this process cancel , and the first non - vanishing effects are suppressed by the high scale ( planck mass @xmath56 ) ; thus the theory in this sector appears to be safe up to the gravity scale @xmath56 , and as such is well - behaved at the intermediate scale @xmath86 . for the cancellation to occur it is however essential that there is only a single scalar non - minimally coupled to gravity @xcite . so , the theory coupled to gravity as defined by its lagrangian in the jordan frame appears to be safe up to the high uv scale @xmath56 and the appearance of the relatively large non - minimal coupling @xmath76 does not immediately invalidate the theory at lower intermediate scales . we will return to this point in what follows . we now perform a metric transformation to the einstein frame where the non - minimal scalar to gravity interaction is removed , @xmath92 so that the lagrangian becomes , @xmath93 as expected , the non - minimal interaction of @xmath60 with the scalar curvature @xmath75 has disappeared . the price for this is the non - canonical normalisation of the kinetic term for @xmath60 . the factor @xmath94 arises from the transformation of @xmath75 , @xmath95 followed by the integration by parts in the second term . this indeed gives @xmath96 which appears in . the expression describes a dimension-6 interaction suppressed by the scale @xmath86 . this appears to be a dangerously low scale for the successful investigation of inflation which as we will see occurs at the larger scale @xmath97 . of course this is the einstein frame reflection of the same phenomenon we have already considered in the jordan frame ( see the paragraph above ) . for the same reason , in the class of models considered here , where the non - minimal coupling to gravity involves a _ real scalar , the appearance of this scale in the kinetic term is harmless . it turns out that the contribution to @xmath89 scattering of the 4-point vertex is vanishing when the external states are put on - shell @xcite . this is true at tree - level and continues to hold to arbitrary loop order . the underlying reason for why the theory continues to be well - behaved at the intermediate scale ( at least in the scalar - kinetic sector and in the scalar - gravity sector ) is the existence of the field redefinition in the einstein frame which will render the kinetic term canonical . we will now proceed along these lines . to bring the kinetic term for our singlet field @xmath60 into the canonical form we perform the field redefinition @xmath98 following @xcite and define the singlet field @xmath99 via @xmath100 so that @xmath101 the kinetic term for the singlet in is now brought to the canonical form , @xmath102 we note that as pointed out in @xcite , this field redefinition is possible and works only for the case of a _ single _ real scalar field coupled non - minimally to gravity which is the case we are considering . ( 2 real components ) , no field redefinition would have amounted to the canonical normalisation of all field components . as a result , the perturbative unitarity would have been lost at the scale @xmath103 which is lower than the inflation scale @xmath104 we want to study . the use of the unitary gauge where only the radial component remains and the extra degrees of freedom are gauged away only shifts the unitarity problem to the gauge - higgs interactions and the longitudinal vector boson sector . ] at small field values , for example at the electroweak scale and all the way up to @xmath105 gev , the field redefinition amounts to @xmath106 at higher values of @xmath60 , it is easy to see that the solution for @xmath107 is well approximated by the analytic expression @xcite @xmath108\ , \quad { \rm for } \ , \,\,s \gg \frac{m}{\xi_s } \,,\ ] ] or @xmath109 at an even higher scale @xmath110 this gives , @xmath111 from and we can compute @xmath112\right)\,\simeq\,\frac{m^2}{\xi_s } \ , , \quad { \rm for } \ , \,\,s \gg \frac{m}{\xi_s } \,,\ ] ] and thus we find that at large field values , the singlet self - interaction potential in is exponentially flat when expressed in terms of the canonically normalised @xmath99 , and is well - suited for the slow - roll inflation @xcite , @xmath113\right)^2\ , , \quad { \rm for } \ , \,\,s \gg \frac{m}{\xi_s } \,.\ ] ] next we need to take care of the inverse powers of @xmath114 in the lagrangian . the appeared because the transformation was a pure metric transformation , rather then the full weyl scaling transformation , @xmath115 we would now like to excise inverse factors of @xmath114 from the einstein - frame lagrangian in to achieve the canonical normalisation for all the remaining kinetic terms . to do this , we now transform the scalar fields @xmath80 and @xmath6 ( but not @xmath60 or @xmath107 ! ) as well as the fermions , as prescribed by the weyl transformation . for the kinetic terms of non - singlet scalar fields in ( here we introduce the simplifying notation @xmath116 for all six real components of the higgs and the cw scalar ) , with the transformation @xmath117 we have , @xmath118 where we have used the substitution @xmath119 valid at the singlet field values @xmath120 . the key point we want to make is that the two terms on the second line of are suppressed by the planck scale @xmath121 . the planck scale is treated as the _ large _ scale @xmath122 than the uv cutoff in our effective field theory in the gravitational background , and we can neglect the terms suppressed by @xmath56 . the inflation will happen at the parametrically lower scale @xmath123 dictated but the exponential fall - off of the potential , and is under control in the theory with the canonically normalised kinetic terms , given by the last line of . the terms in involving portal interactions or the singlet @xmath60 with other scalars , at large values of the singlet field ( where @xmath124 ) become @xmath125 giving the mass terms for the higgs and the cw field in this large-@xmath60-field regime , with the singlet @xmath60 being ( exponentially ) decoupled from the rest of the theory , _ _ eq . . finally , the higgs portal potential @xmath126 is the homogeneous degree-4 expression in terms of scalars , given by the classical potential plus the radiative corrections also involving @xmath127 , such as the 1-loop cw expression , ( see the appendix ) , @xmath128\,.\ ] ] since @xmath22 goes as a sclar field to the 4th power , the factors of @xmath114 cancel , @xmath129 to summarise , the einstein - frame lagrangian after the weyl transformation of all the fields , except for the singlet @xmath49 which is substituted by the field redefinition , at large values of the singlet field takes the canonical form . the singlet in this regime is effectively decoupled from the sm degrees of freedom and the exponentially flat potential , @xmath130\right)^2 \,-\,\frac{1}{4}f^{\mu\nu}f_{\mu\nu } \,+\ , { \rm fermions } \nonumber \\ & & + \ , ( d^\mu h)^\dagger d_\mu h \,+\ , \frac{1}{2 } \,(d^\mu \phi)^\dagger d_\mu \phi \,-\ , \frac{\lambda_{hs}m^2}{2\xi_s } \ , \|h\|^2 \,-\ , \frac{\lambda_{\phi s } m^2}{4\xi_s } \| \phi\|^{2 } \,-\ , v(h,\phi ) \,+\ , { \rm yukawas}\ , . \nonumber \\ \label{gravw } \end{aligned}\ ] ] in the equation above we have used the exponential form potential for the canonically normalised @xmath107-singlet which is the correct description at large field values . at the same time , to simplify the formulae , in the terms involving portal interactions between the singlet and other scalars , we used the asymptotic values for @xmath131 as in , emphasising the emerging large mass terms for @xmath80 and @xmath6 , but neglecting exponentially suppressed corrections involving @xmath107 . these are small exponentially suppressed effects which are not going to modify the analysis of inflation . the inflaton is the @xmath107-field and the inflation is generated by the self - interacting part of the potential in , @xmath132\right)^2\ , .\ ] ] it is exponentially flat for large field values and is suitable for the slow - roll inflation @xcite . the slow - roll inflation parameter is @xmath133 inflation ends when @xmath134 which corresponds to @xmath135 or @xmath136 . inflation starts at the singlet field value @xcite @xmath137 . the cmb normalisation condition , @xmath138 which determines the non - minimal singlet coupling to gravity @xcite , @xmath139 the spectral index and the tensor - to - scalar perturbation ratios in this model are the same as computed in the bezrukov - shaposhnikov higgs - inflation model @xcite . they are found to be in excellent agreement @xcite with the latest planck measurements , @xcite . our realisation of inflation in is a one - field slow - roll inflation model . the singlet field @xmath107 plays the role of the inflaton , while the other degrees of freedom , such as the higgs and the cw scalar at the inflation scale being much heavier than the hubble expansion parameter during inflation , @xmath140 @xmath141 the main technical advantage of the singlet field realisation of inflation presented here , compared to the original higgs inflation model @xcite and the subsequent implementations with other multi - component non - minimally coupled scalars , as in @xcite , is that our model does not require introduction of new physics effects associated with the ` low ' @xmath142 and ` intermediate ' scale @xmath143 . indeed , the appearance of the low - scale dimension-6 operator in was rendered harmless by the field redefinition of the one - component field @xmath49 . at the same time we should note that the field redefinition - when applied in the regime of _ small _ field values , @xmath144 , does introduce higher - dimensional corrections in the einstein - frame potential for the singlet , _ cf _ , @xmath145 if one considers multi - particle ( 2-to-4 and higher ) scattering processes , this appears to break unitarity at scales @xmath146 in the perturbative regime of the theory when expanding at small values of @xmath107 . these scattering processes are relevant for the theory in the ` collider regime ' i.e. in the regime where we are using perturbation theory around vanishing background field values . however in the collider regime the theory is safe : scales @xmath147 are of course inaccessibly high for any collider experiments . on the other hand , in order to probe the theory around the scale of inflation , the more appropriate approach is to expand around a non - trivial background . in this case , as argued in ref . @xcite , the effective uv - cutoff scale of the theory itself depends on the background field values ( i.e. in our case it depends on @xmath148 or @xmath149 ) and rises with these values appropriately . the ( slightly simplified ) results of the analysis in @xcite imply for the effective uv cut - off of theory , @xmath150 for @xmath151 ; @xmath152 for @xmath153 , and finally , @xmath154 for @xmath155 in the inflationary regime . we conclude that the effective potential of the theory is likely to remain safe at all relevant energy scales ( below the planck scale ) when evaluated at the appropriate background field value ( see ref . @xcite and references therein for the detailed analysis of quantum corrections ) . we also note that the fact that all the terms for the non - singlet scalars , @xmath80 and @xmath6 , are already canonically normalised in the lagrangian implies that there are no non - renormalisable interactions in the theory involving the sub - planckian scales . indeed , the only terms we dropped in were suppressed by the high scale @xmath56 . this is different from the case of the higgs - driven inflation where the gauge and fermion interactions of the higgs are modified and have to be cut - off at the intermediate scale @xmath143 , as pointed out in @xcite . it is important to note that if the new physics was required to be included at the scales below the uv cutoff @xmath56 , this would have destroyed the classical scale invariance of our model and induced large threshold contributions to the masses of the higgs and the cw scalar , reintroducing the fine - tinning problem into the sm . for this reason the elegant construction of ref . @xcite which integrates in a massive linear sigma - model field at the scale @xmath142 to restore the perturbative unitarity of the higgs inflation model does not work in our classically scale - invariant case . finally , we would like to briefly comment on the role of loop - induced quantum corrections to the effective potential . as explained in the introduction , our calculational approach is based on the use of dimensional regularisation where no power - like divergencies can appear . at the 1-loop level we have the coleman - weinberg corrected potential , @xmath156 which entails the derivatives of the exponentially slowly varying potential in the background of a large @xmath107-field . higher loop orders will involve higher derivatives of the flat potential enhanced by only a logarithmic dependence on the uv - cutoff . as a result , it is quite likely that quantum corrections computed in the dimensional regularisation scheme in the background perturbation theory around a large @xmath107-field will be small and not affect the picture of inflation at the leading order , see also @xcite for a more comprehensive discussion of this point . singlet scalar dark matter ( dm ) models @xcite are the simplest possible uv - complete models of dark matter containing a weakly interacting massive particle ( wimp ) in this case a scalar interacting with the higgs via the portal interaction . in the classically scale - invariant sm@xmath13u(1)@xmath14 theory with the real singlet @xmath49 and the scalar potential , the stability of the singlet is protected by a @xmath50 symmetry , @xmath157 giving a natural dark matter candidate . indeed , the @xmath50 symmetry of the potential is an automatic consequence of the renormalisability ( dimension 4 ) , scale - invariance and gauge invariance ( which does not allow odd powers of @xmath80 and @xmath6 ) of the theory . as shown in the previous section , at large field values , @xmath158 , the singlet ( or more precisely its log in ) plays the role of the inflaton field which during the inflation slowly rolls in an exponentially flat potential . in this regime , the inflaton gives large masses to the sm fields and otherwise decouples from the rest of the theory . after inflation is completed , the singlet enters the regime @xmath159 where it is canonically normalised , its potential is no longer flat and given by , and the large masses of the sm fields are no longer there . the singlet now assumes the role of the dark matter in the classically scale - invariant theory . there are only two phenomenologically relevant parameters of this singlet dark matter model , the coupling @xmath67 of the dark matter particle to the sm higgs , and the induced mass of the singlet , @xmath160 in eq . which gives the dark matter mass . expressed in terms of these two parameters , the dm phenomenology of this classically scale - invariant model is the same as that of the usual singlet scalar dm model . papers @xcite contain the recent comprehensive studies of the singlet scalar dm , see also references therein . here we give only a very brief summary of their results . the allowed / excluded portions of the @xmath67 , @xmath161 parameter space are shown on figure [ fig : dm ] taken from ref . @xcite . the relic density of singlet scalar dark matter , @xmath162 , should not exceed the full dm relic density @xmath163 the dm relic density is determined by the ( in-)efficiency of the dm particles annihilating into the sm degrees of freedom . these processes are dominated by the higgs - mediated s - channel . essentially the constraint @xmath164 amounts to a lower bound on the coupling @xmath67 . the excluded region is the lower portion of the parameter space shown in dark grey on both plots of fig . [ fig : dm ] . \2 . indirect detection of dark matter constraints arise from annihilation of dm to sm particles . for singlet scalar dm these constraints work much in the same way as the relic density constraint above , and for @xmath165 provide a slight improvement on the relic density contours . [ cols="^,^ " , ] -.4 cm \3 . when @xmath166 the dm can be produced via invisible higgs decays at the lhc with the higgs partial decay widths into the singlet dm particles given by @xmath167 the upper bounds on the invisible higgs decays from the lhc give the exclusion contour shown in red on the upper - left corner of the left plot of fig . [ fig : dm ] which zooms into the singlet mass region between 45 and 70 gev . not surprisingly , the bounds from invisible higgs decays and from the dm relic density work in the opposite way , with the former excluding larger values of the coupling @xmath67 in the kinematically allowed phase space . \4 . direct detection experiments provide upper bounds on the dm - nucleon cross - sections . here dm interacts elastically with nuclei through the higgs boson exchange . the resulting nuclear recoil is then interpreted in terms of the dm mass and dm - nucleon cross section . the spin - independent cross - section is @xmath168 where @xmath169 parameterises the higgs - nucleon coupling . these bounds are translated into exclusion contours on the ( @xmath67 , @xmath161 ) plane , which essentially are lower bounds on the scalar dm mass @xmath161 . constraints from the current xenon100 direct detection experiment @xcite are shown in dark blue on the left of second plot in fig . [ fig : dm ] together with the future reach of xenon upgrades in lighter shades of blue , as computed in ref . @xcite . to summarise , the singlet scalar dm model is perfectly viable at present on a large portion of its parameter space for @xmath170 . at lower values of @xmath16 in the region between 55 gev and @xmath171 , there is only a small triangle ( in white ) left unexplored , as can be seen from the plot on the left . more generally , the xenon100 upgrade and the xenon1 t direct detection experiment should be able to probe the entire parameter space of the singlet dm model @xcite . in this paper we considered a classically scale - invariant sm theory extended with the coleman - weinberg sector and an additional real singlet field , both coupled to the sm via the higgs portal interactions . the singlet classically massless field @xmath49 was shown to give rise to the inflaton and to the dark matter candidate . we have presented for the first time an implementation of the slow - roll inflation mechanism in a bsm theory with classical scale invariance . our model is a single - real - field inflationary model which does not suffer from a breakdown of unitarity at scales below or comparable to the scale of inflation . the model maintains its classical scale invariance and is self - consistent all the way up to the planck - scale uv cutoff . the sm higgs potential is stabilised by the higgs portal interactions with the coleman - weinberg scalar . the higgs mass and the scale of the electroweak symmetry breaking are unaffected by the inflaton and its interactions . furthermore , the singlet field responsible for inflation also gives a viable scalar dark matter candidate in our model . when discussing cosmological inflation in section 2 , our classically scale invariant quantum field theory was coupled to a gravitational background with an explicit dependence on the reduced planck scale @xmath56 . how does this fit with classical scale invariance of the non - gravitational theory ? the planck scale @xmath56 , as it appears in our effective lagrangians and does not cause problems to the effective field theory . at the scale @xmath56 and above , the theory coupled to gravity of course becomes non - renormalisable , but @xmath56 is above the uv cutoff scale of our theory and gravity is treated as an external background . one can continue using dimensional regularisation in the curved background for all qft processes . the interesting for us phenomena , such as the cosmological inflation , happen at large values of the singlet field set by the intermediate scale @xmath172 , at this scale the theory is robust , the scalar inflaton @xmath99 is exponentially decoupled from the sm degrees of freedom ( in the relevant for inflation regime of large field values ) and there are no heavy degrees of freedom coupled to the sm sector present in the theory . still it is desirable to not have an explicit scale @xmath56 present in the theory which is supposed to become classically scale - less when gravity decouples . one can do this and embed the classical scale - invariance of the theory in the ir into the scale invariance as a fundamental symmetry of nature - including gravity - with the planck scale being set by the vev of the dilaton . we stress first that this is not a necessary requirement , the classical scale invariance of the ir theory can in principle exist even if the uv - complete theory is not scale - invariant . secondly , as we will explain below , the classical scale invariance of the effective qft does not automatically follow from the ( broken ) scale invariance of the gravitational theory . it has to be seen as an additional constraint . to proceed , we impose the scale - invariance on the full theory including gravity and assume that this is a _ local _ symmetry which is non - linearly realised i.e. spontaneously broken by the dilaton field with the vev @xmath173 . very recently there has been a resurgence of interest in such an approach to gravity and cosmology , see refs . @xcite and also @xcite and references to earlier work therein for other related approaches which assume or utilise full quantum scale invariance of the underlying theory . thus our original lagrangian can be written in a manifestly locally scale - invariant ( weyl - invariant ) form if we retrofit the planck scale in terms of the dilaton vev , @xmath173 and introduce the kinetic term for the dilaton ( see section iii of @xcite for a review of a general construction ) , @xmath174 where the dots indicate omitted kinetic terms of other fields , and @xmath22 is the scalar potential which must scale as the 4th power of the scalars , which our potential does ( see the discussion above ) . the dilaton can be expressed as , @xmath175 where @xmath176 transforms linearly under local scale transformations . note that the dilaton kinetic term in has the ` wrong ' sign , but this is not a problem @xcite , since the dilaton is not a physical degree of freedom - it can be gauged away . ( this also means that there are no additional massless or light degrees of freedom arising from the dilaton , thus no complications for the sm . ) in the unitary gauge we have @xmath177 the dilaton spontaneously breaks local scale invariance of the gravitational theory at the scale @xmath56 , and gives the mass gap @xmath178 to heavy gravitational states . in order to maintain the classical scale invariance of the sm theory , the graviton should not couple to any of the sm states in the potential @xmath22 in . this should be seen as an additional requirement ( rather than an automatic consequence ) to the local scale invariance of the microscopic theory . when weyl transforming to the einstein frame , as in section 2 , we now promote @xmath56 to the dilaton field and maintain the manifest local scale invariance , or gauge fix to the unitary gauge and keep @xmath56 as in section 2 . quantum effects are computed in dimensional regularisation in the manifestly scale - invariant way . in the flat background and without scale invariance , the dimensionless coupling constants acquire the dependence on the rg scale @xmath179 in @xmath180 dimensions , @xmath181 . in the gravitational background , one treats the metric tensor @xmath182 as a @xmath183 matrix , and in the einstein frame , it introduces the dependence on the dilaton . the result is that in our theory with spontaneously broken ( or non - linearly realised scale - invariance ) , the every appearance of the rg scale @xmath179 in the effective potential is now substituted by , see ref . @xcite , @xmath184 which now transforms under scale - transformations . this gives a manifestly scale - invariant formulation in quantum theory regularised in dimensional regularisation . the message we take from the above discussion is that the classical scale invariance of the effective sm theory can be compatible with , and form a natural part of the full local scale invariance of the full theory . this , however , is not automatic and requires an additional constraint that the dilaton vev @xmath185 does not appear in the sm potential , though @xmath186 can appear in the logarithms . after the gravity is decoupled , together with the heavy states whose masses are set by the dilaton mss gap , the sm @xmath13 cw @xmath13 @xmath60 theory in the ir is classically scale invariant . this classical scale invariance has its own pseudo - dilaton the cw field @xmath6 , which is a physical degree of freedom and has nothing to do with the heavy dilaton of the gravitational theory . we acknowledge the hospitality of the aspen center for physics and thank participants for useful discussions . we also thank christoph englert , joerg jaeckel , gunnar ro and michael spannowsky for exchange of ideas and collaboration on related topics . this material is based upon work supported in part by stfc through the ippp grant st / g000905/1 , in part by the wolfson foundation and royal society and in part by the national science foundation grant no . phy-1066293 . the simplest theory where the cw mechanism @xcite is realised is the massless scalar qed with the classical potential @xmath187 the field @xmath6 is the coleman - weinberg complex scalar , it is charged under the u(1)@xmath14 gauge symmetry with the usual covariant derivative coupling , @xmath188 . normalisation conditions for @xmath6 are chosen to be the same as in @xcite , @xmath189 , with canonical kinetic terms for its real and imaginary part components @xmath190 . this is related to the canonically normalised complex scalar @xmath191 via a simple rescaling , @xmath192 . the 1-loop corrected scalar potential is given by the usual coleman - weinberg expression @xcite , @xmath193\,,\ ] ] where @xmath179 is the rg scale . the effective potential above is computed in the massless theory , in the uv subtraction scheme is chosen to set the renormalised mass at the origin of the field space to zero , @xmath194 we use dimensional regularisation which preserves classical scale - invariance of the theory . in dimensional regularisation the equation is satisfied automatically and is independent of the rg scale @xmath179 . the minimum of the coleman - weinberg effective potential occurs at @xmath71 where the first derivative of vanishes . validity of weakly - coupled perturbative approach requires that @xmath195 in the minimum ( rather than the more usual relation @xmath196 ) where we defined @xmath197 . this is because the minimum of the effective potential is determined @xcite by ballancing the tree - level contribution @xmath198 against the 1-loop term @xmath199 . thus one can neglect in the @xmath200 1-loop contribution on the right hand side of which is subleading relative to the @xmath201 term . choosing the value of the rg scale @xmath202 , the minimum of the 1-loop corrected effective potential occurs at @xcite @xmath203 and the vev @xmath10 is determined by the condition on the couplings renormalised at the scale of the vev , @xmath204 an rg point can always be found ( in the right theory ) where this relation between the two couplings holds . in the cw settings the beta function for @xmath205 is positive , @xmath206 thus starting in the uv at a positive initial value of @xmath62 and following the rg running to the ir , one reaches the critical value of the rg scale where @xmath205 crosses zero . before this happens , at @xmath207 and positive , the @xmath205 trajectory necessarily crosses the @xmath208 trajectory . at the rg point where the constraint is satisfied , the vacuum is reached , the condensate @xmath31 develops and the coupling freezes . since the couplings run only logarithmically , the vev fixed by the condition depends exponentially on the coupling constants . in fact , in weakly coupled perturbation theory the vev is naturally generated at the scale which is exponentially smaller than the uv cutoff @xmath209 . by solving the leading - order rg - running equation for the coupling @xmath12 , @xmath210 we find @xmath211 \,\ll m_{\sst uv}\,.\ ] ] the cw gauge group is spontaneously broken by @xmath31 and the gauge boson acquires the mass , @xmath212 . the mass of the cw scalar is follows from expanding to the second order in @xmath213 , @xmath214 it is parametrically ( 1-loop ) suppressed relative to the corresponding vector boson mass , @xmath215 . to adapt the expressions above to a more general cw theory , e.g. the b - l theory where the cw field @xmath6 has charge @xmath216 under the u(1)@xmath14 gauge group , and other matter fields are coupled to the latter with their @xmath32 charges , or other models , we can re - write , and as @xmath217\right)\,,\ ] ] @xmath218 @xmath219 note that eq . is manifestly rg - invariant but it also incorporates the finite contribution @xmath220 , as in . the last two equations take these finite contributions into account , e.g. originates from @xmath221 and both , the log - dependent and the finite effects are important . generalising now this simple model to the sm @xmath13 u(1)@xmath14 theory , the classical scalar potential in conventions of @xcite reads , @xmath222 here @xmath80 is the usual sm higgs complex doublet , which in the unitary gauge takes the form @xmath223 . the field cw field @xmath6 is normalised as before , and the higgs is canonically normalised , @xmath224 . all scalars have vanishing renormalised masses , @xmath225 once these conditions are enforced at one scale , they hold @xcite at all rg scales in dimensional regularisation maintaining classical scale invariance of the model . the value of @xmath31 is given by and it induces the higgs vev required to minimise the potential such that @xmath228 and triggers the electroweak symmetry breaking . the higgs self - coupling and the portal coupling responsible for the emergence of ewsb are then determined from @xmath10 through @xcite , @xmath229 where we have set @xmath19 and @xmath18 . of course one can argue that the theory under consideration ( or for this matter any theory which does not include quantum gravity or in addition has landau poles in the non - asymptotically free couplings , e.g. the hyper charge , even if these landau poles occur at above the planck scale ) must have a uv cutoff above which it breaks down . in this case the unknown more microscopic theory above the cutoff , could ( or should ( ? ) ) contain heavy @xmath230 degrees of freedom which after being integrated out produce large threshold contributions to the higgs mass . the quadratic cutoff contributions to the higgs mass in non - dimensional regularisation schemes is an effective realisation of these physical effects and implies that the masslessness requirement is the fine - tuning . the vanishing masses in in this reading correspond to an exact cancellation of all the @xmath231 terms between the bare masses terms and the counterterms . one can also take the view that the unknown microscopic theory might after all not generate such large corrections to the higgs in order to preserve the classical scale invariance of its sub - planckian effective theory . the more practical outcome of this discussion is that from the perspective of our classically scale invariant effective theory which should be valid below and up to the planck scale , there are no cancellations between large scales in dimensional regularisation of this theory and no associated problems or uncertainties . technically the sm@xmath13cw theory is natural and there is no fine - tuning . the computational approach we take is to first define the maximal scale above which the theory should not be used , effectively the planck scale , and in this theory regulate all integrals using the analytic continuation prescribed by the dimensional regularisation . with that , the maximal uv scale becomes irrelevant , if one wishes it can be safely taken to infinity , the integrals are convergent in the uv . no powers of the cutoff scale appear in the integrals , the poles in epsilon are subtracted in the usual way in the ms or @xmath232 scheme and only the logarithms of the rg scale remain . g. aad _ et al . _ [ atlas collaboration ] , `` observation of a new particle in the search for the standard model higgs boson with the atlas detector at the lhc , '' phys . b * 716 * ( 2012 ) 1 , arxiv:1207.7214 [ hep - ex ] . s. chatrchyan _ et al . _ [ cms collaboration ] , `` observation of a new boson at a mass of 125 gev with the cms experiment at the lhc , '' phys . b * 716 * ( 2012 ) 30 , arxiv:1207.7235 [ hep - ex ] . s. r. coleman and e. j. weinberg , `` radiative corrections as the origin of spontaneous symmetry breaking , '' phys . d * 7 * ( 1973 ) 1888 . r. hempfling , `` the next - to - minimal coleman - weinberg model , '' phys . b * 379 * ( 1996 ) 153 , hep - ph/9604278 . k. a. meissner and h. nicolai , `` conformal symmetry and the standard model , '' phys . b * 648 * ( 2007 ) 312 , hep - th/0612165 . w. a. bardeen , `` on naturalness in the standard model , '' fermilab - conf-95 - 391-t . c. englert , j. jaeckel , v. v. khoze and m. spannowsky , `` emergence of the electroweak scale through the higgs portal , '' jhep * 1304 * ( 2013 ) 060 , arxiv:1301.4224 [ hep - ph ] . w. -f . chang , j. n. ng and j. m. s. wu , `` shadow higgs from a scale - invariant hidden u(1)(s ) model , '' phys . d * 75 * ( 2007 ) 115016 , hep - ph/0701254 . r. foot , a. kobakhidze and r. r. volkas , `` electroweak higgs as a pseudo - goldstone boson of broken scale invariance , '' phys . b * 655 * ( 2007 ) 156 , arxiv:0704.1165 [ hep - ph ] . r. foot , a. kobakhidze , k. l. mcdonald and r. r. volkas , `` a solution to the hierarchy problem from an almost decoupled hidden sector within a classically scale invariant theory , '' phys . d * 77 * ( 2008 ) 035006 arxiv:0709.2750 [ hep - ph ] . s. iso , n. okada and y. orikasa , `` classically conformal @xmath233 l extended standard model , '' phys . b * 676 * ( 2009 ) 81 , arxiv:0902.4050 [ hep - ph ] . l. alexander - nunneley and a. pilaftsis , `` the minimal scale invariant extension of the standard model , '' jhep * 1009 * ( 2010 ) 021 [ arxiv:1006.5916 [ hep - ph ] ] . s. iso and y. orikasa , `` tev scale b - l model with a flat higgs potential at the planck scale - in view of the hierarchy problem - , '' arxiv:1210.2848 [ hep - ph ] . e. j. chun , s. jung and h. m. lee , `` radiative generation of the higgs potential , '' phys . b * 725 * ( 2013 ) 158 , arxiv:1304.5815 [ hep - ph ] . m. heikinheimo , a. racioppi , m. raidal , c. spethmann and k. tuominen , `` physical naturalness and dynamical breaking of classical scale invariance , '' arxiv:1304.7006 [ hep - ph ] . t. hambye and a. strumia , `` dynamical generation of the weak and dark matter scale , '' arxiv:1306.2329 [ hep - ph ] . v. v. khoze and g. ro , `` leptogenesis and neutrino oscillations in the classically conformal standard model with the higgs portal , '' arxiv:1307.3764 [ hep - ph ] . c. d. carone and r. ramos , `` classical scale - invariance , the electroweak scale and vector dark matter , '' arxiv:1307.8428 [ hep - ph ] . r. n. mohapatra and r. e. marshak , `` local b - l symmetry of electroweak interactions , majorana neutrinos and neutrono oscillations , '' phys . * 44 * ( 1980 ) 1316 [ erratum - ibid . * 44 * ( 1980 ) 1643 ] ; + r. e. marshak and r. n. mohapatra , `` quark - lepton symmetry and b - l as the u(1 ) generator of the electroweak symmetry group , '' phys . b * 91 * , 222 ( 1980 ) ; + c. wetterich , `` neutrino masses and the scale of b - l violation , '' nucl . b * 187 * ( 1981 ) 343 . m. fukugita and t. yanagida , `` baryogenesis without grand unification , '' phys . b * 174 * ( 1986 ) 45 . e. k. akhmedov , v. a. rubakov and a. y. .smirnov , `` baryogenesis via neutrino oscillations , '' phys . * 81 * ( 1998 ) 1359 , hep - ph/9803255 . t. asaka and m. shaposhnikov , `` the numsm , dark matter and baryon asymmetry of the universe , '' phys . b * 620 * ( 2005 ) 17 , hep - ph/0505013 . m. drewes and b. garbrecht , `` leptogenesis from a gev seesaw without mass degeneracy , '' jhep * 1303 * ( 2013 ) 096 , arxiv:1206.5537 [ hep - ph ] . g. degrassi , s. di vita , j. elias - miro , j. r. espinosa , g. f. giudice , g. isidori and a. strumia , `` higgs mass and vacuum stability in the standard model at nnlo , '' jhep * 1208 * ( 2012 ) 098 arxiv:1205.6497 [ hep - ph ] . d. buttazzo , g. degrassi , p. p. giardino , g. f. giudice , f. sala , a. salvio and a. strumia , `` investigating the near - criticality of the higgs boson , '' arxiv:1307.3536 [ hep - ph ] . o. lebedev , `` on stability of the electroweak vacuum and the higgs portal , '' eur . j. c * 72 * ( 2012 ) 2058 arxiv:1203.0156 [ hep - ph ] . j. elias - miro , j. r. espinosa , g. f. giudice , h. m. lee and a. strumia , `` stabilization of the electroweak vacuum by a scalar threshold effect , '' jhep * 1206 * ( 2012 ) 031 , arxiv:1203.0237 [ hep - ph ] . a. h. guth , `` the inflationary universe : a possible solution to the horizon and flatness problems , '' phys . d * 23 * ( 1981 ) 347 ; + a. d. linde , `` a new inflationary universe scenario : a possible solution of the horizon , flatness , homogeneity , isotropy and primordial monopole problems , '' phys . b * 108 * ( 1982 ) 389 ; + a. albrecht and p. j. steinhardt , `` cosmology for grand unified theories with radiatively induced symmetry breaking , '' phys . * 48 * ( 1982 ) 1220 ; + a. d. linde , `` chaotic inflation , '' phys . b * 129 * ( 1983 ) 177 . p. a. r. ade _ et al . _ [ planck collaboration ] , `` planck 2013 results . xxii . constraints on inflation , '' arxiv:1303.5082 [ astro-ph.co ] . d. s. salopek , j. r. bond and j. m. bardeen , `` designing density fluctuation spectra in inflation , '' phys . d * 40 * ( 1989 ) 1753 . f. l. bezrukov and m. shaposhnikov , `` the standard model higgs boson as the inflaton , '' phys . b * 659 * ( 2008 ) 703 arxiv:0710.3755 [ hep - th ] . j. l. f. barbon and j. r. espinosa , `` on the naturalness of higgs inflation , '' phys . d * 79 * ( 2009 ) 081302 arxiv:0903.0355 [ hep - ph ] . c. p. burgess , h. m. lee and m. trott , `` power - counting and the validity of the classical approximation during inflation , '' jhep * 0909 * ( 2009 ) 103 arxiv:0902.4465 [ hep - ph ] . r. n. lerner and j. mcdonald , `` higgs inflation and naturalness , '' jcap * 1004 * ( 2010 ) 015 arxiv:0912.5463 [ hep - ph ] . c. p. burgess , h. m. lee and m. trott , `` comment on higgs inflation and naturalness , '' jhep * 1007 * ( 2010 ) 007 arxiv:1002.2730 [ hep - ph ] . m. p. hertzberg , `` on inflation with non - minimal coupling , '' jhep * 1011 * ( 2010 ) 023 arxiv:1002.2995 [ hep - ph ] . f. bezrukov , a. magnin , m. shaposhnikov and s. sibiryakov , `` higgs inflation : consistency and generalisations , '' jhep * 1101 * ( 2011 ) 016 arxiv:1008.5157 [ hep - ph ] . g. f. giudice and h. m. lee , `` unitarizing higgs inflation , '' phys . b * 694 * ( 2011 ) 294 arxiv:1010.1417 [ hep - ph ] . o. lebedev and h. m. lee , `` higgs portal inflation , '' eur . j. c * 71 * ( 2011 ) 1821 arxiv:1105.2284 [ hep - ph ] . m. shaposhnikov and i. tkachev , `` the numsm , inflation , and dark matter , '' phys . b * 639 * ( 2006 ) 414 [ hep - ph/0604236 ] . t. e. clark , b. liu , s. t. love and t. ter veldhuis , `` the standard model higgs boson - inflaton and dark matter , '' phys . d * 80 * ( 2009 ) 075019 [ arxiv:0906.5595 [ hep - ph ] ] . r. n. lerner and j. mcdonald , `` gauge singlet scalar as inflaton and thermal relic dark matter , '' phys . d * 80 * ( 2009 ) 123507 [ arxiv:0909.0520 [ hep - ph ] ] . f. bezrukov and d. gorbunov , `` light inflaton hunter s guide , '' jhep * 1005 * ( 2010 ) 010 [ arxiv:0912.0390 [ hep - ph ] ] . r. n. lerner and j. mcdonald , `` distinguishing higgs inflation and its variants , '' phys . d * 83 * ( 2011 ) 123522 [ arxiv:1104.2468 [ hep - ph ] ] . f. bezrukov , `` the higgs field as an inflaton , '' arxiv:1307.0708 [ hep - ph ] . v. silveira and a. zee , `` scalar phantoms , '' phys . b * 161 * ( 1985 ) 136 ; + j. mcdonald , `` gauge singlet scalars as cold dark matter , '' phys . d * 50 * ( 1994 ) 3637 hep - ph/0702143 [ hep - ph ] ; + c. p. burgess , m. pospelov and t. ter veldhuis , `` the minimal model of nonbaryonic dark matter : a singlet scalar , '' nucl . b * 619 * ( 2001 ) 709 hep - ph/0011335 . y. mambrini , `` higgs searches and singlet scalar dark matter : combined constraints from xenon 100 and the lhc , '' phys . d * 84 * ( 2011 ) 115017 arxiv:1108.0671 [ hep - ph ] . c. englert , t. plehn , m. rauch , d. zerwas and p. m. zerwas , `` lhc : standard higgs and hidden higgs , '' phys . b * 707 * ( 2012 ) 512 arxiv:1112.3007 [ hep - ph ] . a. djouadi , o. lebedev , y. mambrini and j. quevillon , `` implications of lhc searches for higgs portal dark matter , '' phys . b * 709 * ( 2012 ) 65 arxiv:1112.3299 [ hep - ph ] . i. low , p. schwaller , g. shaughnessy and c. e. m. wagner , `` the dark side of the higgs boson , '' phys . d * 85 * ( 2012 ) 015009 arxiv:1110.4405 [ hep - ph ] . k. cheung , y. -l . s. tsai , p. -y . tseng , t. -c . yuan and a. zee , `` global study of the simplest scalar phantom dark matter model , '' jcap * 1210 * ( 2012 ) 042 arxiv:1207.4930 [ hep - ph ] . j. m. cline , k. kainulainen , p. scott and c. weniger , `` update on scalar singlet dark matter , '' arxiv:1306.4710 [ hep - ph ] . e. aprile _ et al . _ [ xenon100 collaboration ] , `` dark matter results from 225 live days of xenon100 data , '' phys . rev . * 109 * ( 2012 ) 181301 arxiv:1207.5988 [ astro-ph.co ] . r. kallosh and a. linde , `` superconformal generalization of the chaotic inflation model @xmath234 , '' jcap * 1306 * ( 2013 ) 027 arxiv:1306.3211 [ hep - th ] . i. bars , p. steinhardt and n. turok , `` local conformal symmetry in physics and cosmology , '' arxiv:1307.1848 [ hep - th ] . f. englert , c. truffin and r. gastmans , `` conformal invariance in quantum gravity , '' nucl . phys . b * 117 * ( 1976 ) 407 . m. shaposhnikov and d. zenhausern , `` quantum scale invariance , cosmological constant and hierarchy problem , '' phys . b * 671 * ( 2009 ) 162 [ arxiv:0809.3406 [ hep - th ] ] . j. garcia - bellido , j. rubio , m. shaposhnikov and d. zenhausern , `` higgs - dilaton cosmology : from the early to the late universe , '' phys . d * 84 * ( 2011 ) 123504 [ arxiv:1107.2163 [ hep - ph ] ] . r. sundrum , `` gravity s scalar cousin , '' hep - th/0312212 .
the coherence property plays an essential role in the wave nature of bose - einstein condensates ( becs ) , and it has been investigated intensively @xcite after the experimental realizations of becs in dilute gases @xcite . for a single bec which is a perfect quantum fluid at zero temperature , the existence of a macroscopic wave function ( or order parameter ) means that spatial coherence is an intriguing property of the condensate . in the dilute bose - condensed gases , the spatial coherence can be investigated directly by interfering two bose condensates . for two coherently separated becs , because the relative phase of two sub - condensates is locked , it is not surprising that there is a clear interference pattern when the two sub - condensates are allowed to overlap . in the celebrated experiment by andrews _ et al _ @xcite , however , high - contrast fringes were observed even for two completely independent condensates at an initial time . in this experiment , to prepare two initially independent condensates , the dilute bose gases were evaporatively cooled in a double - well potential created by splitting a magnetic trap in half with a far - off blue - detuned laser beam ( see also @xcite ) . in particular , the height of the external potential due to the laser beam is much larger than the chemical potential of two separated condensates , which means that the tunneling current can be safely omitted . after switching off the double - well potential , the two initially independent condensates overlapped and high - contrast fringes were observed in ref . this experimental result shows clearly that there is a spatial coherence property after the overlapping between two independent interacting condensates . although high - contrast fringes were observed for two initially independent condensates , in many literature ( see for example leggett , pethick , stringari and references therein ) , it is shown that there is no interference term in the density expectation value for two initially independent condensates . to solve the contradiction between this result and the observed high - contrast fringes @xcite for two initially independent condensates , the observed high - contrast fringes were interpreted with the aid of the high - order correlation function @xmath0 ( which is an oscillation function of @xmath1 ) and quantum measurement process in the present popular viewpoint . several theories have been proposed to interpret the observed high - contrast fringes for two initially independent condensates such as the stochastic simulations of the photon detection for atoms jav , the expansion of fock state by the linear superposition of coherent states @xcite , and the continuous measurement theory @xcite . in the present work , we calculate the density expectation value for two initially independent condensates by including carefully the interatomic interaction . quite different from the simple derivation in other theories , it is found that _ for the case of two initially independent condensates , upon expansion , there is a nonzero interference term in the density expectation value when the interatomic interaction and exchange symmetry of identical bosons are both taken into account . _ to exclude any possible error in the cumbersome derivations , we calculate the density expectation value and evolution equations by both the first - quantization and second - quantization methods , and the same results are obtained . after removing the double - well potential trapping the two initially independent condensates , we give the theoretical result of the density expectation value which agrees with the interference pattern observed in @xcite . to show the interaction - induced coherence process between two initially independent bose condensates when both the interatomic interaction and exchange symmetry of identical bosons are considered carefully , the paper is organized as follows . in sec . ii , we give the general expression of the density expectation value for two initially independent condensates based on the many - body wave function of the whole system . it is shown that there is a nonzero interference term when the wave functions of two initially independent condensates are no more orthogonal after their overlapping . in sec . iii , we give the general expression of the overall energy of the whole system , and the evolution equations are given by the action principle . in sec . iv , we give a brief proof why the wave functions of two initially independent condensates become non - orthogonal after their overlapping when the interatomic interaction is considered . in this section , an effective order parameter is introduced to show further the interaction - induced coherence formation process for two initially independent condensates . in sec . v , we give the numerical results for the evolution of the density expectation value according to the experimental parameters in ref . it is shown clearly that our theoretical results of the density expectation value agree with the experimental results of the interference patterns . in sec . vi , the evolution of the density expectation value is given for different coupling constants . it is shown that increasing the coupling constant has the effect of enhancing the interference effect for two independent condensates . to show further the physical mechanism of the interaction - induced coherence process , in sec . vii , we prove that the theoretical results based on the second - quantization method is the same as the results derived from the many - body wave function in the previous sections . in sec . viii , the role of quantum fluctations and the orthogonality of the many - body wave function are discussed . finally , we give a brief summary and discussion in the last section . first , we give a brief introduction to the scheme of observing the interference effect of two separated condensates . as shown in fig . 1 , the double - well potential can be created by superposing a far - off blue - detuned laser beam ( which generating a repulsive optical dipole force for atoms ) upon the magnetic trap . the combined double - well potential is@xmath2 + u_{0}e^{-\left ( x - x_{0}\right ) ^{2}/w_{x}^{2}-\left ( y^{2}+z^{2}\right ) /w_{\perp } ^{2 } } , \label{external potential}\]]where the first term is the trapping potential due to the magnetic trap , while the second term represents the potential due to the laser beam . for the bose gases confined in this double - well potential , there are two quite different cases . ( i ) if the height @xmath3 of the external potential due to the laser beam is smaller than the chemical potential of the system , the two sub - condensates can be regarded to be coherently separated after the evaporative cooling . in this situation , the two coherently - separated sub - condensates can be regarded as a single condensate . the relative phase of two sub - condensates is thus locked . ( ii ) if the height @xmath3 of the laser beam is much larger than the chemical potential of the system so that the tunneling effect can be omitted safely , the two condensates are completely independent after the evaporative cooling . for two coherently separated sub - condensates , every atom is described by the following normalization wave function @xmath4 /\sqrt{n},\]]where @xmath5 and @xmath6 are the normalization wave functions accounting for the two sub - condensates . @xmath7 is the total number of particles . before removing the double - well potential , the average particle numbers in each condensate are @xmath8 and @xmath9 , respectively . after removing the double - well potential , the evolution of the wave function @xmath10 can be obtained based on the gross - pitaevskii ( gp ) equation @xcite . the density expectation value is then @xmath11 + n_{2}\left\vert \phi _ { c2}\left ( \mathbf{r},t\right ) \right\vert ^{2}. \label{coherent - density}\end{aligned}\]]the second term in the above equation accounts for the interference effect when there is an overlapping between two sub - condensates upon expansion . for @xmath12 satisfying the gp equation , the nonlinear effects in the interference pattern of two coherently separated sub - condensates were investigated in refs @xcite . as shown in fig . 1(b ) , if the intensity of the blue - detuned laser beam is sufficiently high so that the tunneling effect can be omitted , the two condensates can be regarded to be completely independent . in this situation , the number of particles @xmath8 and @xmath9 in each of the two condensates are fixed . we assume that there are @xmath8 particles described by the wave function @xmath13 in the left well , while there are @xmath9 particles described by the wave function @xmath14 in the right well . to investigate clearly the role of the exchange symmetry of identical particles and interatomic interaction , we calculate in the following the density expectation value @xmath15 directly from the many - body wave function . in sec . vii , we will also calculate the density expectation value based on the second - quantization method , and the same result is obtained . we will prove in sec . iv that in the presence of the interatomic interaction , @xmath16 and @xmath17 will become non - orthogonal after the overlapping between two initially independent condensates . thus we consider here the following general case from the beginning@xmath18after removing the double - well potential , the two initially independent condensates will overlap , and thus one should consider the indistinguishability of identical particles . when the exchange symmetry of identical bosons is considered , the many - body wave function is@xmath19 , \label{wave - function}\end{aligned}\]]where @xmath20 denotes the @xmath21 permutations for the particles in different single - particle state @xmath13 or @xmath14 . @xmath22 is a normalization factor to assure @xmath23 . @xmath22 is determined by the following equation : @xmath24 ^{1/2}=1.\]]in this paper , to give a concise expression for various coefficients such as @xmath22 , we have introduced the rule @xmath25 . when the non - orthogonal property between @xmath13 and @xmath14 is considered , one should note that the normalization constant @xmath22 is relevant to the parameter @xmath26 . from the form of the many - body wave function ( [ wave - function ] ) , the quantum depletion originating from interparticle interaction is omitted . thus , this form of quantum state is valid when @xmath27 with @xmath28 and @xmath29 being respectively the scattering length and mean distance between particles . the role of quantum depletion will be discussed in sec . viii . from the above many - body wave function , after straightforward derivations , the exact expression of the density expectation value takes the following form : @xmath30 + c_{d}\left\vert \phi _ { 2}\left ( \mathbf{r},t\right ) \right\vert ^{2 } , \label{density}\end{aligned}\]]where the coefficients are@xmath31 in the above summations,@xmath32 for two independent ideal condensates , before the overlapping between the two condensates , we have @xmath33 . based on the schrdinger equation , it is easy to verify that after the double - well potential separating the condensates is removed , we have @xmath34 at any further time . thus @xmath35 , and the density expectation value is given by@xmath36 in this situation , the interference term is zero in the density expectation value . in the presence of the interatomic interaction , @xmath37 can be a nonzero value after the two initially independent condensates begin to overlap . for @xmath37 being nonzero , we see clearly from the coefficient @xmath38 that the interference term in ( [ density ] ) gives a contribution to the density expectation value . based on eqs . ( [ adi ] ) , ( [ bdi ] ) and ( [ cdi ] ) , we have@xmath39and@xmath40it is easy to understand that when @xmath41 and @xmath42 , @xmath43 can not be omitted , and thus there would be clear interference patterns . this shows clearly that for large particle number , a small @xmath44 can play important role in the density expectation value when @xmath45 and @xmath46 . generally speaking , even in the presence of the interatomic interaction , @xmath44 is much smaller than @xmath47 because @xmath48 is an oscillation function about the space coordinate . as shown above , however , a nonzero value of @xmath49 can give significant contribution to the density expectation value for large @xmath8 and @xmath9 . in fig . 2(a ) , we give the relation between @xmath50 and @xmath26 for @xmath51 based on eqs . ( [ bd ] ) and ( [ cd ] ) . the relation between @xmath50 and @xmath52 for @xmath53 is shown in fig . generally speaking , for @xmath54 and @xmath55 , one has @xmath56 . ) , ( [ bd ] ) and ( [ cd ] ) , shown in fig . 2(a ) is the relation between @xmath50 and @xmath57 , while shown in fig . 2(b ) is the relation between @xmath50 and @xmath58 for @xmath59 . it is shown clearly that the interference term can play an important role in the density expectation value for @xmath60 being larger than @xmath61 . ] for the many - body wave function @xmath62 , in the presence of the interatomic interaction , the evolution equation can be obtained based on the standard quantum mechanical principle . after the double - well potential is removed , the evolution equation is given by@xmath63where the hamiltonian of the whole system in the first - quantization method is@xmath64here the coupling constant @xmath65 with @xmath28 being the scattering length . in the above expression of the hamiltonian , we have used the well - known two - body pseudopotentials . it is well known that the action principle is quite useful to derive the gp equation for a single condensate ( see for example ref . similarly , we consider here the evolution of @xmath62 based on the action principle . to get the evolution equation based on the action principle , we first give the general expression of the overall energy of the whole system . after removing the double - well potential , the overall energy of the whole system is @xmath66after straightforward derivations , the exact expression of the overall energy is given by@xmath67where the kinetic energy @xmath68 is@xmath69 in addition , the interaction energy @xmath70 of the whole system is given by@xmath71 , \label{exact - inter}\end{aligned}\]]where the coefficients are given by@xmath72 by using the ordinary action principle and the energy of the whole system , one can get the following coupled evolution equations for @xmath13 and @xmath73 : @xmath74where @xmath75 and @xmath76 are functional derivatives . ii , we have shown that the nonzero interference term in the density expectation value origins from the assumption that @xmath37 can be a nonzero value after the overlapping between two initially independent condensates . in the last section , we have given the evolution equations about @xmath13 and @xmath14 . thus , an important question emerges naturally : is it physical for @xmath37 being nonzero with the development of time ? based on the evolution equations ( [ gp1 ] ) and ( [ gp2 ] ) given in the last section , one can understand easily that @xmath37 becomes nonzero after the overlapping between two initially independent condensates in the presence of the interatomic interaction . before removing the double - well potential , @xmath79 and @xmath80 because there is no overlapping between @xmath13 and @xmath14 . after removing the double - well potential and at the beginning of the overlapping between two condensates , @xmath81 and @xmath82 . for @xmath83 and @xmath84 , the overall energy can be approximated as@xmath85the above overall energy is obtained by setting @xmath86 in eq . ( s - overallenergy ) . based on eqs . ( [ gp1 ] ) and ( [ gp2 ] ) , the approximate evolution equations for @xmath16 and @xmath17 are then @xmath87and @xmath88where @xmath89and @xmath90 we see that @xmath91 is not equal to @xmath92 for @xmath93 . this leads to an important result that @xmath37 can be nonzero when there is an overlapping between two initially independent condensates . one can understand this result further through the following equation which determines the evolution of @xmath37 : @xmath94where @xmath95 is a nonzero function given by @xmath96 the above analyses show clearly that why @xmath37 becomes nonzero after the overlapping between two initially independent condensates for @xmath93 . obviously , for @xmath97 being zero , @xmath34 at any further time because @xmath98 . our numerical calculations in the following section also show that @xmath37 can be a nonzero value in the presence of the interatomic interaction . in the presence of the interatomic interaction , it is inconsistent to assume that @xmath37 is always zero with the development of time . if @xmath34 , the evolution equations eqs . ( [ gp1 ] ) and ( [ gp2 ] ) are exact . however , based on these two evolution equations , @xmath99 will increase from zero after the overlapping . as shown above , the evolution of @xmath13 and @xmath14 is determined by the expression of the overall energy and the evolution equations ( gp1 ) and ( [ gp2 ] ) . it seems that there is no exact solution for these evolution equations considering the fact that even there is no exact solution for three - dimensional nonlinear schrdinger equation . however , for @xmath100 and @xmath101 , we find that there is a quite simple evolution equation by introducing an effective parameter order . this would also contribute to our understanding of the interaction - induced coherence process for two initially independent condensates . for the cases of @xmath54 , @xmath102 and @xmath103 , first we introduce the following effective order parameter @xmath104 which is given by @xmath105based on this effective order parameter , the density expectation value can be approximated well as@xmath106 + c_{d}^{\prime } \left\vert \phi _ { 2}\left ( \mathbf{r},t\right ) \right\vert ^{2 } , \label{app - density}\end{aligned}\]]where@xmath107 to compare with the exact expression of the density expectation value given by eq . ( [ density ] ) , fig . 3 shows the ratio @xmath108 , @xmath109 and @xmath110 . we see that for @xmath111 , @xmath112 , the approximate density expectation value @xmath113 agrees very well with the exact expression of the density expectation value . , @xmath114 and @xmath115 . we see that for @xmath116 , @xmath117 and @xmath118 , the density expectation value given by the effective order parameter agrees well with the exact expression of the density expectation value given by eq . ( [ density ] ) . ] based on the effective order parameter , for @xmath111 , @xmath112 and @xmath119 , we also find that the overall energy of the whole system can be approximated very well as @xmath120where@xmath121and@xmath122 for the cases of @xmath54 , @xmath102 and @xmath103 , it is easy to verify that @xmath123 based on the analogous analyses about the density expectation value . for @xmath54 , @xmath124 and @xmath103 , one can also prove the result of @xmath125 . based on eq . ( eappr - inter ) , @xmath126 can be expanded as : @xmath127 , \end{aligned}\]]where @xmath128to compare with the exact expression of the interaction energy given by eq . ( [ exact - inter ] ) , fig . 4 shows the relation between @xmath129 ( @xmath130 ) and @xmath26 for @xmath51 . it is shown clearly that for @xmath54 and @xmath55 , @xmath131 , and thus @xmath132 . in this situation , the overall energy can be approximated well as @xmath133the evolution of the effective order parameter can be then obtained by using the action principle . based on this approximate energy , after removing the double - well potential , it is quite interesting to note that the evolution of the effective order parameter can be described very well by the ordinary gross - pitaevskii equation : @xmath134 ( @xmath130 ) and @xmath57 for @xmath51 . we see that for @xmath135 and @xmath136 , @xmath137 , which means that @xmath125 . ] based on eqs . ( [ app - density ] ) and ( [ s - app - equation ] ) , we see that the emergence of the effective order parameter @xmath138 gives us strong evidence that the coherence is formed in the interaction process between two initially independent condensates , and thus results in the emergence of high - contrast interference fringes . the effective order parameter and the approximate density expectation value suggest strongly that a full coherence is formed between two initially independent condensates for the cases of @xmath139 and @xmath55 . now we turn to give the theoretical results of the density expectation value according to the experimental parameters in ref . @xcite where clear interference patterns were observed for two initially independent condensates . in the experiment of ref . @xcite , @xmath140 condensed sodium atoms were confined in a magnetic trap with @xmath141 , @xmath142 . a blue - detuned laser beam of wavelength @xmath143 was focused into a light sheet with a cross section of @xmath144 @xmath145 by @xmath146 @xmath145 . the long axis of the laser beam was perpendicular to the long @xmath147axis of the condensate . for a laser power of @xmath148 , the barrier height is about @xmath149 @xmath150 , which is much larger than the chemical potential @xmath151 @xmath150 . for this laser beam , the two condensates can be regarded to be independent because they are well separated and the tunneling effect can be omitted . with these experimental parameters and the @xmath152wave scattering length @xmath153 , the initial profile of the two condensates is shown in fig . 5 . at the initial time , the overlapping between two condensates can be omitted safely , and thus @xmath33 . after the double - well potential is removed , the evolution of the density expectation value @xmath154 ( in unit of @xmath155 ) is given in fig . 5 through the numerical calculations of eqs . ( [ gp1 ] ) , ( [ gp2 ] ) and ( [ density ] ) . we see that there is a clear interference pattern in the density expectation value which agrees with the experimental result . shown in the inset is the evolution of @xmath156 for these parameters . we see that @xmath157 increases from zero with the development of time . for the expansion time of @xmath158 , the numerical result of @xmath156 shows that @xmath159 . to check further that the nonzero value of @xmath160 does not originate from numerical error , with the same initial conditions , we have verified in the numerical calculations that @xmath160 is always zero ( smaller than @xmath161 ) if the scattering length is assumed as zero . for the experimental parameters in ref . ] based on the evolution of the density expectation value shown in fig . 5 , we see that the overall width ( about @xmath162 @xmath145 ) of the ultracold gases in the @xmath147direction does not increase obviously . this is due to the fact that the initial density distribution is cigar - shaped , and thus the expansion in the @xmath147direction is very slow , while the expansion in @xmath163 and @xmath164 directions is much quick . in the experiment of ref . @xcite , one can see clearly that there is no obvious expansion in the long @xmath147axis for cigar - shaped condensate . for two initially independent condensates , when the double - well potential is switched off , one should note that in the regime close to @xmath165 @xmath145 shown in fig . 5 , the ultracold gases expand rapidly in the @xmath147direction because in this regime the ultracold gases have higher kinetic energy . thus , although the total width of the system in the @xmath147direction does not increase obviously , the rapid expansion in the central regime leads to the overlapping between two initially independent condensates , and results in the interference effect . the overall width of the system shown in fig . 5 is smaller than the experimental result of about @xmath166 @xmath145 @xcite . this difference may come from the expansion of thermal cloud in this experiment @xcite . after @xmath158 expansion , the numerical result in @xcite for two coherently separated condensates also showed that the overall width of the system is about @xmath162 @xmath145 . we see that the interatomic interaction plays an essential role in the emergence of the interference effect for two initially independent condensates . generally speaking , increasing the particle number will enhance the effect of the interference term in the density expectation value . based on eqs . ( [ gp1 ] ) and ( [ gp2 ] ) , increasing the coupling constant @xmath97has the effect of increasing @xmath37 . together with the relation between @xmath50 and @xmath26 illustrated in fig . 2(a ) , this shows that increasing the interatomic interaction will enhance the effect of the interference term . to show more clearly the interaction - induced coherence process between two initially independent condensates , in this section we consider the density expectation value for different coupling constants . we consider here the evolution of the density expectation value for one - dimensional case . at @xmath167 , to give a general comparison , the initial wave functions for two independent condensates are assumed to be identical for different coupling constants . the initial wave functions are respectively given by @xmath168 , \label{inital1 } \\ \phi _ { 2}\left ( x_{l},t=0\right ) & = & \frac{1}{\pi ^{1/4}\sqrt{\delta _ { 2}}}% \exp \left [ -\frac{\left ( x_{l}-x_{l2}\right ) ^{2}}{2\delta _ { 2}^{2}}\right ] . \label{inital2}\end{aligned}\]]in the above wave functions , we have introduced a dimensionless variable @xmath169 with @xmath170 being a length . in the present work , we assume that @xmath171 and @xmath172 . for these parameters , at @xmath167 , the two condensates are well separated . in the numerical calculations of the coupled equations given by eqs . ( [ gp1 ] ) and ( gp2 ) , it is useful to introduce the dimensionless variable @xmath173 with @xmath174 , and dimensionless coupling constant @xmath175 . in addition , the particle number is assumed as @xmath176 . in real experiments , interatomic interaction plays very important role in the initial ground - state wave function of the condensates . however , in principle , one can prepare the state given by eqs . ( [ inital1 ] ) and ( [ inital2 ] ) by adjusting the trapping potential for different coupling constants . in this section , the identical initial wave functions for different coupling constants would be helpful in the comparison of the density expectation value for different coupling constants . for @xmath177 , we consider the evolution of the density expectation value in free space . the evolution of @xmath13 and @xmath14 is obtained based on the numerical calculations of eqs . ( [ gp1 ] ) and ( [ gp2 ] ) . from @xmath178 and @xmath14 , we can get @xmath26 , and thus the density expectation value based on eq . ( [ density ] ) . shown in fig . 6 is the evolution of the density expectation value for different coupling constants . it is shown clearly that increasing the coupling constant has the effect of enhancing the coherence effect , and results in higher contrast in the interference patterns . ( in unit of @xmath179 ) for different coupling constants @xmath180 . shown in the inset of each figure is the relation between @xmath181 and dimensionless time @xmath182 . for two ideal condensates shown in fig . 6(a ) , we see that there is no interference pattern even there is an overlapping between two condensates . for the case of @xmath183 shown in fig . 6(b ) , we see that low - contrast interference patterns begin to emerge due to the interaction - induced coherence process . in fig . 6(c ) and fig . 6(d ) , we see that there are high - contrast interference patterns . in particular , in fig . 6(d ) for @xmath184 , two initially independent condensates can be regarded to be fully coherent near @xmath185 . ] the essential reason for the emergence of the interference term of two initially independent condensates lies in that because of the exchange symmetry of identical bosons and interatomic interaction , the two initially independent condensates become coherent after the overlapping between the two condensates . the physical mechanism of this interaction - induced coherence can be understood further based on the second quantization method . thus , in this section , we give the density expectation value and evolution equations based on the second - quantization method . the merit of the second - quantization method lies in that the indistinguishability of identical bosons is satisfied when the correct commutation relation of the field operator is used . for two initially independent condensates comprising particle number @xmath8 and @xmath9 , the corresponding quantum state is ( see for example refs . @xcite):@xmath186where @xmath187 is a normalization constant to assure @xmath188 . @xmath189 ( @xmath190 ) is a creation operator which creates a particle described by the single - particle state @xmath13 ( @xmath14 ) in the left ( right ) condensate . similarly to eq . ( [ wave - function ] ) , the quantum depletion is omitted in this sort of quantum state . for two initially independent condensates , for the state given by eq . ( initial - state ) , it seems that there is no interference term in the density expectation value with the following simple calculation : @xmath191 in the above equation , the field operator @xmath192 is expanded as @xmath193 with @xmath194 and @xmath195 being the annihilation operators . one should note that , to get ( [ nd - field(2 ) ] ) from ( nd - center ) , there is an implicit assumption that @xmath194 and @xmath190 are commutative . this holds when @xmath196 . when @xmath197 = 0 $ ] , it is easy to understand that the interference term ( the last term in eq . ( nd - center ) ) is zero in @xmath15 . as shown in the previous sections , @xmath13 and @xmath14 will become non - orthogonal in the presence of the interatomic interaction . the operators @xmath194 and @xmath195 can be written as@xmath198and@xmath199here @xmath192 is the field operator . by using the commutation relations of the field operators @xmath200=0 $ ] and @xmath201=\delta \left ( \mathbf{r}_{1}-\mathbf{r}% _ { 2}\right ) $ ] , it is easy to get the commutation relation@xmath202=\zeta ^{\ast } . \label{commre}\]]we see that @xmath194 and @xmath203 are not commutative any more for @xmath204 being a nonzero value . in this situation , it is obvious that one can not get the result ( nd - field(2 ) ) from ( [ nd - center ] ) any more . this means that one should be very careful to get the correct density expectation value for two initially independent condensates . it is well - known that the field operator should be expanded in terms of a complete and orthogonal basis set . generally speaking , the field operator @xmath205 can be expanded as : @xmath206where @xmath13 and @xmath207 are orthogonal normalization wave functions . assuming that @xmath208 , based on the conditions @xmath209 and @xmath210 , we have @xmath211 and @xmath212 @xmath213 . based on@xmath214we have@xmath215it is easy to get the following commutation relations : @xmath216 & = & [ \widehat{k}^{\dagger } , \widehat{k}% ^{\dagger } ] = 0,[\widehat{k},\widehat{k}^{\dagger } ] = 1 , \nonumber \\ \left [ \widehat{a}_{1},\widehat{a}_{1}\right ] & = & [ \widehat{a}_{1}^{\dagger } , % \widehat{a}_{1}^{\dagger } ] = 0,[\widehat{a}_{1},\widehat{a}_{1}^{\dagger } ] = 1 , \nonumber \\ \lbrack \widehat{k},\widehat{a}_{1 } ] & = & [ \widehat{k},\widehat{a}% _ { 1}^{\dagger } ] = 0 . \label{kcom}\end{aligned}\ ] ] because @xmath217 and @xmath218 are commutative , it is convenient to calculate the density expectation value @xmath219 by using the operators @xmath217 and @xmath220 . after straightforward derivations , the exact expression of the density expectation value is @xmath221 ^{n_{2}}\widehat{a}_{1}^{n_{1}}\left [ \widehat{a}% _ { 1}^{\dag } \phi _ { 1}^{\ast } + \widehat{k}^{\dag } \left ( \phi _ { 2}^{\prime } \right ) ^{\ast } \right ] \nonumber \\ & & \left [ \widehat{a}_{1}\phi _ { 1}+\widehat{k}\phi _ { 2}^{\prime } \right ] \left ( \widehat{a}_{1}^{\dagger } \right ) ^{n_{1}}\left [ \frac{\widehat{k}% ^{\dagger } } { \beta } + \zeta ^{\ast } \widehat{a}_{1}^{\dagger } \right ] ^{n_{2}}\left\vert 0\right\rangle \nonumber \\ & = & \alpha _ { d}\left\vert \phi _ { 1}\left ( \mathbf{r},t\right ) \right\vert ^{2}+2\beta _ { d}\times \mathrm{re}\left ( e^{i\varphi _ { c}}\phi _ { 1}^{\ast } \left ( \mathbf{r},t\right ) \phi _ { 2}\left ( \mathbf{r},t\right ) \right ) + \gamma _ { d}\left\vert \phi _ { 2}\left ( \mathbf{r},t\right ) \right\vert ^{2 } , \label{ndensity}\end{aligned}\]]where the coefficients are @xmath222 in addition , the normalization constant @xmath187 is determined by@xmath223 the above density expectation value is obtained based on the second quantization method . although it seems that the expressions of the coefficients given by eqs . ( [ ads ] ) , ( [ bds ] ) , ( [ cds ] ) and ( normconsts ) are quite different from the results calculated from the many - body wave function , we have proven that @xmath224 , @xmath225 , @xmath226 and @xmath227 . thus , the density expectation value given by eq . ( [ ndensity ] ) is the same as the result calculated from the many - body wave function @xmath62 which satisfies the exchange symmetry of identical bosons . based on the second quantization method , after removing the double - well potential , the hamiltonian in the second - quantization method is@xmath228after straightforward calculations , the overall energy of the whole system is @xmath229where the kinetic energy @xmath68 is given by@xmath230 in addition , the interaction energy @xmath70 of the whole system is given by@xmath231 , \end{aligned}\]]where the coefficients are given by@xmath232we have proven that the above overall energy is equal to the results based on the many - body wave function , by checking that @xmath233 for @xmath234 , @xmath235 , @xmath236 , @xmath237 . when the non - orthogonal property between @xmath13 and @xmath14 are considered , because the derivations of the density expectation value and overall energy are quite cumbersome , the same results based on the first - quantization method and second - quantization method give us strong evidence that our derivations are correct . in both the first - quantization and second - quantization methods of the previous calculations , the quantum depletion originating from the elementary excitations at zero temperature is omitted in the quantum state of the whole system . based on the bogoliubov theory of the elementary excitations , for the initial quantum state , the number of particles due to the quantum depletion is of the order of @xmath238 and thus the quantum depletion is negligible for bose condensate in dilute gases considered in the present work . with the development of time , the role of quantum depletion can also be omitted . because of the factor @xmath239 ( @xmath240 is the wave number of the elementary excitations ) in the wave function @xmath241 of the elementary excitations , a simple analysis shows qualitatively that @xmath242 and @xmath243 are of the order of @xmath244 with @xmath245 being the spatial size of the system . this exponential decay of @xmath246 and @xmath247 originates from the integral where there is spatially oscillating phase factor in the wave functions of the elementary excitations and condensates . thus , the contribution to the effective order parameter and density expectation value due to elementary excitations can be omitted safely . in the preceding paragraph , it is shown that in calculating the evolution of the density expectation value , the role of quantum depletion can be omitted . however , the quantum depletion plays important role in the consistency of our theory . here we discuss mainly the orthogonality for the quantum state of the whole system , especially about the subtle problem that whether the nonorthogonality between @xmath13 and @xmath14 violates the orthogonality of the whole system which must be satisfied . this sort of problem about orthogonality exists also for a single condensate . for a single condensate , if the quantum depletion is omitted , the many - body wave function is@xmath248based on the many - body schrdinger equation and the action principle , it is easy to get the following gp equation@xmath249assume that there are two orthogonally initial quantum states @xmath250 and @xmath251 . with the development of time , because of the nonlinearity of the gp equation , @xmath252 and @xmath253 may become non - orthogonal in the presence of interatomic interaction . it is obvious that @xmath254 ^{n}$ ] . although for large @xmath255 , this integral can be approximated as zero , based on the consideration of consistency , the physical mechanism of the exact orthogonality between @xmath256 and @xmath257 is an interesting problem . we think that the effect of quantum depletion omitted in eq . ( many - body - single ) is the essential reason for this sort of unphysical nonorthogonality . for the hamiltonian ( [ hamiltonian - first ] ) of the whole system , the term @xmath258 is non - factorable about the coordinate , and thus the exact solution of the many - body wave function is non - factorable too . for a single condensate , the exact many - body wave function can be assumed as@xmath259here @xmath260 accounts for the non - factorable component , and @xmath261 is a normalization constant . in the hard - sphere approximation , @xmath262 if @xmath263 for any @xmath264 ( @xmath265 is the hard - sphere radius ) , and @xmath266 for otherwise situation . @xmath267 represents the quantum depletion , and was successfully used to calculate the quantum depletion of superfluid liquid @xmath268 in ref . the omission of the quantum depletion means that we approximate @xmath260 as @xmath47 . for dilute bose condensed gases , @xmath269 . thus , the wave function ( [ many - body - single ] ) can describe well the condensate . when the orthogonality of the whole system is considered , we stress here that the wave function @xmath270 should be used . the non - factorable factor @xmath260 assures the orthogonality of the many - body quantum state , because @xmath270 is the exact solution of the many - body schrdinger equation . for dilute bose condensed gases , these analyses lead to two results : when the dynamic evolution of dilute bose condensed gases is considered , omitting the quantum depletion can give us quite good description ; when the consistency especially the orthogonality is considered , one should consider the role of quantum depletion . it is natural to generalize the above analyses to two initially independent condensates . the nonorthogonality between @xmath13 and @xmath14 does not mean in any sense the violation of the orthogonality of the quantum state of the whole system . we stress here again that when orthogonality is considered , we should check whether the quantum state of the whole system satisfies the orthogonality . when the quantum depletion is considered , the many - body wave function takes the following form@xmath271here @xmath62 is given by eq . ( [ wave - function ] ) . @xmath272 satisfies exactly the many - body schrdinger equation , and thus this sort of wave function satisfies the orthogonality condition . assume that @xmath273 is constructed by @xmath274 and @xmath275 , while @xmath276 is constructed by @xmath277 and @xmath278 . assume further that @xmath274 , @xmath279 , @xmath280 and @xmath278 are orthogonal to each other at the initial time , so that @xmath273 and @xmath276 are orthogonal at the initial time . omitting the quantum depletion and roughly speaking , the integral between @xmath281 and @xmath282 is of the order of @xmath283 ^{\alpha } \left [ \int \left ( \phi _ { 1}^{a}\right ) ^{\ast } \phi _ { 2}^{b}d\mathbf{r}\right ] ^{\beta } \left [ \int \left ( \phi _ { 2}^{a}\right ) ^{\ast } \phi _ { 1}^{b}d% \mathbf{r}\right ] ^{\gamma } \left [ \int \left ( \phi _ { 2}^{a}\right ) ^{\ast } \phi _ { 2}^{b}d\mathbf{r}\right ] ^{\delta } $ ] ( here @xmath284 ) . we see that similarly to a single condensate , the orthogonality between @xmath281 and @xmath282 is quite good for large @xmath255 when the quantum depletion is omitted . but to assure the exact orthogonality of the whole quantum state , we must consider the role of quantum depletion . although we do not provide here a fully rigorous proof , these analyses lead to the result that : although for dilute bose condensed gases , the omission of the quantum depletion will not play important contribution to the evolution of @xmath285 and @xmath14 , it s the quantum depletion that assure the orthogonality of the whole quantum state . up to date , we mainly use the bases @xmath13 and @xmath14 to study the dynamic evolution and density expectation value . considering the nonorthogonality between @xmath13 and @xmath14 in the presence of interparticle interaction , it is attractive to discuss the physical picture using the orthogonal bases @xmath13 and @xmath207 introduced in sec . vii . if we use the orthogonal bases @xmath13 and @xmath286 , the quantum state of the whole system is @xmath287 . we see that the number of particles in the orthogonal modes @xmath13 and @xmath207 are no more definite . this quantum state becomes a superposition of different number of particles in the orthogonal modes @xmath13 and @xmath207 . this result is very natural because the coupling ( interaction ) between two initially independent condensates leads to the coherent transfer of particles between two condensates . in ref . @xcite , it was shown that for a single condensate , a coherent state description of the bose condensed system is a robust state in the presence of the interactions between the condensate and its environment . in the present work , our research shows that the interaction between two condensates makes each condensate become a coherent superposition of different particle number . the results in ref . barnett about the robustness of the coherent state imply that @xmath288 is a robust quantum state even when the coupling with the environment exists . in summary , upon expansion , we calculate the density expectation value of two initially independent condensates . it is found that there is a nonzero interference term when the interatomic interaction and the exchange symmetry of identical bosons are both considered carefully . in fact , it is well - known that the interaction plays an essential role in the formation of the order parameter of bose - condensed gases , i.e. the formation of a stable coherent property . here , we provide an example in which the interaction induces coherent evolution between two initially independent condensates . in the present popular viewpoint , the high - order correlation function and quantum measurement process are used to interpret the interference patterns observed for two initially independent condensates . although we find that there is already high - contrast interference patterns in the density expectation value for the experimental parameters in ref . @xcite , it is still possible that the high - order correlation function and thus the mechanism of the interference due to the measurement process play an important role in the emergence of the interference patterns . thus , to investigate more clearly the interference pattern due to the measurement process alone , we believe an experimental investigation of two independent ideal condensates would be very interesting , because there is no interference term for ideal condensates in the density expectation value . in the last few years , the rapid experimental advances of feshbach resonance where the scattering length can be tuned from positive to negative make this sort of experiment be feasible . note added : the present paper is an expansion of our two previous unpublished works @xcite . in addition , after the submission of the present paper , we noticed a paper @xcite by l. s. cederbaum et al . , where the role of interaction in the emergence of interference patterns for two initially independent bose condensates is stressed . ketterle w , durfee d s and stamper - kurn d m 1999 in _ making , probing and understanding bose - einstein condensates , _ proceedings of the international school of physics , enrico fermi course cxl , edited by inguscio m , stringari s and wieman c e ( ios press : amsterdam ) , pp . 67359 xiong h w , liu s j and zhan m s 2005 _ preprint _ cond - mat/0507354 xiong h w and liu s j 2005 _ preprint _ cond - mat/0507509 cederbaum l s , streltsov a i , band y b and alon o e 2006 _ preprint _ cond - mat/0607556
an inflationary epoch of the early universe is widely accepted as the most efficient mechanism to solve the flatness and the horizon problem , while providing a nearly scale invariant spectrum of scalar perturbations , in agreement with observations . however , the embedding of inflation driven by a single scalar field within string theory ( presently , the only known self - consistent theory of quantum gravity ) has proven to be challenging , since it requires an extremely flat potential . in addition , the presence of many dynamic fields in string theory renders single field models less appealing because all but one degree of freedom need to be already stabilized at the onset of inflation . though far from being simple or generic , a partially successful implementation is the kklmmt construction @xcite . to ameliorate the problem of fine tuning the potential , liddle et . al . proposed assisted inflation in @xcite ( see also @xcite and follow ups ) , a certain type of multi - field inflation whereby the presence of many fields increases the hubble friction , allowing for steeper potentials to still drive slow roll inflation . one shortcoming , however , is the need for fine tuned initial conditions for particular potentials , especially in the absence of an attractor solution ( for instance in n - flation @xcite ; see @xcite for a more general stability analysis of assisted inflation ) . the presence of many degrees of freedom in string theory has , over the years , galvanized the emergence of a variety of models implementing assisted inflation ; these include inflation from multiple tachyons @xcite , from multiple m5-branes within m - theory @xcite or from axions @xcite among others , see e.g. @xcite for a small selection . to make predictions in these models , it is common to introduce a single effective degree of freedom , visualized as the length of the trajectory in field space , see e.g. the review @xcite . utilizing this approach correctly recovers adiabatic perturbations , but not isocurvature / entropy modes . the latter ones can be pictured as perturbations perpendicular to the trajectory and develop whenever several degrees of freedom are present @xcite . although non - gaussianities may be produced ( see the review @xcite for details ) , it is possible to show that they are generically slow roll suppressed @xcite ( see however @xcite ) . with this in mind , is it then feasible to discriminate a given multi - field model from its single field analog ? in this article , we investigate a , so far largely ignored ( see however @xcite ) , discriminating property of multi - field models , namely the possibility that fields decay during inflation . by `` decay '' we mean that individual fields suddenly become obsolete , while their energy is converted into other forms , such as radiation . this is a widespread feature of models within string theory : consider , for instance , an association of inflatons with distances between branes that are located in some internal space . if a brane annihilates during inflation , for example as a result of its dissolution into a boundary brane or a collision with an anti - brane , the distance to the just mentioned brane becomes meaningless . of course , energy does not vanish , but is converted into a different type during the annihilation . a further example is inflation driven by tachyons , which can condense during inflation in a staggered fashion . the disappearance of fields causes an additional decrease of the potential energy during inflation , which can be even more important than the reduction of potential energy induced by slow roll . hence , leading order corrections to observable parameters that are sensitive to the slow roll parameters , such as the scalar spectral index , are possible . indeed , one can construct models of inflation entirely without a slow roll phase , similar to inflation without inflatons as proposed in @xcite . in order to compute these effects we assume a simple separable potential @xmath0 and promote the number of fields @xmath1 to a time dependent function . we further smooth out @xmath1 so that we can introduce a continuous decay rate @xmath2 . this approach is only justified if the number of fields and @xmath3 are large enough so that several fields disappear in any given hubble time . further , any signal due to a rapid drop in the potential energy , such as a ringing in the power spectrum or additional non - gaussianities @xcite , can not be recovered by this approach . to guarantee energy - momentum conservation ( @xmath4 ) , we are forced to introduce an additional component of the total energy momentum tensor that takes over the energy of the disappearing fields . the ratio of its energy density to the one in the remaining inflatons yields an additional small parameter @xmath5 appearing alongside the usual slow roll parameters . as a consequence , the equation of motion of the effective field is modified . the resulting set - up is reminiscent of warm inflation @xcite , and might indeed be seen as a new , less problematic implementation of warm inflation within string theory . at the perturbed level , we consider adiabatic and entropy perturbations and show that the latter ones are suppressed in the models of interest . focusing on the mukhanov variable , we derive the scalar spectral index and its running for general decay rates and number of fields . we recover the usual slow roll result as well as @xcite ( a dynamically relaxing cosmological constant ) in the appropriate limits , but we also find leading order corrections to the slow roll result in general . the running remains second order in small parameters and is , therefore , too small to be observed . to exemplify the applicability of the formalism we make a detailed study of three concrete models ; firstly , as an instructive toy model , we consider assisted inflation with exponential potentials and a nonzero , constant decay rate , which we insert by hand . @xmath6 provides a graceful exit to inflation at the cost of requiring somewhat flatter potentials or more fields so that the model remains consistent with observations . if the decay rate is too big ( @xmath7 ) the scalar spectral index becomes unacceptably red , but all smaller rates work well . secondly , we investigate inflation from multiple tachyons , as proposed in @xcite . here , tachyons get displaced from the top of their potential by thermal or quantum mechanical fluctuations , causing them to condense during inflation . we find that the best motivated and least fine tuned setup consists of a constant decay rate and tachyons close to the top of their potential so that their slow roll evolution yields negligible contributions to observable parameters . the resulting spectral index is @xmath8 , where @xmath9 is the number efolds at which we evaluate @xmath10 . this is in good agreement with the observations of the cmbr , @xmath11 @xcite . lastly , we look at inflation from multiple m-5 branes within m - theroy @xcite . here , inflatons are associated with the inter - brane distance between m5-branes located within an orbifold @xmath12 . whenever a brane comes close to a boundary brane it dissolves into the boundary via a small instanton transition . given the model s parameter ranges , which encompass , for instance , the concrete potential and the maximal number of branes , we show that inflation comes to an end within a few efolds after the first outermost brane disintegrates . hence , only an initial narrow stacking of branes ( involves fine tuning ) permits a consistent scenario , and the usual slow roll expressions apply . based upon these findings , cascade inflation as investigated in @xcite is irrelevant for the cosmological scales that are observed in the cmbr , given the parameters put forward in @xcite . it is evident that whether or not corrections due to @xmath6 are important is model - dependent , ranging from being the primary ingredient , as in the tachyon case , to being not pertinent , as in the m5-brane case . thus , the many implementations of multi - field inflation within string theory should be thoroughly re - investigated . the outline of this paper is as follows : in section [ sec : main ] , we derive the formalism to deal with dacaying fields , at the background level ( sec . [ sec : background ] ) and the perturbed one ( sec . [ sec : spectralindex ] ) . we compute the spectral index and the running in section [ sec : spectralindex ] and [ sec : running ] . along the way , we draw comparisons to warm inflation in section [ sec : warm ] and comment on isocurvature modes as well as non - gaussianities in section [ sec : entropy ] . we then shift gears and focus on applications , that is we investigate concrete models : assisted inflation ( sec . [ sec : staggered ] ) , inflation from tachyons ( sec . [ sec : acdmm ] ) and inflation from multiple m5-branes ( sec . [ sec : bbk ] ) ; summaries of the conclusions for each model can be found at the end of their respective subsection . we conclude in section [ sec : concl ] . consider @xmath1 scalar fields with canonical kinetic terms and a separable potential @xmath0 , is not the susy superpotential . ] so that the action reads @xmath13 for ease of notation , we set the reduced planck mass equal to one throughout @xmath14 . assume that the fields all evolve according to the same potential @xmath15 and that they start out from an identical initial value , so that @xmath16 in this case , an effective single field model with @xmath17 and @xmath18 describes correctly the inflationary phase ( if @xmath19 is sufficiently flat ) as well as the production of adiabatic perturbations ( we comment on entropy perturbations in section [ sec : entropy ] ) . to guarantee slow roll inflation , we demand @xmath20 where a prime on @xmath21 or @xmath22 denotes a derivative with respect to @xmath23 or @xmath24 respectively . so far , this is merely a simple model of assisted inflation @xcite , see also @xcite and follow ups . given this setup , we would like to investigate the consequences of individual fields @xmath23 dropping out of the model , more or less instantaneously . by dropping out we mean that a field decays while its energy is converted into a different form , for example radiation . this may seem artificial at first glance , but it is actually quite generic in multi - field models of inflation within string theory . for instance , in the model of @xcite ( see also @xcite ) the inflatons are related to the distances between adjacent m5-branes . these branes are located along an orbifold @xmath25 and slowly separate from each other in the orbifold direction , corresponding to the inflationary phase in the effective four dimensional description . since the orbifold is quasi - static during this regime , the outermost branes will collide with the orbifold - fixed planes at some point in time , dissolving into the boundaries through small instanton transitions . thus , the degrees of freedom associated with the distances to these just dissolved branes become obsolete during inflation . of course , the energy associated with the branes does not vanish , but gets converted to other degrees of freedom , such as radiation . another example is inflation driven by multiple tachyons as proposed in @xcite ; here , tachyons roll slowly away from the top of their potential , but fluctuations may displace a given field far enough to condense suddenly , again making obsolete this degree of freedom during inflation . we come back to these two concrete models later on , after developing the formalism to deal with decaying fields . for additional multi - field models see e.g. @xcite and @xcite for a review . it is worth stressing at this point that the disappearance of a given field in the models considered in this paper is a sudden , but continuous process . this means potential problems associated with first order phase transitions , as discussed in some detail in @xcite , are absent . to model the disappearance of individual fields we promote @xmath1 to a time dependent function . we further smooth out , we will not be able to recover additional features in the power - spectrum that are directly related to sudden drops in the potential . for instance , features could consist of a ringing or additional non - gaussianities @xcite . however , such signals depend crucially on the detailed physics of the fields dissapearance , a rather badly understood and model dependent issue for the current generation of setups . as a consequence , these features will be quite hard to estimate within an effective single field description and we will not address them further in this article . ] @xmath26 so that we can introduce a continuous decay rate @xmath27 , which is to be determined from the underlying model . this smoothing is the key simplifying assumption in our approach , and it is only viable if the number of fields is large and the decay rate is such that within any given hubble time a few fields become obsolete . for the cases we are interested in , this rate is small compared to the hubble parameter @xmath28 during inflation ( inflation ends quickly otherwise ) . this means we can introduce the small parameter @xmath29 the time dependence of @xmath1 induces an additional decrease in the energy of the effective inflaton . to be concrete , the continuity equation of @xmath24 needs to be modified to account for the energy loss in the inflaton sector due to decaying fields @xmath30 , so that @xmath31 here and throughout the subscript @xmath24 denotes the entire inflaton sector . it is worthwhile mentioning that the individual @xmath23 evolve according to the standard slow roll equation of motion @xmath32 , as long as they are present . to retain @xmath33 , we must allow for an additional component @xmath34 within the energy budget , so that energy always flows from the inflaton field sector into @xmath35 . ] satisfying @xmath36 we use the subscript `` @xmath37 '' , since we have radiation in mind ( @xmath38 with @xmath39 ) ; this choice seems most natural to us , considering the models we are interested in , but to remain as general as possible we keep @xmath40 arbitrary throughout . the condition @xmath41 implies that @xmath34 makes up only a small fraction of the total energy density during inflation . thus , we may introduce the small parameter @xmath42 naturally , @xmath5 is not independent of @xmath43 , and one can show under mild assumptions that @xmath44 during inflation ( see below ) . further , using slow roll of the individual fields @xmath23 , one can show that the total energy and pressure of the effective inflaton obey @xmath45 . this , along with @xmath38 and the definitions above , leads to @xmath46 during slow roll ( the `` @xmath47 '' always denotes equality to first order in small parameters such as @xmath48 or @xmath5 ) . taking the derivative with respect to time of the friedmann equation @xmath49 and using ( [ cont1 ] ) as well as ( [ cont2 ] ) , yields the _ hubble slow roll parameter _ @xmath50 this explains the chosen pre - factors in ( [ defepsilonn ] ) and ( [ defbarepsilon ] ) , as well as our demand that all epsilons should be small . we have carefully avoided the use of the klein gordon equation for @xmath24 , since it gets modified by the presence of @xmath6 . to derive this modification , one can use @xmath51 in the continuity equation ( [ cont1 ] ) and make the usual slow roll approximations , or take the time derivative of @xmath52 directly with @xmath53 so that @xmath54 where we introduced the short hand notation @xmath55 as expected , the usual slow roll equation of motion is recovered from ( [ eomsrvarphi ] ) in the limit @xmath56 . the limit @xmath57 , that is @xmath58 , corresponds to a dynamically relaxing cosmological constant , which is discussed in great detail in @xcite . during inflation , @xmath28 , @xmath43 as well as @xmath59 change very slowly , so that ( [ cont2 ] ) can be integrated to @xmath60 with @xmath61 . this means that the additional component of the energy budget approaches a scaling solution @xmath62 for which @xmath63 and @xmath64 . thus , we can use @xmath65 during inflation . at this point , we would like to comment on similarities to warm inflation @xcite , where the scalar field s interaction with other particles ( through which the scalar field transfers some of its energy into a thermal bath ) prevents the temperature from rapidly reaching zero . within warm inflation , the motion of the scalar field is described by the modified klein - gordon equation ( assuming slow roll ) @xmath66 the extra friction term @xmath67 represents an additional energy loss of the scalar field @xmath24 stemming from particle creation . this equation needs to be compared to our equation ( [ eomsrvarphi ] ) : in our case , an increase / decrease of the right hand side is present , depending on the sign of @xmath68 , while in ( [ eom_wi ] ) , the right hand side is always decreased by @xmath69 . a more concrete realization of warm inflation based on thermal viscosity has been ( critically ) examined in @xcite . in this study , part of the inflaton s energy is converted into radiation through a viscosity term . as a consequence , one can show that the energy density satisfies @xmath70 with @xmath71 . comparing ( [ energy_density_wi ] ) to our case ( [ energy_density ] ) we see a similar modification : an additional positive term , potentially counterbalancing the dilution of @xmath34 due to redshifting ; however , the origin of this term differs : in our case it is due to the transfer of potential energy , whereas that of ( [ energy_density_wi ] ) arises from infusing kinetic energy . though the friction term in ( [ energy_density_wi ] ) could at first glance be large enough to allow for successful warm inflation , a more careful examination by yokoyama and linde revealed @xcite that warm inflation is not feasible in this framework , since @xmath24 changes significantly over the relaxation time of the relevant particles , violating an adiabaticity condition required for the validity of ( [ energy_density_wi ] ) . nevertheless , albeit the presence of many problems like the one just mentioned , warm inflation remains an active field of research . staggered multi - field inflation , as introduced in the present paper , might be seen as an independent ( less problematic ) realization of warm inflation , which can be embedded into string theory . in order to compute the scalar spectral index we focus on adiabatic perturbations , that is , for the time being we neglect isocurvature perturbations the latter ones arise due to the presence of the many scalar fields and @xmath34 , but one can show that they are suppressed during inflation . the applicability of this approximation is discussed in section [ sec : entropy ] . the canonical degree of freedom that diagonalizes the action of adiabatic scalar perturbations is the mukhanov variable @xmath72 @xcite . it is related to the curvature perturbation on uniform density hyper - surfaces @xmath73 via @xmath74 where @xmath75 here , @xmath76 with @xmath77 and @xmath78 is the equation of state parameter , while @xmath79 is the adiabatic sound speed , which becomes @xmath80 on large scales . in terms of @xmath81 , the power spectrum is given by @xmath82 and the scalar spectral index is @xmath83 the mukhanov variable satisfies the simple equation of motion @xcite @xmath84 where a prime denotes a derivative with respect to conformal time @xmath85 , @xmath86 . if the hubble slow roll parameter @xmath87 in ( [ defhatepsilon0 ] ) is evolving slowly , which is the case during inflation , we can approximate @xmath88 and solve ( [ eomvk ] ) analytically in terms of hankel functions ( see e.g. @xcite or @xcite ) . writing @xmath89 treating @xmath90 as a constant and imposing the bunch - davies vacuum at @xmath91 ( @xmath92 ) , we get @xmath93 where @xmath94 is an irrelevant phase factor , @xmath95 . hence , the curvature perturbation reads @xmath96 ) and taking the logarithmic derivative , one can read off the scalar spectral index to @xmath97 thus , we only have to compute @xmath98 and identify @xmath90 in order to get @xmath10 . since @xmath99 is second order in small parameters , we need to focus on @xmath100 from ( [ deftheta ] ) only . to leading order in @xmath101 and @xmath102 , the equation of state parameter reads @xmath103 here , we used the equation of motion for @xmath24 in ( [ eomsrvarphi ] ) , @xmath104 from ( [ defgamma ] ) and @xmath105 from ( [ barepsilonandepsilonn ] ) . consequently @xmath106 in order to take the derivatives with respect to conformal time , we need @xmath107-\varepsilon_{\mathcal{n}}\gamma\right]\ , , \label{dergam}\end{aligned}\ ] ] where we introduced @xmath108 note that @xmath109 corresponds to @xmath110 , whereas @xmath111 corresponds to @xmath112 . deriving ( [ derep])-([dergam ] ) based on section [ sec : background ] is straightforward , albeit somewhat tedious . after some more algebra , we arrive at @xmath113-\varepsilon_{\mathcal{n}}\gamma\right ] + \frac{1}{2}\varepsilon_{\mathcal{n}}(\varepsilon+\varepsilon_{\mathcal{n}})(1-\delta ) \big)\bigg]\,,\end{aligned}\ ] ] to leading order in small parameters ( we assume that @xmath114 is of the same order as @xmath43 ) . using @xmath115 from ( [ aoftau ] ) so that @xmath116 we can read off @xmath90 . the resulting scalar spectral index is @xmath117.\label{finalns } % % % % % the below is my old checked expression , which we use in the maple codes . do not erase ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! % % \nonumber n_s-1&\simeq&-2(\varepsilon+\varepsilon_{\mathcal{n}})-\frac{2}{\varepsilon\gamma^2+\varepsilon_{\mathcal{n } } } % \bigg [ % \gamma^3\left(2\varepsilon^2-\varepsilon\eta\right)\\ % & & + \varepsilon\gamma\left[(\gamma-1)\left[(\varepsilon+\varepsilon_{\mathcal{n}})(1-\delta)+\gamma(\eta-2\varepsilon)\right]-\varepsilon_{\mathcal{n}}\gamma\right ] % + \frac{1}{2}\varepsilon_{\mathcal{n}}(\varepsilon+\varepsilon_{\mathcal{n}})(1-\delta ) % \bigg]\,.\label{finalns } % \end{aligned}\ ] ] this is our first major result . in the limit @xmath118 , so that @xmath119 and @xmath120 , we recover the slow roll result @xmath121 . on the other hand , if @xmath110 ( that is @xmath109 ) and @xmath122 , we recover the case of a dynamically relaxing cosmological constant of @xcite @xmath123 . the running @xmath124 can be computed by applying , resulting in @xmath125 ; we use that @xmath28 is evolving slowly during inflation , that is @xmath126 . ] @xmath127 to ( [ finalns ] ) . using again @xmath63 and @xmath128 where @xmath129 , as well as @xmath130 , @xmath131 and @xmath132 from ( [ derep])-([dergam ] ) the running reads @xmath133\\ & & + \nonumber(\varepsilon+\varepsilon_{\mathcal n})\left[(1-\delta)\left[(\gamma-1)(\varepsilon'\gamma+\varepsilon\gamma')+\varepsilon\gamma\gamma'+\frac{\varepsilon_{\mathcal n}'}{2}\right]-\delta'\left[\varepsilon\gamma(\gamma-1)+\frac{\varepsilon_{\mathcal n}}{2}\right]\right]\bigg]\\ & & -\nonumber\frac{2[\varepsilon'\gamma^2 + 2\varepsilon\gamma\gamma'+\varepsilon_{\mathcal n}']}{(\varepsilon\gamma^2+\varepsilon_{\mathcal n})^2}\left[\varepsilon\gamma^2(\eta+\varepsilon_{\mathcal n}-2\varepsilon)-(\varepsilon+\varepsilon_{\mathcal n})(1-\delta)\left[\varepsilon\gamma(\gamma-1)+\frac{\varepsilon_{\mathcal n}}{2}\right]\right]\,.\end{aligned}\ ] ] in the limit @xmath56 , we recover the standard slow roll result @xmath134 . on the other hand , if @xmath135 , the running reduces to @xmath136 . in any event , the running remains second order in small parameters , rendering it un - observably small . so far we focused on adiabatic perturbations only , neglecting entropy perturbations entirely . to check if entropy perturbations are indeed negligible , we follow @xcite ( see also @xcite ) . the perturbation of the total pressure is in general @xmath137 where @xmath138 is the entropy density , @xmath139 ( we use cosmic time in this section in order to avoid confusion of this @xmath140 with conformal time ) and @xmath141 is the adiabatic sound speed , which reduces to @xmath142 on large scales . the non - adiabatic pressure perturbation may thus be defined as @xmath143 in a multi - component fluid , there are two contributions to @xmath144 , a relative one @xmath145 between the fluid components and an intrinsic one within each fluid @xmath146 @xcite . in our case , we have two main components , the effective inflaton ( composed of @xmath1 fields ) with @xmath59 , and the additional component @xmath34 , for instance given by radiation , with a constant equation of state parameter @xmath147 . let us first discuss @xmath148 , which can be written as @xcite @xmath149 where @xmath150 and the relative entropy perturbation is defined as @xmath151 we have already seen in the discussion before ( [ barepsilonandepsilonn ] ) that @xmath34 approaches a scaling solution @xmath62 during inflation for which @xmath152 is small ( @xmath64 ) . since @xmath153 , the relative non - adiabatic pressure perturbation is heavily suppressed and can safely be ignored during inflation . remains small and finite in the limit of small @xmath152 , because the curvature perturbations @xmath154 and @xmath155 remain small ; see @xcite . ] second , consider the intrinsic contributions @xmath156 since @xmath157 we have @xmath158 and @xmath159 . therefore @xmath160 , that is the intrinsic non - adiabatic pressure @xmath161 vanishes identically . however , for the effective inflaton the case is less clear , since its equation of state ( and thus the intrinsic sound speed ) changes and @xmath24 is actually composed of @xmath1 components . we first note that the individual fields approach an equation of state @xmath162 during inflation , so that the equation of state parameter becomes nearly constant for each of the fields . thus we have @xmath163 where @xmath164 . therefore , the intrinsic non - adiabatic pressure within each inflaton field is @xmath165 . second , the @xmath1 components relative non - adiabatic pressure perturbations contribute to @xmath166 . but if we look at the relative non - adiabatic pressure contributions between the fields @xmath167 we conclude that @xmath168 since @xmath169 . thus , the total non - adiabatic pressure is negligible during inflation and we are justified to focus on adiabatic perturbations . @xcite . ] these are correctly recovered by focusing on the mukhanov variable @xmath72 , as we did in section [ sec : spectralindex ] . one further comment might be in order : due to the suppression of entropy perturbations during inflation , we do not expect any large additional non - gaussianities ( ng ) caused by @xmath6 . further , since multi - field inflationary models generically yield comparable ng to their single field analogs @xcite , and ng are suppressed during slow roll inflation ( see however @xcite ) , we do not expect any measurable ng within the setups discussed in this paper . one caveat to this argument consists of the short intervals whenever one of the inflaton fields decays ( see e.g. @xcite for ng from steps in a potential in single field inflation ) . we can not exclude the production of ng at these instances , but we do not anticipate them either , because the process is very similar to preheating and generically , preheating ( e.g. instant preheating ) does not cause large ng @xcite ( see however the possibility of larger ng in tachyonic preheating @xcite ) . nevertheless , given a better microphysical understanding of how inflatons become obsolete , e.g. by investigating the brane annihilation in @xcite , one can and should check the validity of this expectation . we would like to compute the scalar spectral index in ( [ finalns ] ) and the running in ( [ finalrunning ] ) within a couple of models . first , we extend assisted inflation with exponential potentials @xcite by incorporating a non - zero @xmath3 , which provides a graceful exit of inflation ; this phenomenological model has the advantage of being instructive and simple . next , we consider two concrete models @xcite which have the feature of decaying fields during inflation already build in , at the price of being more complicated to treat . our approach consists of extracting the potential slow roll parameters @xmath101 , @xmath170 and @xmath171 , as well as @xmath172 with @xmath173 in order to apply ( [ finalns ] ) and ( [ finalrunning ] ) . consider the original proposal of assisted inflation @xcite , where the @xmath1 scalar fields have identical exponential potentials @xmath174 so that the potential for the single effective field @xmath175 reads @xmath176 with @xmath177 ( we assume identical initial values for all @xmath23 ) . power law inflation ( @xmath178 ) results for large enough @xmath179 , which can be achieved even with steep potentials if @xmath180 . note that the single field solution is an attractor during inflation @xcite ( see @xcite for a general discussion of stability in multi - field inflation ) . the slow roll parameters in the above model are @xmath181 so that the scalar spectral index is @xmath182 if @xmath118 ( see @xcite or equation ( [ finalns ] ) ) ; in addition , the running is zero since @xmath183 . however , the above model has a graceful exit problem : inflation never ends because @xmath101 and @xmath170 are constant . this problem can be alleviated by introducing a non - zero @xmath3 , so that the number of fields decreases during inflation . let s for simplicity take a constant rate @xmath112 so that @xmath111 . consequently , inflation comes to an end when @xmath184 becomes of order one , at which point @xmath1 decreases rapidly during a hubble time so that the assistance effect diminishes . since a shift in the individual fields @xmath185 can be absorbed into a redefinition of @xmath186 , we can set @xmath187 at @xmath188 efolds before the end of inflation , without loss of generality . hence @xmath120 , and the scalar spectral index in ( [ finalns ] ) becomes since @xmath24 changes , and a shift in @xmath43 since @xmath186 changes ; however , an observable such as the scalar spectral index remains unaffected . ] @xmath189 where we used @xmath190 and @xmath111 . similarly , the running in ( [ finalrunning ] ) reads @xmath191 where we also used @xmath192 and @xmath193 . these predictions differ from the corresponding slow roll ones : the additional energy loss in the inflaton sector due to @xmath6 causes a redder spectrum with a running that is second order in the epsilons ( such a running is well below current observational limits , we will not comment on it further in section [ sec : acdmm ] and [ sec : bbk ] . ] ) . the physical reason for the difference in the spectrum is the smooth graceful exit caused by a non - zero decay rate , resulting in @xmath194 . note that for a sharp exit from inflation to the reheating era , which could be modeled by a step function @xmath195 with @xmath196 , we anticipate no corrections . using @xmath190 ( assisted inflation with exponential potentials ) we plot @xmath197 lines in the @xmath101-@xmath43 plane using ( [ assistedns ] ) , showing the @xmath198 interval of wmap5 @xcite @xmath11 ( solid and dotted lines ) . thus , a non - zero @xmath43 requires a lower value of @xmath101 in order to remain consistent with observations . ] in figure [ pic : assistedns ] , we plot @xmath197 lines in the @xmath101-@xmath43 plane , showing the @xmath198 interval of wmap5 @xcite @xmath11 . given @xmath199 , it is evident that in order to fit @xmath10 from observations in assisted inflation we need a smaller @xmath101 . to put it another way , the smooth graceful exit of inflation introduced by @xmath6 comes at the price of either requiring flatter potentials or more fields . further , the spectral index lies outside the @xmath198 region if @xmath43 is too large ( @xmath200 ) , even if @xmath101 were identical to zero . this means the half - time @xmath201 needs to be considerably larger than the hubble time @xmath202 during inflation , in order to be consistent with observations ( @xmath203 ) . in the phenomenological model above , we inserted a non - zero decay rate by hand . as we shall see in the next sections , more concrete models within string theory actually force us to have @xmath6 , rendering the above model more natural . in @xcite , @xmath1 brane anti - brane pairs ( @xmath204 ) where considered , giving rise to @xmath205 tachyons due to a @xmath206 gauge symmetry . however , these tachyons are generically all coupled to each other , making them not well suited for assisted inflation . to alleviate this problem , davis and majumdar proposed to focus on the abelian part of @xmath206 , resulting in @xmath1 uncoupled tachyons @xmath23 , and time by @xmath140 . ] @xmath207 ; admittedly , this choice is to some extent unphysical @xcite , but offers an instructive model . within this setup , the potential has the form @xcite @xmath208 which is valid in close proximity to @xmath209 . here , @xmath210 and @xmath211 @xcite , but the brane tension @xmath212 is model dependent . for large @xmath1 sufficient inflation can result if the tachyons are initially close to zero . for simplicity , we assume that they all start out from the same initial value , in line with our assumptions in section [ sec : main ] . whenever a tachyon gets displaced far enough from the origin ( be it due to its slow roll evolution or a dislocation caused by either thermal or quantum mechanical fluctuations @xcite ) the perturbative potential in ( [ potentialacd ] ) becomes unreliable , and the tachyon condenses quickly . this condensation corresponds to the annihilation of a brane anti - brane pair . inflation ends when all tachyons condense . this condensation is also expected to be responsible for reheating , but a concrete study is lacking in the literature . giving the model the benefit of the doubt , we assume that whenever a brane anti - brane pair annihilates , its energy is indeed converted into some type of relativistic matter with @xmath213 ( from the four dimensional point of view ) . thus , the model of @xcite is exactly of the type we examined in section [ sec : main ] . next , we extract the slow roll parameters as well as specify the rate @xmath214 at which the tachyons condense in order to compute @xmath10 . if all tachyons condense at the same time , we have @xmath118 during inflation and the standard slow roll expressions apply . however , given the sensitivity of condensation to thermal and quantum mechanical dislocations of the tachyons , a staggered fashion of condensation is expected . in @xcite three types of such a staggered condensation were proposed : 1 . the number of tachyons decreases exponentially so that @xmath112 ( @xmath215 in the notation of @xcite ) , similar to the model in section [ sec : staggered ] . 2 . tachyons condense serially so that @xmath216 and @xmath217 ( @xmath218 in the notation of @xcite ) . this is the case if tachyons condense at wildly different times . 3 . tachyons condense in a staggered fashion , but a handful survive and drive an extended phase of slow roll inflation , rendering again @xmath118 during the cosmological relevant phase of inflation . ( we do not examine this case further . ) because the applicability of the potential ( and thus slow roll ) is questionable for large values of @xmath24 , we first examine cases one and two in a simplified setup with all tachyons sitting very close to the top of their potentials ( @xmath219 , @xmath220 ) . whenever a tachyon gets displaced by a fluctuation , it is assumed to condense immediately . in later sections , we incorporate slow roll . * case 1 ) @xmath112 * + similar to section [ sec : staggered ] , we consider a constant condensation rate so that the number of tachyons decreases exponentially in time @xmath221 where we take @xmath222 at @xmath9 efolds before the end of inflation . once the number of fields is depleted , that is once @xmath223 , inflation ends . to be concrete , we take @xmath224 . then , using @xmath225 and @xmath226 , the number of efolds becomes @xmath227 further , since @xmath228 we get from ( [ finalns ] ) @xmath229 which is within the @xmath198 error bars of wmap5 @xcite . if we fine - tune the brane tension such that @xmath22 vanishes at the minimum , @xmath230 , we obtain @xmath231 thus , for @xmath232 , we need @xmath233 tachyons , which is somewhat large ; however , for @xmath234 we achieve the desired amount of inflation and a spectral index within observational bounds with a few hundred tachyons . it should be noted that neglecting slow roll might actually be the best motivated case : we only trust the tachyon potential close to @xmath235 ; then the potential is indeed very flat so that @xmath101 can be neglected . ( just as in @xcite ) , and not by @xmath101 . ] as soon as an individual tachyon gets dislodged a bit , it should quickly condense and drop out of the model , leading to the above estimate for @xmath10 . * case 2 ) @xmath236 * + here , the number of fields is decreasing linearly @xmath237 so that inflation ends around @xmath238 . analogous to ( [ efoldswsrconstgamma ] ) , the number of efolds becomes @xmath239 so that @xmath240 . since @xmath241 we get from ( [ finalns ] ) @xmath242 smaller than in the @xmath112 case and close to the @xmath198 boundary of wmap5 @xcite . tuning the brane tension again to @xmath230 , we obtain @xmath243 as a result , in order to achieve the desired @xmath188 efolds of inflation with a few hundred fields we need @xmath244 . in case all tachyons condense at once we have @xmath118 during inflation and the usual slow roll expressions apply . in order to provide a concrete example we assume @xmath246 , tune the brane tension to @xmath247 and take @xmath248 even for @xmath249 , which is stretching the applicability of the potential . corresponds to individual field values of order @xmath250 if @xmath251 . since @xmath252 in order for ( [ potsr ] ) to apply , we are reaching the limit of its applicability . ] inflation ends when either @xmath101 or @xmath170 become of order one . for our potential , @xmath253 first at @xmath254 the value of @xmath24 at @xmath188 efolds before the end of inflation can then be computed numerically from @xmath255 given @xmath256 , the slow roll parameters and @xmath257 follow straightforwardly . since the only free parameter is the number of fields , we can fine - tune @xmath1 to yield the desired scalar spectral index . to be concrete , for @xmath258 we get @xmath259 and @xmath260 so that @xmath261 , @xmath262 resulting in @xmath263 , matching wmap5 @xcite . if the full tachyon potential should become steeper before @xmath264 or the tachyons condense collectively at some smaller value @xmath265 , the corresponding @xmath256 shifts to lower values , causing @xmath101 and @xmath266 to decrease further . as a consequence , even less fields @xmath267 are needed to match @xmath268 . based on the last two sections , we expect the contributions to @xmath10 in ( [ finalns ] ) by the slow roll parameters and @xmath43 to be of comparable magnitude if the tachyons do not start out too close to the origin and we have @xmath269 as well as a decay rate of order @xmath270 ( or @xmath271 ) . for even smaller rates , @xmath43 becomes negligible and the usual slow roll results apply , whereas larger values cause a premature end of slow roll inflation . on the other hand , if the tachyons start out very close to the origin , we can neglect the slow roll contribution altogether , just as in section [ sec : acdwithoutslowroll ] . to quantify these statements , we take the slow roll setup with @xmath247 as well as @xmath272 fields and slowly turn on @xmath3 in order to show its effect on @xmath10 . * case 1 ) @xmath112 * + first , we determine the end of slow roll inflation , which occurs whenever @xmath101 , @xmath170 or @xmath43 becomes of order one ( or @xmath1 itself becomes of order one ) . assuming @xmath180 is still valid at the end of inflation , one can show that @xmath253 before @xmath170 and @xmath43 if @xmath273 , where @xmath274 since @xmath275 for @xmath258 fields and we are primarily interested in small decay rates , we determine @xmath276 from @xmath253 . given @xmath264 , we can determine the time @xmath277 , and thus the field value @xmath278 from the requirement @xmath279 using @xmath280 and a numerical solution to the equation of motion for @xmath24 in ( [ eomsrvarphi ] ) . once @xmath256 is known , we can straightforwardly compute @xmath101 , @xmath170 and @xmath43 at @xmath188 efolds before the end of inflation . using these in ( [ finalns ] ) , we plot the resulting spectral index over @xmath3 in figure [ pic : nsm1overgamma ] . as expected , the decay rate becomes important for @xmath281 , quickly driving the scalar spectral index outside the observationally favored region @xmath11 @xcite as @xmath3 increases further . setting @xmath247 so that the potential in ( [ potentialacd ] ) vanishes at the minimum , we plot @xmath282 over @xmath283 , for @xmath112 ( solid line ) and @xmath236 ( dashed line ) . we choose @xmath284 , so that @xmath285 ( the wmap5 value @xcite ) if @xmath286 ( or @xmath287 ) . as anticipated form section [ sec : acdwithoutslowroll ] , the decay rate becomes important for @xmath281 . ] * case 2 ) @xmath236 * + we compute @xmath282 analogously to the previous section , but using @xmath288 instead of @xmath112 . since @xmath3 is time dependent , we get a non - zero @xmath289 , leading to @xmath290 which also needs to be inserted into ( [ finalns ] ) . the resulting scalar spectral index is plotted over @xmath291 in figure [ pic : nsm1overgamma ] . @xmath10 is nearly indistinguishable from the @xmath112 case for the small @xmath292 under consideration . allowing for @xmath6 yields the interesting possibility to drive inflation and generate observational viable values for the scalar spectral index with negligible contributions from slow roll . one may start out with a few hundred tachyons very close to the top of their potential ( the potential is known in this regime ) . as soon as a fluctuation , be it quantum mechanical or thermal , dislodges a field , it should condense quickly and drop out of the effective description , while contributing to @xmath34 . this is the physical origin for @xmath6 . if the combination @xmath293 is chosen appropriately , for instance close to @xmath294 if @xmath112 , then @xmath188 efolds of inflation result with @xmath295 from ( [ ns3overn ] ) , in agreement with observations ( @xmath296 @xcite ) . admittedly , there is still some tuning involved , but it does not seem overly contrived to us : a few hundred tachyons are certainly possible and @xmath270 does not seem far fetched either . if the tachyons start to roll down the potential during inflation , the usual slow roll conditions must be satisfied too , so that @xmath282 in ( [ finalns ] ) remains small . however , the applicability of the potential in ( [ potentialacd ] ) is questionable in the regime where @xmath101 and @xmath170 become of interest , since @xmath297 . further , having both , @xmath101 and @xmath43 , of similar magnitude so that their contributions to @xmath10 are comparable constitutes an additional fine - tuning , which is not needed . we conclude that with minimal fine - tuning of @xmath298 and using @xmath299 , the model of @xcite predicts @xmath300 , with a proportionality factor depending on the exact form of @xmath301 ( for instance @xmath302 if @xmath112 and @xmath303 if @xmath304 ) . thus , a constant decay rate is in good agreement with observations . in the model of @xcite inflation is driven by an effective field @xmath24 , that evolves according to the exponential potential @xmath305 where @xmath306 with @xmath307 and @xmath308 ; here we used @xmath309 and @xmath310 , as in @xcite is denoted by @xmath311 . ] ( see also @xcite ) , but other combinations are possible @xcite . @xmath1 is the number of m5-branes , which are ( equidistantly ) stacked somewhere within an orbifold @xmath25 of length @xmath312 . its value is constraint to @xmath313 , justifying the use of @xmath180 above . ( that is @xmath314 ) initially . ] if @xmath315 , the scalar spectral index reads @xmath316 , just like in the case of assisted inflation in section [ sec : staggered ] . for @xmath317 we need @xmath318 m5-branes , a number which lies well within the allowed interval . note that this prediction is only valid as long as branes are not dissolving into the boundary branes , that is as long as the distance between adjacent branes satisfies @xmath319 . in @xcite it was also suggested that one could distribute the m5-branes uniformly over the interval in order to avoid the fine - tuning associated with the narrow stacking of branes . the number of branes is bounded by the brane separation via @xmath320 , and whenever @xmath321 a brane dissoves into a boundary brane , decreasing @xmath1 by one . in terms of @xmath322 the effective field is defined via @xmath323 where @xmath324 . note that whenever @xmath1 decreases , @xmath24 makes a jump due to the decreased @xmath179 , whereas @xmath322 is continuous ( we do not perform any smoothing of @xmath26 in this section ) . let us consider the following question in order to assess whether or not we need to consider effects caused by a decreasing number of m5-branes : if we have indeed a narrow stack of @xmath325 branes initially , how many efolds follow after the outermost m5-brane hits one of the boundary branes ? if the number of efolds @xmath326 that follow up until @xmath327 ( at which point @xmath101 becomes of order one and slow roll inflation ends ) is small , the above fixed-@xmath1 result is applicable . on the other hand , if @xmath328 , one can not ignore the time dependence of @xmath1 and cascade - inflation @xcite results . to address this question , we assume that the stack of branes is located near the center of the orbifold , so that the branes are uniformly distributed over the whole orbifold when the first brane dissolves into a boundary brane . during slow roll , @xmath326 is easily computed from @xmath329 where @xmath330 and the integration boundaries follow from @xmath331 with @xmath332 , @xmath333 and @xmath334 ( which remains constant during each integration interval ) as @xmath335 since the integrand is constant during each interval , we get @xmath336\,,\end{aligned}\ ] ] where @xmath337 is the digamma function ( the logarithmic derivative of the gamma function ) . for large @xmath338 , this can be approximated by @xmath339 , that is @xmath340 for @xmath325 . hence , inflation comes to an end shortly after the first m5-brane dissolves into the boundary brane and we are justified to use the fixed-@xmath1 result for @xmath10 . further , since all branes dissolve within the last few efolds , we expect all of them to contribute to reheating . one might wonder how many branes are needed to achieve @xmath188 if one starts out with a uniform distribution of m5-branes over the entire orbifold ( instead of a narrow stack of branes ) as mentioned in @xcite . solving @xmath341 for the number of branes yields @xmath342 , which is unfortunately well outside the allowed interval for @xmath1 . thus , such an initial uniform distribution is not feasible within the framework of @xcite , and the investigation of @xcite ( where @xmath343 and @xmath344 as well as @xmath345 are considered ) is strictly speaking not applicable either : any imprint onto fluctuations within the last few efolds of inflation does not correspond to the large scales observed in , e.g. the cmbr . however , it should be noted that other values of @xmath346 and @xmath347 seem possible @xcite so that cascade - inflation might be important in different scenarios . we leave this interesting possibility to future studies . were within the allowed interval , one could not simply invoke the formalism of section [ sec : main ] , since the potential differs from the type we considered in that it is not a simple sum , for instance @xmath348 instead of @xmath349 . therefore , one would have to adjust the formalism accordingly first . ] to summarize , in the model of @xcite and adhering to the prefered parameter ranges , successful inflation can only be driven if the branes are stacked narrowly ( this involves fine - tuning ) ; then @xmath350 , so that about @xmath351 branes are needed in order to match observations . as soon as the outermost m5-branes dissolve into the boundary branes , inflation comes to an end within a few efolds due to the rapidly decreasing number of branes . as a consequence , all branes should contribute to reheating . in this study , we investigated the consequences of disappearing inflatons during multi - field inflation . such a process occurs quite naturally in string theoretical models of multi - field inflation , for instance if the inflatons are associated with inter - brane distances and said branes start to annihilate during inflation , or if inflatons are identified with tachyons that condense in a staggered fashion . focusing on inflationary models with separable potentials , we promoted the number of fields to a time dependent , continuous variable and derived the general formalism at the background and perturbed level to allow for a non - zero decay rate . this approach is only valid if the number of fields is large and the decay rate is big enough so that at least a few fields drop out of the model in a given hubble time . to satisfy conservation of the energy momentum tensor , an additional component to the energy budget needs to be introduced , which takes over the energy of the decaying fields . a non - zero decay rate can cause leading order corrections in observable quantities such as the scalar spectral index or the running , for which we derived general expressions . we recovered standard slow roll results and the case of a dynamically relaxing cosmological constant of @xcite in the appropriate limits . in addition , we showed that isocurvature modes are generically suppressed during inflation for the models of interest ; as a consequence , perturbations are adiabatic and nearly gaussian . we also point out that staggered multi - field inflation might be seen as a new , less problematic type of warm inflation , which can be implemented within string theory . based on our formalism , we investigated three concrete models : assisted inflation with exponential potentials @xcite , inflation from multiple tachyons @xcite , and inflation from multiple m5-branes @xcite . for assisted inflation , the presence of a non - zero decay rate offers a graceful exit to inflation , while remaining observationally viable . for inflation from multiple tachyons , the use of our formalism offers a reliable method to compute the scalar spectral index ; we concluded that the setup with best computational control , minimal fine - tuning and good agreement with observations consists of a constant decay rate slightly below one , a few hundred tachyons and a negligible slow roll contribution . for inflation from multiple m5-branes , we showed that the dissolution of m5-branes into boundary branes is quite rapid within the allowed parameter ranges . therefore , inflation comes to an end within a few efolds after the first m5-branes disappear . hence , the only self consistent setup employs a ( fine - tuned ) narrow stack of branes , whose number is constant around the pivot scale for which the scalar spectral index is measured , and the standard slow roll result is applicable . we did not investigate further the connection to warm inflation , since it would go beyond the scope of this article , but it is surely an avenue worth investigating more thoroughly . further , our selection of models is not extensive , and had the primary purpose of exemplifying the applicability of the formalism in a few concrete , distinct scenarios . variations of the models discussed here ( and also entirely different ones ) are abundant in the literature and should be re - examined . we thank paul steinhardt and scott watson for comments on the darft . d. b. would like to thank damtp and a. c. davis for support when this project was started , as well as princeton university for hospitality during the final stages of this work . d. b. is supported by the eu fp6 marie curie research and training network `` universenet '' ( mrtn - ct-2006 - 035863 ) . t. b. would like to thank kari enquist and the helsinki institute of physics for hospitality . t. b. is supported by the council on science and technology at princeton university . this work is supported in part by pparc grant pp / d507366/1 and the stfc at damtp . s. kachru , r. kallosh , a. linde , j. m. maldacena , l. p. mcallister and s. p. trivedi , `` towards inflation in string theory , '' jcap * 0310 * , 013 ( 2003 ) [ arxiv : hep - th/0308055 ] . a. r. liddle , a. mazumdar and f. e. schunck , `` assisted inflation , '' phys . d * 58 * , 061301 ( 1998 ) [ arxiv : astro - ph/9804177 ] . k. a. malik and d. wands , `` dynamics of assisted inflation , '' phys . d * 59 * , 123501 ( 1999 ) [ arxiv : astro - ph/9812204 ] . p. kanti and k. a. olive , `` on the realization of assisted inflation , '' phys . d * 60 * , 043502 ( 1999 ) [ arxiv : hep - ph/9903524 ] . s. dimopoulos , s. kachru , j. mcgreevy and j. g. wacker , `` n - flation , '' arxiv : hep - th/0507205 . g. calcagni and a. r. liddle , `` stability of multi - field cosmological solutions , '' phys . d * 77 * , 023522 ( 2008 ) [ arxiv:0711.3360 [ astro - ph ] ] . m. majumdar and a. c. davis , `` inflation from tachyon condensation , large n effects , '' phys . d * 69 * , 103504 ( 2004 ) [ arxiv : hep - th/0304226 ] . k. becker , m. becker and a. krause , `` m - theory inflation from multi m5-brane dynamics , '' nucl . b * 715 * , 349 ( 2005 ) [ arxiv : hep - th/0501130 ] . j. ward , `` dbi n - flation , '' jhep * 0712 * , 045 ( 2007 ) [ arxiv:0711.0760 [ hep - th ] ] . t. w. grimm , `` axion inflation in type ii string theory , '' arxiv:0710.3883 [ hep - th ] . k. l. panigrahi and h. singh , `` assisted inflation from geometric tachyon , '' jhep * 0711 * , 017 ( 2007 ) [ arxiv:0708.1679 [ hep - th ] ] . y. s. piao , r. g. cai , x. m. zhang and y. z. zhang , `` assisted tachyonic inflation , '' phys . rev . d * 66 * , 121301 ( 2002 ) [ arxiv : hep - ph/0207143 ] . a. mazumdar , s. panda and a. perez - lorenzana , `` assisted inflation via tachyon condensation , '' nucl . b * 614 * , 101 ( 2001 ) [ arxiv : hep - ph/0107058 ] . j. m. cline and h. stoica , `` multibrane inflation and dynamical flattening of the inflaton potential , '' phys . d * 72 * , 126004 ( 2005 ) [ arxiv : hep - th/0508029 ] . d. wands , `` multiple field inflation , '' lect . notes phys . * 738 * , 275 ( 2008 ) [ arxiv : astro - ph/0702187 ] . c. gordon , d. wands , b. a. bassett and r. maartens , `` adiabatic and entropy perturbations from inflation , '' phys . d * 63 * , 023506 ( 2001 ) [ arxiv : astro - ph/0009131 ] . t. battefeld and r. easther , `` non - gaussianities in multi - field inflation , '' jcap * 0703 * , 020 ( 2007 ) [ arxiv : astro - ph/0610296 ] . d. battefeld and t. battefeld , `` non - gaussianities in n - flation , '' jcap * 0705 * , 012 ( 2007 ) [ arxiv : hep - th/0703012 ] . c. t. byrnes , k. y. choi and l. m. h. hall , `` conditions for large non - gaussianity in two - field slow - roll inflation , '' arxiv:0807.1101 [ astro - ph ] . a. ashoorioon and a. krause , `` power spectrum and signatures for cascade inflation , '' arxiv : hep - th/0607001 . s. watson , m. j. perry , g. l. kane and f. c. adams , `` inflation without inflaton(s ) , '' jcap * 0711 * , 017 ( 2007 ) [ arxiv : hep - th/0610054 ] . x. chen , r. easther and e. a. lim , `` large non - gaussianities in single field inflation , '' jcap * 0706 * , 023 ( 2007 ) [ arxiv : astro - ph/0611645 ] . l. covi , j. hamann , a. melchiorri , a. slosar and i. sorbera , `` inflation and wmap three year data : features have a future ! , '' phys . d * 74 * , 083509 ( 2006 ) [ arxiv : astro - ph/0606452 ] . a. berera , `` warm inflation , '' phys . lett . * 75 * , 3218 ( 1995 ) [ arxiv : astro - ph/9509049 ] . e. komatsu _ et al . _ , `` five - year wilkinson microwave anisotropy probe ( wmap ) observations : cosmological interpretation , '' arxiv:0803.0547 [ astro - ph ] . d. chialva and u. h. danielsson , `` chain inflation revisited , '' arxiv:0804.2846 [ hep - th ] . a. krause , `` large gravitational waves and lyth bound in multi brane inflation , '' arxiv:0708.4414 [ hep - th ] . j. yokoyama and a. d. linde , `` is warm inflation possible ? , '' phys . d * 60 * , 083509 ( 1999 ) [ arxiv : hep - ph/9809409 ] . v. f. mukhanov , h. a. feldman and r. h. brandenberger , `` theory of cosmological perturbations . classical perturbations . quantum theory of perturbations . extensions , '' phys . rept . * 215 * , 203 ( 1992 ) . d. h. lyth and a. riotto , `` particle physics models of inflation and the cosmological density perturbation , '' phys . * 314 * , 1 ( 1999 ) [ arxiv : hep - ph/9807278 ] . a. r. liddle and d. h. lyth , `` cosmological inflation and large - scale structure , '' _ cambridge , uk : univ . pr . ( 2000 ) 400 p _ a. kosowsky and m. s. turner , `` cbr anisotropy and the running of the scalar spectral index , '' phys . rev . d * 52 * , 1739 ( 1995 ) [ arxiv : astro - ph/9504071 ] . k. a. malik , d. wands and c. ungarelli , `` large - scale curvature and entropy perturbations for multiple interacting fluids , '' phys . d * 67 * , 063516 ( 2003 ) [ arxiv : astro - ph/0211602 ] . h. kodama and m. sasaki , `` cosmological perturbation theory , '' prog . * 78 * , 1 ( 1984 ) . a. ashoorioon , private discussion at cosmo 2008 . k. enqvist , a. jokinen , a. mazumdar , t. multamaki and a. vaihkonen , `` non - gaussianity from instant and tachyonic preheating , '' jcap * 0503 * , 010 ( 2005 ) [ arxiv : hep - ph/0501076 ] . n. barnaby and j. m. cline , `` nongaussian and nonscale - invariant perturbations from tachyonic preheating in hybrid inflation , '' phys . d * 73 * , 106012 ( 2006 ) [ arxiv : astro - ph/0601481 ] . n. barnaby and j. m. cline , `` nongaussianity from tachyonic preheating in hybrid inflation , '' phys . d * 75 * , 086004 ( 2007 ) [ arxiv : astro - ph/0611750 ] . k. ohmori , `` a review on tachyon condensation in open string field theories , '' arxiv : hep - th/0102085 . m. becker , g. curio and a. krause , `` de sitter vacua from heterotic m - theory , '' nucl . b * 693 * , 223 ( 2004 ) [ arxiv : hep - th/0403027 ] .
we are living in the age of `` big data '' , generally characterized by a series of `` vs '' . data are generated at an unprecedented _ volume _ by scientific instruments observing the macrocosm @xcite and the microcosm @xcite , or by humans connected around - the - clock to mobile platforms such as facebook and twitter . these data come in a _ variety _ of formats , ranging from delimited text to semi - structured json and multi - dimensional binaries such as fits . the volume and variety of `` big data '' pose serious problems to traditional database systems . before it is even possible to execute queries over a dataset , a relational schema has to be defined and data have to be loaded inside the database . schema definition imposes a strict structure on the data , which is expected to remain stable . however , this is rarely the case for rapidly evolving datasets represented using key - value and other semi - structured data formats , e.g. , json . data loading is a schema - driven process in which data are duplicated in the internal database representation to allow for efficient processing . even though storage is relatively cheap , generating and storing multiple copies of the same data can easily become a bottleneck for massive datasets . moreover , it is quite often the case that many of the attributes in the schema are never used in queries . motivated by the flexibility of nosql systems to access schema - less data and by the hadoop functionality to directly process data in any format , we have recently witnessed a sustained effort to bring these capabilities inside relational database management systems ( rdbms ) . starting with version 9.3 , postgresql includes support for json data type and corresponding functions . vertica flex zone and sinew @xcite implement flex table and column reservoir , respectively , for storing key - value data serialized as maps in a blob column . in both systems , certain keys can be promoted to individual columns , in storage as well as in a dynamically evolving schema . with regards to directly processing raw data , several query - driven extensions have been proposed to the loading and external table @xcite mechanisms . instead of loading all the columns before querying , in adaptive partial loading @xcite data are loaded only at query time , and only the attributes required by the query . this idea is further extended in invisible loading @xcite , where only a fragment of the queried columns are loaded , and in nodb @xcite , data vaults @xcite , sds / q @xcite , and raw @xcite , where columns are loaded only in memory , but not into the database . scanraw @xcite is a super - scalar pipeline operator that loads data speculatively , only when spare i / o resources are available . while these techniques enhance the rdbms flexibility to process schema - less raw data , they have several shortcomings , as the following examples show . * example 1 : twitter data . * the twitter api provides access to several objects in json format through a well - defined interface . the schema of the objects is , however , not well - defined , since it includes `` nullable '' attributes and nested objects . the state - of - the - art rdbms solution to process semi - structured json data @xcite is to first load the objects as tuples in a blob column . essentially , this entails complete data duplication , even though many of the object attributes are never used . the internal representation consists of a map of key - values that is serialized / deserialized into / from persistent storage . the map can be directly queried from sql based on the keys , treated as virtual attributes . as an optimization , certain columns chosen by the user or by the system based on appearance frequency are promoted to physical status . the decision on which columns to materialize is only an heuristic , quite often sub - optimal . * example 2 : sloan digital sky survey ( sdss ) data . * sdss is a decade - long astronomy project having the goal to build a catalog of all the astrophysical objects in the observable universe . images of the sky are taken by a high - resolution telescope , typically in binary fits format . the catalog data summarize quantities measured from the images for every detected object . the catalog is stored as binary fits tables . additionally , the catalog data are loaded into an rdbms and made available through standard sql queries . the loading process replicates multi - terabyte data three times in ascii csv and internal database representation and it can take several days if not weeks @xcite . in order to evaluate the effectiveness of the loading , we extract a workload of 1 million sql queries executed over the sdss catalog in 2014 . the most frequent table in the workload is ` photoprimary ` , which appears in more than 70% of the queries . ` photoprimary ` has 509 attributes , out of which only 74 are referenced in queries . this means that 435 attributes are replicated three times without ever being used a significantly sub - optimal storage utilization . * problem statement . * inspired by the above examples , in this paper , we study the raw data processing with partial loading problem . _ given a dataset in some raw format , a query workload , and a limited database storage budget , find what data to load in the database such that the overall workload execution time is minimized . _ this is a standard database optimization problem with bounded constraints , similar to vertical partitioning in physical database design @xcite . however , while physical design investigates what non - overlapping partitions to build over internal database data , we focus on what data to load , i.e. , replicate , in a columnar database with support for multiple storage formats . existing solutions for loading and raw data processing are not adequate for our problem . complete loading not only requires a significant amount of storage and takes a prohibitively long time , but is also unnecessary for many workloads . pure raw data processing solutions @xcite are not adequate either , because parsing semi - structured json data repeatedly is time - consuming . moreover , accessing data from the database is clearly optimal in the case of workloads with tens of queries . the drawback of query - driven , adaptive loading methods @xcite is that they are greedy , workload - agnostic . loading is decided based upon each query individually . it is easy to imagine a query order in which the first queries access non - frequent attributes that fill the storage budget , but have limited impact on the overall workload execution time . * contributions . * to the best of our knowledge , this is the first paper that incorporates query workload in raw data processing . this allows us to model raw data processing with partial loading as fully - replicated binary vertical partitioning . our contributions are guided by this equivalence . they can be summarized as follows : we provide a linear mixed integer programming optimization formulation that we prove to be np - hard and inapproximable . we design a two - stage heuristic that combines the concepts of query coverage and attribute usage frequency . the heuristic comes within close range of the optimal solution in a fraction of the time . we extend the optimization formulation and the heuristic to a restricted type of pipelined raw data processing . in the pipelined scenario , data access and extraction are executed concurrently . we evaluate the performance of the heuristic and the accuracy of the optimization formulation over three real data formats csv , fits , and json processed with a state - of - the - art pipelined operator for raw data processing . the results confirm the superior performance of the proposed heuristic over related vertical partitioning algorithms and the accuracy of the formulation in capturing the execution details of a real operator . * outline . * the paper is organized as follows . raw data processing , the formal statement of the problem , and an illustrative example are introduced in the preliminaries ( section [ sec : preliminaries ] ) . the mixed integer programming formulation and the proof that the formulation is np - hard are given in section [ sec : mip ] . the proposed heuristic is presented in detail in section [ sec : heuristic ] . the extension to pipelined raw data processing is discussed in section [ sec : pipeline ] . extensive experiments that evaluate the heuristic and verify the accuracy of the optimization formulation over three real data formats are presented in section [ sec : experiments ] . related work on vertical partitioning and raw data processing is briefly discussed in section [ sec : rel - work ] , while section [ sec : conclusions ] concludes the paper . in this section , we introduce query processing over raw data . then , we provide a formal problem statement and an illustrative example . query processing over raw data is depicted in figure [ fig : scanraw ] . the input to the process is a raw file from a non - volatile storage device , e.g. , disk or ssd , a schema that can include optional attributes , a procedure to extract tuples with the given schema from the raw file , and a driver query . the output is a tuple representation that can be processed by the query engine and , possibly , is materialized ( i.e. , loaded ) on the same storage device . in the ` read ` stage , data are read from the original raw file , page - by - page , using the file system s functionality . without additional information about the structure or the content stored inside the file or in some external structure the entire file has to be read the first time it is accessed . ` extract ` transforms tuples one per line from raw format into the processing representation , based on the schema provided and using the extraction procedure given as input to the process . there are two stages in ` extract``tokenize ` and ` parse ` . [ fig : scanraw ] ` tokenize ` identifies the schema attributes and outputs a vector containing the starting position for every attribute in the tuple or a subset , if the driver query does not access all the attributes . in ` parse ` , attributes are converted from raw format to the corresponding binary type and mapped to the processing representation of the tuple the record in a row - store , or the array in column - stores , respectively . multiple records or column arrays are grouped into a chunk the unit of processing . at the end of ` extract ` , data are loaded in memory and ready for query processing . multiple paths can be taken at this point . in external tables @xcite , data are passed to the query engine and discarded afterwards . in nodb @xcite and in - memory databases @xcite , data are kept in memory for subsequent processing . in standard database loading @xcite , data are first written to the database and only then query processing starts . scanraw @xcite invokes ` write ` concurrently with the query execution , only when spare i / o - bandwidth is available . the interaction between ` read ` and ` write ` is carefully scheduled in order to minimize interference . consider a relational schema @xmath0 and an instantiation of it that contains @xmath1 tuples . semi - structured json data can be mapped to the relational model by linearizing nested constructs @xcite . in order to execute queries over @xmath2 , tuples have to be read in memory and converted from the storage format into the processing representation . two timing components correspond to this process . @xmath3 is the time to read data from storage into memory . @xmath3 can be computed straightforwardly for a given schema and storage bandwidth @xmath4 . a constraint specific to raw file processing and row - store databases , for that matter is that all the attributes are read in a query even when not required . @xmath5 is the second timing component . it corresponds to the conversion time . for every attribute @xmath6 in the schema , the conversion is characterized by two parameters , defined at tuple level . the _ tokenizing time @xmath7 _ is the time to locate the attribute in a tuple in storage format . the _ parsing time @xmath8 _ is the time to convert the attribute from storage format into processing representation . a limited amount of storage @xmath9 is available for storing data converted into the processing representation . this eliminates the conversion and replaces it with an i / o process that operates at column level only complete columns can be saved in the processing format . the time to read an attribute @xmath6 in processing representation , @xmath10 , can be determined when the type of the attribute and @xmath1 are known . .query access pattern to raw data attributes . [ cols="<,^,^,^,^,^,^,^,^",options="header " , ] [ tbl : mip : params ] while written as a single constraint , @xmath11 decomposes into three separate constraints one corresponding to each `` @xmath12 '' operator for a total of @xmath13 constraints . @xmath14 is a reduced form of @xmath11 , applicable to query processing . the largest number of constraints , i.e. , @xmath15 , in the mip formulation are of type @xmath16 . they enforce that it is not possible to tokenize an attribute in a tuple without tokenizing all the preceding schema attributes in the same tuple . @xmath16 applies strictly to raw files without direct access to individual attributes . constraint @xmath17 guarantees that every attribute accessed in a query is either extracted from the raw file or read from the processing representation . there are @xmath18 binary 0/1 variables in the linear mip formulation , where @xmath19 is the number of queries in the workload and @xmath20 is the number of attributes in the schema . solving the mip directly is , thus , impractical for workloads with tens of queries over schemas with hundreds of attributes , unless the number of variables in the search space can be reduced . we prove that this is not possible by providing a reduction from a well - known np - hard problem to a restricted instance of the mip formulation . moreover , we also show that no approximation exists . [ def : k - elem - cover ] given a set of @xmath20 elements @xmath21 , @xmath19 subsets @xmath22 of @xmath2 , such that @xmath23 , and a value @xmath24 , the objective in the k - element cover problem is to find a size @xmath24 subset @xmath25 of @xmath2 that covers the largest number of subsets @xmath26 , i.e. , @xmath27 , @xmath28 . for the example in table [ tbl : prelim : problem : example ] , @xmath29 is the single 2-element cover solution ( covering @xmath30 ) . while many 3-element cover solutions exist , they all cover only one query . the k - element cover problem is a restricted instance of the mip formulation , in which parameters @xmath7 , @xmath8 , and the loading and reading time to / from database are set to zero , i.e. , @xmath31 , while the raw data reading time is set to one , i.e. , @xmath32 . the objective function is reduced to counting how many times raw data have to be accessed . the bounding constraint limits the number of attributes that can be loaded , i.e. , @xmath33 , while the functional constraints determine the value of the other variables . the optimal solution is given by the configuration that minimizes the number of queries accessing raw data . a query does not access raw data when the @xmath34 variables corresponding to its attributes are all set to one . when the entire workload is considered , this equates to finding those attributes that cover the largest number of queries , i.e. , finding the k - attribute cover of the workload . given this reduction , it suffices to prove that k - element cover is np - hard for the mip formulation to have only exponential - time solutions . we provide a reduction to the well - known minimum k - set coverage problem @xcite that proves k - element cover is np - hard . [ def : min - k - set - cover ] given a set of @xmath20 elements @xmath21 , @xmath19 subsets @xmath22 of @xmath2 , such that @xmath23 , and a value @xmath24 , the objective in the minimum k - set coverage problem is to choose @xmath24 sets @xmath35 from @xmath36 whose union has the smallest cardinality , i.e. , @xmath37 . set @xmath21 and @xmath19 subsets @xmath22 of @xmath2 ; number @xmath38 of sets @xmath26 to choose in minimum set coverage minimum number @xmath24 of elements from @xmath2 covered by choosing @xmath38 subsets from @xmath36 @xmath39 = * _ k - element cover_(@xmath36 , @xmath40 ) if * @xmath41 * then * * return * @xmath40 algorithm [ alg : reduction - cover - coverage ] gives a reduction from k - element cover to minimum k - set coverage . the solution to minimum k - set coverage is obtained by invoking k - element cover for any number of elements in @xmath2 and returning the smallest such number for which the solution to k - element cover contains at least @xmath38 subsets of @xmath36 . since we know that minimum k - set coverage is np - hard @xcite and the solution is obtained by solving k - element cover , it implies that k - element cover can not be any simpler , i.e. , k - element cover is also np - hard . the following theorem formalizes this argument . [ thm : reduction ] the reduction from k - element cover to minimum k - set coverage given in algorithm [ alg : reduction - cover - coverage ] is correct and complete . * _ proof . _ * in order to prove the theorem , we have to show that if the answer to the minimum k - set coverage problem is @xmath42 , algorithm [ alg : reduction - cover - coverage ] returns @xmath42 and if algorithm [ alg : reduction - cover - coverage ] returns @xmath42 , the answer to minimum k - set coverage problem is @xmath42 . we start with the first implication . let the optimal solution to the minimum k - set coverage problem be @xmath43 and @xmath44 be the set of elements in @xmath2 covered by @xmath45 , where @xmath46 and @xmath47 . suppose there exists a subset @xmath48 of @xmath2 and let the sets covered by @xmath48 be @xmath49 , where @xmath50 and @xmath51 . when @xmath51 , the union of any @xmath38 sets in @xmath49 is no larger than @xmath52 , which is smaller than @xmath42 . we get a contradiction . thus , there is no subset of @xmath2 whose size is smaller than @xmath42 that covers at least @xmath38 sets in @xmath36 . as a result , algorithm [ alg : reduction - cover - coverage ] does not return when @xmath53 . by the problem definition , we can use @xmath54 to cover at least @xmath38 sets @xmath45 . therefore , algorithm [ alg : reduction - cover - coverage ] returns when @xmath55 . for the second implication , let the elements chosen by the _ * k - element cover * _ function be @xmath54 and @xmath45 be the sets covered by @xmath54 , where @xmath47 . suppose the optimal solution to minimum k - set coverage is @xmath49 , which covers elements @xmath48 , where @xmath56 and @xmath57 . then , @xmath49 is the answer to the function @xmath58 . in this case , algorithm [ alg : reduction - cover - coverage ] returns @xmath52 before @xmath42 . we have a contradiction . therefore , we know that the optimal solution to minimum k - set coverage can not be smaller than @xmath59 . we can choose any @xmath38 sets in @xmath45 as the solution to the minimum k - set coverage problem . the union of the sets in @xmath45 is not larger than @xmath59 . therefore , the second implication holds . based on these two implications , we conclude that the reduction is correct . the fact that algorithm [ alg : reduction - cover - coverage ] has linear time complexity @xmath60 guarantees the completeness of the reduction . @xmath61 [ thm : reduction : corolar ] the mip formulation is np - hard and can not be approximated unless np - complete problems can be solved in randomized sub - exponential time . the np - hardness is a direct consequence of the reduction to the k - element cover problem and theorem [ thm : reduction ] . in addition , @xcite and @xcite prove that minimum k - set coverage can not be approximated within an absolute error of @xmath62 , for any @xmath63 , unless p = np . consequently , the mip formulation can not be approximated . in this section , we propose a novel heuristic algorithm for raw data processing with partial loading that has as a starting point a greedy solution for the k - element cover problem . the algorithm also includes elements from vertical partitioning a connection we establish in the paper . the central idea is to combine _ query coverage _ with _ attribute usage frequency _ in order to determine the best attributes to load . at a high level , query coverage aims at reducing the number of queries that require access to the raw data , while usage frequency aims at eliminating the repetitive extraction of the heavily - used attributes . our algorithm reconciles between these two conflicting criteria by optimally dividing the available loading budget across them , based on the format of the raw data and the query workload . the solution found by the algorithm is guaranteed to be as good as the solution corresponding to each criterion , considered separately . in the following , we make the connection with vertical partitioning clear . then , we present separate algorithms based on query coverage and attribute usage frequency . these algorithms are combined into the proposed heuristic algorithm for raw data processing with partial loading . we conclude the section with a detailed comparison between the proposed heuristic and algorithms designed specifically for vertical partitioning . vertical partitioning @xcite of a relational schema @xmath64 splits the schema into multiple schemas possibly overlapping each containing a subset of the columns in @xmath2 . for example , @xmath65 is the atomic non - overlapping vertical partitioning of @xmath2 in which each column is associated with a separate partition . tuple integrity can be maintained either by sorting all the partitions in the same order , i.e. , positional equivalence , or by pre - pending a tuple identifier ( @xmath66 ) column to every partition . vertical partitioning reduces the amount of data that have to be accessed by queries that operate on a small subset of columns since only the required columns have to be scanned when they form a partition . however , tuple reconstruction @xcite can become problematic when integrity is enforced through @xmath66 values because of joins between partitions . this interplay between having partitions that contain only the required columns and access confined to a minimum number of partitions , i.e. , a minimum number of joins , is the objective function to minimize in vertical partitioning . the process is always workload - driven . raw data processing with partial loading can be mapped to _ fully - replicated binary vertical partitioning _ as follows . the complete raw data containing all the attributes in schema @xmath2 represent the _ raw partition_. the second partition _ loaded partition _ is given by the attributes loaded in processing representation . these are a subset of the attributes in @xmath2 . the storage allocated to the loaded partition is bounded . the asymmetric nature of the two partitions differentiates raw data processing from standard vertical partitioning . the raw partition provides access to all the attributes , at the cost of tokenizing and parsing . the loaded partition provides faster access to a reduced set of attributes . in vertical partitioning , all the partitions are equivalent . while having only two partitions may be regarded as a simplification , all the top - down algorithms we are aware of @xcite apply binary splits recursively in order to find the optimal partitions . the structure of raw data processing with partial loading limits the number of splits to one . workload @xmath67 ; storage budget @xmath9 set of attributes @xmath68 to be loaded in processing representation [ a2:l1 ] @xmath69 ; @xmath70 [ a2:l2 ] [ a2:l3 ] @xmath71 [ a2:l4 ] * if * @xmath72 * then * * break * [ a2:l5 ] @xmath73 [ a2:l6 ] @xmath74 [ a2:l8 ] @xmath75 a query that can be processed without accessing the raw data is said to be _ covered_. in other words , all the attributes accessed by the query are loaded in processing representation . these are the queries whose attributes are contained in the solution to the k - element cover problem . intuitively , increasing the number of covered queries results in a reduction to the objective function , i.e. , total query workload execution time , since only the required attributes are accessed . moreover , access to the raw data and conversion are completely eliminated . however , given a limited storage budget , it is computationally infeasible to find the optimal set of attributes to load the k - element cover problem is np - hard and can not be approximated ( corollary [ thm : reduction : corolar ] ) . thus , heuristic algorithms are required . we design a standard greedy algorithm for the k - element cover problem that maximizes the number of covered queries within a limited storage budget . the pseudo - code is given in algorithm [ alg : query - coverage ] . the solution @xmath76 and the covered queries @xmath77 are initialized with the empty set in line [ a2:l1 ] . as long as the storage budget is not exhausted ( line [ a2:l2 ] ) and the value of the objective function @xmath78 decreases ( line [ a2:l4 ] ) , a query to be covered is selected at each step of the algorithm ( line [ a2:l3 ] ) . the criterion we use for selection is the reduction in the cost function normalized by the storage budget , i.e. , we select the query that provides the largest reduction in cost , while using the smallest storage . this criterion gives preference to queries that access a smaller number of attributes and is consistent with our idea of maximizing the number of covered queries . an alternative selection criterion is to drop the cost function and select the query that requires the least number of attributes to be added to the solution . the algorithm is guaranteed to stop when no storage budget is available or all the queries are covered . * example . * we illustrate how the _ * query coverage * _ algorithm works on the workload in table [ tbl : prelim : problem : example ] . without loss of generality , assume that all the attributes have the same size and the time to access raw data is considerably larger than the extraction time and the time to read data from processing representation , respectively . these is a common situation in practice , specific to delimited text file formats , e.g. , csv . let the storage budget be large enough to load three attributes , i.e. , @xmath79 . in the first step , only queries @xmath30 , @xmath80 , and @xmath81 are considered for coverage in line [ a2:l3 ] , due to the storage constraint . while the same objective function value is obtained for each query , @xmath30 is selected for loading because it provides the largest normalized reduction , i.e. , @xmath82 . the other two queries have a normalized reduction of @xmath83 , where @xmath3 is the time to read the raw data . in the second step of the algorithm , @xmath84 . this also turns to be the last step since no other query can be covered in the given storage budget . notice that , although @xmath80 and @xmath81 make better use of the budget , the overall objective function value is hardly different , as long as reading raw data is the dominating cost component . workload @xmath22 of @xmath2 ; storage budget @xmath9 ; set of loaded attributes @xmath85 set of attributes @xmath86 to be loaded in processing representation @xmath87 [ a3:l2 ] [ a3:l3 ] @xmath88 @xmath89 @xmath75 the query coverage strategy operates at query granularity . an attribute is always considered as part of the subset of attributes accessed by the query . it is never considered individually . this is problematic for at least two reasons . first , the storage budget can be under - utilized , since a situation where storage is available but no query can be covered , can appear during execution . second , a frequently - used attribute or an attribute with a time - consuming extraction may not get loaded if , for example , is part of only long queries . the assumption that accessing raw data is the dominant cost factor does not hold in this case . we address these deficiencies of the query coverage strategy by introducing a simple greedy algorithm that handles attributes individually . as the name implies , the intuition behind the attribute usage frequency algorithm is to load those attributes that appear frequently in queries . the rationale is to eliminate the extraction stages that incur the largest cost in the objective function . the pseudo - code for the attribute usage frequency strategy is given in algorithm [ alg : freq ] . in addition to the workload and the storage budget , a set of attributes already loaded in the processing representation is passed as argument . at each step ( line [ a3:l3 ] ) , the attribute that generates the largest decrease in the objective function is loaded . in this case , the algorithm stops only when the entire storage budget is exhausted ( line [ a3:l2 ] ) . * example . * we illustrate how the _ * attribute usage frequency * _ algorithm works by continuing the example started in the query coverage section . recall that only two attributes @xmath90 out of a total of three are loaded . @xmath91 is chosen as the remaining attribute to be loaded since it appears in five queries , the largest number between unloaded attributes . given that all the attributes have the same size and there is no cost for tuple reconstruction , @xmath92 is the optimal loading configuration for the example in table [ tbl : prelim : problem : example ] . workload @xmath67 ; storage budget @xmath9 set of attributes @xmath68 to be loaded in processing representation @xmath93 [ a4:l2 ] [ a4:l3 ] @xmath94 [ a4:l4 ] @xmath95 @xmath96 @xmath97 @xmath98 @xmath99 @xmath100 the heuristic algorithm for raw data processing with partial loading unifies the query coverage and attribute usage frequency algorithms . the pseudo - code is depicted in algorithm [ alg : combined ] . given a storage budget @xmath9 , _ * query coverage * _ is invoked first ( line [ a4:l3 ] ) . _ * attribute usage frequency * _ ( line [ a4:l4 ] ) takes as input the result produced by _ * query coverage * _ and the unused budget @xmath101 . instead of invoking these algorithms only once , with the given storage budget @xmath9 , we consider a series of allocations . @xmath9 is divided in @xmath102 increments ( line [ a4:l2 ] ) . each algorithm is assigned anywhere from @xmath103 to @xmath9 storage , in @xmath102 increments . a solution is computed for each of these configurations . the heuristic algorithm returns the solution with the minimum objective . the increment @xmath102 controls the complexity of the algorithm . specifically , the smaller @xmath102 is , the larger the number of invocations to the component algorithms . notice , though , that as long as @xmath104 remains constant with respect to @xmath19 and @xmath20 , the complexity of the heuristic remains @xmath105 . the rationale for using several budget allocations between query coverage and attribute usage frequency lies in the limited view they take for solving the optimization formulation . query coverage assumes that the access to the raw data is the most expensive cost component , i.e. , processing is i / o - bound , while attribute usage frequency focuses exclusively on the extraction , i.e. , processing is cpu - bound . however , the actual processing is heavily - dependent on the format of the data and the characteristics of the system . for example , binary formats , e.g. , fits , do not require extraction , while hierarchical text formats , e.g. , json , require complex parsing . moreover , the extraction complexity varies largely across data types . the proposed heuristic algorithm recognizes these impediments and solves many instances of the optimization formulation in order to identify the optimal solution . as discussed in section [ sec : heuristic : vertical - part ] , raw data processing with partial loading is a special case of vertical partitioning binary vertical partitioning with full replication . however , there is a fundamental difference between the problem addressed in this paper and standard vertical partitioning . the amount of storage allocated to partitions is not a parameter in vertical partitioning because it is constant and independent of the layout all the partitions use the same storage , plus - minus metadata . the bounded storage constraint is what makes raw data processing with partial loading a considerably more complicated problem , to which standard vertical partitioning algorithms are not directly applicable . a comprehensive comparison of vertical partitioning methods is given in @xcite . with few exceptions @xcite , vertical partitioning algorithms consider only the non - replicated case . when replication is considered , it is only partial replication . the bounded scenario limited storage budget for replicated attributes is discussed only in @xcite . at a high level , vertical partitioning algorithms can be classified along several axes @xcite . we discuss the two most relevant axes for the proposed heuristic . based on the direction in which partitions are built , we have top - down and bottom - up algorithms . a top - down algorithm @xcite starts with the complete schema and , at each step , splits it into two partitioned schemas . the process is repeated recursively for each resulting schema . a bottom - up algorithm @xcite starts with a series of schemas , e.g. , one for each attribute or one for each subset of attributes accessed in a query , and , at each step , merges a pair of schemas into a new single schema . in both cases , the process stops when the objective function can not be improved further . a second classification axis is given by the granularity at which the algorithm works . an attribute - level algorithm @xcite considers the attributes independent of the queries in which they appear . the interaction between attributes across queries still plays a significant role , though . a query or transaction - level algorithm @xcite works at query granularity . a partition contains either all or none of the attributes accessed in a query . based on the classification of vertical partitioning algorithms , the proposed heuristic qualifies primarily as a top - down query - level attribute - level algorithm . however , the recursion is only one - level deep , with the loaded partition at the bottom . the partitioning process consists of multiple steps , though . at each step , a new partition extracted from the raw data is merged into the loaded partition similar to a bottom - up algorithm . the query coverage algorithm gives the query granularity characteristic to the proposed heuristic , while attribute usage frequency provides the attribute - level property . overall , the proposed heuristic combines ideas from several classes of vertical partitioning algorithms , adapting their optimal behavior to raw data processing with partial loading . an experimental comparison with specific algorithms is presented in the experiments ( section [ sec : experiments ] ) and a discussion on their differences in the related work ( section [ sec : rel - work ] ) . in this section , we discuss on the feasibility of mip optimization in the case of pipelined raw data processing with partial loading . we consider a super - scalar pipeline architecture in which raw data access and the extraction stages tokenize and parse can be executed concurrently by overlapping disk i / o and cpu processing . this architecture is introduced in @xcite , where it is shown that , with a sufficiently large number of threads , raw data processing is an i / o - bound task . loading and accessing data from the processing representation are not considered as part of the pipeline since they can not be overlapped with raw data access due to i / o interference . we show that , in general , pipelined raw data processing with partial loading can not be modeled as a linear mip . however , we provide a linear formulation for a scenario that is common in practice , e.g. , binary fits and json format . in these cases , tokenization is atomic . it is executed for all or none of the attributes . this lets parsing as the single variable in the extraction stage . the mip formulation can not be solved efficiently , due to the large number of variables and constraints much larger than in the sequential formulation . we handle this problem by applying a simple modification to the heuristic introduced in section [ sec : heuristic ] that makes the algorithm feasible for pipelined processing . since raw data access and extraction are executed concurrently , the objective function corresponding to pipelined query processing has to include only the maximum of the two : @xmath106 this is the only modification to the mip formulation for sequential processing given in section [ sec : mip ] . since the @xmath107 function is non - linear , solving the modified formulation becomes impossible with standard mip solvers , e.g. , cplex , which work only for linear problems . the only alternative is to eliminate the @xmath107 function and linearize the objective . however , this can not be achieved in the general case . it can be achieved , though , for specific types of raw data binary formats that do not require tokenization , e.g. , fits , and text formats that require complete tuple tokenization , e.g. , json . as discussed in the introduction , these formats are used extensively in practice . queries over raw data can be classified into two categories based on the pipelined objective function in eq . ( [ eq : obj : pipeline : max ] ) . in i / o - bound queries , the time to access raw data is the dominant factor , i.e. , @xmath107 returns the first argument . in cpu - bound queries , the extraction time dominates , i.e. , @xmath107 returns the second argument . if the category of the query is known , @xmath107 can be immediately replaced with the correct argument and the mip formulation becomes linear . our approach is to incorporate the category of the query in the optimization as 0/1 variables . for each query @xmath40 , there is a variable for cpu - bound ( @xmath108 ) and one for io - bound ( @xmath109 ) . only one of them can take value 1 . moreover , these variables have to be paired with the variables for raw data access and extraction , respectively . variables of the form @xmath110 , @xmath111 , and @xmath112 correspond to the variables in table [ tbl : mip : vars ] , in the case of a cpu - bound query . variables @xmath113 , @xmath114 , and @xmath115 are for the io - bound case , respectively . with these variables , we can define the functional * constraints * for pipelined raw data processing : @xmath116 constraint @xmath117 forces a query to be either cpu - bound or io - bound . constraints @xmath118 tie the new family of cpu / io variables to their originals in the serial formulation . for example , the raw data is accessed in a cpu / io query only if it is accessed in the stand - alone query . the same holds for tokenizing / parsing a column @xmath119 in query @xmath40 . constraints @xmath120 and @xmath121 , respectively , tie the value of the cpu / io variables to the value of the corresponding query variable . for example , only when a query @xmath40 is cpu - bound , it makes sense for @xmath111 and @xmath112 to be allowed to take value 1 . if the query is io - bound , @xmath114 and @xmath115 can be set , but not @xmath111 and @xmath112 . at this point , we have still not defined when a query is cpu - bound and when is io - bound . this depends on the relationship between the time to access the raw data and the time to extract the referenced attributes . while the parsing time is completely determined by the attributes accessed in the query , the tokenizing time is problematic since it depends not only on the attributes , but also on their position in the schema . for example , in the sdss ` photoprimary ` table containing 509 attributes , the time to tokenize the @xmath122 attribute is considerably smaller than the time to tokenize the @xmath123 attribute . moreover , there is no linear relationship between the position in the schema and the tokenize time . for this reason , we can not distinguish between cpu- and io - bound queries in the general case . however , if there is no tokenization the case for binary formats such as fits or the tokenization involves all the attributes in the schema the case for hierarchical json format we can define a threshold @xmath124 that allows us to classify queries . @xmath125 is given by the ratio between the time to access raw data less the constant tokenize time and the average time to parse an attribute . intuitively , @xmath125 gives the number of attributes that can be parsed in the time required to access the raw data . if a query has to parse more than @xmath125 attributes , it is cpu - bound . otherwise , it is io - bound . the threshold * constraints * @xmath126 and @xmath127 make these definitions formal : @xmath128 for the atomic tokenization to hold , constraint @xmath16 in the serial formulation has to be replaced with @xmath129 . the complete pipelined mip includes the constraints in the serial formulation ( eq . ( [ eq : mip : formulation ] ) ) and the constraints @xmath130 . the linear * objective function * corresponding to query processing is re - written using the newly introduced variables as follows : @xmath131 since the number of variables and constraints increases with respect to the serial mip formulation , the task of a direct linear solver becomes even harder . it is also important to notice that the problem remains np - hard and can not be approximated since the reduction to the k - element cover still applies . in these conditions , heuristic algorithms are the only solution . we design a simple modification to the heuristic introduced in section [ sec : heuristic ] specifically targeted at pipelined raw data processing . given a configuration of attributes loaded in processing representation , the category of a query can be determined by evaluating the objective function . what is more important , though , is that the evolution of the query can be traced precisely as attributes get loaded . an i / o - bound query remains i / o - bound as long as not all of its corresponding attributes are loaded . at that point , it is not considered by the heuristic anymore . a cpu - bound query has the potential to become i / o - bound if the attributes that dominate the extraction get loaded . once i / o - bound , a query can not reverse to the cpu - bound state . thus , the only transitions a query can make are from cpu - bound to i / o - bound , and to loaded from there . if an io - bound query is not covered in the _ * query coverage * _ section of the heuristic , its contribution to the objective function can not be improved since it can not be completely covered by _ * attribute usage frequency_. based on this observation , the only strategy to reduce the cost is to select attributes that appear in cpu - bound queries . we enforce this by limiting the selection of the attributes considered in line [ a3:l3 ] of _ attribute usage frequency * _ to those attributes that appear in at least one cpu - bound query . the objective of the experimental evaluation is to investigate the accuracy and performance of the proposed heuristic across a variety of datasets and workloads executed sequentially and pipelined . to this end , we explore the accuracy of predicting the execution time for complex workloads over three raw data formats csv , fits , and json . additionally , the sensitivity of the heuristic is quantified with respect to the various configuration parameters . specifically , the experiments we design are targeted to answer the following questions : what is the impact of each stage in the overall behavior of the heuristic ? how accurate is the heuristic with respect to the optimal solution ? with respect to vertical partitioning algorithms ? how much faster is the heuristic compared to directly solving the mip formulation ? compared to other vertical partitioning algorithms ? can the heuristic exploit pipeline processing in partitioning ? do the mip model and the heuristic reflect reality across a variety of raw data formats ? * implementation . * we implement the heuristic and all the other algorithms referenced in the paper in ` c++ ` . we follow the description and the parameter settings given in the original paper as closely as possible . the loading and query execution plans returned by the optimization routine are executed with the scanraw @xcite operator for raw data processing . scanraw supports serial and pipelined execution . the real results returned by scanraw are used as reference . we use ibm cplex 12.6.1 to implement and solve the mip formulations . cplex supports parallel processing . the number of threads used in the optimization is determined dynamically at runtime . * we execute the experiments on a standard server with 2 amd opteron 6128 series 8-core processors ( 64 bit ) 16 cores 64 gb of memory , and four 2 tb 7200 rpm sas hard - drives configured raid-0 in software . each processor has 12 mb l3 cache while each core has 128 kb l1 and 512 kb l2 local caches . the storage system supports 240 , 436 and 1600 mb / second minimum , average , and maximum read rates , respectively based on the ubuntu disk utility . the cached and buffered read rates are 3 gb / second and 565 mb / second , respectively . ubuntu 14.04.2 smp @xmath132-bit with linux kernel 3.13.0 - 43 is the operating system . * methodology . * we perform all experiments at least 3 times and report the average value as the result . we enforce data to be read from disk by cleaning the file system buffers before the execution of every query in the workload . this is necessary in order to maintain the validity of the modeling parameters . * we use three types of real data formats in our experiments csv , fits , and json . the csv and fits data are downloaded from the sdss project using the cas tool . they correspond to the complete schema of the ` photoprimary ` table , which contains 509 attributes . the csv and fits data are identical . only their representation is different . csv is delimited text , while fits is in binary format . there are 5 million rows in each of these files . csv is 22 gb in size , while fits is only 19 gb . json is a lightweight semi - structured key - value data format . the twitter api provides access to user tweets in this format . tweets have a hierarchical structure that can be flattened into a relational schema . we acquire 5,420,000 tweets by making requests to the twitter api . there are at most 155 attributes in a tweet . the size of the data is 19 gb . * workloads . * we extract a real workload of 1 million sql queries executed over the sdss catalog in 2014 . out of these , we select the most popular 100 queries over table ` photoprimary ` and their corresponding frequency . these represent approximately 70% of the 1 million queries . we use these 100 queries as our workload in the experiments over csv and fits data . the weight of a query is given by its relative frequency . furthermore , we extract a subset of the 32 most popular queries and generate a second workload . the maximum number of attributes referenced in both workloads is 74 . we create the workload for the tweets data synthetically since we can not find a real workload that accesses more than a dozen of attributes . the number of attributes in a query is sampled from a normal distribution centered at 20 and having a standard deviation of 20 . the attributes in a query are randomly selected out of all the attributes in the schema or , alternatively , out of a subset of the attributes . the smaller the subset , the more attributes are not accessed in any query . the same weight is assigned to all the queries in the workload . in this set of experiments , we evaluate the sensitivity of the proposed heuristic with respect to the parameters of the problem , specifically , the number of queries in the workload and the storage budget . we study the impact each stage in the heuristic has on the overall accuracy . we measure the error incurred by the heuristic with respect to the optimal solution computed by cplex and the decrease in execution time . we also compare against several top - down vertical partitioning algorithms . we use the sdss data and workload in our evaluation . we consider the following vertical partitioning algorithms in our comparison : agrawal @xcite , navathe @xcite , and chu @xcite . the _ agrawal algorithm _ @xcite is a pruning - based algorithm in which all the possible column groups are generated based on the attribute co - occurrence in the query workload . for each column group , an interestingness measure is computed . since there is an exponential number of such column groups , only the `` interesting '' ones are considered as possible partitions . a column group is interesting if the interestingness measure , i.e. , cg - cost , is larger than a specified threshold . the interesting column groups are further ranked based on another measure , i.e. , vp - confidence , which quantifies the frequency with which the entire column group is referenced in queries . the attributes to load are determined by selecting column groups in the order given by vp - confidence , as long as the storage budget is not filled . while many strategies can be envisioned , our implementation is greedy . it chooses those attributes in a column group that are not already loaded and that minimize the objective function , one - at - a - time . the agrawal algorithm has exponential complexity @xmath133 since this is the number of potential column groups . this can be reduced by selecting the cg - cost threshold intelligently . however , this results in a corresponding accuracy decrease . trojan layouts @xcite are a newer version of the agrawal algorithm in which a different interestingness measure is defined and the selection of the partitions from the interesting column groups is done using an optimal exponential algorithm . since the authors admit that `` finding the right trojan layouts for scientific data sets ( having hundreds of attributes ) , like sdss , becomes a difficult task to achieve '' @xcite , we use the original agrawal algorithm in our implementation . the _ navathe algorithm _ @xcite starts with an affinity matrix that quantifies the frequency with which any pair of two attributes appear together in a query . the main step of the algorithm consists in finding a permutation of the rows and columns that groups attributes that co - occur together in queries . this extends upon the affinity of two attributes to a larger number of attributes . while finding the optimal permutation is exponential in the number of attributes , a quadratic greedy algorithm that starts with two random attributes and then chooses the best attribute to add and the best position , one - at - a - time , is given . these are computed based on a benefit function that is independent of the objective . the attributes are ordered on the benefit function in the resulting matrix . the final step of the algorithm consists in finding a split point along the attribute axis that generates two partitions with minimum objective function value across the query workload . an additional condition that we have to consider in our implementation is the storage budget we find the optimal partition that also fits in the available storage space . the _ chu algorithm _ @xcite considers only those partitions supported by at least one query in the workload , i.e. , a column group can be a partition only if it is accessed entirely by a query . moreover , a column group supported by a query is never split into smaller sub - parts . the algorithm enumerates all the column groups supported by any number of queries in the workload from a single query to all the queries and chooses the partition that minimizes the objective function . the remaining attributes not supported by the query form the second partition . this algorithm is exponential in the number of queries in the workload @xmath134 . the solution proposed in @xcite is to limit the number number of query combinations to a relatively small constant , e.g. , 5 . in our implementation , we let the algorithm run for a limited amount of time , e.g. , one hour , and report the best result at that time if the algorithm has not finished by that time . * heuristic stage analysis . * figure [ fig : combined_compare ] and [ fig : error ] depict the impact each stage in the heuristic query * coverage * and attribute usage * frequency * has on the accuracy , when taken separately and together , i.e. , * heuristic. we measure both the absolute value ( figure [ fig : combined_compare ] ) and the relative error with respect to the optimal value ( figure [ fig : error ] ) . we depict these values as a function of the storage budget , given as the number of attributes that can be loaded . we use the 32 queries workload . as expected , when the budget increases , the objective decreases . in terms of relative error , though , the heuristic is more accurate at the extremes small budget or large budget . when the budget is medium , the error is the highest the reason for this behavior is that , at the extremes , the number of choices for loading is considerably smaller and the heuristic finds a good enough solution . when the storage budget is medium , there are many loading choices and the heuristic makes only local optimal decisions that do not necessarily add - up to a good global solution . the two - stage heuristic has better accuracy than each stage taken separately . this is more clear in the case of the difficult problems with medium budget . between the two separate stages , none of them is dominating the other in all the cases . this proves that our integrated heuristic is the right choice since it always improves upon the best stage taken individually . serial heuristic accuracy . * figure [ fig : sequential_100_obj ] depicts the accuracy as a function of the storage budget for several algorithms in the case of serial raw data processing . the workload composed of 100 queries is used . out of the heuristic algorithms , the proposed heuristic is the most accurate . as already mentioned , the largest error is incurred when the budget is medium . between the vertical partitioning algorithms , the query - level granularity algorithm @xcite is the most accurate . the other two algorithms @xcite do not improve as the storage budget increases . this is because they are attribute - level algorithms that are not optimized for covering queries . * serial heuristic execution time . * figure [ fig : sequential_100_time ] depicts the execution time for the same scenario as in figure [ fig : sequential_100_obj ] . it is clear that the proposed heuristic is always the fastest , even by three orders of magnitude in the best case . surprisingly , calculating the exact solution using cplex is faster than all the vertical partitioning algorithms almost in all the cases . if an algorithm does not finish after one hour , we stop it and take the best solution at that moment . this is the case for chu @xcite and agrawal @xcite . however , the solution returned by chu is accurate a known fact from the original paper . * pipelined heuristic accuracy . * the objective function value for pipelined processing over fits data is depicted in figure [ fig : pipeline_100_obj ] . the same 100 query workload is used . the only difference compared to the serial case is that cplex can not find the optimal solution in less than one hour . however , it manages to find a good - enough solution in most cases . the proposed heuristic achieves the best accuracy for all the storage budgets . * pipelined heuristic execution time . * the proposed heuristic is the only solution that achieves sub - second execution time for all the storage budgets ( figure [ fig : pipeline_100_time ] ) . cplex finishes execution in the alloted time only when the budget is large . the number of variables and constraints in the pipeline mip formulation increase the search space beyond what the cplex algorithms can handle . we provide a series of case studies over different data formats in order to validate that the raw data processing architecture depicted in figure [ fig : scanraw ] is general and the mip models corresponding to this architecture fit reality . we use the implementation of the architecture in the scanraw operator @xcite as a baseline . for a given workload and loading plan , we measure the cumulative execution time after each query and compare the result with the estimation computed by the mip formulation . if the two match , this is a good indication that the mip formulation models reality accurately . the csv format maps directly to the raw data processing architecture . in order to apply the mip formulation , the value of the parameters has to be calibrated for a given system and a given input file . the time to tokenize @xmath7 and parse @xmath8 an attribute are the only parameters that require discussion . this can be done by executing the two stages on a sample of the data and measuring the average value of the parameter for each attribute . as long as accurate estimates are obtained , the model will be accurate . figure [ fig : csv ] confirms this on the sdss workload of 32 queries . in this case , there is a perfect match between the model and the scanraw execution . [ fig : csv ] since fits is a binary format , there is no extraction phase , i.e. , tokenizing and parsing , in the architecture . moreover , data can be read directly in the processing representation , as long as the file access library provides such a functionality . cfitsio the library we use in our implementation can read a range of values of an attribute in a pre - allocated memory buffer . however , we observed experimentally that , in order to access any attribute , there is a high startup time . essentially , the entire data are read in order to extract the attribute . the additional time is linear in the number of attributes . [ fig : fits ] based on these observations that may be specific to cfitsio the following parameters have to be calibrated : the time to read the raw data corresponds to the startup time ; an extraction time proportional with the number of attributes in the query is the equivalent of @xmath8 . @xmath7 is set to zero . although pipelining is an option for fits data , due to the specifics of the cfitsio library , it is impossible to apply it . the result for the sdss data confirms that the model is a good fit for fits data since there is almost complete overlap in figure [ fig : fits ] . at first sight , it seems impossible to map json data on the raw data processing architecture and the mip model . looking deeper , we observe that json data processing is even simpler than csv processing . this is because every object is fully - tokenized and parsed in an internal map data structure , independent of the requested attributes . at least this is how the jsoncpp library works . + once the map is built , it can be queried for any key in the schema . for schemas with a reduced number of hierarchical levels the case for tweets there is no difference in query time across levels . essentially , the query time is proportional only with the number of requested keys , independent of their existence or not . based on these observations , we set the model parameters as follows . @xmath7 is set to the average time to build the map divided by the maximum number of attributes in the schema . @xmath8 is set to the map data structure query time . since @xmath7 is a constant , the pipelined mip formulation applies to the json format . the results in figure [ fig : json ] confirm the accuracy of the model over a workload of 32 queries executed in scanraw . [ fig : json ] the experimental evaluation provides answers to each of the questions raised at the beginning of the section . the two - stage heuristic improves over each of the component parts . it is not clear which of the query coverage and attribute usage frequency is more accurate . using them together guarantees the best results . the proposed heuristic comes close to the optimal solution whenever the storage budget is either small or large . when many choices are available the case for a medium budget the accuracy decreases , but remains superior to the accuracy of the other vertical partitioning methods . in terms of execution time , the proposed heuristic is the clear winner by as much as three orders of magnitude . surprisingly , cplex outperforms the other heuristics in the serial case . this is not necessarily unexpected , given that these algorithms have been introduced more than two decades ago . the case studies confirm the applicability of the mip formulation model to several raw data formats . the mip model fits the reality almost perfectly both for serial and pipelined raw data processing . two lines of research are most relevant to the work presented in this paper raw data processing and vertical partitioning as a physical database design technique . our contribution is to integrate workload information in raw data processing and model the problem as vertical partitioning optimization . to the best of our knowledge , this is the first paper to consider the problem of optimal vertical partitioning for raw data processing with partial loading . * raw data processing . * several methods have been proposed for processing raw data within a database engine . the vast majority of them bring enhancements to the external table functionality , already supported by several major database servers @xcite . a common factor across many of these methods is that they do not consider loading converted data inside the database . at most , data are cached in memory on a query - by - query basis . this is the approach taken in nodb @xcite , data vaults @xcite , sds / q @xcite , raw @xcite , and impala @xcite . even when loading is an option , for example in adaptive partial loading @xcite , invisible loading @xcite , and scanraw @xcite , the workload is not taken into account and the storage budget is unlimited . the decision on what to load is local to every query , thus , prone to be acutely sub - optimal over the entire workload . the heuristic developed in this paper requires workload knowledge and aims to identify the optimal data to load such that the execution time of the entire workload is minimized . as in standard database processing , loading is executed offline , before query execution . however , the decision on what data to load is intelligent and the time spent on loading is limited by the allocated storage budget . notice that the heuristic is applicable both to secondary storage - based loading as well as to one - time in - memory caching without subsequent replacement . * vertical partitioning . * vertical partitioning has a long - standing history as a physical database design strategy , dating back to the 1970 s . many types of solutions have been proposed over the years , ranging from integer programming formulations to top - down and bottom - up heuristics that operate at the granularity of a query or of an attribute . a comparative analysis of several vertical partitioning algorithms is presented in @xcite . the serial mip formulation for raw data processing is inspired from the formulations for vertical partitioning given in @xcite . while both are non - linear , none of these formulations considers pipeline processing . we prove that even the linear mip formulation is np - hard . the scale of the previous results for solving mip optimizations have to be taken with a grain of salt , given the extensive enhancements to integer programming solvers over the past two decades . as explained in section [ sec : heuristic : comparison - vertical ] , the proposed heuristic combines ideas from several classes of vertical partitioning algorithms , adapting their optimal behavior to raw data processing with partial loading . the top - down transaction - level algorithm given in @xcite is the closest to the query coverage stage . while query coverage is a greedy algorithm , @xcite employs exhaustive search to find the solution . as the experimental results show , this is time - consuming . other top - down heuristics @xcite consider the interaction between attributes across the queries in the workload . the partitioning is guided by a quantitative parameter that measures the strength of the interaction . in @xcite , only the interaction between pairs of attributes is considered . the attribute usage frequency phase of the proposed heuristic treats each attribute individually , but only after query coverage is executed . the objective in @xcite is to find a set of vertical partitions that are subsequently evaluated for index creation . since we select a single partitioning scheme , the process is less time - consuming . finally , the difference between the proposed heuristic and bottom - up algorithms @xcite is that the latter can not guarantee that only two partitions are generated at the end . this is a requirement for raw data processing with partial loading . all these algorithms are offline . they are executed only once , before query processing , over a known workload . online vertical partitioning algorithms form a separate class . in @xcite , the entire workload is known in advance . however , the order of the queries is fixed and the vertical partitioning evolves . another series of algorithms @xcite operates over an unknown workload , given one query at a time . their goal is to gather evidence from the past workload in order to determine the optimal vertical partitioning at each query . in this paper , we study the problem of workload - driven raw data processing with partial loading . we model loading as binary vertical partitioning with full replication . based on this equivalence , we provide a linear mixed integer programming optimization formulation that we prove to be np - hard and inapproximable . we design a two - stage heuristic that combines the concepts of query coverage and attribute usage frequency . the heuristic comes within close range of the optimal solution in a fraction of the time . we extend the optimization formulation and the heuristic to a restricted type of pipelined raw data processing . in the pipelined scenario , data access and extraction are executed concurrently . we evaluate the performance of the heuristic and the accuracy of the optimization formulation over three real data formats csv , fits , and json processed with a state - of - the - art pipelined operator for raw data processing . the results confirm the superior performance of the proposed heuristic over related vertical partitioning algorithms and the accuracy of the formulation in capturing the execution details of a real operator . following the steps of database physical design , we envision several avenues to extend the proposed research in the future . we can move from the offline loading setting to online loading , where query processing and loading are intertwined . we can assume that the workload is known beforehand or it is given one query at a time . we can drop the strict requirement of atomic attribute loading and allow for portions horizontal partitions of an attribute to be loaded . finally , we can consider a multi - query processing environment in which raw data access and attribute extraction can be shared across several queries .
in the standard scenario of planet formation , planetesimals are the precursors of planets . their formation process is one of the unsolved problems of the planet formation theory . beginning with micron - sized dust grains , they grow to centimeter size in a protoplanetary disk via collisional agglomeration @xcite . the least understood growth phase is that of growth from centimeter size to kilometer size . since the gas drag of meter - sized aggregates is weak , they are slightly decoupled from gas with the sub - keplerian velocity . the resulting head - wind decreases their angular momentum and causes inward drift , whose timescale is about a few hundred years @xcite . the growth process in this stage must be faster than this drift timescale . another problem is the sticking probability of the aggregates . in this stage , since the threshold velocity for the destruction is low , their collisions may lead to destruction ( e.g. , * ? ? ? * ; * ? ? ? a very thin and dense layer of settled dust aggregates in the mid - plane of the protoplanetary disk is gravitationally unstable . then the gravitational collapse of the dust layer occurs , and kilometer - sized planetesimals are formed directly @xcite . this scenario has the advantage of a very rapid formation timescale , which is on the order of the keplerian time , thus avoiding the migration of meter - sized aggregates . however , turbulence in the protoplanetary disk may prevent dust from settling to the mid - plane . as dust settles in the protoplanetary disk , a vertical shear velocity develops because the dust - rich gas in the mid - plane rotates with the velocity closer to the keplerian velocity than that of the dust - depleted gas , which may cause turbulence . the turbulence can mix dust with gas and prevent the dust layer from settling into a dense enough sheet for gravitational instability @xcite . problems concerning the gravitational instability of the dust layer still remain unsolved . there are two methods for calculating the dynamics of dust particles in gas : fluid simulation and @xmath0-body simulation . one can consider the particles as virtually fluid and treat them with the euler equation or navier - stokes equation . for example , @xcite performed the two - dimensional numerical simulation to investigate the density evolution of the dust layer due to gravitational instability . they assumed that the dust layer is axisymmetric with respect to the rotational axis . they treated the dust component as a pressure - free fluid because the velocity dispersion of dust is negligible if the stopping time @xmath2 is sufficiently short , where the stopping time @xmath2 is the characteristic timescale for a particle to stop in gas . they found that the dust layer becomes extremely thin if the stopping time is long . @xcite performed the numerical simulation of the gravitational instability of a two - dimensional thin disk and investigated the non - axisymmetric modes . on the other hand , we can investigate the dynamics of particles in gas by using @xmath0-body simulations @xcite . this is straightforward and precise , but the calculation is very time - consuming and the practical number of particles is limited . @xcite performed @xmath0-body simulations of a gravitationally unstable disk with a local - shearing box . they investigated the formation of clumps of planetesimals at t30 au by the gravitational instability . they considered the drag force from the background gas . they showed that the planetesimal clumps form owing to the gravitational instability , which correspond to planetesimals in our calculation . in their simulation , the optical depth is smaller than that in our calculation and the particle disk is unstable even initially . we consider the formation of planetesimals at 1 au , thus we use the large optical depth , and assume the dust layer is initially stable . @xcite performed a set of local simulations of the self - gravitating collisional particle disks without gas using @xmath0-body simulation ( hereafter paper i ) . dust particle dynamics is calculated with the hill equations in a local - shearing box @xcite . they adopted the rubble pile model ( hard and soft sphere models ) as collision models . they found that the formation process is divided into three stages : the formation of non - axisymmetric wake - like structures @xcite , the creation of aggregates , and the rapid collisional growth of the aggregates . the mass of the largest aggregates is larger than the mass predicted by the linear perturbation theory . @xcite adopted the alternative model of collisions , ` accretion model ' ( paper ii ) . in the accretion model , the number of particles decreases as the calculation proceeds ; thereby this model enables us to perform large - scale and long - term simulations . they obtained the final mass of planetesimals as a function of the size of the computational box . @xcite performed the numerical simulations to investigate the validity of the super - particle approximation . they considered the various physics , such as the gas drag , the self - gravity , the physical collision and the turbulence . they treated the collisions as the hard sphere model . they investigated the numerical requirements to study the gravitational instability . to understand the basic dynamics of gravitational instability , we neglected the effect of gas in papers i and ii . however , it is obvious that the gas plays an important role in planetesimal formation when dust particles are small . the interaction with gas through drag is dominant in the dynamical evolution of particles . many authors investigated the effect of gas . the particles drift radially due to gas drag @xcite . as the sedimentation of dust aggregates towards the midplane proceeds , the vertical velocity shear increases . thus , the kelvin - helmholtz instability occurs @xcite . the kelvin - helmholtz instability makes the dust layer turbulent . the turbulent gas prevents the concentration of dust @xcite , and helps the concentration @xcite . the interaction between gas and dust causes the streaming instability @xcite . by the streaming instability , the dust grains concentrate strongly and the large ceres - sized planetesimals form @xcite . the loss of angular momentum due to the gas drag helps gravitational instability @xcite . the time - scale for the dissipative instability is relatively slow , but the dust layer may be unstable even when the toomre s q value is larger than unity . as a first step toward understanding the effect of gas on gravitational instability , we introduce gas as a background laminar flow in the present paper . the aim of the present paper is to investigate the effect of the laminar gas on the planetesimal formation through gravitational instability . the laminar flow causes the dissipation of the kinetic energy and thus radial migration of dust . in 2 , we summarize the results of the linear analysis of the gravitational instability under gas drag . in 3 , we describe the simulation method and the initial condition . in 4 , we present our numerical results . 5 is devoted to discussions . we summarize the results in 6 . the linear stability analyses of the dust layer with a finite thickness were performed by @xcite and @xcite . they assumed that the size of dust aggregates is sufficiently small , in other words , dust is fully coupled with gas , thus they treated gas and dust as one fluid . that introduces no dissipation due to gas drag . they considered the vertical structure of the dust layer , and studied the roche criterion . the linear stability analyses of dissipative gravitational instability were performed by @xcite , @xcite , and @xcite . in this section , we summarize the essence of their results . @xcite and @xcite used one component model , i.e. , they neglected the dynamics of gas flow , but they consider the gas drag from the stationary background gas flow . using their results , we discuss conditions for the gravitational instability under the influence of gas drag . we adopt the isothermal equation of state for the dust layer : @xmath3 where @xmath4 is the pressure , @xmath5 is the isothermal sound speed , and @xmath6 is the density of the dust layer . in this equation of state , the velocity dispersion of dust is always constant . clearly , the velocity dispersion must not be constant . it changes owing to several processes such as gas drag , inelastic collisions , and gravitational scattering . this assumption gives us the simple analytical expression of the dispersion relation , with which we can grasp the nature of dissipative gravitational instability . we assume that the gas velocity is equal to the kepler velocity . we consider the reference point that is at the semi - major axis @xmath7 . the kepler angular velocity of the reference point is @xmath8 where @xmath9 is the mass of the central star and @xmath10 is the gravitational constant . we introduce the local cartesian coordinates in which the @xmath11-axis is directed radially , the @xmath12-axis follows the direction of rotation , and the @xmath13-axis follows the direction perpendicular to mid - plane . we define @xmath14 as the deviation velocity field from the local kepler velocity . the basic equations for the dust layer in this frame are given by @xmath15 @xmath16 @xmath17 where @xmath18 is the gravitational potential due to the self - gravity . the simulation results in 4 show that the density fluctuation is not axisymmetric . however , for the sake of simplicity , restricting ourselves to the axisymmetric mode ( @xmath19 ) , we carry out a normal mode analysis in the form of @xmath20 where @xmath21 is the wave number in the @xmath11-direction and @xmath22 is the growth rate . integrating toward @xmath13 , we obtain the equation of perturbed quantities @xmath23 : @xmath24 @xmath25 @xmath26 @xmath27 where @xmath28 is the surface density of the dust layer . @xcite considered the finite thickness model using the softening term @xcite . poisson equation with the softening parameter is @xmath29 where @xmath30 is the scale height of the dust layer and @xmath31 is the softening parameter : @xmath32 for nontrivial solutions , the determinant of the coefficient matrix must vanish . thus the following dispersion relation is obtained @xcite : @xmath33 where @xmath34 . the mode for @xmath35 is unstable , where @xmath36 is the real part of a complex number @xmath37 . in the gas - free limit ( @xmath38 ) and the thin disk limit ( @xmath39 ) , the dispersion relation converges to the conventional dispersion relation of a thin disk @xcite . equation ( [ eq : disp1 ] ) can be written in a non - dimensional form scaled by the length @xmath40 and the time @xmath41 : @xmath42 where we used toomre s @xmath1 given by @xmath43 . a tilde denotes non - dimensional quantity hereafter . a dissipative dust layer is secularly unstable for long - wavelength modes although @xmath44 . the physical mechanism of the instability is explained as follows @xcite . we consider a thin axisymmetric overdense ring and the thick back - ground gas . from the force balance of the self - gravity , we find that the dust rotates faster at the outer edge of the ring . thus , the drag force at the outer edge is strong . this causes the inward drift . conversely , the dust at the inner edge drifts outward . these effects shrink ring radially . figure [ fig : disper.eps ] shows the dispersion relation of the dust layer under gas drag for @xmath45 . we used the thin disk model ( @xmath46 ) . the stopping time is @xmath47 , @xmath48 , @xmath49 , and @xmath50 ( gas - free ) . all models have the maximum growth rate at @xmath51 . we will see that the growth rate has the maximum value at @xmath52 in [ sec : mugr ] . as the stopping time lengthens , the growth rate shrinks . in the gas - free model , long - wavelength modes are stable , while in the gas model , they are unstable . as discussed in [ sec : dispersion ] , the dust layer is always unstable for long - wavelength modes . however , the timescale of the instability can be longer than the dynamical timescale , because this is the secular instability . therefore , in order to understand what happens , we have to compare the timescale of the instability with those of other processes such as the sedimentation , the radial migration , and the growth of dust aggregates . here , we assume the thin disk approximation ( @xmath46 ) in order to obtain the simple analytical formula . for @xmath53 , @xmath54 is a real positive value . the growth rate for the most unstable mode @xmath54 is a function of @xmath55 , @xmath56 , and @xmath1 . from equation ( [ eq : disp2 ] ) , the partial derivative of @xmath54 with respect to @xmath55 is given as @xmath57 using equation ( [ eq : disp2 ] ) , we can show that the denominator on the right hand side of equation ( [ eq : disp3 ] ) is positive when @xmath58 . hence , the @xmath54 has the maximum value at @xmath59 for a fixed @xmath56 and @xmath1 . the substitution of @xmath59 into equation ( [ eq : disp2 ] ) provides the equation of the maximum growth rate @xmath60 : @xmath61 to obtain the condition that the gravitational instability dominates over the other process ( symbolically denoted by `` x '' ) , we equate @xmath62 with @xmath63 where @xmath64 is the characteristic timescale of a process x , ans solve the resultant equation with respect to q to obtain @xmath1 value : @xmath65 when @xmath66 , gravitational instability is a dominant process compared to the process x. on the other hand , when @xmath67 , the timescale for the process x is shorter than the timescale of gravitational instability . although the dust layer is unstable , the process x is dominant . we can derive the leading term of the power series expansion with respect to @xmath2 for equation ( 13 ) and find the asymptotic solution @xcite . when the gas drag is strong @xmath68 , the growth rate for the most unstable mode is @xmath69 conversely , when the gas drag force is weak @xmath70 and @xmath44 , the growth rate for the most unstable mode is @xmath71 when @xmath72 , the growth rate for the most unstable mode converges to the finite value @xmath73 as @xmath74 , which is the maximum growth rate for the gas - free gravitational instability . the method of calculation is the same as those used in papers i and ii except for including gas , which is based on the method of the local simulation of planetary rings ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? we describe quantities in the non - dimensional form independent of the semi - major axis @xmath7 , and the mass of the central star @xmath9 by scaling the time by @xmath75 , the length by the hill radius @xmath76 , and the mass by @xmath77 , where @xmath78 and @xmath79 is the initial mass of particles @xcite . the initial mass of particles @xmath79 is assumed to be identical . the non - dimensional hill equations for the particle @xmath80 are given by ( e.g. , * ? ? ? * ) : @xmath81 where @xmath82 , @xmath83 , and @xmath84 are the position , mass , and the stopping time of the particle @xmath80 , @xmath85 is the distance between particles @xmath80 and @xmath86 , and @xmath87 is the velocity field of gas . we neglect the dust back - reaction on gas for the sake of simplicity . we assume that the laminar gas flow , which has the sub - keplerian velocity field : @xmath88 where @xmath89 is the keplerian shear velocity , @xmath90 is the difference from the keplerian velocity . the stopping time @xmath84 depends on the size of particle : @xmath91 where @xmath92 is the power - law index , and @xmath93 is the stopping time of the particle with @xmath94 . in this paper , we adopt @xmath95 and 2/3 . the index @xmath96 corresponds to stokes law ( e.g. , * ? ? ? the calculation box has the periodic boundary condition ( e.g. , * ? ? ? the size of the box is taken as a square @xmath97 where @xmath98 is the most unstable wavelength of gravitational instability and @xmath99 is the non - dimensional parameter . the system is characterized by two non - dimensional parameters , the optical depth and the ratio of the hill radius to the diameter of a particle @xcite : @xmath100 @xmath101 where @xmath102 and @xmath103 are the internal density and initial radius of the particle , respectively . the realistic value for `` solid '' dust particles is @xmath104 for @xmath105 and @xmath106 . however , we adopt @xmath107 in this paper owing to the computational limit ( see papers i and ii ) . we set the optical depth @xmath108 , which is approximately equal to the realistic value @xmath109 for @xmath110 , @xmath111 , and @xmath105 . since our choice of @xmath112 is smaller than the realistic value , a particle used in the simulation is not a realistic dust particle but a ` super - particle ' that represents a group of many small particles that have the same position and velocity . thus the restitution coefficient @xmath113 corresponds to the rate of the dissipation due to collisions between super - particles . the low value of @xmath112 means that the physical size of a particle in our simulation is very large . this implies that the effect of the gravitational scattering is relatively weaker than that of collisions . we investigated the @xmath112 value dependence in paper i and ii in the narrow range ( @xmath114 ) and we confirmed that the formation process does not change in this range qualitatively . however , we show that the collisional growth is important in the final stage where @xmath115 controls the growth rate . thus , we should perform the large @xmath112 value . in the future work , we will concentrate on the collisional growth and perform the @xmath0-body simulations with a realistic @xmath112 value . in the hill coordinates , the keplerian orbit is determined by six parameters : the position of the guiding center , eccentricity @xmath116 , inclination @xmath80 , and two phases for epicyclic and vertical oscillations @xcite . the initial eccentricity and inclination of particles are assumed to follow the rayleigh distribution . we fix @xmath117 according to @xcite . the other parameters are uniformly distributed , avoiding overlapping . we adopt the hard - sphere and accretion models as a collision model ( papers i and ii ) . in the hard - sphere model , the penetration of particles is not possible ( e.g. , paper i , * ? ? ? when a collision occurs , the particle velocity changes instantly . the relative tangential velocity is conserved and the magnitude of the relative normal velocity is reduced by a factor of @xmath113 . we use the constant @xmath113 . we also use the accretion model ( e.g. , paper ii , * ? ? ? * ) . in this model , when two particles collide , if the binding condition is satisfied , the two particles merge into one particle . the binding condition is described by the jacobi integral @xmath118 ( e.g. , paper ii , * ? ? ? * ) . in the hard - sphere model , since a planetesimal consists of many small particles , it can be distorted in shape and break up . on the other hand , in the accretion model , a planetesimal can not be distorted or break up . in this respect , the hard - sphere model is more realistic than the accretion model . however the accretion model has an advantage . first , since the number of particles decreases as the calculation proceeds , this model enables us to reduce the calculation cost . second , we can easily handle size - dependent drag coefficient . in general , the drag coefficient depends on the size of a particle . in the hard sphere model , it is complicated to handle the size - dependent drag coefficient . in the accretion model , we can adopt the size - dependent drag without any difficulty . we have shown that results in the accretion model are the same as those in the hard sphere model in the gas - free model ( paper ii ) . we compare the accretion model with the hard sphere model in the laminar gas flow . we performed simulations with the same model parameters except for collision models ( models 100h0 , 100ac0 , and 100as0 ) in order to check the validity of the accretion model . the initial stopping time of particles is @xmath119 . in the accretion models , we calculate the drag coefficient using the constant model ( @xmath120 ) and the stokes law model ( @xmath96 ) . the left panel of figure [ fig : q_value_accretion.eps ] shows the time evolution of @xmath1 . in the phase where @xmath1 decreases , there are no clear differences among all models . on the other hand , in the phase where @xmath1 increases , the discrepancies appear among them . @xmath1 value in the accretion model with @xmath96 is larger than those in the hard sphere model and the accretion model with @xmath120 . in the stokes law model , as particles grow , the drag coefficient becomes small and thus the dissipation becomes inefficient , which leads to the relatively large velocity dispersion . the right panel of figure [ fig : q_value_accretion.eps ] shows the maximum mass of planetesimals for all models . in the early stage , particles or aggregates do not grow . there the difference in collision models is not important . the difference can be seen in the late stage . in the stokes law model @xmath96 , the drag coefficient changes as particles grow , which affects the evolution of the maximum mass . in the hard - sphere model , the drag coefficient is constant although aggregates grow , therefore the evolution in the hard - sphere model must be similar to that in the accretion model with @xmath120 . however , the difference appears at @xmath121 . the maximum mass of the accretion model with @xmath120 increases , but that of the hard - sphere model does not change . at @xmath121 , there are only a few particles or aggregates . the maximum mass of planetesimals is sensitive to the initial condition and noise . but , at @xmath122 , the maximum mass converges to the same value . the final mass of the planetesimal depends on the size of the computational domain @xcite . the dispersion of the mass of the final state is relatively small . in the early stage where no planetesimals form , the time evolution is quite similar in all models . in the late stage , once planetesimals form , there are no remarkable differences among all models . if we focus on the the evolution of statistical value , such as the velocity dispersion , the number of planetesimals , we can investigate the gravitational instability and planetesimal formation using the accretion model . although the time evolution of the maximum mass of the planetesimal in the late stage depends on the initial noise , the maximum mass finally converge to the same value . therefore , we adopt the accretion model to investigate the gravitational instability from now on . we performed 14 simulations with different disk models . the model parameters are summarized in table @xmath123 . we fixed the following parameters : the optical depth , @xmath124 , the restitution coefficient @xmath125 , the ratio of the hill radius to the diameter @xmath126 , the initial toomre s @xmath1 value @xmath127 , and the size of the computational domain @xmath128 . the basic dependence of planetesimal formation of these parameters was investigated in papers i and ii . ccccc 010as0 & 0.10 & accretion & 2/3 & 0 + 025as0 & 0.25 & accretion & 2/3 & 0 + 050as0 & 0.50 & accretion & 2/3 & 0 + 100as0 & 1.00 & accretion & 2/3 & 0 + 200as0 & 2.00 & accretion & 2/3 & 0 + 400as0 & 4.00 & accretion & 2/3 & 0 + 1000as0 & 10.00 & accretion & 2/3 & 0 + infa & @xmath129 & accretion & - & 0 + 100ac0 & 1.00 & accretion & 0 & 0 + 100h0 & 1.00 & hard & 0 & 0 + 100ac1 & 1.00 & accretion & 0 & 10 + 100as1 & 1.00 & accretion & 2/3 & 10 + 100ac2 & 1.00 & accretion & 0 & 20 + 100as2 & 1.00 & accretion & 2/3 & 20 + [ table : model2 ] figure [ fig : snap_shot ] shows the typical evolution of the simulation ( model 100as0 ) . at @xmath130 , and @xmath47 , particles are distributed randomly and uniformly in the computational box . no gravitational instability occurs , thus we can see no remarkable structures . at @xmath131 , the large non - axisymmetric density structures appear . then , the gravitational instability seems to start . in the model where ( @xmath132 ) , the structure is not clear . for the short stopping time models , the velocity dispersion is small and the critical wave length is short . thus , we can not observe the large density structure clearly . at @xmath133 , particles begin to grow in the dense region . at @xmath134 and @xmath48 , particles continue to grow , and the number of small particles decreases rapidly . the small particles are absorbed by large particles . the rapid collisional growth of particles continues until the one large planetesimal finally forms . the formation process of planetesimals in the laminar gas model is similar to those in the gas - free model ( papers i and ii ) . the gas drag from the laminar gas does not qualitatively change formation process . this process is divided into three stages : the formation of density structures , the creation of planetesimal seeds , and their collisional growth . the linear stability analysis of gas - free gravitational instability shows that the disk is stable when @xmath44 @xcite . as discussed in 3 , the dust layer in the laminar gas disk is secularly unstable although @xmath44 . however , the growth time of the dissipative gravitational instability is longer than the sedimentation in the initial stage . thus , the disk shrinks vertically . as the velocity dispersion and the scale height decrease , the growth rate of the dissipative gravitational instability increases . then , the gravitational instability becomes a dominant process , and the wake - like structures appear . these structures fragment into the seeds of planetesimals . the seeds grow rapidly owing to mutual collisions . finally , almost all mass in the computational domain is absorbed by only one planetesimal in this calculation . the size of the final planetesimal depends on the size of the computational domain ( paper ii ) . [ cols="^,^,^,^ " , ] we assume that the difference in the rotational velocity from the kepler velocity is equal to zero ( @xmath135 ) ( models infa , 010as0 - 1000as0 ) . in the model infa , @xmath136 , i.e. , the drag term is neglected . in other models , the stopping times are @xmath137 and @xmath138 . we adopt the stokes law model , @xmath139 . the top panel of figure [ fig : q_value_hard.eps ] shows the time evolution of toomre s @xmath1 values . in all models , toomre s @xmath1 value initially decreases . in the gas - free model ( model infa ) , the kinetic energy is dissipated only by inelastic collisions . the dissipation timescale due to inelastic collisions is approximately estimated by @xmath140 @xcite . in the laminar gas models , the decrease of toomre s @xmath1 value is caused by the dissipation due to both the gas drag and inelastic collisions . the top panel of figure [ fig : min_q_value.eps ] shows the dependence of the decay time @xmath141 on @xmath142 , where the decay time of toomre s @xmath1 value @xmath141 is defined by @xmath143 where @xmath144 is the minimum @xmath1 value at @xmath145 . the decay time @xmath141 is proportional to @xmath142 for @xmath146 . @xmath1 value is proportional to the velocity dispersion @xmath147 . the inelastic collisions and the gas drag decrease the velocity dispersion . if the drag is sufficiently strong , the decay time scale of the velocity dispersion is on the order of the stopping time . the timescale of the decay of the velocity dispersion due to inelastic collisions is about @xmath148 . thus , when @xmath149 as @xmath150 increases , we can not neglect collisions . therefore the decay timescale of the velocity dispersion is smaller than the stopping time for @xmath149 . in all models , toomre s @xmath1 value has the minimum value @xmath144 at @xmath151 . in the gas - free model , the minimum @xmath1 value is about @xmath152 ( papers i , and ii ) , and according to the linear theory , the critical value of a thin disk is @xmath153 . this is because the decrease in the velocity dispersion due to inelastic collisions is sufficiently slow . however , as shown in figure [ fig : q_value_hard.eps ] , if the gas drag is strong , @xmath1 value becomes smaller than unity . as the stopping time shortens , the minimum @xmath1 value decreases . for example , the minimum @xmath1 value is 0.15 for @xmath154 . the axisymmetric structure starts forming when @xmath1 value is minimum . for example , in the model where @xmath155 , @xmath1 value is minimum at @xmath156 , and we can see the non - axisymmetric structure at @xmath157 . particles are scattered by the non - axisymmetric structures , and the velocity dispersion of particles increases . if the decrease of the @xmath1 value is more rapid than the gravitational instability , the dust layer can not collapse although the dust layer is gravitationally unstable . as @xmath1 value decreases , the timescale of the gravitational instability becomes short , therefore , the critical @xmath1 value is determined by @xmath158 , where @xmath159 is the timescale of the gravitational instability . here , we assume @xmath160 because particles do not grow much before gravitational instability . from equation ( [ eq : crit_q ] ) , we obtain the following condition : @xmath161 the minimum toomre s @xmath1 value as a function of @xmath142 is shown in the bottom panel of figure [ fig : min_q_value.eps ] . strictly speaking , equation ( [ eq : qmin ] ) is valid only when the disk is thin . nevertheless , the analytical expression agrees with the numerical result . we define the scale height of a dust layer as the mean square root of @xmath13 , @xmath162 . we can calculate the mean density of the dust layer from the scale height , @xmath163 . the density at which the self - gravity is equal to the tidal force is called the roche density , which is defined by the following equation ( e.g. , * ? ? ? * ) : @xmath164 the dust layer may collapse if the dust layer density exceeds the roche density . the linear analysis gives the more precise criterion @xcite . however , the criterion does not change very much . thus , we introduce @xmath165 given by @xmath166 which corresponds to the non - dimensional scale height . the bottom panel of figure [ fig : q_value_hard.eps ] shows the time evolution of @xmath167 . the time evolution of @xmath167 is similar to toomre s @xmath1 . the value @xmath167 has the minimum value @xmath168 at @xmath169 in all models . but , the time evolution of @xmath167 is slower than that of @xmath1 value . this result indicates that the standard hydrostatic relation @xmath170 is not satisfied . figure [ fig : v_over_h ] shows the time evolution of the ratio @xmath171 , which is on the order of the sound crossing time across a scale height . in the gas - free model , the ratio @xmath172 is nearly constant in @xmath173 . the hydrostatic relation is satisfied . in a laminar gas model , the ratio @xmath171 is not constant . as the stopping time decreases , the variation of the ratio becomes steep . especially , those for @xmath174 is prominent . in the model @xmath132 , the ratio @xmath171 increases for @xmath175 . the maximum value is about @xmath176 . the sound crossing time is @xmath177 in the gas - free model . when @xmath146 , the dissipation is faster than the sound crossing . therefore , the velocity dispersion decays before the scale height decays , and the ratio becomes larger than unity . in figure [ fig : v_over_h ] , we can see the oscillation . the period is about @xmath178 , and the amplitude decreases gradually . the reason for the oscillation is not clear . this may be due to some kind of the relaxation process . we can not use the roche density as the collapse criterion for @xmath179 . for @xmath180 , @xmath167 is always larger than unity . but the large particles form . for @xmath181 , @xmath167 becomes unity at @xmath182 , but the large particles start forming at @xmath133 . for @xmath183 , the time when @xmath184 corresponds to the time when the large particles start forming . when we derive the roche criterion , we simply compare the self - gravity with the tidal force . however , we showed that the non - axisymmetric structure develops . the azimuthal motion may be important to discuss the formation of the wake - like structure . therefore , the roche criterion is not applicable . the decay time of @xmath167 value @xmath185 is defined by @xmath186 where @xmath187 is the initial scale height of a dust layer . figure [ fig : min_height_time.eps ] shows @xmath185 as a function of @xmath142 . the decay time of @xmath167 has a minimum value at @xmath188 . for @xmath189 , the decay time shortens with the stopping time . for @xmath189 , the strong coupled particles with gas fall to the midplane at the terminal velocity . the settling time is proportional to @xmath190 . on the other hand , for @xmath191 , the decay time lengthens with longer stopping time . in this regime , the particles oscillate around the equatorial plane @xcite . the decay time scale of the amplitude is proportional to the stopping time . figure [ fig : cluster_number.eps ] shows the time evolution of the number of planetesimal seeds per the area @xmath192 , @xmath193 . we define a particle whose mass is larger than @xmath194 as a seed of planetesimals , where @xmath194 is the planetesimal mass predicted by the linear theory @xmath195 . in all models , @xmath193 has a maximum value at @xmath196 . the number @xmath193 increases rapidly in the early stage , and decreases in the late stage . in the early stage , gravitational instability occurs and the small density fluctuation grows and many planetesimal seeds form . in this stage , the number of seeds increases . they form through the gravitational instability . once many planetesimal seeds form , the mutual collisions among them start . the number of planetesimals decreases rapidly in the late stage . in the final state , almost all particles are absorbed by a few particles . the number of planetesimals in the final state depends on the size of the computational domain ( paper ii ) . the peak of @xmath193 depends on the stopping time @xmath150 . figure [ fig : max_number.eps ] shows the maximum number of the planetesimal @xmath197 as a function of the stopping time @xmath150 . the maximum number of planetesimals @xmath197 has a peak at @xmath198 . as discussed in [ sec : scaleheight ] , the scale height has the minimum value at @xmath198 . the characteristic length scale of gravitational instability corresponds to the scale height . as the characteristic length of gravitational instability shortens , the number of the planetesimals formed increases . figure [ fig : largest_planetesimal.eps ] shows the time evolution of the mass of the largest planetesimal . in all models , the largest planetesimal grows rapidly in @xmath199 , and the growth is stalled at @xmath200 . the growth in the gas - free model is slower than those in the gas models . as discussed in [ sec : qvalue ] , the gas drag enhances the dissipation of the kinetic energy of dust . as the drag becomes stronger , the gravitational instability occurs earlier and the growth becomes more rapid . the final mass of the largest planetesimal is @xmath201 . the clear dependence of the largest mass on the initial stopping time can not be seen . as seeds grow , the stopping time lengthens . in the final state , we can neglect the effect of the drag . therefore , the final mass and the number do not depend on the initial stopping time . large planetesimals sweep small particles in the rotational direction . therefore the final mass depends on the size of the computational domain ( paper ii ) . the difference in rotational velocity between dust and gas causes the radial migration of dust . by neglecting the self - gravity and the time derivatives of equation ( [ eq : hilleqx ] ) and ( [ eq : hilleqy ] ) , we obtain the steady solution : @xcite @xmath202 figure [ fig : maxmass_vx.eps ] shows the time evolution of the radial velocity @xmath203 of the largest particle . in these models , the initial stopping time is @xmath155 . the power - law index and the velocity difference are @xmath204 , @xmath205 , @xmath206 , and @xmath207 ( models 100ac1 , 100as1 , 100ac2 , and 100as2 ) . in all models , the radial velocity @xmath203 is negative at @xmath208 . this corresponds to the radially inward drift . the rotational velocity of gas is slower than the kepler velocity ; dust experiences a headwind , which causes its inward migration . therefore their radial velocities become negative . although @xmath92 is different , if @xmath209 is same , the time evolutions of @xmath210 are similar in the early phase . this is because the planetesimals do not grow sufficiently in @xmath211 . if we estimate the terminal velocity from equation ( [ eq : driftvelo ] ) by the initial stopping time , they are the radial velocity @xmath212 for @xmath213 and @xmath214 for @xmath215 . in models for @xmath120 , the drag coefficient of a particle is independent of its size , thus the radial velocity converges to @xmath216 for @xmath213 and @xmath217 for @xmath215 ; the radial migration does not stop . on the other hand , in models for @xmath96 , as planetesimals grow , the drag coefficient becomes small ; the radial migration stops finally . the radial velocity oscillates with the period of about @xmath218 . this oscillation is caused by the epicycle motion . the gas drag is so weak that the oscillation is not damped . @xcite performed linear analyses of the dissipative gravitational instability of a dust layer . they studied the effect of the stopping time , the thickness of the dust layer , and the velocity dispersion . we summarized the essence of their results in [ sec : disprel ] . the dust layer is always unstable for long wavelength modes . thus , we should compare the timescale of the gravitational instability with those of other processes . under the thin disk approximation ( @xmath46 ) , we derived that simple condition that the gravitational instability dominates the other processes . though the disk thickness can not be neglected and the gravitational wakes is not axisymmetric , this condition corresponds to the occurrence of the gravitational wakes , as shown in [ sec : result ] . to discuss the formation of the self - gravity wakes , toomre s @xmath1 value is more important than the roche criterion @xmath219 . they applied the growth rate of the gravitational instability to the gravitational collapse in the turbulent disks . they assumed the simple turbulence model and studied the particle response to the turbulence flow . the protoplanetary disk can be turbulent due to magnetorotational instability or kelvin - helmholtz instability . they suggested that the standard hydrostatic relation @xmath170 is not satisfied in the turbulent disk and the dust disk can be very thick . the thin disk approximation is clearly not valid . the criterion of the formation of the self - gravitational wakes may change . we plan to study the effect of the turbulence in the next paper . @xcite performed the two - dimensional numerical simulation of the gravitational instability of a dust layer . they found that the dust layer becomes extremely thin if the stopping time is long , such as @xmath220 . this is because the dust settling is faster than the growth of the gravitational instability . however , our results show that the gravitational instability occurs although the stopping time is long . there are some differences between our simulations and theirs . we treated the dust component as particles . but , @xcite treated it as a pressure - free fluid . if the stopping time is short and the velocity dispersion is small , we can treat dust as a pressure - free fluid . the pressure - free fluid model becomes irrelevant for longer stopping times . for @xmath221 , we can use the pressure - free fluid , thus this should not cause the difference . in addition , the velocity dispersion stabilizes the gravitational instability generally . we neglected the dust back - reaction on the gas , but @xcite adopted two - component model and solved the interaction between them . if the dust density is high and the stopping time is short , the dust back - reaction on the gas should be important . the resulting gas flow may change the criterion . we performed three - dimensional local calculation . on the other hand , @xcite assumed that the the dust layer to be axisymmetric with respect to the rotational axis . the gravitational instability forms the non - axisymmetric structures such as the gravitational wakes ( e.g. , * ? ? ? * ; * ? ? ? the motion in the azimuthal direction is important to study the structure of the gravitationally unstable disk . to understand the dynamics of planetesimal formation in gas , we need to perform three - dimensional two - component simulations . we should consider the dust back - reaction on the gas and solve the hydrodynamic equations consistently . we will treat the dust component as particles because this method is applicable although the dust is loosely coupled to gas . we plan to study this point in the following papers of this series . we performed local @xmath0-body simulations of the planetesimal formation through gravitational instability . in papers i and ii , we adopted the gas - free model . however the gas is not negligible and is important to the formation of planetesimals . as a first step to understand the effect of gas , we considered gas as the background laminar flow . we neglected the dust back - reaction on the gas for the sake of simplicity . laminar flow causes the dissipation of the kinetic energy and thus radial migration of dust . in 2 , we summarized the results of the linear analysis of the dissipative gravitational instability . to handle the dispersion relation analytically , we imposed the assumptions : the isothermal equation of state , the infinitely thin disk , and the axisymmetric perturbation . simulations show that these assumptions are not valid , but they help us to understand the physical nature of the dissipative gravitational instability . we can use the growth rate to understand the simulation results . the long wavelength modes are always secularly unstable although @xmath44 . therefore , we must compare with the timescale of the gravitational instability with other processes . we provided the equation ( [ eq : crit_q ] ) for the critical @xmath1 value . if @xmath1 is smaller than the critical value , the gravitational instability is a dominant process . the numerical simulations show that the formation process of planetesimals is the same as that in the gas - free model . the formation process is divided into three stages qualitatively : the formation of wake - like structures , the creation of planetesimal seeds , and their collisional growth . by the linear stability analysis , the dust layer in the laminar gas disk is secularly unstable although @xmath44 . the growth time of the dissipative gravitational instability is slower than the dust sedimentation and the decrease of the velocity dispersion when the initial @xmath1 value is sufficiently large . thus , the disk shrinks vertically and @xmath1 decreases . as the velocity dispersion and the scale height decrease , the growth rate of the gravitational instability increases . finally , the gravitational instability becomes dominant . then wake - like structures are formed by the gravitational instability . these structures fragment into planetesimal seeds . seeds grow rapidly owing to the mutual collisions . finally , almost all mass in the computational domain is absorbed by only one planetesimal in the calculation . we investigated the dependence of results on the initial stopping time . the @xmath1 value decreases in the initial stage , and it reaches the minimum . the decay time scale of @xmath1 is proportional to the stopping time for @xmath222 , which is the decay timescale of the velocity dispersion . the minimum @xmath1 is determined by the balance between the dissipation and gravitational instability . at the minimum q , the wake - like structures are formed . the time evolution of the ratio of the dust layer density to the roche density @xmath167 is similar to that of @xmath1 . however the decay time scale of @xmath167 is longer than that of @xmath1 . the time when @xmath184 or @xmath223 does not correspond to the onset of the gravitational instability , such as the formation of the wake - like structures . this indicates that the roche criterion is not applicable in this system . we investigated the time evolution of the number of planetesimal seeds . we define a particle whose mass is larger than @xmath194 as a planetesimal seed , where @xmath194 is the planetesimal mass predicted by the linear theory @xmath195 . the number of planetesimals has a maximum value . after the gravitational instability becomes a dominant process , planetesimal seeds form . in this stage , the number of planetesimal seeds increases . however , in the late stage , the number of seeds decreases by the mutual collisions among them . the maximum of the number of the planetesimal seeds depends on the initial stopping time . the maximum number of planetesimal seeds has a peak at @xmath198 . we are unable to identify a clear difference in the mass of the largest particle in the final state . the final mass of the planetesimal depends on the size of the computational domain @xcite . we also investigated the models where the gas rotational velocity is slower than the kepler velocity . we confirmed that the planetesimal is able to form in the sub - keplerian gas disk . in the initial stage , particles experience a headwind , which causes its inward migration . if we adopt stokes gas drag law , as planetesimals grow , the drag coefficient becomes small ; the radial migration finally slows . almost all mass is absorbed by the largest planetesimal . in the gas - free models , we showed that the final mass of planetesimals depends on the size of computational domain . to investigate the final mass of planetesimals in detail , we should perform larger scale simulations . this remains to be discussed further . in addition , in this paper , we assumed gas to be laminar flow . however , the gas may be turbulent . if the turbulence is strong , the particles are stirred up and the gravitational instability may be prevented . the timescale for the dissipative gravitational instability is long @xcite . however , though the velocity dispersion of dust particles is large and the timescale for the instability is long , the dust layer may disrupt owing to the dissipative gravitational instability over time in the turbulent disk . in the next paper , we will investigate the particle response to the turbulence and the gravitational instability in the turbulent disk .
energetic electromagnetic radiation is emitted by a wide range of astrophysical objects , such as supernova remnants ( snrs ) @xcite , microquasars @xcite and gamma ray bursts ( grbs ) @xcite . this radiation is at least partially attributed to the synchrotron- or synchrotron jitter emissions @xcite of ultra - relativistic electrons , which gyrate in strong magnetic fields . such extreme conditions do not exist in the interstellar medium and probably not in astrophysical jets or in snr blast shells , if their plasma would be in a steady state . efficient electron acceleration and magnetic field amplification mechanisms must be at work and they are driven primarily by collisionless plasma shocks @xcite . through their interaction with the electromagnetic fields in the shock transition layer , the electrons may gain enough energy to be injected into the diffusive shock acceleration process @xcite . the latter is probably capable of accelerating charged particles to the highest observed cosmic ray energies and it should supply a population of highly relativistic electrons . the apparent magnetization of snr shocks @xcite is attributed to cosmic ray - driven instabilities , which can operate on large scales @xcite . filamentation instabilities of counterstreaming beams of charged particles are thought to be the origin of the micro - scale magnetic fields , in particular of those in grb jets @xcite , although the lifetimes of these fields might be too short @xcite . filamentation or beam - weibel instabilities develop in an initially spatially uniform system of two counter - streaming particle beams . typically one examines this instability in the reference frame , in which the net charge and current vanish . the beams consist of electrons and ions and they may contain a sizeable fraction of positrons in grb jets . the filamentation instability transforms the spatially uniform beams into a system of spatially separated current flux tubes , which are enwrapped and separated by electromagnetic fields [ 13 - 17 ] . observations of the filamentation instability in laboratory settings and the therefrom resulting electromagnetic radiation exist @xcite and they are likely to occur close to shocks @xcite in astrophysical settings . the weibel instability @xcite is another magnetic instability . it is driven by a thermal anisotropy of a single electron species . weibel considered in his original work a spatially uniform bi - maxwellian electron velocity distribution . the electrons were cooler along one direction than along the other two . the weibel instability gives rise to the aperiodic growth of magnetowaves with a wavevector that is parallel to the cool direction @xcite and with wavelengths that are comparable to an electron skin depth . the weibel instability has been widely examined in the past both analytically and numerically [ 21 - 30 ] . its importance for the generation of magnetic fields in jet outflows and on a cosmological scale has been pointed out @xcite . it has , however , previously been unclear how a strong thermal anisotropy could form and survive over spatial scales that exceed the wavelength of the modes , which are destabilized by the weibel instability in its original form @xcite . such a process has been identified recently @xcite . a plasma density gradient together with the high electron mobility gives rise to an ambipolar electrostatic field , which confines the electrons and accelerates the ions . the electric field vector is antiparallel to the density gradient . the ion acceleration launches a rarefaction wave [ 33 - 38 ] . the electrons , which move from the dense plasma to the front of the rarefaction wave , are slowed down and reflected inelastically by the electric field of the expanding wave , through which they provide the energy for the ion acceleration . the modulus of the electron momentum along the density gradient is reduced by these processes , while both orthogonal momentum components are unchanged . the thermal anisotropy can give rise to the weibel instability . in what follows we refer to it as the thermal anisotropy - driven weibel instability ( tawi ) . particle - in - cell ( pic ) simulations @xcite have shown that the tawi can magnetize a planar plasma cloud that expands into a vacuum . a shock develops , if the rarefaction wave expands into a dilute ambient medium . the tawi is suppressed if a shock is limiting the extent of the rarefaction wave to a length that is shorter than the wavelength of its modes @xcite . this process , which requires ion density gradients and hot electrons , can be important for energetic astrophysical plasma . the internal shock model for the prompt grb emissions assumes that the jet consists of plasma clouds with varying densities and mildly relativistic relative speeds @xcite . we can also assume that the ion density close to snr shocks will not be spatially uniform . the bulk electrons in the grb jets are thought to be relativistically hot and the bulk electrons close to snr shocks have kev temperatures . the magnetic energy generated by the tawi can reach a sizeable fraction of the electron thermal energy , which could result in strong magnetic fields . the tawi has previously been examined numerically in a two - dimensional planar geometry . a plasma slab is centred in these studies in the simulation box and confined by two parallel boundaries , across which the density jumps from the constant high density within the slab to a low density or a vacuum outside of it . the plasma density is constant along the second simulation direction and the boundary conditions are periodic . the rarefaction waves launched at both boundaries move into opposite directions , causing an expansion of the slab plasma . the density gradients of both rarefaction waves are antiparallel . an electron can cross the slab many times and its momentum component along the density gradients remains decoupled from both orthogonal ones . multiple electron reflections , which reduce only the electron momentum component along the plasma density gradient , result in an increasingly strong thermal anisotropy , which may exaggerate the strength of the tawi . we examine here with a pic simulation the expansion of an initially unmagnetized circular plasma cloud . the electron momentum can in this case not be subdivided into a component that is always parallel to the density gradient and into two components orthogonal to it , unless the electron moves exactly in the radial direction . this case is unlikely . consecutive electron reflections by the electric field of the rarefaction wave will diminish different momentum components and the anisotropy between the temperature in the radial and azimuthal directions is not boosted by multiple reflections . the purpose of our simulation study is to assess if the slowdown of the electrons in the sheath potential alone is still strong enough to trigger the tawi . our results are as follows . the ambipolar electric field accelerates the protons in the simulations to a few percent of @xmath0 . a tawi develops within the density gradient of the rarefaction wave . this instability is similar to that observed in ref . it drives radial current channels , which yield the growth of a transverse magnetic ( tm ) field component . the growth of the tm wave is followed by that of a secondary te magnetowave . the rotational polarization of this te wave is concluded from the simultaneous presence of in - plane magnetic vortices and out - of - plane magnetic fields that interfere with those driven by the tawi . the combined magnetic field of both waves breaks the circular symmetry of the rarefaction wave and scatters rapidly the electrons within , thereby reducing their thermal anisotropy . the instability saturates in our simulation , when the total magnetic energy density is about 1% of the electron thermal energy . the te and tm magnetowaves develop in the same radial interval , which suggests their connection . we propose that the superposition of the electric field , which is induced by the magnetic field of the tawi , with the radial electric field of the rarefaction wave triggers the growth of the te wave by the misalignment of the plasma density gradient and the electric field vector . such a system is unstable @xcite and it results in an instability similar to a thermoelectric instability . the structure of this paper is as follows . section 2 describes the pic simulation code and the initial conditions . our results are presented in section 3 . section 4 summarizes our findings and it discusses how these could be verified experimentally . the 2d3v ( resolves two spatial and three momentum dimensions ) particle - in - cell simulation code we employ here is based on the virtual particle - mesh numerical scheme @xcite . faraday s and ampre s laws @xmath1 are solved on a grid and evolve the electromagnetic fields in time , while @xmath2 and @xmath3 are fulfilled as constraints . the code solves all three field components as a function of @xmath4 . the relativistic force equation @xmath5 + \mathbf{v}_i \times \mathbf{b } [ \mathbf{x}_i ] \right ) \label{eq3}\ ] ] is used to update the momentum of the @xmath6 computational particle ( cp ) of species @xmath7 , with charge @xmath8 and mass @xmath9 , that is located at the position @xmath10 . the electric @xmath11 and magnetic @xmath12 fields are connected with the cps as follows . the micro - current of each cp is interpolated to the grid . the sum of the interpolated micro - currents of all cps gives the macroscopic current @xmath13 , which is used to update the electromagnetic fields through eq . the new fields are interpolated back to the positions of the individual cps to update their momentum in time through eq . the position of each cp is updated with @xmath14 . a detailed review of the pic method is given elsewhere @xcite . the plasma cloud with radius @xmath15 is surrounded by vacuum and it consists of electrons with the number density @xmath16 and protons with the same density , which are initially distributed uniformly across the cloud . figure [ fig1 ] illustrates the setup . . a circular plasma cloud has its center at @xmath17 and the radius @xmath18 . it is filled with electrons and protons with the density @xmath16 . the boundary conditions are periodic for the fields and for the particles . we will in some cases exploit the circular symmetry and use the circular coordinates @xmath19 instead of @xmath20.,width=302 ] the electron s relativistic maxwellian velocity distribution has a temperature of 32 kev or a thermal speed @xmath21 . the proton thermal speed in both simulations is @xmath22 m@xmath23 and corresponds to a thermal energy of 10 ev . the mean speeds of the electrons and of the protons vanish initially . the electrons and protons are approximated each by @xmath24 2500 cps per cell . the electron mass is @xmath25 and the proton mass @xmath26 . the @xmath27 is the initial electron skin depth and @xmath28 is the electron plasma frequency . the time is expressed in units of @xmath29 . the time step is @xmath30 . the simulation time is @xmath31 . the cloud radius @xmath32 . an electron with the thermal speed @xmath33 will cross the cloud diameter @xmath34 about 8 times during @xmath35 , resulting in multiple bounces at the electrostatic sheath field . the simulation box with periodic boundary conditions in all directions uses @xmath36 cells for a square domain with the side length @xmath37 . we examine the plasma phase space distributions and the electromagnetic field distributions , the loss of electron thermal energy to the protons and the thermal anisotropy . we consider in more detail the simulation times @xmath38 , @xmath39 and @xmath40 . this section concludes with a discussion of the time - evolution of the plasma quantities and effects due to the periodic boundary conditions . we transform the distributions from a representation in @xmath20-space into one in @xmath19-space , where @xmath41 is the center of the cloud and of the simulation box . we exploit the approximate circular symmetry and average these quantities over the azimuth angle @xmath42 . this average will become less accurate at late times , when @xmath12 breaks this symmetry . the electron number density @xmath43 . the phase space densities @xmath44 are integrated as @xmath45 , where @xmath46 and @xmath47 stand for electrons and protons . the phase space densities at all times are normalized to the maximum of the respective distribution at @xmath48 . the azimuthally averaged radial thermal energy @xmath49 and azimuthal thermal energy @xmath50 define the thermal anisotropy @xmath51 . the initial electron thermal energy per degree of freedom is @xmath52 . the in - plane electric field @xmath53 and the orthogonal @xmath54 field are expressed in units of @xmath55 , while the magnetic field components @xmath56 and @xmath57 are expressed in units of @xmath58 . the fields are azimuthally averaged as @xmath59 , where @xmath60 serves as a place holder for @xmath61 and @xmath57 . their mobility implies that some electrons can escape from the cloud and build up a negatively charged sheath outside of the now positively charged cloud . an ambipolar electrostatic field develops that points along the surface normal . its modulus is displayed in fig . [ fig2 ] for the time @xmath62 . : ( a ) shows the modulus of the in - plane electric field @xmath63 . ( b ) displays the in - plane magnetic field @xmath64 and ( c ) the orthogonal field @xmath65 . a strong electric field points radially outwards and outlines the cloud s perimeter . the magnetic fields are at noise levels . overplotted is a black circle with @xmath66.,title="fig:",width=302 ] : ( a ) shows the modulus of the in - plane electric field @xmath63 . ( b ) displays the in - plane magnetic field @xmath64 and ( c ) the orthogonal field @xmath65 . a strong electric field points radially outwards and outlines the cloud s perimeter . the magnetic fields are at noise levels . overplotted is a black circle with @xmath66.,title="fig:",width=302 ] the electric field outlines the cloud radius and has a thickness comparable to @xmath67 . the magnetic @xmath64 and @xmath57 components show noise that is strongest for @xmath68 and weaker in the rarefaction wave . the noise originates from statistical fluctuations of the current density . the initial plasma evolution is dominated by the electrostatic fields because their amplitude is much larger than the amplitude of the magnetic fields and because the latter are incoherent . the protons are accelerated radially outwards by the electric field , which confines at the same time the electrons . this is confirmed by fig . the electrons lose speed in the radial direction and their density decreases drastically as they cross the strong in - plane electric field @xmath63 at @xmath66 . this field is not strong enough to hold back all electrons and some escape into the vacuum . the fastest electrons have expanded to a radius @xmath69 , while the proton expansion is still negligible . the protons with @xmath70 are accelerated by @xmath63 and some reach already at the time @xmath62 a peak speed that exceeds their initial thermal speed @xmath71 by an order of magnitude . normalized to its initial value : panels ( a ) and ( b ) show @xmath72 and @xmath73 , respectively . the color scale is 10-logarithmic and @xmath15 is the initial cloud radius . [ fig3],width=302 ] the protons with @xmath74 still show their initial velocity distribution . their low thermal speed implies , that they could not yet interact with the surface electric field . figure [ fig4 ] shows the particle distributions at the time @xmath75 . the protons have been accelerated by the ambipolar electric field to about @xmath76 or to 2% of the speed of light @xmath0 . the protons form a narrow beam with a speed , which grows approximately linearly with @xmath77 for @xmath78 , and a density that decreases with an increasing radius . this high speed has been reached by the protons , because there is no shock that limits their acceleration . a shock would form , if the rarefaction wave would expand into an ambient plasma . normalized to its initial value : panels ( a ) and ( b ) show @xmath72 and @xmath73 , respectively . the color scale is 10-logarithmic and @xmath15 is the initial cloud radius.[fig4],width=302 ] the electron distribution has been altered significantly at this time . the peak speed of the electron s bulk population in the cloud s core , which has shrunk to @xmath79 at this time , has been decreased compared to that in fig . [ fig3](a ) , which is indicative of an energy loss . this energy loss can be attributed to the electron s inelastic reflection by the ambipolar electric field of the expanding rarefaction wave . the electrons fill up the entire simulation box , although their phase space density close to the boundary is 4 orders of magnitude less than the peak density within @xmath79 . the erosion of the cloud implies , that the initial distribution of @xmath63 in fig . [ fig2](a ) could not have been upheld . this is confirmed by fig . [ fig5 ] , which shows the @xmath63 , @xmath64 and the modulus of @xmath57 at the time @xmath75 . : panel ( a ) shows the modulus of the in - plane electric field @xmath63 and ( b ) shows that of the in - plane magnetic field @xmath64 . panel ( c ) shows the modulus @xmath65 . the black circles with radius @xmath15 outline the initial cloud perimeter.[fig5],title="fig:",width=302 ] : panel ( a ) shows the modulus of the in - plane electric field @xmath63 and ( b ) shows that of the in - plane magnetic field @xmath64 . panel ( c ) shows the modulus @xmath65 . the black circles with radius @xmath15 outline the initial cloud perimeter.[fig5],title="fig:",width=302 ] the electric field maximum is located at @xmath80 , which is at the tip of the expanding protons in fig . [ fig4](b ) . the electric field distribution expands to lower @xmath77 . it thus continues to accelerate the protons and sustains the increase of the proton mean speed with the radius . the electric fields at @xmath70 are a consequence of the plasma density gradient in this interval and of the higher electron mobility , which implies a space charge . figure [ fig5](a ) reveals , that the electric field in the interval @xmath70 does not show an exact circular symmetry . the amplitude is a function of @xmath42 . this break of the circular symmetry and its origin become evident in fig . [ fig6 ] , which shows the amplitude modulus of @xmath63 and the electron density distribution in the relevant spatial interval and compares both quantities to @xmath57 . the amplitude of the azimuthal oscillation of @xmath63 at @xmath70 is about 10% of the mean value . . the contour lines correspond to the values -0.01 , -0.005 , 0.005 and 0.01 of the @xmath57 field . the @xmath57 field has maxima in the upper right and lower left quadrant and minima in the other two quadrants . panel ( b ) compares the same contour lines of @xmath57 with the contour lines 2.7 , 2.9 , 3.1 and 3.3 of the 10-logarithmic electron number density . the density increases with decreasing radius . the number density is expressed in computational particles per cell and @xmath81.,title="fig:",width=291 ] . the contour lines correspond to the values -0.01 , -0.005 , 0.005 and 0.01 of the @xmath57 field . the @xmath57 field has maxima in the upper right and lower left quadrant and minima in the other two quadrants . panel ( b ) compares the same contour lines of @xmath57 with the contour lines 2.7 , 2.9 , 3.1 and 3.3 of the 10-logarithmic electron number density . the density increases with decreasing radius . the number density is expressed in computational particles per cell and @xmath81.,title="fig:",width=291 ] figure [ fig6](a ) suggests a correlation of @xmath63 with @xmath57 for two reasons . firstly the azimuthal oscillation of @xmath63 and the strong @xmath57 are observed at approximately the same radius . secondly the @xmath63 field shows two amplitude minima and two maxima , which are shifted by 45@xmath82 relative to those of @xmath57 . we find one maximum of @xmath63 at @xmath83 and @xmath84 ( @xmath85 ) , while one maximum of @xmath57 is located at @xmath86 ( @xmath87 ) . the circular symmetry of @xmath63 can , in principle , be broken by an electrostatic drift wave along the azimuthal direction . they could be generated by the lower - hybrid drift instability @xcite , which is caused by gradients in the plasma density and in the magnetic field amplitude . a second possibility is the simon - hoh instability that develops , if a radial electrostatic field and a spatially uniform orthogonal magnetic field are present @xcite . the electron @xmath88 drift destabilizes the plasma . the electric field driving the simon - hoh instability would be the @xmath63 and the orthogonal magnetic field would be @xmath57 . however , the characteristic wave lengths of the waves driven by both instabilities would be much less than that of the modulation of @xmath63 observed in fig . the lower hybrid instability destabilizes wavelengths , which are shorter than the characteristic length scale of the magnetic field gradient and the simon - hoh instability requires that @xmath57 be uniform over a scale much larger than the wave length of the unstable wave . figure [ fig6](b ) furthermore demonstrates that the electron density distribution is still circularly symmetric . the azimuthal oscillation of @xmath63 is thus not correlated with a charge density modulation and it is probably not an electrostatic wave . we can understand the modulation of @xmath63 , if we take into account the in - plane current @xmath89 and electric field @xmath90 that sustain the localized @xmath57 . we neglect the weak @xmath64 in fig . [ fig5](b ) . ampre s law ( eq . [ eq2 ] ) connects @xmath57 with the in - plane electric field and the in - plane current . consider the right part of the simulation box @xmath91 in fig . [ fig6](a ) . the contour lines with @xmath92 denote a maximum of @xmath57 and those with @xmath93 a minimum . at @xmath94 and @xmath84 we find with the right hand rule that the electric field induced by @xmath95 gives @xmath96 and @xmath97 . the electric field amplitude should grow to a sizeable amplitude before a balance between the in - plane current and @xmath57 cancels any further growth of @xmath63 . the superposition of the ambipolar electric field from the rarefaction wave and the practically parallel induced @xmath98 at @xmath94 and @xmath99 imply that @xmath63 is strong . the induced electric field will also amplify the ambipolar electric field at @xmath100 and @xmath99 , because both fields reverse their sign . the induced electric field weakens the ambipolar electric field at @xmath101 and @xmath102 . a magnetically induced electric field is not tied to charge density modulations , which would explain why the circular symmetry is maintained for the electron density in fig . [ fig6](b ) . we turn now to the identification of the source mechanism of the magnetic @xmath57 component in fig . [ fig5](c ) . figure [ fig7 ] compares the azimuthally averaged energy density @xmath103 of the magnetic field , the thermal anisotropy @xmath104 and the azimuthally averaged electron number density @xmath105 . the division of @xmath106 by @xmath77 ensures that a uniform density within the cloud results in a constant value of @xmath105 and the multiplication of the azimuthally averaged energy density by @xmath77 gives us a curve that is proportional to the magnetic energy . these quantities are computed at the time @xmath75 . : the ( dashed ) curve @xmath107 corresponds to the radially averaged electron number density @xmath105 , which is normalized to @xmath16 . the curve @xmath108 equals @xmath109 and denotes the local plasma frequency . the curve @xmath110 is the total magnetic energy density @xmath111 . the thermal anisotropy @xmath112 is the second dashed curve.,width=302 ] the onset of the rarefaction wave is found at @xmath113 in fig . [ fig7 ] , where the electron density and plasma frequency start to decrease . a sidelobe of the magnetic energy @xmath110 is found within this radius , but most energy is concentrated within @xmath114 . the thermal anisotropy curve is a steadily decreasing function up to @xmath115 and it is below 1 everywhere . the mean kinetic energy of the electrons in the radial direction is thus lower than their azimuthal one , which we attribute to the radial proton acceleration . the tawi destabilizes waves with wave vectors along the cool direction . figure [ fig5](c ) shows that @xmath57 oscillates fastest in the radial direction . one oscillation is visible with the wavelength @xmath116 . the magnetic @xmath57 component performs two oscillations in the azimuthal direction , each with a wavelength @xmath117 . a long oscillation orthogonal to the density gradient has also been observed in simulations of a planar rarefaction wave @xcite , which is probably a consequence of the density gradient that has not been taken into account by the existing analytic studies . the simulation data suggests that the large anisotropy observed in fig . [ fig7 ] triggers the tawi . the @xmath110 peaks at @xmath118 , where @xmath119 . one may think that @xmath110 should peak in the interval , where @xmath120 is lowest . however , both the growth rate of the tawi in units of @xmath121 and the magnetic field amplitude the plasma can support increase with @xmath108 . the latter can be concluded from the fact , that we can normalize the vlasov - maxwell set of equations with the help of @xmath108 . a higher @xmath108 results in a higher @xmath12 for the same plasma configuration . the magnetic amplitude decreases to noise levels at about @xmath122 . it is thus confined to within the rarefaction wave and it does not extend into the surrounding electron plasma at this time . the electrons have spread out in significant numbers up to @xmath123 at the time @xmath124 . normalized to its initial value : panels ( a ) and ( b ) show @xmath72 and @xmath73 , respectively . the color scale is 10-logarithmic and @xmath15 is the initial cloud radius.,width=302 ] the radial electric field has accelerated the protons to about 260@xmath71 or 2.6% of c at the time @xmath124 , as we can see from fig . [ fig8](b ) . the protons form a cool beam with one well defined mean speed for any @xmath125 . this compact distribution justifies its approximation by a single cold fluid @xcite . the mean speed of the protons increases linearly within @xmath126 . figure [ fig9 ] shows @xmath63 , @xmath64 and @xmath65 at the time @xmath124 . : panel ( a ) shows the in - plane electric field @xmath63 and ( b ) shows the in - plane magnetic field @xmath64 . panel ( c ) shows the modulus @xmath65 . the black circles with radius @xmath15 outline the initial cloud perimeter . , title="fig:",width=302 ] : panel ( a ) shows the in - plane electric field @xmath63 and ( b ) shows the in - plane magnetic field @xmath64 . panel ( c ) shows the modulus @xmath65 . the black circles with radius @xmath15 outline the initial cloud perimeter . , title="fig:",width=302 ] the in - plane electric field has weakened while the magnetic fields have grown further . the plasma density gradient and the ambipolar electric field in a rarefaction wave decrease with time . the influence on the particle dynamics of the magnetic field and its induced electric field thus become more important . the circular symmetry of @xmath63 imposed by the rarefaction wave is no longer dominant in fig . [ fig9](a ) . the amplitude of the in - plane @xmath64 in fig . [ fig9](b ) has increased by an order of magnitude compared to that in fig . [ fig5](b ) and it is now comparable to the amplitude of @xmath57 . the distribution of @xmath65 in fig . [ fig9](c ) does no longer show four almost equally important peaks as it did in fig . [ fig5](c ) . we would expect such a pattern , if the wave responsible for the @xmath64 is driven by a separate secondary instability and if it results in a circularly or elliptically polarized magnetowave . an instability is discussed in the next subsection that can generate such a wave . the @xmath64 field oscillates much faster in the radial than in the azimuthal direction in fig . [ fig9](b ) and its wave vector is thus practically aligned with the radius vector . we call this magnetowave _ wave 2_. an elliptically polarized wave will have a @xmath57 component that is phase shifted along @xmath77 with respect to its @xmath64 component . the @xmath57 component of wave 2 would interfere with the @xmath57 component of the wave that is driven by the tawi ( wave 1 ) . wave 2 performs one oscillation in the azimuthal direction according to fig . [ fig9 ] and we may assume that the same is true for its @xmath57 component . the @xmath57 field of wave 1 oscillates twice along @xmath42 . so we expect that wave 2 interferes constructively with two half - oscillations of wave 1 and that it interferes destructively with the other two . interference between both waves could thus explain the pattern observed in fig . [ fig9](b ) . figure [ fig7 ] has demonstrated that the magnetic field energy density is increased in the radial interval , in which we find a strong thermal anisotropy in the electron distribution . the anisotropy is caused by the loss of electron thermal energy to the radially accelerating protons . the supplementary movie 1 animates in time this energy exchange . it shows the electron phase space distribution in ( a ) and the proton phase space distribution in ( b ) , which are displayed in the form of snapshots and discussed in figs . [ fig3 ] , [ fig4 ] and [ fig8 ] . movie 1 confirms that electrons , which cross the boundaries on one side , re - enter on the opposite side , as we expect from the periodic boundary conditions of the simulation . the fastest electrons that escaped from the plasma cloud reach the boundaries at @xmath127 and they encounter the electrons that crossed the boundary from the other direction . note that movie 1 ( a ) shows the velocity modulus and does thus not reveal that both electron populations move into opposite directions . the electron two - stream configuration can result in instabilities . however , the two - stream configuration is rapidly destroyed by the arrival of the slower electrons at the boundaries . the supplementary movie 2 demonstrates that the life - time of the two - stream configuration is too short and the electron density too low to drive detectable instabilities at @xmath127 . it animates in time the 10-logarithmic electron density in the simulation box , which is averaged over blocks of @xmath128 cells and given in particles per cell . an average density of less than one computational particle per cell is found close to the boundaries . the local plasma frequency is @xmath129 , which implies a low growth rate for any beam instability @xcite . movie 2 reveals faint electron structures at a time @xmath130 that flow parallel to the x - axis and y - axis . they are probably caused by an instability between the electrons of the plasma cloud and the electrons , which return from the boundary and are accelerated by the ambipolar electric field when they re - enter the cloud @xcite . figure [ fig10 ] quantifies both the energy loss of the electrons to the accelerating protons and it follows the thermal anisotropy in time . figure [ fig10](a ) is a measure for the average kinetic energy per electron , which corresponds to their velocity in the simulation plane . this energy is obtained by summing @xmath131 and @xmath132 and by dividing the result by @xmath133 . @xmath52 is the initial thermal energy contributed by each velocity component and the division by @xmath134 gives us a measure of the kinetic energy per electron . the thermal anisotropy @xmath135 is displayed in fig . [ fig10](b ) . and panel ( b ) the thermal anisotropy @xmath136 . the colour scale is linear . the horizontal black line corresponds to the time @xmath75,title="fig:",width=291 ] and panel ( b ) the thermal anisotropy @xmath136 . the colour scale is linear . the horizontal black line corresponds to the time @xmath75,title="fig:",width=291 ] the electrons within the cloud lose about 30% of their kinetic energy to the accelerating protons . the electron kinetic energy is apparently increased outside of the rarefaction wave . this is a selection process , which can be seen in the movie 1 . the bulk of the slow electrons is contained by the rarefaction wave , while the most energetic electrons , which have a kinetic energy that far exceeds the potential energy of the rarefaction wave , can escape . the thermal anisotropy is negligible for the radial interval without a plasma density gradient , e.g. for @xmath137 and @xmath138 . @xmath139 reveals that the electrons in the interval with the proton density gradient lose more energy in the radial than in the azimuthal direction . the radial electron energy is less than half of the azimuthal energy over an extended radial interval . the interval with a value of @xmath140 is largest at the time @xmath75 in fig . [ fig10](b ) , when it covers @xmath141 . the @xmath57 saturates at around this time . the tawi starts to grow at @xmath142 , when the anisotropy is sufficiently strong . the saturation is demonstrated by fig . [ fig11 ] , which compares the time - evolution of the mean amplitude of the azimuthally averaged fields @xmath143 , @xmath144 , @xmath145 and @xmath146 . the supplementary movie 3 animates in time the spatial distributions of the in - plane electric field @xmath63 ( a ) , the in - plane magnetic field @xmath64 ( b ) and the @xmath57 . ( a ) , @xmath144 ( b ) , @xmath145 ( c ) and @xmath147 ( d ) . the color scale is 10-logarithmic . the black horizontal line corresponds to the time @xmath75.,title="fig:",width=302 ] ( a ) , @xmath144 ( b ) , @xmath145 ( c ) and @xmath147 ( d ) . the color scale is 10-logarithmic . the black horizontal line corresponds to the time @xmath75.,title="fig:",width=302 ] the in - plane electric field @xmath63 in fig . [ fig11](a ) peaks initially at @xmath66 . the build - up of this electrostatic field launches electrostatic waves into the cloud , which yields the amplitude peaks for @xmath68 . the electric field that expands to larger @xmath148 and reaches @xmath149 at @xmath150 is induced by the electrons that escape from the cloud , e.g. those with @xmath151 in fig . [ fig3](a ) . the electron current is not balanced by an ionic current and a radial electric field grows . the density of the escaped electrons decreases as they spread over the simulation box and the electric amplitude eventually drops below the threshold of fig . [ fig11](a ) . the position associated with the maximum of @xmath63 accelerates in time . this maximum coincides with the tip of the rarefaction wave . the electric field with the uniform amplitude of @xmath152 behind the tip of the ion front roughly outlines the extent of the rarefaction wave . it expands to lower @xmath68 as the cloud is progressively eroded . the electric field component @xmath54 in fig . [ fig11](b ) remains at noise levels , except for a peak at @xmath153 . this peak coincides spatially and temporally with the growth of the rotational magnetowave , which is responsible for the @xmath154 in fig . [ fig11](c ) . it is related to the displacement current @xmath155 connected to @xmath64 . a comparison of the figs . [ fig11 ] ( c ) and ( d ) confirms the observation from the supplementary movie 3 that @xmath57 starts to grow to a large amplitude well before @xmath64 . the latter grows fastest at around the time @xmath153 , when @xmath57 saturates . the presence of a strong @xmath57 is apparently necessary for the growth of @xmath64 , which supports our previous claim that @xmath64 is driven by a secondary instability . figure [ fig11 ] reveals that @xmath57 is confined to within the rarefaction wave , while the magnetic field distribution @xmath64 reaches out into the dilute electron plasma outside of the rarefaction wave . it eventually reaches the boundaries and effects due to the periodic boundary conditions can no longer be neglected at the final time @xmath156 . the development of the strong magnetic fields at @xmath70 coincide with a balancing of @xmath157 and @xmath158 in fig . [ fig10](b ) . this change in @xmath120 is partially due to the electron deflection by @xmath12 , which will contribute to an equilibration of @xmath157 and @xmath158 . we want to determine the source mechanism of the in - plane magnetic field @xmath64 . the ambipolar electric field due to the rarefaction wave and the electric field induced by @xmath57 are both confined to the simulation plane . the electric fields and @xmath57 can excert a force on the electrons only in this plane and no current orthogonal to the simulation plane can be generated . since we have no initial beams moving along the z - axis , we can rule out magnetic beam instabilities @xcite . one possibility to generate an electron flow along z is a coupling of the electron velocity components through the lorentz factor . an electron acceleration in the simulation plane would change @xmath159 . the fastest electrons in fig . [ fig4](a ) have the mildly relativistic speed @xmath160 that gives a lorentz factor @xmath161 . however , their number density is low and we could not expect them to drive strong in - plane magnetic fields . a possible mechanism for the growth of @xmath64 is the electric field induced by @xmath57 . the induced field is strong when @xmath57 saturates and it changes the direction of the in - plane electric field vector by up to 5% relative to the radius vector . the plasma density is according to fig . [ fig6](b ) circularly symmetric at @xmath138 . a misalignment of the in - plane electric field vector and the density gradient can give rise to elliptically polarized magnetowaves @xcite . we have examined here the expansion of an initially unmagnetized circular plasma cloud into a vacuum by means of a 2d3v pic simulation . the purpose of our study has been to test , if a thermal anisotropy - driven weibel instability ( tawi ) can also develop in curved rarefaction waves . previous studies considered planar plasma slabs that expanded under their own thermal pressure @xcite and observed the growth of strong magnetic fields in the rarefaction wave , unless a shock suppressed the tawi @xcite . a thermally anisotropic electron distribution develops due to two reasons . firstly the slow - down of the electrons in the ambipolar electric field implies that their thermal pressure is decreased along the density gradient . this mechanism occurs in planar and curved rarefaction waves . secondly , the inelastic reflection of an electron by the expanding rarefaction wave implies that it looses momentum along the electric field direction . this momentum is transferred to the accelerating ions . the initial plasma distribution used in the refs . @xcite is a plasma slab of infinite extent in one direction , which has a finite width in the second direction . this plasma slab is immersed in a vacuum or in a low - density plasma . both plasma boundaries are parallel . this implies that every reflection of the electron will decrease the same momentum component , namely the component that is aligned with the antiparallel density gradients of both rarefaction waves . the anisotropy increases with every electron reflection . this mechanism can not occur in a circular rarefaction wave , since here consecutive electron reflections tend to decrease different momentum components . our study can thus test if consecutive electron reflections are necessary to achieve a tawi . if the slowdown of the electrons in the ambipolar electric field of a plasma with a spatially non - uniform density is enough to trigger the tawi , then this instability may play a role in the magnetization of astrophysical plasma @xcite besides the filamentation instability @xcite and cosmic ray - driven instabilities @xcite . the magnetic energy that can be generated by the tawi is a fraction of the electron thermal energy , here about 1% . this energy density is well below that of the filamentation instability , which can reach about 10 - 20% of the kinetic energy density of leptonic beams if they are cold . however , the tawi requires only hot electrons and a density gradient to grow , while the filamentation instability is triggered by relativistic particle beams that we may find only close to plasma shocks . the tawi could thus magnetize much larger intervals in energetic astrophysical jets , such as grb jets , which are thought to have a nonuniform plasma density and relativistically hot electrons . the tawi may also generate a seed magnetic field in the turbulent plasma close to snr shocks and speed up the development of the much more powerful cosmic ray - driven instabilities . our findings are as follows . the electron expansion results in a strong radial electrostatic field . it counteracts a further electron expansion and accelerates the protons into the radial direction . a rarefaction wave emerges , which is characterised by an increase of the proton mean speed and by a decrease of the proton density with an increasing radius . the expansion of a rarefaction wave is self - similar @xcite . the absence of an ambient medium has prevented the formation of a shock and the protons could reach a speed exceeding 250 times their thermal speed . the fields within the rarefaction wave have initially been electrostatic . electrons lose momentum to the accelerating protons only in the radial direction , which implies that their radial thermal pressure decreases more quickly than the azimuthal one . a thermal anisotropy - driven weibel instability ( tawi ) @xcite grows once the radial extent of the rarefaction wave is sufficient to accommodate a wavelength of the unstable modes @xcite and once the thermal anisotropy between the radial and the azimuthal direction became sufficiently large . the tawi results in the aperiodic growth of a transverse magnetic ( tm ) wave . this tm wave induces an electric field that is added to that of the electrostatic ambipolar electric field of the rarefaction wave . the total electric field is thus no longer parallel to the plasma density gradient . a misalignment between the electric field and the plasma density gradient can drive circularly or elliptically polarized magnetowaves @xcite . we propose that these waves are responsible for the in - plane magnetic field we observe in our simulation , which can not be explained in terms of the tawi . the magnetic fields of this wave reach an amplitude that is comparable to the waves due to the tawi . our findings can be verified in an experiment , where the absorption of a laser pulse heats up the electrons in a thin wire @xcite . some electrons will leave the now positively charged wire and build up an ambipolar electric field . this ambipolar field accelerates the ions through the transverse normal sheath acceleration ( tnsa ) mechanism @xcite and it results in the formation of a rarefaction wave . the ablation of a wire generates a cylindrical rarefaction wave . our circular plasma cloud corresponds to an idealized cross section of the wire , as long as it is sufficiently far away from the laser impact point @xcite . the cross section is idealized , because we do not consider ionization processes that require more elaborate numerical codes @xcite . the plasma dynamics outside of the wire should be comparable in the simulation and experiment . the strong magnetic fields that develop in the rarefaction wave should be detectable . it is thus possible to investigate systematically the tawi in a controlled laboratory experiment and to determine the strength and life - time of the magnetic field . the interplay of the tawi with the secondary magnetowave we have observed here and other instabilities , such as the rayleigh - taylor instability @xcite , can also be examined . the experiment can reveal the topology of the magnetic fields in a three - dimensional setting , which is currently computationally too expensive . acknowledgments : this work was supported by vetenskapsrdet ( dnr 2010 - 4063 ) , epsrc ( ep / e035728/1 , ep / c003586/1 , ep / d043808/1 ) , consejeria de educacion y ciencia ( ene2009 - 09276 ) , the junta de comunidades de castilla - la mancha ( pai08 - 0182 - 3162 ) , sfi ( 08/rfp / phy1694 ) . computer time and support was provided by the hpc2n in ume and by the ichec in dublin . 99 aharonian fa _ et al _ 2004 _ nature _ * 432 * 75 fender r and belloni t 2004 _ ann . astrophys . _ * 42 * 317 piran t 2004 _ rev phys . _ * 76 * 1143 medvedev mv 2000 _ astrophys . j. _ * 540 * 704 bykov am and treumann ra 2011 _ astron . _ * 19 * 42 blandford rd and eichler d 1987 _ phys . rep . _ * 154 * 1 berezhko eg , ksenofontov lt and volk hj 2003 _ astron . astrophys . _ * 412 * l11 bell ar 2004 _ mon . not . royal astron . _ * 353 * 550 medvedev mv and a. loeb 1999 _ astrophys . j. _ * 526 * 697 brainerd jj 2000 _ astrophys . j. _ * 538 * 628 gruzinov a 2001 _ astrophys . j. _ * 563 * l15 waxman e 2006 _ plasma phys . controll . fusion _ * 48 * b137 kazimura y , sakai ji , neubert t and bulanov sv 1998 _ astrophys . j. _ * 498 * l183 silva lo , fonseca ra , tonge jw , dawson jm , mori wb and medvedev mv 2003 _ astrophys . j. _ * 596 * l121 schlickeiser r and shukla pk 2003 _ ppastrophys . j. _ * 599 * l57 bret a , gremillet l and dieckmann me 2010 _ phys . plasmas _ * 17 * 120501 bret a , firpo mc and deutsch c 2005 _ phys . lett . _ * 94 * 115002 kapetanakos ca 1974 _ appl . _ * 25 * 484 weibel es 1959 _ phys . * 2 * 83 kalman bg , montes c and quemada d 1968 _ phys . fluids _ * 11 * 1797 morse rl and nielson cw 1971 _ phys . fluids _ * 14 * 830 davidson rc , hammer da , haber i and wagner ce 1972 _ phys . fluids . _ * 15 * 317 kaang hh , ryu cm and yoon ph 2009 _ phys . plasmas _ * 16 * 082103 bret a 2007 _ contrib . plasma phys . _ * 47 * 133 stockem a , dieckmann me and schlickeiser r 2009 _ plasma phys . controll . fusion _ * 51 * 075014 stockem a , dieckmann me and schlickeiser r 2010 _ plasma phys . fusion _ * 52 * 085009 palodhi l , califano f and pegoraro f 2009 _ plasma phys . controll . fusion _ * 51 * 125006 lazar m , schlickeiser r and poedts s 2010 _ phys . plasmas _ * 17 * 062112 tautz rc 2011 _ phys . plasmas _ * 18 * 012101 shaisultanov r and eichler d 2011 _ phys . plasmas _ * 18 * 034501 schlickeiser r 2005 _ plasma phys . controll . fusion _ * 47 * a205 thaury c , mora p , heron a and adam jc _ phys . e _ * 82 * 016408 sack c and schamel h 1987 _ phys . rep . _ * 156 * 311 grismayer t , mora p , adam jc and heron a 2008 _ phys . e _ * 77 * 066407 lyutikov m 2010 _ phys . e _ * 82 * 056305 gargate l , fonseca ra , bingham r and silva lo 2008 _ ieee trans . plasma sci . _ * 36 * 1168 sarri g , dieckmann me , kourakis i and borghesi m 2011 _ phys . _ , * 107 * 025003 sarri g , murphy gc , dieckmann me , bret a , quinn k , kourakis i , borghesi m , drury loc and ynnerman a 2011 _ new j. phys . _ * 13 * 073023 saleem h 1996 _ phys . e _ * 54 * 4469 saleem h 2009 _ phys . plasmas _ * 16 * 082102 eastwood jw 1991 _ comput * 64 * 252 dawson jm 1983 _ rev . mod . * 55 * 403 daughton w , lapenta g and ricci p 2004 _ phys . rev . lett . _ * 93 * 105004 lapenta g and brackbill ju 2002 _ phys . plasmas _ * 9 * 1544 simon a 1963 _ phys . fluids _ * 6 * 382 sakawa y , joshi c , kaw pk , jain vk , johnston tw , chen ff and dawson jm 1992 _ phys . lett . _ * 69 * 85 dieckmann me , sarri g , romagnani l , kourakis i and borghesi m 2010 _ plasma phys . controll . fusion _ * 52 * 025001 quinn k _ et al _ 2009 _ phys . * 102 * 194801 snavely ra _ et al _ 2000 _ phys . lett _ * 85 * 2945 dhumieres e , lefebvre e , gremillet l and malkav 2005 _ phys . plasmas _ * 12 * 062704 gibbon p , beg fn , clark el , evans rg and zepf m 2004 _ phys . plasmas _ * 11 * 4032 nuter r , gremillet l , lefebvre e , levy a , ceccotti t and martin p 2011 _ phys . plasmas _ * 18 * 033107 piriz ar , di lucchio l and prieto gr 2011 _ phys . plasmas _ * 18 * 012702 piriz ar , sanz j and ibanez lf 1997 _ phys . plasmas _ * 4 * 1117
neutron stars provide a natural laboratory to study extremely dense matter . in the interiors of such stars , the density can reach up to several times the nuclear saturation density , @xmath8 . at such high densities quarks could be squeezed out of nucleons to form quark matter @xcite . the true ground state of dense quark matter at high densities and low temperatures remains an open problem due to the difficulty of solving non - perturbative quantum chromodynamics ( qcd ) . it has been suggested that strange quark matter that consists of comparable numbers of @xmath9 , @xmath10 , and @xmath11 quarks may be the stable ground state of normal quark matter @xcite . this led to the conjecture that the family of compact stars may have members consisting entirely of quark matter ( so - called strange stars ) and/or members featuring quark cores surrounded by a hadronic shell ( hybrid stars ) @xcite . observationally , it is very challenging to distinguish the various types of compact objects , such as the strange stars , hybrid stars , and ordinary neutron stars . their early cooling behavior is dominated by neutrino emission which is a useful probe of the internal composition of compact stars . thus , cooling simulations provide an effective test of the nature of compact stars @xcite . however , many theoretical uncertainties and the current amount of data on the surface temperatures of neutron stars leave sufficient room for speculations @xcite . another useful avenue for testing the internal structure and composition of compact stars is astro - seismology , i.e. , the study of the phenomena related to stellar vibrations @xcite . in particular , there are a number of instabilities which are associated with the oscillations of rotating stars . here we will be concerned with the so - called r - mode instability [ see refs . @xcite for reviews ] . this instability is known to limit the angular velocity of rapidly rotating compact stars . the r - mode and related instabilities in rotating neutron stars are damped by the shear and bulk viscosities of matter , therefore these are important ingredients of theoretically modelling rapidly rotating stars . such models and their microscopic input can then be constrained via the observations of rapidly rotating pulsars , such as the crab pulsar and the millisecond pulsars . for quark matter in chemical equilibrium , the shear viscosity is dominated by strong interactions between quarks . the bulk viscosity , however , is dominated by flavor - changing weak processes , whereas strong interactions play a secondary role . for normal ( non - superconducting ) strange quark matter , the bulk viscosity is dominated by the non - leptonic process @xcite @xmath12 @xmath13 since the contributions of the leptonic processes @xmath14 and @xmath15 are suppressed due to much smaller phase spaces . the bulk viscosity of various phases of quark matter has been studied extensively , see refs . @xcite . compact stars are strongly magnetized . neutron star observations indicate that the magnetic field is of the order of @xmath16 g at the surface of ordinary pulsars . magnetars - strongly magnetized neutron stars - may feature even stronger magnetic fields of the order of @xmath17 g @xcite . an upper limit on the magnetic field can be set through the virial theorem . gravitational equilibrium of stars is compatible with magnetic fields of the order of @xmath18 g @xcite . in such a strong magnetic field , not only the thermodynamical but also the hydrodynamical properties of stellar matter will be significantly affected . in particular , due to the large magnetization of strange quark matter the fluid will be strongly anisotropic in a strong magnetic field ( we note here that the magnetization of ordinary neutron matter is small @xcite ) . therefore , there is need to develop an anisotropic hydrodynamic theory to describe strongly magnetized matter in compact stars . as we show below , the matter is completely described in terms of eight viscosity coefficients , which include six shear viscosities and two bulk viscosities . in this paper , we will carry out a theoretical study of the anisotropic hydrodynamics of magnetized strange quark matter and will calculate the two bulk viscosities . we will also discuss the implications of the anisotropic bulk viscosities on the r - mode instability in rotating quark stars . the paper is organized as follows . the formalism of anisotropic hydrodynamics for magnetized strange quark matter is developed in sec . [ 2 ] . in sec . [ 3 ] we apply the local linear response method to derive explicit expressions for bulk viscosities . the stability of the fluid under strong magnetic field is analyzed in sec . [ 4 ] . section [ 5 ] contains our numerical results for the bulk viscosities . the damping of the r - mode instability in rotating quark stars by the bulk viscosity is studied in sec . section [ 7 ] contains our summary . we use natural units @xmath19 . the metric tensor is @xmath20 . we will use the si system of units in our equations involving electromagnetism , however we will quote the strength of the magnetic field in cgs units ( gauss ) , as is common in the literature on compact stars . hydrodynamics arises as an effective theory valid in the long - wavelength , low - frequency limit where the energy - momentum tensor @xmath21 , the conserved baryon current @xmath22 , the conserved electric current @xmath23 , and the entropy density flux @xmath24 , _ etc . _ , are expanded in terms of gradients of the 4-velocity @xmath25 and the thermodynamic parameters of the system , such as the temperature @xmath26 , baryon chemical potential @xmath27 , _ etc._. the hydrodynamic equations can be expressed as conservation laws for the total energy - momentum tensor @xmath21 , as well as baryon and electric currents , @xmath22 and @xmath23 . the zeroth order terms in the expansion correspond to an ideal fluid and we shall use the index 0 to label them . in the presence of an electromagnetic field , the zeroth - order terms can be generally written as @xcite , @xmath28 where @xmath29 , @xmath30 , @xmath31 , @xmath32 , and @xmath11 are the local energy density , thermodynamic pressure , baryon number density , electric charge density , and entropy density , respectively measured in the rest frame of the fluid . @xmath33 is the projector on the directions orthogonal to @xmath25 . here @xmath34 is the energy - momentum tensor of the electromagnetic field . @xmath35 is the field - strength tensor which can be decomposed into components parallel and perpendicular to @xmath25 as @xmath36 where in the second line we have introduced the 4-vectors @xmath37 and @xmath38 with @xmath39 being the totally anti - symmetric levi - civita tensor . in the rest frame of the fluid , @xmath40 , we have @xmath41 , @xmath42 and @xmath43 , which are precisely the electric and magnetic fields in this frame . therefore , @xmath44 and @xmath45 are nothing but the electric and magnetic fields measured in the frame where the fluid moves with a velocity @xmath25 . the antisymmetric tensor @xmath46 is the polarization tensor which describes the response to the applied field strength @xmath35 . for example , if @xmath47 is the thermodynamic potential of the system , @xmath48 . for later use , we also define the in - medium field strength tensor @xmath49 . in analogy to @xmath35 we can decompose @xmath46 and @xmath50 as @xmath51 with @xmath52 , @xmath53 , @xmath54 and @xmath55 . in the rest frame of the fluid , the non - trivial components of these tensors are @xmath56 , @xmath57 , @xmath58 , @xmath59 , @xmath60 , and @xmath61 . here @xmath62 and @xmath63 are the electric polarization vector and magnetization vector , respectively . in the linear approximation they are related to the fields @xmath64 and @xmath65 by @xmath66 and @xmath67 , with @xmath68 and @xmath69 being the electric and magnetic susceptibilities . the 4-vectors @xmath70 are all space - like , @xmath71 , and normalized as @xmath72 , where @xmath73 and @xmath74 . since the electric field is much weaker than the magnetic field in the interior of a neutron star , we will neglect it in most of the following discussion . upon introducing the 4-vector @xmath75 , which is parallel to @xmath45 and is normalized by the condition @xmath76 , and the antisymmetric tensor @xmath77 , we can write @xmath78 with @xmath79 and @xmath80 . the maxwell equation @xmath81 takes the form @xmath82 its non - relativistic form , which is known as the induction equation , is given by @xmath83 where @xmath84 is the 3-velocity of the fluid . contracting eq . ( [ maxwell0 ] ) with @xmath85 gives @xmath86 where @xmath87 and @xmath88 . the second maxwell equation can be written as @xmath89 whose non - relativistic form is @xmath90 where @xmath91 is the electric charge density and @xmath92 is the corresponding current . it is useful to re - write the energy - momentum tensor in the following form @xcite , @xmath93 where @xmath94 is the projection tensor on the direction perpendicular to both @xmath25 and @xmath95 . we have defined the transverse and longitudinal pressures @xmath96 and @xmath97 relative to @xmath95 ; here @xmath30 is the thermodynamic pressure . in the absence of a magnetic field , the fluid is isotropic and @xmath98 . in the local rest frame of fluid , we have @xmath99 ( without loss of generality , we choose the @xmath100-axis along the direction of the magnetic field ) , hence the electromagnetic tensor takes the usual form , while @xmath101 . next we would like to check the consistency of the terms that appear in @xmath102 with the formulae of standard thermodynamics involving electromagnetic fields . by using the thermodynamic relation @xmath103 and the conservation equations for @xmath104 , @xmath105 , and @xmath106 in ideal hydrodynamics , one can show that the hydrodynamic equation @xmath107 together with the maxwell equation ( [ maxwell ] ) implies @xmath108 which is consistent with the standard thermodynamic relation @xmath109 one should note that the potential energy @xmath110 has already been included in our definition of @xmath29 . otherwise , new terms @xmath110 , @xmath111 and @xmath112 should be added to the left - hand sides of eq . ( [ thermo ] ) , eq . ( [ de ] ) , and eq . ( [ de2 ] ) , respectively . thus , we conclude that our hydrodynamical equations are consistent with well - known thermodynamic relations . by keeping the first - order terms of the derivative expansion of conserved quantities one obtains the navier - stokes - fourier - ohm theory . in this theory , @xmath21 , @xmath22 , @xmath23 , and @xmath24 can be generally expressed as @xmath113 where @xmath114 , and @xmath115 are the dissipative fluxes . they all are orthogonal to @xmath25 ; this reflects the fact that the dissipation in the fluid should be spatial . we shall assume that @xmath115 can be expressed as a linear combination of @xmath116 , and @xmath117 @xcite . this allows us to incorporate the fact that the entropy flux is determined by the energy - momentum and baryon number diffusion fluxes . thus , @xmath118 with the coefficients @xmath119 , @xmath120 , and @xmath121 being functions of thermodynamic variables . next , the hydrodynamic equations are specified by utilizing the conservation laws of the total energy - momentum @xmath21 , the baryon number density flow @xmath22 , electric current @xmath23 , and the second law of thermodynamics , @xmath122 to discuss the dissipative parts , let us first define the 4-velocity @xmath25 , since it is not unique when energy exchange by thermal conduction is allowed for . we will use the landau - lifshitz frame in which @xmath25 is chosen to be parallel to the energy density flow , so that @xmath123 . upon projecting the first equation of eq . ( [ hydroeqn ] ) on @xmath124 and after some straightforward manipulations , we find @xmath125 combining eq . ( [ energyeqn ] ) , eq . ( [ thermo ] ) , and the second equation in eq . ( [ hydroeqn ] ) , we arrive at @xmath126 where @xmath127 and @xmath128 . for a thermodynamically and hydrodynamically stable system , ( [ divs ] ) should be non - negative . this implies @xmath129 where @xmath130 , @xmath131 is the rank - four tensor of viscosity coefficients , and @xmath132 and @xmath133 are thermal and electrical conductivity tensors with respect to the diffusion fluxes of baryon number density and electric charge density . by definition , @xmath131 is symmetric in the pairs of indices @xmath134 and @xmath135 . it necessarily satisfies the condition @xmath136 , which is onsager s symmetry principle for transport coefficients . similarly , the tensors @xmath132 and @xmath133 should satisfy the conditions @xmath137 and @xmath138 . furthermore , all the tensors of transport coefficients @xmath131 , @xmath132 , and @xmath133 must be orthogonal to @xmath25 by definition . as we have seen , the appearance of the magnetic field makes the system anisotropic . such anisotropy is specified by the vector @xmath95 , so that the tensors @xmath131 , @xmath133 , and @xmath132 should be in general expressed in terms of @xmath25 , @xmath95 , @xmath139 , and @xmath140 . all independent irreducible tensor combinations having the symmetry of @xmath131 and which are orthogonal to @xmath25 are @xcite(see also appendix [ proof ] ) @xmath141 all independent irreducible tensor combinations having the symmetry of @xmath132 and @xmath133 and which are orthogonal to @xmath25 are @xmath142 in accordance with the number of tensors ( [ combination ] ) and ( [ combination - k ] ) , a fluid in a magnetic field in general has eight independent viscosity coefficients , three independent thermal conduction coefficients and three independent electrical conductivities . they may be defined as the coefficients in the following decompositions for the viscous stress tensor , heat flux , and electric charge flux @xmath143 @xmath144 @xmath145 where @xmath146 , @xmath147 and @xmath148 is constructed so that the @xmath149 s are the coefficients of its traceless parts , _ i.e. _ , they can be regarded as shear viscosities ; @xmath150 s are the coefficients of the parts with non - zero trace and can be considered as bulk viscosities . the @xmath151 s and @xmath152 s are thermal and electrical conductivities , respectively . now the divergence of entropy density flux ( [ divs ] ) can be explicitly written as @xmath153 one should note that the terms corresponding to the transport coefficients @xmath154 , @xmath155 , and @xmath156 in eqs . ( [ pmn])-([jbe ] ) do not contribute to the divergence of the entropy density flux . for stable systems , all the other transport coefficients must be positive definite according to the second law of thermodynamics . in sec.[4 ] we will demonstrate explicitly that negative bulk viscosities @xmath0 or / and @xmath1 indeed cause an instability in the hydrodynamic evolution of strange stars . to conclude this section , we compare our definition of the viscosity coefficients in eq . ( [ pmn ] ) with the definition given in ref . @xcite for non - relativistic fluid , which reads @xmath157 where @xmath158 and the remaining notations are self - explanatory . our viscosity coefficients in eq . ( [ pmn ] ) are related to the coefficients in eq . ( [ pij ] ) by @xmath159 in ref . @xcite there is no term that corresponds to our @xmath160 . the reason is that ref . @xcite considers the combination of vectors ( viii ) in eq . ( [ combination ] ) as dependent on the others ; in the appendix we will show that ( at least for relativistic fluids ) all the combinations ( i)-(viii ) are linearly independent . note that the transport coefficients in eq . ( [ divs2 ] ) appear as prefactors of quadratic forms , therefore the second law of thermodynamics requires that these coefficients must be positive definite for stable ensembles . this is not manifest in eq . ( [ pij ] ) . the typical oscillation frequency of neutron stars is of the order of magnitude of the rotation frequency , @xmath161s@xmath162s@xmath163 . the most important microscopic processes which dissipate energy on the corresponding timescales are the weak processes . the compression and expansion of strange quark matter with nonzero strange quark mass will drive the system out of equilibrium . the processes ( [ weak1 ] ) and ( [ weak2 ] ) are the most efficient microscopic processes that restore local chemical equilibrium . therefore , the bulk viscosities are determined mainly by the processes ( [ weak1 ] ) and ( [ weak2 ] ) . in this section we will derive analytical expressions for the bulk viscosities @xmath0 and @xmath1 . , these processes involve quark - quark scattering via an interchange of a single gluon @xcite . however , these processes are too fast to contribute to the bulk viscosity at the frequency of the order of the stellar oscillation frequency . ] let us imagine an isotropic flow @xmath164 which characterizes the stellar oscillation . if there are no dissipative processes , such an oscillation will drive the system from one instantaneous equilibrium state to another instantaneous equilibrium state . the appearance of dissipation changes the picture : during the oscillations the thermodynamic quantities will differ from their equilibrium values . let us explore how the thermodynamic quantities evolve during the flow oscillation . in general , we can write the change of baryon density @xmath165 induced by the oscillation of the fluid as @xmath166 where @xmath167 is the static ( time - independent ) equilibrium value , @xmath168 denotes the equilibrium value shift from @xmath167 due to the volume change and @xmath169 denotes the instantaneous departure from the equilibrium value . because processes ( [ weak1 ] ) and ( [ weak2 ] ) conserve baryon number , @xmath170 can be set to zero , if we neglect other microscopic processes . then @xmath171 can be determined through the continuity equation of ideal hydrodynamics , @xmath172 since processes ( [ weak1 ] ) and ( [ weak2 ] ) also conserve the sum @xmath173 , a similar argument leads to the relation @xmath174 when the system is driven out of chemical equilibrium , the chemical potential of the @xmath11 quark will be slightly different from that of the @xmath10 quark . let us denote this difference by @xmath175 , with @xmath176 being the deviation of @xmath177 from its static equilibrium value . up to linear order in the deviation we find @xmath178 where @xmath179 denotes the number density of quarks of flavor @xmath180 and the subscript 0 indicates that the quantity in the bracket is computed in static equilibrium state . @xmath181 and @xmath182 are the deviations of @xmath11-quark and @xmath10-quark densities from their static equilibrium value . in the final expressions they should be functions of @xmath183 . the instantaneous departure from equilibrium is restored by the weak processes ( [ weak1 ] ) and ( [ weak2 ] ) . adopting the linear approximation , this can be described by @xmath184 where @xmath185 and @xmath186 are the rates of processes ( [ weak1 ] ) and ( [ weak2 ] ) , respectively . if the weak processes are turned off , one should have @xmath187 where the dot denotes the time derivative . after turning on the weak processes , we have @xmath188 this system of coupled linear first - order equations is closed by substituting eqs . ( [ deltanb])-([deltamu ] ) . it is then easy to obtain the solution , @xmath189 where the coefficient @xmath190 is defined by @xmath191 the parallel and transverse components of the pressure @xmath192 and @xmath193 can be written as @xmath194 and @xmath195 where @xmath196 the small departure of the magnetic field @xmath197 can be calculated by the variation of eq . ( [ maxwell ] ) . one finds @xmath198 a direct calculation then gives , @xmath199 where we introduced the coefficient @xmath200 as @xmath201 now we obtain @xmath202 with @xmath203 defined as @xmath204 then the deviation of @xmath21 from its equilibrium value can be written as @xmath205 for isotropic flows , we have @xmath206 by comparing the above two expressions , we obtain @xmath207 and @xmath208 expressions ( [ vis1 ] ) and ( [ vis2 ] ) show that the bulk viscosities @xmath0 and @xmath1 are functions of the perturbation frequency @xmath209 , the weak rate @xmath210 and the thermodynamic quantities @xmath211 . from the derivation above we can convince ourselves that these expressions should be valid also in the case of color - superconducting matter . for zero magnetic field , ( [ vis2 ] ) reduces to eq . ( [ vis1 ] ) which , with parameters @xmath212 , and @xmath210 taken in the absence of magnetic field gives the expression for the usual bulk viscosity @xmath213 defined in isotropic hydrodynamics . both @xmath0 and @xmath1 attain their maxima in the limit of zero frequency , @xmath214 ; and the maxima are inversely proportional to the weak interaction rate . at high frequency , @xmath215 , @xmath0 and @xmath1 fall off as @xmath216 . for practical applications to cold strange stars , where the chemical potential is much larger than the temperature , the quantities @xmath217 , and @xmath190 can be evaluated in the zero - temperature limit . their dependence on temperature is weak . contrary to this , the coefficient @xmath210 depends strongly on temperature : for normal quark matter , @xmath210 has a power - law dependence on @xmath26 ( see sec.[5 ] ) ; for a fully paired color - superconducting phase , the weak rate is exponentially suppressed by a boltzmann factor @xmath218 with @xmath219 being the superconducting gap . consequently , @xmath0 and @xmath1 depend exponentially on @xmath26 @xcite . before we find the numerical values for the bulk viscosities , we first need to analyze the stability of magnetized strange quark matter . we observe that according to eq . ( [ vis2 ] ) negative values of @xmath0 are a priori not excluded . in the following we analyze the consequences and implications of negative @xmath0 on the stability of the system . stable equilibrium in a self - gravitating fluid , such as in strange quark stars , is attained through the balance of gravity and pressure . the gravitational equilibrium requires that both components of the pressure @xmath192 and @xmath193 should be positive ( otherwise the star will undergo a gravitational collapse ) . at zero temperature , the one - loop thermodynamic pressure @xmath220 of non - interacting strange quark matter , where @xmath221 is the grand partition function , is given by @xmath222,{\nonumber\\}\end{aligned}\ ] ] where @xmath223 is the absolute value of electric charge , @xmath177 and @xmath224 are the chemical potential and the mass of quark of flavor @xmath180 , @xmath225 labels the landau levels , @xmath226 is the degree of degeneracy of each landau level and @xmath227 $ ] is the highest landau level for quarks of flavor @xmath180 . by differentiating eq . ( [ pstrong ] ) with respect to @xmath2 one can easily get the magnetization as @xmath228.{\nonumber\\}\end{aligned}\ ] ] in fig . [ mag_u ] we illustrate the magnetization as a function of magnetic field at zero temperature . the parameters are chosen as @xmath229 on average , the magnetization increases when @xmath2 grows and eventually becomes constant when @xmath230 , where @xmath231 however , the detailed structure of the magnetization exhibits strong de haas - van alphen oscillations @xcite . this oscillatory behavior is of the same origin as the de haas - van alphen oscillations of the magnetization in metals and originates from the quantization of the energy levels associated with the orbital motion of charged particles in a magnetic field . the irregularity of this oscillation shown in fig . [ mag_u ] is due to the unequal masses and charges of @xmath232 , and @xmath11 quarks . when @xmath230 , all quarks are confined to their lowest landau level and their transverse motions are frozen . in this case , the longitudinal pressure @xmath233 , so the magnetization @xmath234 is independent of @xmath2 , and the transverse pressure @xmath96 of the system vanishes . this behavior is evident in fig . [ pressure_u ] . .,width=283 ] ( dashed , black online ) and transverse @xmath193 ( solid , red online ) pressures of strange quark matter as functions of the magnetic field @xmath2 in units of the pressure @xmath235 for zero magnetic field.,width=283 ] thus , we conclude that when all the quarks are confined to their lowest landau level , the transverse pressure vanishes . the system therefore becomes mechanically unstable and would collapse due to the gravity @xcite . this phenomenon establishes an upper limit on the magnetic field sustained by a quark star . given @xmath236 mev , this upper limit is roughly @xmath237 g as shown in fig . [ pressure_u ] . the thermodynamic stability requires that the local entropy density should reach its maximum in the equilibrium state @xcite . the total energy density of the fluid and the magnetic field is @xmath238 and the corresponding first law of thermodynamics , in variational form , is @xmath239 where @xmath240 is the strength of the magnetic field and @xmath241 stands for a small departure of a given quantity from its equilibrium value . varying eq . ( [ dtve ] ) on both sides and taking into account that @xmath242 , and @xmath2 are independent variational variables , one obtains @xmath243 where @xmath244 and @xmath245 the thermodynamical stability criteria require that @xmath246 , or , equivalently , @xmath247 is positive definite . taking into account the relation @xmath248 , it is easy to show that these criteria are equivalent to the requirement @xmath249 from eq . ( [ pstrong ] ) we obtain @xmath250 then it is evident that the condition @xmath251 is always satisfied . however , @xmath252 is divergent when @xmath2 approaches a @xmath253 landau level for each flavor quark from below , @xmath254 this shows that strange quark matter will be thermodynamically unstable just below each landau level @xmath255 ( @xmath253 ) for every flavor @xmath180 . the first three thermodynamically unstable windows ( tuw ) associated , respectively , with @xmath256 , and @xmath257 are illustrated in the @xmath258 plane in fig . [ tiw ] for our parameters ( [ paras ] ) . the tuw is actually very narrow . one may conjecture that such an instability may lead to formation of magnetic domains @xcite . the presence of possible magnetic domains in neutron star crusts was discussed in ref . @xcite ; furthermore , such a possibility for color - flavor - locked quark matter was pointed out in ref . we will not pursue here the study of domain structure and related physics , since among other things , this will require us to specify the geometry of the system . plane for @xmath259 and @xmath260 . , width=283 ] in this subsection we address the problem of hydrodynamic stability within the theory presented in sec.[2 ] . the fluid is said to be stable if it returns to its initial state after a transient perturbation . otherwise , _ i.e. _ , when the perturbation grows and takes the fluid into another state , the fluid is unstable . our particular goal here is to determine whether a small , plane - wave perturbation around a homogeneous equilibrium state grows for nonzero @xmath0 and @xmath1 @xcite . all the other transport coefficients are set to zero . for this purpose , it is sufficient to solve the hydrodynamic and maxwell equations that are linearized around the homogeneous equilibrium state @xmath261 where @xmath262 upon linearizing the normalization conditions @xmath263 , one finds that the perturbed variables need to satisfy the constraints @xmath264 in the equations above the perturbations are assumed to have the form @xmath265 , where @xmath266 is constant , and the unperturbed quantities are independent of space and time . the hydrodynamic and maxwell equations need to be supplemented by an equation of state ( eos ) in order to close the system . the linearized eos is given by @xmath267 where @xmath268 is the speed of sound . in the most general case eqs . ( [ linear])-([eos ] ) constitute 15 independent equations , but the equations associated with the conservation of @xmath31 and @xmath32 are decoupled from the others if the eos is taken in the form ( [ eos ] ) . therefore , we are left with 13 equations . we work in the rest frame of the equilibrium fluid , @xmath269 and @xmath99 and choose as independent variables @xmath270 ( n.b . one can choose other independent variables , but the results do not change ) . the thirteen linear equations can be collected into the following matrix form @xmath271 the matrix @xmath272 has the following form @xmath273 where @xmath274 is the total enthalpy . the exponential plane - wave solutions for frequencies @xmath275 and wave - vectors @xmath276 satisfy the dispersion relations given by @xmath277 for the modes propagating parallel ( longitudinal modes ) or perpendicular ( transverse modes ) to the magnetic field , eq . ( [ detg ] ) has simple solutions . \1 ) transverse modes , @xmath278 . there are two types of transverse modes . one is solely determined by the maxwell equations and has the following dispersion relation @xmath279 where @xmath280 , and describes simply an electromagnetic wave . another solution has the dispersion relation @xmath281 where the second approximate relation is valid in the long - wavelength limit . this solution represents a sound wave propagating perpendicular to the magnetic field . the speed of this sonic wave is @xmath282 and is smaller than the speed @xmath283 of a ordinary sound wave . it is seen that positive @xmath0 implies dissipation of the sonic wave , _ i.e. _ , a decay of the initial disturbance . we conclude that the fluid flow is stable in the case . however , we see that for negative @xmath0 , the initial disturbance grows and the fluid is unstable . thus , we conclude that negative transverse bulk viscosity implies hydrodynamic instability via growth of transverse sound waves . \2 ) longitudinal modes , @xmath284 . we find three types of longitudinal modes . the first one is again the electromagnetic wave with the dispersion relation @xmath285 the second one is a transverse wave oscillating perpendicularly to the magnetic field , but traveling along the magnetic field lines . it has the dispersion relation @xmath286 this mode is the alfven wave whose speed is equal @xmath287 . the third longitudinal mode has the following dispersion relation @xmath288 this mode represents an ordinary sound wave with dissipation due to the longitudinal bulk viscosity @xmath1 . it is obvious that if @xmath289 this mode will not decay , rather grow , thus leading to hydrodynamic instability . in the next section , we will show that for certain values of the parameters , the transverse bulk viscosity @xmath0 could be indeed negative . we emphasize here that this does not imply a violation of the second law of thermodynamics , rather this manifests a hydrodynamic instability of the ground state , _ i.e. _ , small perturbations will take the system via this hydrodynamic instability to a new state . a candidate state is the one which has inhomogeneous ( domain ) structure . both the structure of the new state and the transition from the homogeneous to the inhomogeneous state are interesting problems which are beyond the scope of this study . however , we would like to point out a number analogous cases where a negative transport coefficient indicates instability towards formation of a new state with domain structure . one such case is the negative resistivity ( also known as the gunn effect ) in certain semiconducting materials @xcite . another case is the negative ( effective ) shear viscosity , which is extensively studied in the literature @xcite . finally , negative bulk viscosity has been investigated in different contexts in @xcite . in order to calculate the bulk viscosities , we need to determine the coefficients @xmath290 and @xmath210 . from eqs . ( [ pstrong ] ) , ( [ m0 ] ) , and ( [ nf0 ] ) we obtain in a straightforward manner @xmath291 where @xmath292 the rate @xmath210 of the weak processes ( [ weak1 ] ) and ( [ weak2 ] ) should also be affected by a strong magnetic field . the major effect of a magnetic field on @xmath210 is to modify the phase space of weak processes ( [ weak1 ] ) and ( [ weak2 ] ) @xcite . taking this into account , one obtains @xmath293 where @xmath294 is the fermi constant . \1 ) when the magnetic field is much smaller than the typical chemical potential , say , @xmath295 , its effect on the bulk viscosities is negligible . for typical parameters ( [ paras ] ) , this condition holds up to @xmath296 g. in this case , the system is practically isotropic , @xmath1 and @xmath0 are effectively degenerate with the isotropic @xmath213 , the bulk viscosity of unmagnetized matter . the zero - magnetic field limit results can be obtained easily by replacing @xmath297 and @xmath298 . the bulk viscosity for zero magnetic field @xmath213 as function of oscillation frequency @xmath209 for various temperature is shown in fig . [ zeta0 ] . the shoulder " structure and the temperature dependence of @xmath213 are easily understood from eq . ( [ vis1 ] ) and have been widely discussed in the literature @xcite . at zero magnetic field as function of the oscillation frequency @xmath209 for @xmath299 mev at @xmath300 ( solid , blue online ) @xmath301 ( dashed , red online ) and @xmath302 ( black , dotted online ) mev . , width=283 ] \2 ) when the magnetic field is extremely large , say , @xmath303 g , for our choice of parameters ( [ paras ] ) all the quarks are confined in their lowest landau level . in this case we obtain @xmath304 and therefore @xmath305 we used the parameters ( [ paras ] ) and assumed physically interesting frequencies @xmath306 s@xmath163 . the bulk viscosity @xmath0 vanishes as a consequence of vanishing @xmath193 when @xmath230 . since @xmath1 is now inversely proportional to @xmath307 it approaches zero for large @xmath2 . therefore , both @xmath0 and @xmath1 are suppressed for large @xmath2 . in contrary , @xmath192 is enhanced by the extremely large magnetic field , see fig . [ pressure_u ] . \3 ) when the magnetic field is strong , but not strong enough to confine all the quarks to their lowest landau level , the situation becomes complicated . for our chosen parameters ( [ paras ] ) , this situation roughly corresponding to the interval @xmath308 g. in this case , a finite number of landau levels is occupied , and the essential observation is that @xmath200 and @xmath203 can be negative . the behaviors of @xmath200 and @xmath203 are shown in fig . [ c_u ] as functions of @xmath2 . let us concentrate on the few levels just above the value @xmath309 g. when @xmath2 grows passing over @xmath310 or @xmath311 for each @xmath225 , both @xmath200 and @xmath203 change their sign . more importantly , they have always opposite signs . therefore , in this region , @xmath0 is negative which leads to hydrodynamic instability ( see the analysis in sec.[43 ] ) . ( dashed , blue online ) and @xmath203 ( dotted , red online ) as functions of @xmath2 at zero temperature.,width=283 ] the numerical values of the bulk viscosities @xmath1 and @xmath0 are shown in fig . [ zeta_u ] as functions of @xmath2 . the parameters are those given in eq . ( [ paras ] ) . we also fix the temperature @xmath312 mev and oscillation frequency @xmath313 . both @xmath1 and @xmath0 have quasi - periodic " oscillatory dependence on the magnetic field . the two boundaries of each period " correspond to a pair of neighboring @xmath314 and @xmath315 , and hence the period is roughly @xmath316 for large @xmath2 . therefore , on average , the period increases as @xmath2 grows . the amplitude of these oscillations also grows with increasing magnetic field until @xmath317 . thereafter all the quarks are confined to their lowest landau levels and @xmath0 vanishes . from fig . [ zeta_u ] we see that the magnitudes of @xmath0 and @xmath1 can be 100 to 200 times larger than their zero field value @xmath213 . due to the unequal masses and charges of @xmath232 and @xmath11 quarks , @xmath1 and @xmath0 behave very irregularly . we illustrate the zoomed - in curves around @xmath318 and @xmath319 g in the sub - panels , which look more regular . the quasi - periodic structures are more evident in these sub - panels . and @xmath0 scaled by the isotropic bulk viscosity @xmath213 as functions of the magnetic field @xmath2 at fixed frequency @xmath320 and temperature @xmath312 mev . the dashed red curves correspond to viscosities lying in the unstable regions and would be not physically reachable . sub - panels show the amplifications around @xmath321 and @xmath322 g. our parameters are given in eq . ( [ paras]).,title="fig:",width=340 ] and @xmath0 scaled by the isotropic bulk viscosity @xmath213 as functions of the magnetic field @xmath2 at fixed frequency @xmath320 and temperature @xmath312 mev . the dashed red curves correspond to viscosities lying in the unstable regions and would be not physically reachable . sub - panels show the amplifications around @xmath321 and @xmath322 g. our parameters are given in eq . ( [ paras]).,title="fig:",width=340 ] the most unusual feature seen in fig . [ zeta_u ] is that for a wide range of field values , the transverse bulk viscosity @xmath0 is negative . therefore , strange quark matter in this region is hydrodynamically unstable . besides this hydrodynamical instability , near each @xmath255 , there is a narrow window in which thermodynamical instability arises . we depict @xmath1 and @xmath0 in these unstable regions by dashed red curves . the solid blue curves correspond to the stable regime . the magnetic field in a compact star need not be homogeneous and may have a complicated structure with poloidal and toroidal components . furthermore , the fields will be functions of position in the star because of the density dependence of the parameters of the theory . furthermore , the instabilities , described above , may lead to fragmentation of matter and formation of domain structures , where the regions with magnetic fields are separated from those without magnetic field by domain walls . accordingly , only the averaged viscosities over some range of magnetic field have practical sense for assessing the large - scale behavior of matter . averaging over many oscillation periods in the stable region , we find that the averaged values of @xmath1 and @xmath0 are much more regular , with their magnitudes restricted from 0 to several @xmath213 , see fig . [ zetaaver ] . in obtaining the curves in fig . [ zetaaver ] , we have eliminated the viscosities lying in the unstable regime . the solid black curves are obtained by averaging over a short period @xmath323 . the period was chosen such that the most rapid fluctuations are smeared out , but the oscillating structures over larger scale are intact . the short - dashed red curves correspond to averaging over an even longer period , @xmath324 . the result of long - period averaging is that @xmath1 first increases slowly and then drops down quickly once @xmath325 g ; similarly , @xmath0 first slowly decreases and then drops down very fast for @xmath326 g. such a dropping behavior reflects the fact that a large number of quarks are beginning to occupy the lowest landau level . and @xmath0 scaled by the isotropic bulk viscosity @xmath213 as functions of the magnetic field @xmath2 at fixed frequency @xmath327 and temperature @xmath312 mev . the black - solid curve corresponds to averaging over a short period ( @xmath323 ) ; while the red short - dashed curve corresponds to averaging over a long period ( @xmath324 ) . the viscosities lying in the unstable regions have been eliminated in the averaging.,title="fig:",width=283 ] and @xmath0 scaled by the isotropic bulk viscosity @xmath213 as functions of the magnetic field @xmath2 at fixed frequency @xmath327 and temperature @xmath312 mev . the black - solid curve corresponds to averaging over a short period ( @xmath323 ) ; while the red short - dashed curve corresponds to averaging over a long period ( @xmath324 ) . the viscosities lying in the unstable regions have been eliminated in the averaging.,title="fig:",width=283 ] we note that the appearances of thermodynamic , mechanical , and hydrodynamical instabilities are all induced by the landau quantization of the quark levels , _ i.e. _ , are quantum mechanical in nature . more precisely , they are all due to the interplay between the landau levels and the fermi momentum ( reflected in the quantity @xmath255 ) . additionally , the hydrodynamical instability requires that the quark matter is para - magnetized . although we did our analysis by using the free quark gas approximation , eq . ( [ pstrong ] ) , it should be valid as long as there are sharp fermi surfaces ( low temperature ) , quantized landau levels ( high magnetic field ) and para - magnetization . the appearances of these instabilities are expected to be robust feature for such systems . we also checked that if one imposes the neutrality condition , _ i.e. _ , the condition @xmath328 , there is only minor quantitative change , while the qualitative conclusions are almost unchanged . our choice of chemical potential @xmath329 mev roughly corresponds to the choice of @xmath330 for neutral strange quark matter . the purpose of this section is to discuss the damping of the r - modes of newtonian models of strange stars by dissipation driven by the bulk viscosities @xmath0 and @xmath1 . as is well known , rotating equilibrium configurations of self - gravitating fluids are susceptible to instabilities at high rotation rates . starting from the mass - shedding limit and going down with the rotation rate , the first instability point corresponds to the dynamical instability of the @xmath331 and @xmath332 mode . this bar - mode instability is independent of the dissipative processes inside the star and occurs at values of the kinetic to potential energy ratio @xmath333 @xcite . for smaller rotation rates two secular instabilities with @xmath331 arise , each corresponding to a sign of @xmath334 . for incompressible fluids at constant density the @xmath335 values for the onset of secular instabilities coincide . one instability is driven by the viscosity , the other instability is driven by the gravitational radiation . for realistic stars the @xmath335 values for the onset of these instabilities do not coincide ; relativity and other factors shift the viscosity - driven instability to higher values of @xmath335 . at the same time the gravitational radiation instability is shifted to lower values of @xmath335 . the gravitational radiation instability arises for the modes which are retrograde in the co - rotating frame , while prograde in the ( distant ) laboratory frame . the underlying mechanism is the well established chandrasekhar - friedman - schutz ( cfs ) mechanism @xcite . the bulk and shear viscosities can prevent the development of the cfs instabilities , except in a certain window in the rotation and temperature plane . in the following we shall concentrate on axial modes of newtonian stars , the so - called r - modes , which are known to undergo a cfs - type instability . our main goal will be to assess the role of strong magnetic fields and bulk viscosity on the stability of these objects . we shall adopt the formalism of refs . @xcite for our study of the damping of the r - modes by bulk viscosity . for the sake of simplicity we shall describe both fluid mechanics and gravity in the newtonian approximation . the equations that describe the dynamical evolution of the star are @xmath336 @xmath337 @xmath338 where @xmath339 is defined by the integral @xmath340 the quantity @xmath341 is the mass density of the fluid which is assumed to satisfy a barotropic equation of state , @xmath342 . @xmath343 is the gravitational potential and @xmath272 is the gravitational constant . the potential @xmath344 is used to determine the velocity field @xmath84 . the oscillation modes of a uniformly rotating star can be completely described in terms of two perturbation potentials @xmath345 and @xmath346 , where @xmath347 and @xmath348 are the potentials that correspond to the equilibrium configuration of the star . we assume that the time and azimuthal angular dependence of any perturbed quantity is described by @xmath349 , where @xmath350 is an integer and @xmath351 is the frequency of the mode in laboratory frame . let @xmath47 denote the rotation frequency of the star and @xmath209 denote the frequency of the perturbed quantity measured in the co - rotating frame [ which corresponds the @xmath209 in eqs . ( [ vis1 ] ) and ( [ vis2 ] ) because we will work in the co - rotating frame ] . for small @xmath47 there is a simple relation between @xmath352 , and @xmath209 @xcite , @xmath353 by linearizing the euler equation ( [ newtonian2 ] ) around the equilibrium configuration , the velocity perturbation @xmath354 is determined by @xcite @xmath355 the tensor @xmath356 is a function of @xmath351 and the rotation frequency @xmath47 of the star , @xmath357},\;\;\;\end{aligned}\ ] ] where @xmath358 is a unit vector pointing along the rotation axis of the equilibrium star , which we assume to be parallel to the magnetic field , _ i.e. _ , @xmath359 in cartesian coordinate system . here @xmath360 is the fluid velocity of the equilibrium star . having the linearized euler equation , one proceeds to the linearization of the mass continuity equation ( [ newtonian1 ] ) and the equation for the gravitational potential ( [ newtonian3 ] ) ; one finds @xmath361 these equations , together with the appropriate boundary conditions at the surface of the star for @xmath362 and at infinity for @xmath363 , determine the potentials @xmath362 and @xmath363 . for slowly rotating stars , eq . ( [ newtonian4 ] ) can be solved order by order in @xmath47 , @xmath364 } , { \nonumber\\}{\delta}{\phi}=r^2{\omega}^2{\left[}{\delta}{\phi}_0+{\delta}{\phi}_2\frac{{\omega}^2}{{\pi}g { \rho}}+o({\omega}^4){\right]},\end{aligned}\ ] ] where @xmath365 is the radius of nonrotating star . since we need only the perturbed velocity , we will focus on @xmath362 in the following discussion . the zeroth - order contribution to the r - mode is generated by the following form of the potential @xmath366 , @xmath367 @xmath368 where @xmath369 is an arbitrary dimensionless constant and @xmath370 are the associated legendre polynomials . it has been shown that the most unstable mode is the one with @xmath332 @xcite , therefore we shall consider only this case in the following discussion . substituting @xmath366 into eq . ( [ dv ] ) one obtains the first - order perturbed velocity , @xmath371 where @xmath372 and @xmath373 is the magnetic - type spherical harmonic function , @xmath374 it is straightforward to check that the first - order perturbed velocity satisfies @xmath375 therefore it does not contribute to the dissipation due to the bulk viscosities @xmath0 and @xmath1 . in order to see how the bulk viscosities damp the r - mode instability one must consider next - to - first order , _ i.e. _ , the third - order perturbed velocity which is generated by the potential @xmath376 . one can not determine analytically @xmath376 from eqs . ( [ newtonian4])-([dudf ] ) , but the angular structure of @xmath376 can be well represented by the following spherical harmonics expansion @xcite , @xmath377 the functions @xmath378 and @xmath379 have been determined numerically in ref . . a useful approximation is provided by the following simple expressions @xmath380 which excellently fit the numerical result . we will use eqs . ( [ f1 ] ) and ( [ f2 ] ) in the following numerical calculation . the energy of r - modes comes both from the velocity perturbation and the perturbation of the gravitational potential . for slowly rotating stars , the main contribution comes from the velocity perturbation @xcite . then , the energy of the r - mode measured in the co - rotating frame is @xmath381 assuming spherical symmetry , we have @xmath382 this energy will be dissipated both by gravitational radiation and by the thermodynamic transport in the fluid @xcite , @xmath383 the dissipation rate due to gravitational radiation is given by @xcite @xmath384},{\nonumber\\}\end{aligned}\ ] ] where @xmath385 ^ 2}.\end{aligned}\ ] ] @xmath386 and @xmath387 are the mass and current multipole moments of the perturbation , @xmath388 taking into account eq . ( [ omega ] ) one obtains @xmath389 which implies that the total sign of @xmath390 is positive : gravitational radiation always increases the energy of the r - modes . in order to compare the relative strengths of different dissipative processes , it is convenient to introduce the dissipative timescales defined by @xmath391 where the index @xmath392 labels the dissipative process . the lowest - order contribution to @xmath393 comes from the current multipole moment @xmath394 . for the most important case @xmath395 , this leads to the following timescale ( derived for a simple polytropic equation of state @xmath396 ) @xcite @xmath397 the bulk viscosities @xmath0 and @xmath1 dissipate the energy of the r - mode according to @xmath398 accordingly , the time scales @xmath399 and @xmath400 are given by @xmath401 currently , the shear viscosities @xmath402 of strange quark matter are not known . in order to determine the damping of the r - mode by shear viscosity , we take as a crude estimate the value of @xmath403 in the absence of a magnetic field @xcite @xmath404 where @xmath405 is the coupling constant of strong interaction . we will choose the value @xmath406 and apply eq . ( [ eta0 ] ) to highly degenerate 3-flavor quark matter with equal chemical potentials of all flavors , ( @xmath407 ) . the contribution to the energy dissipation rate @xmath408 due to shear viscosity @xmath149 now becomes @xmath409 assuming a uniform mass density star , the time scale @xmath410 can be simply expressed as @xcite @xmath411 the total time scale @xmath412 is given by the following sum @xmath413 which characterizes how fast the r - mode decays . most importantly , if the sign of @xmath414 is negative the amplitude of the r - mode will not decay , rather it will increase with time . thus , it is important to determine the critical angular velocity @xmath415 for the onset of instability @xmath416 at a given temperature , stars with @xmath417 will be unstable due to gravitational radiation . of a strange quark star as a function of magnetic field @xmath2 at temperature @xmath418 mev . the red dashed lines correspond to the unstable , while the blue solid lines correspond to the stable regime.,width=283 ] figure.[rb ] shows the critical angular velocity @xmath415 of a strange quark star with mass @xmath419 and radius @xmath420 km as a function of the magnetic field @xmath2 near @xmath322 g. the temperature is fixed as @xmath418 mev and other parameters are taken according to eq . ( [ paras ] ) . in obtaining fig . [ rb ] , we have taken into account the thermodynamical and hydrodynamical stability conditions . the blue - solid curves correspond to the thermodynamically and hydrodynamically stable region , while the red - dashed curves correspond to unstable regions . the critical angular velocity is strongly oscillating with increasing @xmath2 . this behavior is due to the oscillating nature of the bulk viscosities @xmath1 and @xmath0 as shown in fig . [ zeta_u ] . thus , this macroscopic behavior originates from a purely quantum mechanical effect , namely the landau quantization of the energy levels of quarks . as discussed for the bulk viscosities @xmath1 and @xmath0 shown in fig . [ zetaaver ] , averaging is needed to obtain physically relevant quantities . in fig . [ rbaver ] we show the averaged critical angular velocity at various temperatures.the solid black curves are obtained by averaging over a short period @xmath421 , whereas the short - dashed red curves correspond to averaging over a long period @xmath422 . it is seen that after short - period averaging , the critical angular velocity ( solid black curves ) shows regular oscillation , the amplitude of which is growing as the @xmath2-field increases . the critical angular velocity @xmath415 displays a sharp drop for fields @xmath423 g ( short - dashed red curves ) , which is the consequence of the sharp drop of @xmath1 and @xmath0 shown in fig . [ zetaaver ] . thus , we conclude that for extremely large magnetic fields , the critical angular velocity at which the r - mode instability sets in could be significantly lower than in the absence of magnetic field . of a strange quark star as a function of magnetic field @xmath2 at various temperatures @xmath424 , and @xmath425 mev . the black solid curve corresponds to averaging over a short period ( @xmath421 ) ; while the red short - dashed curve corresponds to averaging over a long period ( @xmath422).,title="fig:",width=245 ] of a strange quark star as a function of magnetic field @xmath2 at various temperatures @xmath424 , and @xmath425 mev . the black solid curve corresponds to averaging over a short period ( @xmath421 ) ; while the red short - dashed curve corresponds to averaging over a long period ( @xmath422).,title="fig:",width=245 ] of a strange quark star as a function of magnetic field @xmath2 at various temperatures @xmath424 , and @xmath425 mev . the black solid curve corresponds to averaging over a short period ( @xmath421 ) ; while the red short - dashed curve corresponds to averaging over a long period ( @xmath422).,title="fig:",width=245 ] figure.[rmode ] shows the window of the r - mode instability in the @xmath426 plane for a strange quark star of mass @xmath427 and radius @xmath420 km . the regions above the respective curves correspond to the parameter space where the r - mode oscillations are unstable , _ i.e. _ , a star in this region will rapidly spin down by emission of gravitational waves . the dashed green curve corresponds to vanishing bulk viscosities @xmath428 . the solid black curve represents the ( in)stability window of an un - magnetized strange quark star . the curves with symbols show the typical instability window for magnetic fields around @xmath321 g ( the red curve marked by triangles ) and @xmath429 g ( the blue curve , marked by circles ) . the symbolled curves are obtained by using the bulk viscosities averaged over the period @xmath324 . for low temperatures , @xmath430 kev , the r - mode instability is suppressed mainly by the shear viscosity ; at these low temperatures the bulk viscosities are an insignificant source of damping , independent of how large the magnetic field is . however , for larger temperatures the bulk viscosities dominate the damping of r - mode oscillations . for magnetic fields below @xmath431 g , the critical rotation frequency is almost independent of the @xmath2-field . the r - mode instability window increases as the magnetic field grows . for fields @xmath5 g it is a very sensitive function of the field , as a consequence of the rapid variation of the bulk viscosities with the field . asymptotically , the instability window can become significantly larger than the window at zero magnetic field ( see also fig . [ rbaver ] ) . for completeness , [ rmode ] also shows the observed distribution of low mass x - ray binaries ( lmxbs ) by the shadowed box , which corresponds to the typical temperatures ( @xmath432 k ) and rotation frequencies ( 300 - 700 hz ) of the majority of observed lmxbs @xcite . it is seen that even in the case of extremely large magnetic fields , our instability window is consistent with the current lmxb data . . the black solid curve represents the window of unmagnetized strange quark matter . the curves with symbols show the typical behavior of the instability window when the magnetic fields are around @xmath321 g ( red curve ) and @xmath429 g ( blue curve ) . the shadowed box represents typical temperatures ( @xmath433 k ) and rotation frequencies ( 300 - 700 hz ) of the majority of observed lmxbs @xcite.,width=336 ] in this paper we have studied anisotropic hydrodynamics of strongly magnetized matter in compact stars . we find that there are in general eight viscosity coefficients , six of them are identified as shear viscosities , the other two , @xmath0 and @xmath1 , are bulk viscosities [ see eq . ( [ pmn ] ) ] . we applied our formalism to magnetized strange quark matter and gave explicit expressions for the bulk viscosities [ eq . ( [ vis1 ] ) and eq . ( [ vis2 ] ) ] due to the non - leptonic weak reactions ( [ weak1 ] ) and ( [ weak2 ] ) . due to the landau quantization of the energy levels of charged particles in a strong magnetic field , the magnetic field dependence of @xmath1 and @xmath0 is very complicated and exhibits quasi - periodic " de haas - van alphen type oscillations ( see fig . [ zeta_u ] ) . for a magnetic field @xmath434 g the effect of the magnetic field on the transport coefficients is small and the bulk viscosities can be well approximated by their zero - field values . for large fields @xmath435 g the viscosities are substantially modified , @xmath0 may even become negative for some values of the @xmath2-field . we showed that negative @xmath0 render the fluid hydrodynamically unstable . for a number of reasons ( density dependence of parameters along the star profile , formation of domains , intrinsic multicomponent nature of the magnetic field ) the dependence of the transport coefficients on the magnetic fields are needed at different resolutions , _ i.e. _ , they require some suitable averaging over a range of magnetic field . we have provided such averages over an increasingly larger scale . we find that if the averaging period is small the bulk viscosities show regular oscillations , the amplitudes of which increase with magnetic field ( see the black solid curves in fig . [ zetaaver ] ) . these oscillations are smoothed out if we further increase the averaging scale . at this larger scale the most interesting feature is the rapid drop in the bulk viscosity of the matter due to the confinement of quarks to the lowest landau level ; this occurs for magnetic fields in excess of @xmath436 g ( see the short - dashed red curves in fig . [ zetaaver ] ) . as an application , we utilized our computed anisotropic bulk viscosities to study the problem of damping of r - mode oscillations in rotating newtonian stars . we find that the instability window increases as the magnetic field is increased above the value @xmath437 g. by increasing the field one covers the entire range of parameter space which lies between the two extremes : the case when bulk viscosity vanishes ( dashed green curve in fig . [ rmode ] , which corresponds to extremely large magnetic fields @xmath438 g for which the bulk viscosity drops to zero ) , and the case when the magnetic field is absent ( solid black curve in fig . [ rmode ] ) . the found novel dependence of the r - mode instability window on the magnetic field may help to distinguish quark stars from ordinary neutron stars with strong magnetic fields , since the latter are much more difficult to magnetize . it would be interesting to see whether the objects that lie in between these extremes , _ e.g. _ , hybrid configurations featuring quark cores and hadronic envelopes ( see ref . @xcite and references therein ) , may interpolate smoothly between the physics of ordinary and strange compact objects . we thank t. brauner , t. koide , b. sad , a. schmitt , and i. shovkovy for helpful discussions . this work is supported , in part , by the helmholtz alliance program of the helmholtz association , contract ha216/emmi `` extremes of density and temperature : cosmic matter in the laboratory '' and the helmholtz international center for fair within the framework of the loewe ( landesoffensive zur entwicklung wissenschaftlich - konomischer exzellenz ) program launched by the state of hesse . m. h. is supported by cas program outstanding young scientists abroad brought - in " , cas key project kjcx3-syw - n2 , nsfc10735040 , nsfc10875134 , and by k. c. wong education foundation , hong kong . the purpose of this appendix is to show explicitly that the eight different decompositions in eq . ( [ combination ] ) are linearly independent . to this end , let us write down a general linear combination of the eight different decompositions , @xmath439 if @xmath440 are linearly independent , the coefficients @xmath441 should all vanish for any values of the vectors @xmath25 and @xmath95 . firstly , it is easy to see that the components @xmath442 and @xmath443 are independent of the other components , because they have odd parity under reflection @xmath444 , whereas the other six components have even parity under this transformation . besides that it is obvious that @xmath445 and @xmath443 are independent of each other . therefore , we only need to treat the linear equation @xmath446 by contracting the indices @xmath447 and @xmath448 , we obtain the following three conditions @xmath449 contracting eq . ( [ linear0 ] ) with @xmath95 and @xmath450 we find the following two additional conditions @xmath451 contracting the indices @xmath448 and @xmath369 in eq . ( [ linear ] ) , we obtain one further condition @xmath452 the non - trivial solution of the set of eqs . ( [ linear2])-([linear5 ] ) is @xmath453 then , we have the following condition , @xmath454 the only possible solution is @xmath455 , which thus proves the independence of @xmath440 . 20 d. d. ivanenko and d. f. kurdgelaidze , astrophysics * 1 * , 251 ( 1965 ) [ astrofiz . * 1 * , 479 ( 1965 ) ] . n. itoh , prog . phys . * 44 * , 291 ( 1970 ) . j. c. collins and m. j. perry , phys . lett . * 34 * , 1353 ( 1975 ) . e. witten , phys . d * 30 * , 272 ( 1984 ) . f. weber , prog . phys . * 54 * , 193 ( 2005 ) [ arxiv : astro - ph/0407155 ] . n. iwamoto , phys . * 44 * , 1637 ( 1980 ) . d. blaschke , t. klahn and d. n. voskresensky , astrophys . j. * 533 * , 406 ( 2000 ) [ arxiv : astro - ph/9908334 ] . d. blaschke , h. grigorian and d. n. voskresensky , astron . astrophys . * 368 * , 561 ( 2001 ) [ arxiv : astro - ph/0009120 ] . t. schafer and k. schwenzer , phys . d * 70 * , 114037 ( 2004 ) [ arxiv : astro - ph/0410395 ] . m. alford , p. jotwani , c. kouvaris , j. kundu and k. rajagopal , phys . d * 71 * , 114011 ( 2005 ) [ arxiv : astro - ph/0411560 ] . p. jaikumar , c. d. roberts and a. sedrakian , phys . c * 73 * , 042801 ( 2006 ) [ arxiv : nucl - th/0509093 ] . a. schmitt , i. a. shovkovy and q. wang , phys . d * 73 * , 034012 ( 2006 ) [ arxiv : hep - ph/0510347 ] . r. anglani , g. nardulli , m. ruggieri and m. mannarelli , phys . d * 74 * , 074005 ( 2006 ) [ arxiv : hep - ph/0607341 ] . x. g. huang , q. wang and p. f. zhuang , phys . d * 76 * , 094008 ( 2007 ) [ arxiv:0706.4381 [ hep - ph ] ] . x. g. huang , q. wang and p. f. zhuang , int . j. mod . e * 17 * , 1906 ( 2008 ) [ arxiv:0710.3634 [ hep - ph ] ] . c. kouvaris , phys . d * 79 * , 123008 ( 2009 ) [ arxiv:0901.1028 [ astro-ph.sr ] ] . d. g. yakovlev and c. j. pethick , ann . astrophys . * 42 * , 169 ( 2004 ) [ arxiv : astro - ph/0402143 ] . a. sedrakian , prog . * 58 * , 168 ( 2007 ) [ arxiv : nucl - th/0601086 ] . d. blaschke and h. grigorian , prog . phys . * 59 * , 139 ( 2007 ) [ arxiv : astro - ph/0612092 ] . n. andersson , astrophys . j. * 502 * , 708 ( 1998 ) [ arxiv : gr - qc/9706075 ] . j. l. friedman and s. m. morsink , astrophys . j. * 502 * , 714 ( 1998 ) [ arxiv : gr - qc/9706073 ] . n. andersson and k. d. kokkotas , int . j. mod . d * 10 * , 381 ( 2001 ) [ arxiv : gr - qc/0010102 ] . l. lindblom , arxiv : astro - ph/0101136 . p. jaikumar , g. rupak and a. w. steiner , phys . d * 78 * , 123007 ( 2008 ) [ arxiv:0806.1005 [ nucl - th ] ] . b. knippel and a. sedrakian , phys . d * 79 * , 083007 ( 2009 ) [ arxiv:0901.4637 [ astro-ph.sr ] ] . q. d. wang and t. lu , phys . b * 148 * , 211 ( 1984 ) . r. f. sawyer , phys . b * 233 * , 412 ( 1989 ) [ erratum - ibid . b * 237 * , 605 ( 1990 ) ] . j. madsen , phys . d * 46 * , 3290 ( 1992 ) . x. p. zheng , s. h. yang and j. r. li , phys . b * 548 * , 29 ( 2002 ) [ arxiv : hep - ph/0206187 ] . z. xiaoping , k. miao , l. xuewen and y. shuhua , phys . c * 72 * , 025809 ( 2005 ) [ arxiv : hep - ph/0507216 ] . j. d. anand , n. chandrika devi , v. k. gupta and s. singh , pramana * 54 * , 737 ( 2000 ) . b. a. sad , i. a. shovkovy and d. h. rischke , phys . d * 75 * , 065016 ( 2007 ) [ arxiv : astro - ph/0607643 ] . m. g. alford and a. schmitt , j. phys . g * 34 * , 67 ( 2007 ) [ arxiv : nucl - th/0608019 ] . b. a. sad , i. a. shovkovy and d. h. rischke , phys . d * 75 * , 125004 ( 2007 ) [ arxiv : astro - ph/0703016 ] . h. dong , n. su and q. wang , phys . d * 75 * , 074016 ( 2007 ) [ arxiv : astro - ph/0702104 ] . m. g. alford , m. braby , s. reddy and t. schafer , phys . rev . c * 75 * , 055209 ( 2007 ) [ arxiv : nucl - th/0701067 ] . c. manuel and f. j. llanes - estrada , jcap * 0708 * , 001 ( 2007 ) [ arxiv:0705.3909 [ hep - ph ] ] . m. mannarelli , c. manuel and b. a. sad , phys . * 101 * , 241101 ( 2008 ) [ arxiv:0807.3264 [ hep - ph ] ] . m. mannarelli and c. manuel , arxiv:0909.4486 [ hep - ph ] . h. dong , n. su and q. wang , j. phys . g * 34 * , s643 ( 2007 ) [ arxiv : astro - ph/0702181 ] . m. g. alford , m. braby and a. schmitt , j. phys . g * 35 * , 115007 ( 2008 ) [ arxiv:0806.0285 [ nucl - th ] ] . r. c. duncan and c. thompson , astrophys . j. * 392 * , l9 ( 1992 ) . c. thompson and r. c. duncan , astrophys . j. * 408 * , 194 ( 1993 ) . c. thompson and r. c. duncan , mon . not . soc . * 275 * , 255 ( 1995 ) . c. thompson and r. c. duncan , astrophys . j. * 473 * , 322 ( 1996 ) . c. kouveliotou _ et al . _ , astrophys . j. * 510 * , l115 ( 1999 ) [ arxiv : astro - ph/9809140 ] . c. y. cardall , m. prakash and j. m. lattimer , astrophys . j. * 554 * , 322 ( 2001 ) [ arxiv : astro - ph/0011148 ] . a. e. broderick , m. prakash and j. m. lattimer , phys . b * 531 * , 167 ( 2002 ) [ arxiv : astro - ph/0111516 ] . d. lai and s. l. shapiro , astrophys . j. * 383 * , 745 ( 1991 ) . s. chakrabarty , d. bandyopadhyay and s. pal , phys . * 78 * , 2898 ( 1997 ) [ arxiv : astro - ph/9703034 ] . d. bandyopadhyay , s. chakrabarty and s. pal , phys . lett . * 79 * , 2176 ( 1997 ) [ arxiv : astro - ph/9703066 ] . a. broderick , m. prakash and j. m. lattimer , arxiv : astro - ph/0001537 . s. r. de groot , _ the maxwell equations . non - relativistic and relativistic derivations from electron theory _ , amsterdam , north - holland pub . co. , 1969 . n. sadooghi , arxiv:0905.2097 [ hep - ph ] . w. a. hiscock and l. lindblom , phys . d * 31 * , 725 ( 1985 ) . w. israel and j. m. stewart , annals phys . * 118 * , 341 ( 1979 ) . e. m. lifshitz and l. p. pitaevskii , _ physcial kinetics , course of theoretical physics , vol 10 _ , pergamon press ltd . , 1981 . m. chaichian , s. s. masood , c. montonen , a. perez martinez and h. perez rojas , phys . lett . * 84 * , 5261 ( 2000 ) [ arxiv : hep - ph/9911218 ] . d. kondepudi and i. prigogine , _ modern thermodynamics _ , john wiley and sons , 1998 . s. chakrabarty , d. bandyopadhyay and s. pal , int . j. mod . a * 13 * , 295 ( 1998 ) . z. xiao - ping , z. xia and l. xue - wen , arxiv : astro - ph/0507390 . j. r. ipser and l. lindblom , astrophys . j. * 355 * , 226 ( 1990 ) . l. lindblom , g. mendell and b. j. owen , phys . d * 60 * , 064006 ( 1999 ) [ arxiv : gr - qc/9902052 ] . l. lindblom , b. j. owen and s. m. morsink , phys . lett . * 80 * , 4843 ( 1998 ) [ arxiv : gr - qc/9803053 ] . b. j. owen , l. lindblom , c. cutler , b. f. schutz , a. vecchio and n. andersson , phys . d * 58 * , 084020 ( 1998 ) [ arxiv : gr - qc/9804044 ] . b. a. sad , arxiv:0806.3359 [ astro - ph ] . e. f. brown , l. bildsten and p. chang , astrophys . j. * 574 * , 920 ( 2002 ) .
fundamental physical observables of the gravitational field , such as energy and angular momentum , are notoriously hard to define satisfactorily . in asymptopia , where there exist the well - defined bondi @xcite and arnowitt - deser - misner ( adm ) @xcite masses , the quantities are non - local . the inherent non - locality is usually attributed to the equivalence principle ; one needs at least two observers to distinguish geodesic deviation from acceleration - dependent quantities . nevertheless , given the existence of the bondi and adm energy definitions on 2-surfaces at null and spatial infinities , it is reasonable to expect a degree of localization is possible . one possibility given by penrose @xcite is to associate quasilocal quantities to compact , oriented spatial 2-surfaces . given the difficulty in defining gravitational observables , it is perhaps not surprising that there exists a plethora of proposed quasilocal quantities ( @xcite - @xcite ) . the definitions of quasilocal energy may be distinguished by the ability to satisfy various criteria . these include vanishing in flat space , taking reasonable values for spherically symmetric solutions , and approaching the adm and bondi values in appropriate limits . it is by no means simple to satisfy these criteria @xcite , @xcite . with the complications involved in defining energy even in classical general relativity , it may seem perilous to define quantum quasilocal energy . however , in the context of the canonical ( @xmath0 ) theory there is a clear selection criteria for the correct physical quantity . when defined within a bounded region , functional differentiability of the constraints requires that , generically , surface terms must be added to make the theory well - defined . the surface term required by the hamiltonian constraint is the hamiltonian for the system on shell . in general , the method of functional differentiability generates boundary conditions on the phase space variables , gauge parameters , and surface terms ( as is explored in some depth in ref . @xcite for gravity and bf theory ) . originally , it was noticed that a surface term had to be added to the action so as to match the asymptotic expression . however , regge and teitelboim found that gravitational theory was simply inconsistent without this boundary term @xcite . the gravitational action has derivatives of the phase space functions , so the variation generates surface terms . it is only when these terms vanish ( or are canceled ) that the theory has a well - defined variational principle . the boundary hamiltonian or quasilocal energy is of this form ; it cancels a term arising in the variation of the hamiltonian constraint . this paper provides a quantization of the surface observable @xmath1 this reduces to the quasilocal energy of brown and york @xcite , to the adm energy in asymptopia , and to the misner - sharp mass in spherical symmetry ( @xcite ) . the quantization is carried out in the context of canonical quantum gravity in the real new variables @xcite or , succinctly , `` spin - net gravity '' in which the state space is built from functions of holonomies based on graphs . ( see ref . @xcite for a recent review . ) perhaps the most remarkable result of this study is the discreteness in geometric operators . length @xcite , area ( @xcite ) , volume ( @xcite ) , and angle @xcite have been found to have fully discrete spectra . in this approach to quantum gravity space is discrete . there have been two previous quantizations of eq . ( [ hclass ] ) . the first , by baez , muniain , and piriz , was completed before many of the spin - net techniques were developed @xcite . the action of the operator was qualitatively described as a `` shift . '' in the second , two adm energy operators were introduced in the framework of thiemann s quantum spin dynamics ( @xcite - @xcite ) . the classical expression used in that quantization was different than the one used here . in fact , the expression is only weakly equal to eq . ( [ hclass ] ) @xcite . nevertheless , the resulting quantum operator is very similar and reduces to this adm expression under the same restrictions . a more extensive comparison between the operators is given in section [ conclusion ] . the present work completes the quantization of the boundary hamiltonian , even in a non - asymptotically flat context . the quantum definition is to satisfy modest criteria . i only ask that the operator be well - defined on the hilbert space induced from the full gauge invariant space . important criteria , including the proper behavior on semiclassical states and the appropriate algebra of boundary observables , is left for further investigation . definitions of simple semiclassical states or weaves @xcite are explored in @xcite . the remainder of this paper is organized as follows : spin - net gravity is briefly reviewed in section [ qg ] . this serves to base the definition of the quasilocal energy in the framework of the classical theory and provides an opportunity to fix notation and units . the main developments are contained in section [ qe ] where the operator is regularized in [ reg ] and given final form in [ recouplingnote ] . an alternate quantization is given in [ qevol ] and the full spectra of the resulting operators is presented in [ spectra ] . this paper concludes with a comparison with thiemann s @xmath2 operator and some discussion on wider implications of the quantum quasilocal energy operator . this section sets the classical framework for the operator definitions and provides the basic elements of spin - net quantization . meanwhile , i use the opportunity to fix signs , factors , and units . for an introduction , rather than a brief review , the reader is encouraged to read refs . @xcite . for four - dimensional spacetimes @xmath3 with @xmath4 , where @xmath5 is compact , the ( @xmath0)-action for vacuum , riemannian general relativity is @xmath6 = { 1 \over 8 \pi g } \int_{t_1}^{t_2 } dt \int_{\sigma } d^3x [ \ e^{ai}\dot{a}_a^i - n{\cal h } - n^a{\cal d}_a -\lambda^i g^i\ ] .\ ] ] the action is written in terms of a real , @xmath7-valued connection one - form @xmath8 and its momenta , a densitized inverse triad @xmath9 . i have included the overall factor depending on newton s constant @xmath10 ( @xmath11 throughout the paper ) . the phase space variables satisfy , @xmath12 @xcite . i take @xmath13 until the final section , when it is included in the spectra of the quasilocal energy . ] @xmath14 the lagrange multipliers @xmath15 , @xmath16 , and @xmath17 are , respectively , the lapse , shift , and @xmath18 gauge rotation parameters . varying the action with respect to the last two functions gives the constraints @xmath19 where @xmath20 . varying the lapse gives the hamiltonian constraint . defining @xmath21}^i + \epsilon^{ijk}a_a^j a_b^k$ ] , one may express the constraint in integrated form as @xcite @xmath22}^{b } ( a_{a}^{i } - \gamma_{a}^{i } ) ( a_{b}^{j } - \gamma_{b}^{j } ) \right ) \approx 0.\ ] ] due to the choice of using real variables , the additional term has been added to the constraint @xcite . the triad - compatible connection @xmath23 satisfies @xmath24 . this constraint generates time evolution . a key observation which affects the definition of the quasilocal energy is that the hamiltonian constraint has density weight + 2 so the lapse function has density weight -1 . as bounded spatial regions are the subject of this work , it is best to start by fixing notation . the calculations are in a spacetime of the form @xmath25 , although the space has at least one compact subset @xmath26 such that the boundary of the interior @xmath27 , @xmath28 , is homeomorphic to a 2-sphere . i usually denote the boundary @xmath28 as the surface @xmath29 . the topology of @xmath5 is not specified but two examples are worth keeping in mind . one is a compact spatial slice @xmath5 with one boundary and two `` interiors '' @xmath27 and its complement @xmath30 . the other is the topology of the asymptotically flat spacetimes in which @xmath5 is homeomorphic to @xmath31 with a compact ball cut out . when the theory of eq . ( [ action ] ) is applied to a bounded spatial region @xmath27 , it is no longer well - defined @xcite . the problem arises because the phase space of the compact theory does not contain the physical solutions of the bounded one @xcite ; the variational problem has no solutions . this comes about as boundary terms arise in the variation , making hamilton s equations ill defined ; dynamics takes the system outside of the phase space . the ( @xmath0)-action , or the constraints , must be functionally differentiable . when the theory is defined in a finite region , this requires the addition of surface terms and/or the imposition of boundary conditions . these surface terms , without which the theory would be inconsistent , are the fundamental observables and necessarily satisfy the same algebra as the constraints @xcite . under the variation of the connection , the hamiltonian constraint generates a surface term which must either vanish through the imposition of boundary conditions or be canceled by another surface term . if the lapse is non - vanishing on @xmath28 then the surface term of eq . ( [ hclass ] ) must be added to the constraint . it is this term which is the quasilocal energy and reduces to the adm energy in asymptopia . to ensure that the theory is dynamically well - defined one must impose a complete set of conditions on the boundary @xmath32 . to be concrete , i consider the specific boundary conditions : @xmath33 see ref . @xcite ( or , for a more general setting , ref . @xcite ) for details . this work is devoted to quantizing the observable of eq . ( [ hclass ] ) . before beginning this , i review the quantization program in which the quasilocal energy operation is defined . the mathematically precise formulation is developed in refs . @xcite . the natural framework for a diffeomorphism invariant gauge theory is based on wilson loops @xcite . however , to build three - dimensional geometry it is necessary to use a more general structure based on graphs . the quantum configuration space is most appropriately constructed from holonomies along the edges of a graph . i denote a graph embedded in @xmath5 by @xmath34 . it contains a set of @xmath15 edges @xmath35 and a set of vertices @xmath36 . the key idea is that every smooth connection @xmath37 associates a group element to an edge @xmath38 of @xmath39 via the holonomy , @xmath40 here , @xmath41 with @xmath42 proportional to the pauli matrices via @xmath43 . given an embedded graph , holonomies along the edges , and a complex ( haar - integrable ) function on @xmath44 , one may define a `` cylindrical function '' @xmath45 ( the function @xmath46 is `` cylindrical '' since @xmath47 only depends on a finite number of directions in the space of connections ; it is constant on all other directions . ) with the appropriate norm , the completion of this function space gives the hilbert space @xmath48 . a basis on @xmath48 is given by spin network states @xcite . i denote a spin network @xmath49 by the triple @xmath50 of an oriented graph @xmath51 , labels on the vertices ( or `` intertwiners '' ) , @xmath52 , and integer edge labels , @xmath53 , indexing the representation carried by the edge . the corresponding spin net state @xmath54 in @xmath55 is defined in the connection representation as @xmath56\ ] ] where the holonomy @xmath57 along edge @xmath38 is in the @xmath58 irreducible representation of @xmath18 . when the intertwiners `` tie up '' all the incident edges when they are invariant tensors on the group these states are gauge invariant . as the triads are dual to pseudo two - forms , they most comfortably live on 2-surfaces , denoted by @xmath29 . @xmath59 in which @xmath60 are coordinates on the surface and @xmath61 is the normal . the action of the triads on a function of holonomies may be computed from the poisson brackets , eq . ( [ pb ] ) . a short calculation shows that the bracket with a single edge @xmath62 is given by @xmath63 in which @xmath64 denotes the action of the left ( or right ) invariant vector fields on the group element @xmath62 . does satisfy the algebra of eq . ( [ jalg ] ) , it has an additional property : under orientation reversal of the edge @xmath38 , @xmath65 changes sign @xcite . this may be seen directly from eq . ( [ triadbracket ] ) . ] the bracket will generally have a sum of terms . when a cylindrical function based on a graph @xmath34 , say @xmath66 , is used , the triad defined on a surface , @xmath67 , gives a sum over all intersections , @xmath68 , of the surface and the graph @xmath34 @xmath69 the sum is over all edges @xmath27 incident to @xmath68 ( denoted @xmath70 ) . the geometric factor @xmath71 is defined by @xmath72 this geometrical factor ties the tangent space of the edge ( through @xmath73 ) to the orientation of the surface . due to the behavior of @xmath74 under edge orientation reversal , this factor maintains edge orientation independence and surface orientation dependence . there are two further remarks to make . first , the result is non - vanishing only when there is at least one intersection between the graph @xmath34 and the surface @xmath29 . second , the overall factor of @xmath75 can be seen to arise from a `` thickened surface '' regularization @xcite . with this preparation , one may define the quantum triad operator on a spin network state @xmath76 @xmath77 where the angular momentum - like operator @xmath78 satisfies the usual algebra @xmath79 = i\ , \hbar \ , \epsilon^{ijk } \delta_{e , e ' } \hat{j}_{(e)}^{k}.\ ] ] ( the @xmath80-function restricts the relation to one edge ; @xmath81 on distinct edges commute . ) the triad operator of eq . ( [ eop ] ) is essentially self - adjoint in the hilbert space of quantum gravity @xcite . the diagrammatic form of this operator is the `` one - handed '' @xcite @xmath82 in which the index @xmath83 is the internal space index . the length scale of the theory , @xmath84 , is defined by @xmath85 . the grasping is chosen such that , in the plane of the diagram , the 2-line is on the left when the orientation on the edge points up ( vice versa for downward orientations ) @xcite . though the overall action is orientation independent , such a diagrammatic representation of grasping involves the choice of a sign . the geometric factor @xmath71 in the expression for the `` unclasped hand '' in eq . ( [ ediagram ] ) ensures that the operator is independent of edge orientation @xcite . the quantum configuration space based on graphs and triads which act as signed angular momentum operators are the basic elements of spin - net gravity . the definition of the quasilocal energy makes use of two further results , the geometric operators of area and volume . the area of a surface @xmath29 may be , for simplicity , specified by @xmath86 in an adapted coordinate system . expressed in terms of the triad @xmath9 , the area of the surface only depends on the @xmath87rd vector component via @xcite @xmath88 the quantum operator is defined using the operators of eq . ( [ eop ] ) and by partitioning the surface @xmath29 so that only one edge or vertex threads through each cell in the partition . the integral of eq . ( [ areae ] ) becomes a sum over operators that act only at intersections of the surface with the spin network @xmath89 the spectrum may be computed with recoupling theory @xcite or operator methods @xcite . in both cases it is found by considering all the intersections of the spin network with the surface @xmath29 , including vertices which lie in the surface . the edges incident to a vertex in the surface may be divided into three categories : those which have tangents aligned with the surface normal @xmath90 , anti - aligned @xmath91 , and tangent to the surface @xmath92 . summing over all contributions , eq . ( [ areaop ] ) becomes @xcite @xmath93^{1/2 } { \mid s \rangle}.\ ] ] this result suggests that space is discrete ; measurements of area can only take quantized values . other geometric operators such as length @xcite , angle @xcite , and volume share this property . the construction of the volume operator is more complicated than the area construction . nevertheless , it is possible , starting from the classical expression @xmath94 to regularize the quantity and define the operator @xcite , @xcite - @xcite @xmath95 where the sign factor is given by @xmath96 . ( incident edges are oriented outward for ease of writing . ) the spectra of such an operator can be worked out using recoupling theory as in refs . @xcite , @xcite , and @xcite . introductions to diagrammatic recoupling theory may be found in refs . to begin , i give a regularization of the hamiltonian @xmath97 with a scalar lapse . this term may be made into a well - defined quantum operator on the appropriate state space . however , it does not have the dimensions of energy . two normalizations to fix this problem are explored , one using area and the other using volume . in both cases the full spectrum may be computed . there are differences . the operator which , in the naive classical limit , gives the familiar expression for the adm energy and quasilocal energy requires a restriction on the hilbert space . the alternate operator is defined on the full hilbert space but does not have the correct naive classical limit . this is discussed in sections [ qevol ] and [ conclusion ] . the next section provides a regularization of @xmath98 . a recoupling identity studied in [ recouplingnote ] suggests the final form of the operators . the spectra are given in section [ spectra ] . the classical quantity to be promoted to a quantum operator is given by @xmath99 ( for the remainder of this paper i explicitly give the density weights using tildes . this notation also serves to distinguish the two forms of the quasilocal energy operator . ) the regularization is based on the observation that the boundary hamiltonian has two parts . one part , @xmath100 , is the familiar triad operator integrated over a 2-surface as in eq . ( [ edef ] ) . the second part , @xmath101 , may be roughly described as a projection of the connection along an edge of a spin network . these two parts are regularized by point splitting . the regularization of the classical hamiltonian is in several steps . the surface is thickened , as in the definition of the area operator in ref . this is done by introducing a smooth coordinate @xmath102 over a finite neighborhood of @xmath29 with `` thickness '' @xmath103 . the boundary itself is located at @xmath104 . the thickened surface has a natural foliation in terms of @xmath102 and gives a way to regulate the operator . this portion of the regularization handles the tangent space factor ( the sign @xmath105 ) . the boundary integration is transformed to an integration over the thickened surface , a region @xmath106 : @xmath107 the region @xmath106 is further partitioned into cells @xmath108 which have the property that the coordinate width , denoted by @xmath109 , is tied to the coordinate height so that @xmath110 . the power @xmath111 is restricted to lie between 1 and 2 . by construction , the number of cells @xmath112 is tied to the same limit . the fully regularized operator is averaged over the leaves of the foliation of @xmath106 and point split . the classical expression is defined by @xmath113_{\epsilon } : = \frac{1}{4 \pi g } \frac{1}{\tau } \sum_{c=1}^{n_{\epsilon } } \int \limits_{{\cal r}_{c } \otimes { \cal r}_{c } } \ ! \ ! \ ! \ ! d^{3}x \ , d^{3}y \ , \epsilon^{ijk } ( a_{a}^{i } \tilde{e}^{aj})(y ) \ , n(x ) \ , n_{b}(x ) \tilde{e}^{bk}(x).\ ] ] the regulated hamiltonian is well - defined and goes to eq . ( [ hclass ] ) in the limit as @xmath109 vanishes . in the next steps i re - express the classical variables @xmath114 and @xmath115 in a suitable form for quantization . making use of the identity @xmath116 \tau^{k}\right]$ ] , @xmath117_{\epsilon}$ ] is transformed into the convenient expression @xmath118_{\epsilon } = - \frac{1}{2 \pi g } \frac{1}{\tau } \sum_{c=1}^{n_{\epsilon } } \ , \int \limits_{{\cal r}_{c } \otimes { \cal r}_{c } } \ ! \ ! \ ! d^{3}x \ , d^{3}y \ , n(x ) n_{b}(x ) { \rm tr}\left [ [ a_{a},\tilde{e}^{a}](y ) \tilde{e}^{b}(x ) \right].\ ] ] using the triad integrated over a surface , eq . ( [ edef ] ) , and the action on one edge , eq . ( [ triadbracket ] ) , a short calculation shows that @xmath119_{\epsilon } = - \frac{1}{2 \pi g}(4 \pi g)^{2 } \frac{1}{\tau } \sum_{c=1}^{n_{\epsilon } } \ , \int \limits_{e_{i } \in { \cal r}_{c } } \ ! \ ! \ ! dt \int \limits_{e_{j } \in { \cal r}_{c } } \ ! \ ! \ ! ds n(e_{i}(t ) ) n_{b}(e_{i}(t ) ) \ , \dot{e}_{i}^{b}(t ) \\ \times { \rm tr}\left [ [ a_{a } ( e_{j}(s ) ) \dot{e}_{j}^{a}(s ) , j_{j } ] j_{i } \right ] . \end{split}\ ] ] next , the `` edge - projected connection '' @xmath120 is expressed in terms of holonomies . this can be done in a straightforward manner as smooth connections satisfy @xmath121 the partitioning of the thickened surface splits the integration along the edges in eq . ( [ step4 ] ) into segments of maximum coordinate length @xmath109 . the integral @xmath122 is approximated , in the limit , by @xmath123 . meanwhile , the integral along the edge @xmath124 in @xmath125 contains the geometrical factor @xmath126 this result is a consequence of the construction of the cells . and @xmath109 @xcite . ] as is clear from this expression , the action of the regularized expression is , in the limit , only non - vanishing when an edge or a vertex of the cylindrical function s graph @xmath34 contains a transverse intersection with the bounding surface . for a finite graph , the sum over cells becomes a finite sum over these intersections . as @xmath109 tends to zero @xmath117_{\epsilon}$ ] of eq . ( [ step4 ] ) becomes is or , shortly is promoted to , an operator . this means that the leading order contribution of the factor @xmath127 is linear in @xmath109 . if desired one can include an explicit subtraction term @xmath128 . the resulting operator is identical in action , but the spectrum differs from the one considered here by a factor of 2 . ] @xmath129 + o(\epsilon^{2})\ ] ] in which @xmath130 is the value of the lapse at the vertex @xmath68 . the quantum expression may then be defined as @xmath131 { \mid s \rangle}\ ] ] on a spin network state @xmath54 . the regularization in which the classical hamiltonian is averaged over a thickened surface and point - split produces a gauge invariant , seemingly well - defined boundary hamiltonian . it is a new operator on the kinematic state space . its action has an intriguing `` self - measuring '' property : reading from left to right , the operator grasps an edge then alters the state by adding the edge , @xmath132 . the operator measures the angular momentum of the edge which it modified . before exploring the details of the action , a more obvious problem must be addressed . this operator does not have the correct dimension . the @xmath133 is in the definition of the operator @xmath134 so has the units of area rather than the dimensions @xmath135 of energy . ] the source of this trouble is the ( neglected ) density weight tucked into the lapse ; the expression is missing a factor of dimension inverse volume . there are two ways to fix this problem . first , as the energy is defined on a 2-surface , the natural unit of length is provided by the area operator of eq . ( [ areaop ] ) which may be used to normalize the surface term . the quantum quasilocal energy operator on a surface @xmath29 is defined by @xmath136^{3 } } { \rm tr}\left [ u_{e_{j}}\hat{j}_{j}u_{e_{j}}^{-1 } \hat{j}_{i } \right ] { \mid s \rangle}.\ ] ] the operator @xmath137 acts only on the vertex @xmath68 . as the area operator is a well - defined operator on the state space , it can be replaced with its spectral resolution . the second solution to the correct normalization is to include the volume factor in the regularization . a regularization of the latter form is given in section [ qevol ] . this section is devoted to a recoupling identity which provides a more tractable form of the quasilocal energy operator defined in the last section . the definition simplifies dramatically . the techniques employed are those of diagrammatic recoupling theory introduced to canonical quantum gravity in ref . @xcite and further developed in refs . @xcite . a complete development of the techniques in terms of `` framed spin networks '' or temperley - lieb recoupling theory is in ref . @xcite . starting from the definition of the quasilocal energy operator eq . ( [ qeadef0 ] ) a direct , but moderately lengthy , recoupling calculation yields the full spectrum . however , there is a shorter method . the idea is to use recoupling to investigate and re - express the action of @xmath138.\ ] ] this factor , acting on the edge @xmath139 , can be reduced to a simple form . to see this , note that the two hands , @xmath140 and @xmath141 , grasp edges independently . the unusual action is entirely contained in @xmath142 which has the effect of overlaying a segment of the edge @xmath139 , grasping the altered edge , and then retracing the segment . the overlaying can easily be accounted for using the `` edge addition formula '' @xcite @xmath143 thus , when the holonomy @xmath144 acts on the edge @xmath139 ( labeled by @xmath145 ) , the result is the linear combination @xmath146 the operator @xmath141 acts next . it is tied to the same edge and so measures the angular momentum on the edge that was modified . must act on the edge @xmath139 , there is some ambiguity in where , diagrammatically , it grasps the edge . not surprisingly , the results are the same , even when the operator grasps the edge above or below the modified edge . ] the edge @xmath139 is grasped , yielding a 2-line and an overall factor of @xmath147 @xmath148 finally , the `` loop '' ( @xmath149 ) is closed and the second hand is included to give the result @xmath150 this is the diagrammatic action of the factor @xmath151 . by the wigner - eckart theorem each of these diagrams is equivalent to a single trivalent vertex @xcite . diagrammatically , the reduction is accomplished in two stages , each of which removes a triangular loop . the first is given by @xmath152 the needed recoupling coefficients are given in the appendix ( see eqs . ( [ tetident ] ) and ( [ teteg ] ) ) . inserting these results in eq . ( [ recoupident1 ] ) , one finds @xmath153 { \mathchoice { { \raisebox{-8pt } { \hspace{-1pt } \epsfig{file = figs/2hand.ps , height=24pt , silent= } \hspace{2pt } } } } { { \raisebox{-5pt } { \hspace{1pt } \epsfig{file = figs/2hand.ps , height=20pt , silent= } \hspace{1pt } } } } { { \raisebox{-10pt } { \hspace{-3pt } \epsfig{file = figs/2hand.ps , height=20pt , silent= } \hspace{1pt } } } } { { \raisebox{-8pt } { \hspace{-1pt } \epsfig{file = figs/2hand.ps , height=18pt , silent= } \hspace{0pt } } } } } \\ & = - n { \mathchoice { { \raisebox{-8pt } { \hspace{-1pt } \epsfig{file = figs/2hand.ps , height=24pt , silent= } \hspace{2pt } } } } { { \raisebox{-5pt } { \hspace{1pt } \epsfig{file = figs/2hand.ps , height=20pt , silent= } \hspace{1pt } } } } { { \raisebox{-10pt } { \hspace{-3pt } \epsfig{file = figs/2hand.ps , height=20pt , silent= } \hspace{1pt } } } } { { \raisebox{-8pt } { \hspace{-1pt } \epsfig{file = figs/2hand.ps , height=18pt , silent= } \hspace{0pt}}}}}. \end{split}\ ] ] therefore , we learn that the numerator of the quasilocal energy is most simply expressed in terms of a signed , `` spin - orbit coupling '' term : @xmath138 \equiv \frac{1}{2 } \hat{j}^{i}_{i } \hat{j}^{i}_{j}\ ] ] in which i have used the identity @xmath154 . the additional sign in this expression is due to the way traces are represented in the diagrammatic algebra : whenever a loop is closed , an addition sign is included ( see , for instance , ref . @xcite ) . this identity suggests the final form of the quasilocal energy operator @xmath155 the properties of the operator are explored in section [ spectra ] after the `` volume normalized '' operator is in place . at this stage it is worth commenting on the operator ordering in eq . ( [ qeadef0 ] ) which now can be phrased in terms of the recoupling identity . the natural orderings of the expression are given by the cyclic permutations of the operators in the trace . a short investigation shows that there is one ordering which is distinct from the action given in the last section : the order @xmath156 $ ] differs in that , diagrammatically , @xmath141 grasps the edge @xmath139 before the operator overlays an edge . this produces a diagram distinct from those above . nevertheless , the two terms created with the edge addition formula trivially re - sum to give a loop in the 2-line of the operator . such a loop in a 2-line is equivalent to a 2-line . therefore , this is again proportional to @xmath157 $ ] . we have seen that the regularization of the boundary hamiltonian with scalar lapse led almost directly to this result , eq . ( [ qeadef ] ) . the key element of choice is the `` normalization . '' in the next section i regulate the true boundary hamiltonian . as we will see , this regularization is successful but the resulting operator requires a strong restriction on the state space . there is an alternate regularization of the operator in which the density is no longer absorbed by the lapse . in this case the classical quantity to be quantized is @xmath158 the density factor is also promoted to an operator . to regularize this expression one may follow thiemann @xcite and simultaneously soften the divergence of @xmath159 and point - split with @xmath160 in which the smoothed characteristic function @xmath161 satisfies @xmath162 note that @xmath163 . proceeding in a manner similar to the regularization in section [ reg ] , the two parts are point - split to give the regularized quasilocal energy @xmath164_{\epsilon } : = \frac{-1}{2 \pi g } \int_{s } d^{2}x \ , n_{b}(x ) \frac{n(x)}{\sqrt{q}(x ) } \int d^{3}y \frac{f_{\epsilon}(x , y)}{\epsilon^{3 } } { \rm tr } \left [ [ a_{a } , \tilde{e}^{a}](y ) \tilde{e}^{b}(x ) \right].\ ] ] to reach the quantum operator , the triads and connection are expressed in forms suitable for quantization . replacing the triads with the form in eq . ( [ triadbracket ] ) and the factor @xmath165 with @xmath166 , one finds that @xmath167_{\epsilon } & = - ( 8 \pi g ) \sum_{i , j } \int_{s } d^{2}x \ , n_{b}(x ) \frac{n(x)}{v_{\epsilon}(x ) } \int_{e_{i } } \ ! \ ! ds \ , \dot{e}_{i}(s ) \tilde{\delta}(x , e_{i}(s ) ) \\ & \times \int d^{3 } y \int_{e_{j } } \ ! \ ! dt \ , \tilde{\delta}(y , e_{j}(t ) ) f_{\epsilon}(x , y ) { \rm tr } \left [ [ a_{b}(y ) \dot{e}^{b}_{j}(t ) , j_{j } ] j_{i } \right ] \\ & = - ( 8 \pi g ) \sum_{i , j } \int_{s } d^{2 } x \ , n_{b}(x ) \frac{n(x)}{v_{\epsilon}(x)}\\ & \times \int_{e_{i } } \ ! \ ! dt \ , \dot{e}_{i}^{b}(t ) \tilde{\delta}(x , e_{i}(t ) ) \int_{e_{j } } \ ! \ ! ds \ , f_{\epsilon}(x , e_{j}(s ) ) { \rm tr } \left [ [ a_{b}(e_{j}(s ) ) \dot{e}^{b}_{j}(s ) , j_{j } ] j_{i } \right ] \end{split}\ ] ] with the delta function eliminating the @xmath168 integration in the second step . the connection may be replaced by a holonomy . this is possible as the factors inside the commutator are all evaluated on the edge @xmath139 . it is also convenient to partition the edge @xmath139 using the definition of holonomy @xmath167_{\epsilon } & = \lim_{n \rightarrow \infty } -(8 \pi g ) \sum_{i , j } \int_{s } d^{2}x \ , n_{b}(x ) \frac{n(x)}{v_{\epsilon}(x ) } \int_{e_{i } } \ ! \ ! dt \ , \dot{e}_{i}^{b}(t ) \delta(x , e_{i}(t ) ) \\ & \times \sum_{k=1}^{n } f_{\epsilon}(x , e_{j}(s_{k-1 } ) ) { \rm tr}\left [ u_{e_{j}}(s_{k-1 } , s_{k } ) j_{j } u_{(e_{j})^{-1}}(s_{k-1 } , s_{k } ) j_{i } \right ] \\ & = \lim_{n \rightarrow \infty } - ( 8 \pi g ) \sum_{v \in s \cap { \mathsf g } } \sum_{i\dashv v , j } \chi_{i}^{s } \frac{n_{v}}{v_{\epsilon}(v ) } \sum_{k=1}^{n } f_{\epsilon}(v , e_{j}(s_{k-1}))\\ & \times { \rm tr}\left [ u_{e_{j}}(t_{k-1 } , t_{k } ) j_{j } u_{(e_{j})^{-1}}(t_{k-1 } , t_{k } ) j_{i } \right ] \end{split}\ ] ] where the usual definition for @xmath71 [ eq . ( [ chi ] ) ] is used . i denote the intersection between the edge @xmath124 and the surface @xmath29 by @xmath68 . the volume @xmath169 acts on the vertex @xmath68 . in the limit as @xmath109 vanishes , @xmath170 goes to @xmath171 if , and only if , the edge @xmath139 is incident to @xmath68 . thus , only one term in the sum over partitions of the edge @xmath139 survives and the sum over of @xmath134 is tied to the vertex @xmath68 . also in the limit , the volume goes to @xmath172 , the volume at the vertex @xmath68 . as the regulator is removed , @xmath173_{\epsilon } = - ( 8 \pi g ) \sum_{v \in s \cap { \mathsf g } } \sum_{i , j \dashv v } \chi_{i}^{s } \frac{n_{v}}{v(v ) } { \rm tr}\left [ u_{e_{j } } j_{j } u_{(e_{j})^{-1 } } j_{i } \right]\ ] ] in which the holonomy is defined to start at the incident end of @xmath139 ( the `` germ '' of ref . @xcite ) . the quantization is now immediate . let me pull the units out of the geometric operator so that @xmath174 using the recoupling identity of eq . ( [ recoupleident ] ) , quantum boundary hamiltonian is defined as @xmath175}{\hat{v}_{v}}.\ ] ] the reciprocal of the volume operator is evaluated using its spectral resolution . though the regulation of the boundary hamiltonian is possible , the result suffers from the same difficulty as the @xmath176 operator has in ref . @xcite : the signed angular momentum operator @xmath177 does not commute with the volume . to prevent the whole operator from diverging , one can restrict the state space so that the graph has no edges tangent to the surface . this is the tangle property of ref . @xcite . with this modified state space , denoted by @xmath178 , the quasilocal energy becomes @xmath179 the spin network state @xmath178 satisfies the tangle property on @xmath29 . this is nearly identical to the @xmath176 operator of ref . @xcite . aside from an overall factor , this operator differs from @xmath176 in the sign @xmath71 and the lapse @xmath130 . indeed , under the same asymptotically flat conditions as in ref . @xcite ( unit lapse , tangle property , and all edges outgoing ) , this is @xmath176 . a detailed comparison between @xmath180 and @xmath176 is in section [ conclusion ] , but it is worth offering one observation here . in the context of a quasilocal operator the tangle property seems to be overly restrictive . since the quasilocal quantity may be applied to any surface in @xmath5 , restricting the domain of this operator , and thus the hilbert space , implies that there are no surfaces with edges inside them . since every edge lies within some surface , the property effectively eliminates the entire state space . the tangle property is too strong for the quasilocal energy operator . the operators of eqs . ( [ quanthdef ] ) , ( [ qeadef ] ) , and ( [ qevdef ] ) are quite similar to the geometric operators of spin - net gravity and can be treated with the same methods . the spectra may be computed using the recoupling methods of refs . @xcite or the operator methods of refs . i first give the spectrum of the numerator , the operator `` @xmath181 . '' for completeness , the calculation is in the space of gauge non - invariant spin networks , the `` extended spin networks '' of ref . @xcite . it is convenient to order the edges of each vertex into categories according to the geometric factor @xmath71 . for a vertex of valence @xmath182 , the @xmath183 edges for which @xmath184 is positive are labeled @xmath185 to @xmath186 . the @xmath187 edges for which @xmath184 vanishes are labeled @xmath188 to @xmath189 . finally , the @xmath190 edges for which @xmath184 is negative are labeled @xmath191 to @xmath192 . there is no restriction on the order of edges within these partitions . diagrammatically , this ordering is equivalent to selecting three intertwiner trees which grow from internal edges labeled by @xmath193 , @xmath194 , and @xmath145 , according to the value of @xmath184 . the angular momentum operators are similarly partitioned so that @xmath195 for edges with @xmath196 ; @xmath197 for edges with @xmath198 ; and @xmath199 for edges with @xmath200 . using the methods from the quantum mechanics of angular momentum , one finds @xmath201 two properties of this operator are immediately clear . first , the operator simplifies when evaluated on gauge invariant spin networks . in fact , the spectrum collapses and contains only the value zero . this is obvious from eq . ( [ jform ] ) as this operator contains @xmath202 which is the quantum version of the gauss constraint , eq . ( [ gauss ] ) . operators , is edge orientation invariant . in the current work , the invariance comes from the `` edge - projected connection '' in eq . ( [ step4 ] ) . when the orientation of the edge @xmath134 is reversed , both the operator @xmath141 and the component along the connection change sign . ] of course , one may simply compute the result as well . for general vertices a recoupling calculation of the terms in eq . ( [ jform ] ) shows that the operator vanishes . the recoupling terms are of two forms . the first is diagrammatically @xmath203 making use of identity eq . ( [ bubnident ] ) . the second is @xmath204 making use of identity eq . ( [ trivident ] ) . thus , as a consequence of summing up all the components of the connection along the incident edges ( eq . ( [ acomponent ] ) ) , the quasilocal energy operator vanishes on gauge invariant states . the second property seems equally serious . since the operator of eq . ( [ jform ] ) contains two terms which do not commute , @xmath205 and @xmath206 , the spectrum is not well - defined . these terms can not be simultaneously diagonalized . this result may be seen in the diagrammatic picture as well . in the gauge non - invariant case , vertices of this type have an intertwiner with four edges , one of which is not grasped by either operator . ( this extra edge can be thought of as leaving the 3-dimensional manifold . ) the operator of eq . ( [ jform ] ) requires that all three edges labeled by @xmath193 , @xmath145 , and @xmath194 are part of a single trivalent vertex . this is not possible with a four - valent intertwiner . these properties suggest that the state space be carefully examined . the space on which this operator acts is not characterized by general , gauge invariant vertices . when the spatial manifold is cut into regions with boundary , the hilbert space @xmath207 of a bounded region @xmath27 is based on `` open '' graphs ( graphs with edges ending in vertices of valence 1 ) . given a graph in @xmath5 , i define the open subgraph @xmath208 to be the portion of @xmath34 within the interior @xmath27 and within the boundary @xmath28 , i.e. @xmath209 . thus , all the open edges are incident to the boundary @xmath29 and these gauge non - invariant edges are confined to the intersection of the graph with the boundary ; the `` extended spin networks '' of ref . @xcite are restricted to lie on the boundary . only edges that are incident to @xmath68 and in @xmath210 including tangent edges contribute to the energy . the hilbert space is defined as before , only now the graph used is @xmath211 . with @xmath207 it is possible to give the spectrum . the clearest way to express spectrum of `` @xmath212 '' on @xmath207 is to choose the orientations of the edges incident to the boundary to be outward pointing . this is suggests a form for the intertwiner `` core . '' it is a trivalent vertex labeled by @xmath213 and @xmath194 for the values of @xmath184 . with the orientations outward pointing , then the labels @xmath193 , @xmath145 , and @xmath194 take the meanings outside the surface , inside the surface , and tangent to the surface , respectively . as the @xmath193-edges are not in @xmath207 , they are not seen by the operator . therefore the operator of eq . ( [ jform ] ) reduces to @xmath214 using @xmath215 . for @xmath216 based on a single vertex @xmath68 the spectrum is @xmath217 { \mid s_{i } \rangle}.\ ] ] it is easy to generalize to more than one vertex ; simply sum over all contributing vertices . the spectrum is not positive definite and , acting on a single vertex , is explicitly bounded ( for finite spin ) . there are two cases worth identifying . when the vertex has no tangent edges , the spectrum , which i call type ( i ) , is always positive definite and is proportional to @xmath218 . the general case when there are tangent edges , called type ( ii ) , is given in eq . ( [ hispec ] ) . the form of the operator given in eq . ( [ 2jform ] ) also provides a simple way to check two critical properties of the quasilocal energy , gauge invariance and compatibility with the area operator . gauge invariance is clear from eq . ( [ jform ] ) , may be directly checked , = - i \hbar \epsilon^{ijk } \left ( \hat{j}_{(p)}^{j } \hat{j}_{(n)}^{k } + \hat{j}_{(p)}^{k } \hat{j}_{(n)}^{j } \right ) = 0.\ ] ] ] and can also be made manifest with @xmath219 this expression is also useful to show that the two operators @xmath181 and @xmath137 are compatible on the hilbert space @xmath207 . thus , the terms in the numerator and denominator may be diagonalized simultaneously . it is now possible to assemble all the factors for the full spectrum of the quasilocal energy operator on @xmath207 . for the operator defined in eq . ( [ qeadef ] ) , using the spectrum of the area operator , eq . ( [ areaspec ] ) , one has for type ( i ) @xmath220{n_{v}(n_{v}+2 ) } { \mid s \rangle}. \\ \text{for type ( ii):}\\ \hat{e}_{s}(n ) { \mid s \rangle } & = m \sum_{v \in s \cap { \mathsf g}^{i } } n_{v } \frac{p_{v}(p_{v}+2 ) + n_{v}(n_{v}+2 ) - z_{v}(z_{v}+2 ) } { \left [ 2p_{v}(p_{v}+2 ) + 2n_{v}(n_{v}+2 ) - z_{v}(z_{v}+2 ) \right]^{\frac{3}{4 } } } { \mid s \rangle}. \end{split}\ ] ] the fundamental mass scale is defined as @xmath221 the simple transverse vertices of type ( i ) yield a positive definite spectrum while the others do not . the results in eq . ( [ qeaspec ] ) are the full spectrum of the quasilocal energy operator with area normalization . the operator @xmath222 does not require any restrictions on the hilbert space . it is well - defined on the hilbert space @xmath207 ; all the needed properties of are naturally induced from the full gauge invariant hilbert space in @xmath5 . there is one example of a spin - net state which is particularly interesting . when the intersections between the graph of the spin network state and the surface are entirely transversal ( type ( i ) ) , the quasilocal operator becomes @xmath223{j_{v}(j_{v}+1 ) } { \mid s_{i } \rangle } \equiv m \sqrt{2 } \sum_{v } n_{v } \sqrt{a_{v } } { \mid s \rangle}\ ] ] in which @xmath224 is the eigenvalue of the area operator for the vertex @xmath68 . the operator is the `` square root of the area '' ! this could have been anticipated with a little dimensional analysis . nevertheless , it is remarkable that energy is so closely related to area . ( the close connection is also present in the volume normalization . ) for large spins the energy scales as @xmath225 . the next and final section begins with a summary of the operator definitions and continues with comparisons to earlier work and a discussion of some wider implications . to be well - defined , the action of a theory must be functionally differentiable . this simple observation provides a key to the form of all surface observables and boundary conditions of a given action @xcite . in the case of the ( @xmath0 ) gravitational theory defined in a bounded region , the variation generates a surface observable associated to the hamiltonian constraint . this term is precisely the negative of the hamiltonian of the system . in this paper , this boundary hamiltonian is quantized within the framework of spin - net gravity . spin - net gravity is a background - metric independent , canonical quantization of gravity using real connection variables and the methods of spin networks . under this rubric is also included the assumption that the kinematic state space is rich enough to describe the full physical state space , including dynamics . there are two different , inequivalent expressions for the boundary hamiltonian corresponding to two separate normalizations . both operators are well - defined quantum operators ( but on different spaces ) . the area normalized operator has the following spectrum : for type ( i ) vertices : @xmath226{j_{v}(j_{v } + 1 ) } { \mid s \rangle}\ ] ] and for type ( ii ) vertices : @xmath227 \\ \times \left [ 2 j_{(p_{v})}(j_{(p_{v } ) } + 1 ) + 2 j_{(n_{v})}(j_{(n_{v } ) } + 1 ) - j_{(z_{v})}(j_{(z_{v } ) } + 1 ) \right]^{-3/4 } { \mid s \rangle}. \end{split}\ ] ] all the dimensionful constants and the immirzi parameter are included ; @xmath228 . this operator is defined on the full hilbert space @xmath207 . this comes about through an interplay between the operator and the gauge invariant hilbert space which allows the operator to be defined on the induced hilbert space . as can be easily seen from the spectrum , only transversal edges contribute energy . tangential edges remove energy . the alternate form of the operator , with volume normalization and defined only on states which are of type ( i ) , is @xmath229 where @xmath230 and @xmath231 is the eigenvalue of the volume . on account of the state restriction , the numerator always has the simple form of the area operator ; the operator is proportional to @xmath232 . the volume operator does not have as simple a form as the area operator , so the full spectrum can not be presented , as in eq . ( [ qeaspec ] ) . the eigenvalues @xmath231 can be computed using recoupling theory as in refs . this form of the quasilocal energy has the correct naive classical limit in that it is a direct quantization of the classical surface hamiltonian . this operator requires tight restrictions on the hilbert space : there can be no edges tangent to the surface @xmath29 . as mentioned in section [ qevol ] , since every edge is tangent to some surface , this restriction introduces a contradiction in the construction of this quasilocal energy operator . the operators share some general features . they are defined on bounding 2-surfaces . although they have a strange inclination to be positive , both operators share the property that the spectrum is fully discrete and bounded ( for arbitrarily large but finite graphs and spins ) . the area normalized operator does not have a positive spectrum . the reason may be traced to the geometric factor in the new operator @xmath181 . ( this in turn is a result of the hamiltonian s two factors , the edge - projected connection and the @xmath233 piece . ) on account of this factor , the quasilocal energy depends on the orientation of the surface . it does not , however , depend on the orientation of the edges . the volume normalized operator has a similar form to the @xmath176 energy of ref . @xcite , in which a different , but weakly equivalent , classical expression was quantized . the present operator generalizes the adm operator in one important way . it is a quasilocal operator defined on all bounding surfaces in the spatial manifold . it also shares a key property : to be defined at all , the tangle property of ref . @xcite must be imposed . in the language of section [ spectra ] , all vertices are of type ( i ) . in fact , when the tangle property is satisfied and when the lapse is fixed at unity , the expression for @xmath234 is simply identical to the @xmath176 of eq . ( 3.18 ) in ref . @xcite ( up to an overall numerical factor ) . the volume normalized quasilocal energy reduces to the adm energy . it is remarkable that such different classical expressions yield nearly the same form of the operator . clearly , only one quasilocal energy operator gives physically correct values . while @xmath235 has the correct naive limit , this criterion is not the only physical condition which must be met . more importantly , the expectation value of the quantum energy operator in an appropriate semiclassical state must approximate the classical energy , up to small quantum corrections . in addition , the algebra of boundary observables ought to be anomaly free . until such states and such operators are investigated in full detail , it is hard to definitively select a quasilocal energy operator . to complete this investigation would require quantization of the boundary rotation and boosts a project which will be left to further work ( there is preliminary work already in ref . other approaches may also help fix the correct quantization , perhaps through the matching of quantum to semiclassical results . in advance of a more complete investigation , there is one feature which does differentiate the two operators . the volume normalized operator requires a restriction on the induced hilbert space . while this may be acceptable for the asymptotically flat setting where the property was introduced , this is a too severe _ a priori _ restriction on the whole hilbert space . therefore , for the remainder of this section i restrict the majority of my comments to the area normalized @xmath236 . the previous quantization of the boundary hamiltonian has a qualitatively different action of `` shifting '' edges incident to the boundary @xcite . the quasilocal energy operator does not share this qualitative behavior . from the perspective of the spin - net framework , it is clear that the shift is an artifact of the partial quantization . in fact , there are two terms which shift `` up and back '' as may be seen in eq . ( [ step4 ] ) or in the gauge invariant form of the operator . the `` shift '' is incorporated into the gauge invariant operator , which leads to the `` self - interaction '' measurement of edge addition as shown in section [ recouplingnote ] . there are a few immediate results which follow from the definition of the quasilocal energy operator . gravitational energy takes on quantized values with the smallest gap in energy of @xmath237{3 } m\ ] ] due the the addition or removal of a spin-@xmath238 edge . suppose that a system s bounding surface only intersected spin-@xmath238 edges of the underlying spin network . then , transitions due to adding or removing a spin-@xmath238 edge would result in radiation with the rather energetic fundamental frequency of @xmath239{3 } \ , m / \hbar$ ] . this mechanism is similar to the bekenstein - mukhanov quantization @xcite , in which the area is given by integer multiples of a fundamental area . in this case , however , the quasilocal energy is quantized in integer units , @xmath240 , so area scales as @xmath241 . of course , before we tune our radio receivers to listen to black holes , there is much more to understand . in particular , we need a characterization of semiclassical states and the dynamics . to see that this strongly affects the radiation , we need only study the high spin limit @xcite . the energy gap narrows , as would be expected from the naive semiclassical limit . there are also some wider implications of the quasilocal energy . the number of directions is enormous and i confine my comments to four brief remarks : on a first glance at the quasilocal energy operator of eq . ( [ qeadef0 ] ) it appears that the quasilocal energy operator depends on the orientation of the edges . since the kinematic state space of @xmath18 spin networks is independent of orientation this would be odd . in fact , the quasilocal energy operator does not depend on orientation of the edges . there are two aspects of this property . this is perhaps easiest to see by noting that when the orientation of an edge is changed , both the sign factor @xmath242 and the angular momentum operator @xmath140 change sign @xcite . in an abuse of notation the same @xmath140 are used in the gauss operator @xmath243 , although the gauss operator is independent of edge orientation . on the other hand , the operator definitely does depend on the orientation of the surface . when the orientation of the surface changes , only the geometric factor changes sign . orienting the edges to be outgoing from the vertices simply places the edges in three , simple categories : inside the region , outside the region , and tangent to the surface . one tantalizing aspect to this third category is the interpretation of edges which leave the region @xmath27 . they take energy with them . it remains to be seen whether these could provide structure for the 4-dimensional space `` pillars of time '' in a 4-dimensional spin network model or more exotic manifolds . given the definition of the quasilocal energy operator and the induced hilbert space @xmath207 , the question arises whether the operator is consistent . if the spatial manifold @xmath5 is compact , is it true that @xmath244 vanishes ? the natural language in which to consider such questions is topological quantum field theory ( tqft ) @xcite , in which one studies diffeomorphism invariant theories on manifolds with boundary . while the manifolds are cut and sewn together , the theories associate maps to interiors and vector spaces to boundaries . a more careful study will be left to further work . but the question may addressed directly . on a particular spin network state , say @xmath245 , the two operators act on the two `` halves '' of the state . one is based on @xmath246 ; the other is based on @xmath247 . as @xmath245 is a general , gauge invariant state , there are two types of vertices to consider , types ( i ) and ( ii ) . a short calculation shows that the contributions to the energies are numerically identical . to perform the sum . ] as this quasilocal energy is the boundary hamiltonian , one may describe the time development of the lorentzian theory on @xmath248 @xmath249 ( recall that the boundary conditions fix the lapse and metric on the boundary . a more complete treatment would also include terms with non - vanishing shift @xmath250 , giving rotations and boosts . ) the energy operator is diagonal in the spin network basis , so the unitary evolution operator has a simple , well - defined form ( at least for type ( i ) vertices ) . it describes the system in terms of an observer on the surface @xmath29 . likewise , the quasilocal energy operator gives the partition function @xmath251 . as the quasilocal energy is the true hamiltonian , this function is the partition function for the statistical mechanics of spin - net gravity . this observation is the starting point for a vast range of physical questions . for instance , is the energy quanta @xmath252 statistically favored ? what is the selection criteria for the ground state(s ) ? what is the entropy of a bounded system in spin - net gravity ? what is particularly striking is the delightful number of physical questions that can be addressed with this operator in the current framework on spin - net gravity . this is but one result of the techniques which have been developed since 1995 and which offer methods for detailed study of these questions . in fact , it appears that these techniques are powerful and rich enough to bring us from theoretical modeling to physical predictions . i thank herbert balasin and robert depietri for discussions . thanks also to an anonymous reviewer for helpful , critical remarks . i gratefully acknowledge the support of the austrian science foundation ( fwf ) through a lise meitner fellowship . this appendix contains recoupling formulae needed in the calculations of the spectrum . the conventions are those of kauffman and lins @xcite ( for @xmath253 ) . short introductions and definitions of the basic recoupling quantities can be found in @xcite . the quantities needed in the calculation for the simplification of the quasilocal energy operator are : @xmath264 for instance , @xmath265 this is used in the recoupling calculation of eq . ( [ recoupleident ] ) . carlo rovelli , `` loop quantum gravity , '' _ living reviews in relativity _ at _ http://www.livingreviews.org/articles/volume1/1998-1rovelli_ ; `` strings , loops , and others : a critical survey of the present approaches to quantum gravity , '' in _ gravitation and relativity : at the turn of the millennium _ , proceedings of the gr-15 conference , naresh dadhich and jayant narlikar , ed . ( inter - university center for astronomy and astrophysics , pune , india , 1998 ) , pp . 281 - 331 , online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9803024_. c. rovelli and l. smolin , `` discreteness of area and volume in quantum gravity '' _ nuc . * b 422 * ( 1995 ) 593 ; erratum _ nuc . * b 456 * ( 1995 ) 753 ; online preprint archive : _ http://xxx.lanl.gov / abs / qr - qc/9411005_. s. frittelli , l. lehner , c. rovelli , `` the complete spectrum of the area from recoupling theory in loop quantum gravity '' _ class . quant . grav_. * 13 * ( 1996 ) 2921 - 2932 ; online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9608043_. r. loll , `` the volume operator in discretized quantum gravity '' _ phys . lett . _ * 75 * ( 1995 ) 3048 - 3051 ; online preprint archive : _ http://xxx.lanl.gov/abs/gr-qc/9506014 _ and `` spectrum of the volume operator in quantum gravity '' _ nuc . * b 460 * ( 1996 ) 143 ; online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9511030_. t. thiemann , `` closed formula for the matrix elements of the volume operator in canonical quantum gravity '' _ j. math . * 39 * ( 1998 ) 3347 - 3371 ; online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9606091_. r. borissov , r. de pietri , c , rovelli , `` matrix elements of thiemann s hamiltonian constraint in loop quantum gravity '' _ class . * 14 * ( 1997 ) 2793 - 2823 ; online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9703090_. john baez , javier p. muniain , and dardo piriz , `` quantum gravity hamiltonian for manifolds with boundary '' _ phys . rev _ * d 52 * ( 1995 ) 6840 - 6845 ; online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9501016_. t. thiemann , `` qsd iii : quantum constraint algebra and physical scalar product in quantum general relativity '' _ class . grav . _ * 15 * ( 1998 ) 1207 - 1247 ; online preprint archive : _ http://xxx.lanl.gov/abs/gr-qc/9705017 _ t. thiemann , `` qsd iv : 2 + 1 euclidean quantum gravity as a model to test 3 + 1 lorentzian quantum gravity '' _ class . quant . grav . _ * 15 * ( 1998 ) 1249 - 1280 ; online preprint archive : _ http://xxx.lanl.gov/abs/gr-qc/9705018 _ t. thiemann , `` qsd v : quantum gravity as the natural regulator of matter quantum field theories '' _ class . grav . _ * 15 * ( 1998 ) 1281 - 1314 ; online preprint archive : _ http://xxx.lanl.gov/abs/gr-qc/9705019 _ t. thiemann , `` quantum spin dynamics ( qsd ) : vi . quantum poicar algebra and a quantum positivity of energy theorem for canonical quantum gravity '' _ class . grav . _ * 15 * ( 1998 ) 1463 - 1485 ; online preprint archive : _ http://xxx.lanl.gov/abs/gr-qc/9705020 _ a. ashtekar , a. corichi , and j. a. zapata , `` quantum geometry iii : non - commutativity of riemannian structures '' _ class . grav . _ * 15 * ( 1998 ) 2955 - 2972 ; online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9806041_. roberto depietri and carlo rovelli , `` geometry eigenvalues and the scalar product from recoupling theory in loop quantum gravity '' _ phys . rev . _ * d 54 * ( 1996 ) 2664 - 2690 . online preprint archive ] : _ http://xxx.lanl.gov / abs / gr - qc/9602023_. r. depietri , `` on the relation between the connection and the loop representation of quantum gravity , '' _ class . _ * 14 * ( 1997 ) 53 - 70 ; online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9605064_. a. ashtekar and j. lewandowski , `` representation theory of analytic holonomy @xmath268 algebras '' in _ knots and quantum gravity _ j. baez , ed . ( oxford university press , 1994 ) ; online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9311010_. a. ashtekar and j. lewandowski , `` projective techniques and functional integration for gauge theories '' _ j. math * 36 * ( 1995 ) 2170 - 2191 . online preprint archive : _ http://xxx.lanl.gov / abs / hep - th/9411046_. a. ashtekar and j. lewandowski , `` differential geometry on the space of connections via graphs and projective limits '' _ j. geom . * 17 * ( 1995 ) 191 - 230 ; online preprint archive : _ / abs / hep - th/9412073_. a. ashtekar , j. lewandowski , d. marolf , j. mouro , t. thiemann , `` quantization of diffeomorphism invariant theories of connections with local degrees of freedom '' _ j. math . * 36 * ( 1995 ) 6456 ; online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9504018_. m. barreira , m. carfora , and c. rovelli , `` physics with nonperturbative quantum gravity : radiation from a quantum black hole '' _ gen . _ * 28 * ( 1996 ) 1293 - 1299 ; online preprint archive : _ http://xxx.lanl.gov / abs / gr - qc/9603064_.
iin this paper , we consider a uniform , infinite universe with a spherical hole . ultimately , we intend to develop a new cosmological model for an expanding universe in which holes are the dominant feature . geller and huchra @xcite also indicate the importance of holes in the universe by writing ... it probably makes more physical sense to regard the individual voids as the fundamental structures . " thus motivated , we consider the behavior of a hole in an otherwise infinite and uniform universe . we show first in section 2 , using newtonian theory , that there exists an outward - directed gravitational force acting away from the center of the spherical hole , and we develop a formula for this force . this force is referred to as negative gravity . then , in section 3 , we solve the newtonian differential equations to discover the expansionary behavior of the hole and develop certain formul to be used later . we show the same conclusion using the methods of einstein s general relativity in section 4 . the derivation of the standard schwarzschild solution for an attracting point particle of mass @xmath0 is well known . the results are summarized in section 4.1 , and several consequences are mentioned which will be important later when the techniques used in the solution of the schwarzschild problem are applied to a nonstatic distribution of mass . it is well known that particles and photons are attracted toward the mass . in the geodesic equations describing the motion of particles and photons , if the mass were negative , we would conclude that particles and photons would deflect away from the mass . although negative mass may not have any physical significance , it serves as an introduction to the main results of later sections . the concept of negative mass will be used in terms of superimposing on a uniform universe this schwarzschild solution for negative mass in order to simulate a mass deficiency ( hole ) in an otherwise uniform universe . section 4.2 gives a new interpretation of the solution of einstein s field equations for a uniform distribution of mass . this solution represents a flat geometry and reduces to the flat geometry of special relativity for mass density decreasing to zero . section 4.3 describes the effect of a spherical hole in a uniform universe . in these sections also , the physical significance of the cosmological constant is determined . if this were a linear universe , the two solutions from sections 4.2 and 4.3 could be superimposed . in a physically imaginable universe which is initially uniformly filled with mass except for a spherical hole , particles and photons are deflected away from the hole . this follows since the geometry of the uniform distribution is flat and does not affect the uniform motion of particles and photons , while the negative schwarzschild mass of section 4.2 would repel matter and photons . loosely speaking , the hole repels matter and energy . also in section 4.3 , a remarkably close connection is made between the schwarzschild solution and the newly discovered newtonian potential for holes , which includes the contribution of negative gravity . the purpose of section 4.4 is to show this using the fully nonlinear einstein field equations . finally , the geodesic equations are considered , and it is shown that particles and photons are deflected away from the hole . section 4.5 summarizes the results . section 5 gives a development of a new cosmology , based upon expanding holes in an infinite , uniform universe . this is based upon a paper by r. n. lewis @xcite , in which it is shown how structure could form from the interactions of nearby holes , resulting in the universe we observe today . the microwave background radiation , caused by gravitational redshift , and hubble s law are shown to be natural consequences of the model . we also discuss the relative abundances of elements throughout the universe , olbers paradox , and other observed phenomena . finally , in section 6 , we use some of the ideas and equations previously derived in order to calculate hubble s constant , the average mass density in the universe , and the cosmological constant . certain accurate astronomical data are used in the calculations . the value derived for hubble s constant is shown to lie within the experimentally - determined range for that constant , providing confidence that the underlying theory is correct . iin this section , we consider a newtonian approach to the gravitational force field in an infinite universe filled with a uniform distribution of mass with the exception of a spherical hole . physicists usually consider the potential and the force acting on a particle anywhere in such an infinite universe to be infinite , since a formal integration over space produces this result . however , we apply symmetry in our calculations to produce cancellations , which leads to a finite force at any point . this application of symmetry is mathematically equivalent to a cauchy - type integration , without which the integral would be infinite . whether this approach is valid depends on its agreement with the results of general relativity . using results in later sections of this paper , we show this to be the case . iwe consider an infinite , uniform distribution of matter with mass from a spherical region removed and evaluate the force field at any point . the three main hypotheses we use are : 1 . gravity obeys the ( newtonian ) inverse - square force law , and it acts over arbitrarily large distances and time intervals . the general relativistic aspects of this problem are taken up later in this paper . when necessary , matter will be considered to be distributed continuously rather than discretely throughout space . with this assumption , the calculations become tractable . on the other hand , the difference in effect on a point mass of a continuous distribution of matter versus a discretely distributed ( atomic ) set of matter is negligible . indeed , the only essential aspect of the distribution of matter which is used in the computations leading to the results in section 2.3 is spherical symmetry . also , when convenient , we will assume that the mass density is constant . the universe is spatially infinite . this , along with the constant mass density of hypothesis ( 2 ) , implies that the universe contains an infinite amount of mass . ihere , we present some results which depend only upon the inverse - square law of newtonian theory . see kittel , knight , and rudderman @xcite for appropriate computations involving gravitational forces . the following is a list of the necessary facts and formul : 1 . the gravitational force due to a spherical shell of mass of radius @xmath1 , ( infinitesimal ) thickness @xmath2 , and mass density @xmath3 ( mass per unit volume ) acting on a mass @xmath0 outside the shell ( @xmath4 ) is the same as if all of the mass of the shell were concentrated at the center of the sphere . mathematically , this is expressed as + @xmath5 + where @xmath6 is newton s gravitational constant ( @xmath7 x @xmath8 m@xmath9/kg s@xmath10 ) , @xmath11 is the distance from the center of the sphere to the mass , and @xmath12 is a unit vector from the center of the sphere toward the mass . in the case @xmath13 , a slight modification of the calculations leading to the above result yields @xmath14 2 . if we have the same situation as in ( 1 ) above , but the mass @xmath0 is inside the shell ( @xmath15 ) , then there is no gravitational force acting on the mass . therefore + @xmath16 + this result generalizes to the case of a spherical shell ( not infinitesimally thin ) of inner radius @xmath17 and outer radius @xmath18 . the gravitational force acting on @xmath0 inside @xmath17 is still @xmath19 . furthermore , the above results generalize by integration to a sphere of mass acting on a mass @xmath0 . the force due to a sphere of mass of radius @xmath1 and mass density @xmath3 acting on a mass @xmath0 outside the sphere ( @xmath20 ) is the same as if all the mass were concentrated at the center . thus + @xmath21{{gm}\o{r^2}}\v{e}.\ ] ] 4 . if we have the same situation as in ( 3 ) but the mass @xmath0 is inside the sphere ( @xmath15 ) , then all of the mass outside of @xmath11 contributes nothing to the force , and the mass inside of @xmath11 acts on @xmath0 as if it were concentrated at the center . therefore + @xmath22{{gm}\o{r^2}}\v{e } = -\left({4\o 3}\right)\pi r\rho gm\v{e}.\ ] ] finally , we mention that , for a finite , spherically symmetric ( not necessarily uniform ) mass distribution , the gravitational force of it on a point mass does not depend on the mass farther away from the center than the point mass , and is the same as if all of the mass closer to the center than the point mass were concentrated at the center . we will see in the next section how this result generalizes to infinite mass distributions , with unexpected results . ithe next question we ask is : what is the gravitational force on a point mass @xmath0 situated at a positive distance @xmath11 from the origin due to a uniformly distributed ( infinite ) mass external to a spherical hole with center at the origin and radius @xmath1 ( @xmath15 ) ? that is , what is the gravitational effect on a test particle in a spherical hole in an otherwise uniform universe ? if we consider the situation specified in the generalization of formula ( 2 ) in section 2.2 , and navely let @xmath23 , then we would conclude that the total force acting on @xmath0 is @xmath19 . however , this conclusion is not accurate , since the total mass which lies outside the sphere of radius @xmath18 is not spherically symmetric about @xmath0 ( although it is spherically symmetric about the center of the sphere ) , and does have an effect on @xmath0 . in order to analyze this case , consider the situation depicted in figure 1 . we take the origin of our coordinate system at @xmath0 , with the center of the sphere @xmath24 , of radius @xmath1 , at a point @xmath11 units to the right of @xmath0 . then we draw a sphere , @xmath25 , with center at @xmath0 and radius @xmath26 . we also add a sphere , @xmath27 , concentric with @xmath24 , and with radius @xmath11 . since the mass external to @xmath25 is spherically symmetric with respect to @xmath0 , its total gravitational force on @xmath0 is @xmath19 . thus , the only effective force which will act on @xmath0 will be that due to the mass in the three - dimensional crescent - shaped region ( crescentoid ) , @xmath28 , which lies inside @xmath25 and outside @xmath24 ( see fig . 1 ) . in order to handle this situation , consider a totally uniform distribution of matter ( no hole ) with a test mass @xmath0 at the origin . by symmetry , there will be no force acting on @xmath0 . also by symmetry , the net force on @xmath0 due to all mass inside @xmath25 is @xmath19 . the force on @xmath0 due to all the mass in the spherical shell between @xmath27 and @xmath24 is @xmath19 , from the results of formula ( 2 ) and its generalization in section 2.2 . therefore , the mass in the region @xmath28 will have a force on @xmath0 which will exactly counterbalance the force on @xmath0 due to the mass inside @xmath27 . from section 2.2 , formula ( 4 ) , we find the force on @xmath0 due to the mass inside @xmath27 to be @xmath29 , where @xmath12 is a unit vector from the center of @xmath27 toward @xmath0 . thus , the force on @xmath0 due to the mass in the shaded region ( crescentoid ) is @xmath30 this will also be equal to the force acting on @xmath0 in a uniform universe with a hole ( either @xmath27 or @xmath24 ) , since the counterbalancing mass inside @xmath27 is not present . we note the interesting result that this force does not depend on @xmath1 . because of this , in some applications , we can think of @xmath1 as being arbitrarily large , and the crescentoid region in the figure as being arbitrarily far away from @xmath0 . the matter ( or absence of it ) in the spherical shell between @xmath27 and @xmath24 provides no contribution to the force on @xmath0 . we can therefore say that the force @xmath31 on @xmath0 given by the above formula is due to mass arbitrarily far away from @xmath0 . we also note that , as time passes , matter moves outward from the hole , destroying the constancy of matter but preserving spherical symmetry about the center of the hole . the above comments hold in this case as well , since the crescentoid region referred to above can be considered to be arbitrarily far from the motion and unaffected by it . the formula for force given above still applies , with @xmath11 a function of time . at this point , we mention that a formal mathematical integration over all space , after applying symmetry considerations , yields the same result . we note that the force can be written as @xmath32 where @xmath33 is the mass deficiency of the hole ( out to @xmath11 ) . this shows that newton s inverse - square force law holds for negative masses ( holes ) acting on positive masses . this formula gives the force @xmath31 at any point inside the hole ( @xmath15 ) . however , spherical shells of matter of any radius greater than @xmath11 have no gravitational effect on a mass @xmath0 at @xmath11 , so we can think of @xmath1 as being arbitrarily large . thus , the above formula yields the outward force at any point inside r outside the hole . for a point outside ( @xmath20 ) the hole , however , there will be an additional inward - directed force due to all mass closer to the origin than @xmath11 . this inward force is @xmath34gm}\o{r^2}}\v{e}. \eqno ( 2)\ ] ] the sum of the above two forces yields the force due to a hole of radius @xmath1 in an otherwise infinite and uniform distribution of mass on a mass @xmath0 at distance @xmath11 ( @xmath35 ) from the origin ( that is , the force of a hole acting on a mass outside the hole ) . this force is @xmath36 where @xmath37 is the mass deficiency of the hole . this again shows the applicability of newton s inverse - square force law with one negative and one positive mass . it is tempting to say that newton s inverse - square force law holds for two negative masses as well . however , we would need to define what we mean by a force on a hole . we simply mention that two holes would have an effect on each other due to the above - mentioned force acting on the particles surrounding the holes . the net effect would be that the two holes would move toward each other and become slightly distorted . we conclude that positive masses attract each other , as do negative masses ( holes ) , but positive and negative masses repel each other . the implication of the above result is the following . the infinite amount of matter outside of a hole in an otherwise infinite and uniform distribution of matter will gravitationally draw matter out of the hole . we can loosely say that the hole repels matter , just as a concentration of matter would attract matter ( section 2.2 ) . we refer to this phenomenon as _ negative gravity_. it must be remembered that this term refers to the apparent repulsion of matter by holes . in fact , the actual mechanism causing negative gravity is the positive gravitational attraction of mass at infinity " , that is , the mass in a crescentoid region arbitrarily far from the hole . the term negative gravity should not be taken to imply that the gravitational force between two particles is repulsive ; gravity is always attractive . as we shall see in section 4 , however , the energy - momentum tensor in the neighborhood of a hole can be negative because of the above - described effect . ian obvious generalization to make to the above calculations is the following . in a spherically symmetric ( about the origin ) universe , the net force acting on a particle of mass @xmath0 at a distance @xmath11 from the origin is @xmath38 where @xmath39 , is the difference in the average densities of matter outside and inside @xmath11 , respectively . the density @xmath40 is the mass divided by the volume of the region inside @xmath11 . the density @xmath41 is the average mass density over the region external to the sphere of radius @xmath11 . we note that if @xmath42 , then this formula reduces to the formula in section 2.3 . if @xmath43 , ( i.e. if there is negligible mass outside a sphere of large radius , or if the mass density decreases to 0 as @xmath11 increases ) , then the net force is inward , the force being given by formula ( 3 ) in section 2.2 . thus , the basic underlying idea for negative gravity to exist is that @xmath44 . furthermore , if @xmath44 , we speculate that , even in the absence of total spherical symmetry , there will be a net outward force ( from centers of holes ) tending to pull things apart . inow let us investigate the behavior of a spherical hole in an otherwise infinite and uniform universe and form some conclusions about the rate of expansion . as we point out later , a slow expansion caused by negative gravity does occur . we model the problem mathematically and solve two specific problems ( sections 3.1 and 3.2 ) which assume that the age of the universe under consideration is infinite . the conclusions show the relatively large time intervals that are required for particles to move an appreciable distance . under the assumptions given above , the outwardly - directed negative gravitational force acts at each point of space . for a particle of mass @xmath0 , the force was given above by equation ( 1 ) . this formula holds no matter what @xmath11 is . in other words , it holds both inside ( @xmath15 ) and outside ( @xmath20 ) the hole , although when the point is outside the hole , there is also an inward - directed force @xmath45 [ equ . ( 2 ) ] due to all mass inside @xmath11 . in fact , this formula holds even in a nonuniform universe as long as the universe is spherically symmetric about the origin and @xmath3 represents the average mass density . ilet us suppose that we have a universe of discrete hydrogen atoms . such a universe can not be uniform ( continuous ) except in a statistical sense , but it can be spherically symmetric about the origin . let us suppose that this is the case and that an atom , @xmath46 , lies at the origin . we also assume that the average mass density , @xmath3 , is one atom per cubic meter ( @xmath47 x @xmath48 kg / m@xmath9 ) . this is roughly on the same order of the value that current observations of luminous matter lead us to believe . the presence of unobserved matter would increase this value . the actual value of @xmath3 is immaterial to the conclusion we wish to draw . because of the density assumption , we assume that the nearest other atoms are one meter ( 1 m ) from the origin . to @xmath46 , the other atoms appear to be distributed spherically - symmetrically over all space outside of a hole of radius one meter . now , let @xmath46 be displaced a distance @xmath49 m in any direction . let @xmath50 represent this configuration , with @xmath51 representing the original spherically - symmetric universe . the hypotheses ( a test particle in a hole in a spherically - symmetric universe ) of section 1 are satisfied . the outward force given in that section acts on the atom @xmath46 . newton s second law of motion implies that @xmath52 where @xmath53 is a position vector of the particle . the initial value problem for @xmath11 , therefore , is @xmath54 @xmath55 @xmath56 let @xmath57^{1/2 } = 6.84 $ ] x @xmath58 s@xmath59 . then the solution of the initial value problem is @xmath60 according to this differential equation model , the time , @xmath61 , it takes for the mass @xmath0 ( hydrogen atom ) to reach the edge of the hole ( @xmath62 m ) and its corresponding velocity depend on @xmath63 . for example , if @xmath64 m , @xmath65 x @xmath66 years . if @xmath67 m , @xmath68 x @xmath66 years . in both cases , since @xmath69 , and since @xmath62 m , then @xmath70 x @xmath58 m / s . note that @xmath71/\a \}\bigl)/\beta \approx [ ln(2/\a ) ] /\beta \r \infty$ ] as @xmath72 . the conclusion we arrive at is that it takes infinitely long to transform the universe from configuration @xmath51 to configuration @xmath73 for @xmath74 m. there are two aspects of this model that need clarification . the first is that , as the hydrogen atom @xmath46 moves , it also gravitationally affects all the other atoms in the universe , displacing them slightly and destroying the spherical symmetry of the universe outside the hole . to counter this problem , we argue that the conclusion we reached at the end of the previous paragraph is still valid . the reason to believe this is that the smaller the value of @xmath63 is , the better the initial value problem @xmath75 models the physical situation . in the next section , we present a model in which spherical symmetry is preserved , so that this problem does not arise . the other remark to make is a positive one . this model allows us to start with a uniform universe , @xmath51 , at time @xmath76 ( let @xmath72 ) instead of starting at a particular time , @xmath77 ( big bang " ) , with a particular mass density and having to worry about what happened before that time or about what caused things to happen . inow let us consider a uniform ( rather than a discrete ) distribution of mass [ @xmath78 x @xmath48 kg / m@xmath9 ] and suppose that , at time @xmath80 , a spherical hole of radius @xmath1 is introduced , centered at the origin @xmath81 . @xmath82note : @xmath83^{1/3}$ ] m if we consider the problem of taking away the equivalent mass of one atom at the origin.@xmath84 consider the force acting at point @xmath85 where @xmath86 . at time @xmath87 , we have the situation depicted in figure 2 . there is an outward force ( relative to @xmath81 ) acting on each particle ( see below ) , so at a later time we will have the situation depicted in figure 3 . at time @xmath87 , the force acting on a particle at a distance @xmath88 from @xmath81 is @xmath89 where [ equ . ( 1 ) ] @xmath90 is the outward force due to negative gravity of all mass outside @xmath88 , and [ equ . ( 2 ) ] @xmath91\rho gm}\o{r_1 ^ 2}}\v{e}\ ] ] is the inward force due to mass inside @xmath88 , where @xmath92\rho$ ] is the mass inside the sphere of radius @xmath93 centered at @xmath81 . at time @xmath87 , @xmath94 which is outward - directed . the magnitude of @xmath95 decreases with increasing distance ( @xmath93 ) from @xmath81 . at time @xmath96 , the force acting on a particle at distance @xmath11 from @xmath81 is @xmath89 where @xmath97 is the force due to negative gravity ( it changes with @xmath11 ) , and @xmath45 is the inward force due to positive gravity of all matter closer to the origin than @xmath11 . because of the initial outward - directed force , all particles will be accelerated outward . we will see later that this outward motion will continue forever . at this point , let us make a further assumption about the motion . we will assume that atoms do not pass each other on their outward journey as time progresses . to justify this assumption , consider what would happen if one particle were to pass another . the force @xmath98 on each would not change , but the inward - directed force @xmath45 would . the magnitude of @xmath45 on the overtakng particle would increase , due to the increased mass inside the sphere of radius @xmath11 , thus decreasing the magnitude of the outward force @xmath31 . the magnitude of the force @xmath45 on the overtakn particle would similarly decrease . the net effect of one particle passing another would be a decrease in acceleration on the passing particle and an increase in acceleration on the passed particle . other particles farther away would see no effect of this passing , so we just assume it does not occur . with this assumption , we can write @xmath99\rho gm}\o{r^2}}\v{e}.\ ] ] because of the above assumption , the total amount of matter closer to @xmath81 than @xmath11 remains unchanged ( the term in brackets in the numerator ) , but @xmath45 has changed since @xmath11 in the denominator has increased from its initial value @xmath93 . therefore @xmath100 this is always an outward - directed force since the last parenthesized term is always positive . at time @xmath87 , it equals @xmath101 since @xmath88 . when @xmath102 , then @xmath103 , so the term in parentheses is greater than @xmath104 . since position is given by @xmath53 , then newton s second law gives @xmath105 or , with @xmath106 , @xmath107 but @xmath108 , so @xmath109 therefore @xmath110 the solution satisfying @xmath111 and @xmath88 at @xmath87 is @xmath112\right\ } ^{1/2},\ ] ] or @xmath113 this is difficult to integrate except when @xmath114 ( a point at the edge of the hole ) . in that case , @xmath115 the solution satisfying @xmath116 , found by trigonometric substitution in the integral of the above expression or by direct solution of the second order differential equation [ with @xmath117 , is @xmath118 @xmath119 note that , for @xmath120 ( the same formula was derived in section 3.1 for a discrete distribution of matter ) , which shows that hubble s law holds in this ( accelerating ) expanding universe , with no need for introducing a singularity ( big bang " ) at some finite time in the past . in addition , the value of @xmath121 given here as the proportionality constant is surprisingly close to the estimated value of the hubble constant @xmath122 . i propose that this value for @xmath121 is the one that should be used for the hubble constant , not necessarily the numerical value which is dependent upon a more accurate estimate of the average density in the universe , but the formula given before equation ( 4 ) . see section 6 for a more accurate calculation of @xmath3 and hence of @xmath121 . now we consider equation ( 5 ) when @xmath123 . in equation ( 5 ) , we let @xmath124 and @xmath125 and integrate , using @xmath126 , to obtain @xmath127^{1/2}}}ds.\ ] ] this can be written in terms of elliptic integrals of the first and third kinds ( see @xcite ) . we will not pursue this integral solution further since we are able to derive the desired results directly from the differential equation . now we consider the velocity of the particles . we have already shown that the magnitude of @xmath95 ( and therefore of acceleration ) is a decreasing function of initial position . thus , for a certain time interval after @xmath87 , the speed of particles will be a decreasing function of position also . we will assume that this situation holds for all @xmath128 . since particles that start farther out always move slower , the particles which began closer to the origin will begin to catch up to , but never overtake , the more distant particles . thus , an outward moving spherical wave of matter starts to build up at the edge of the hole . now let us find out how long the above process takes . using values for the constants given above , the time @xmath61 required for the spherical wave of matter to reach one light year in radius is given by @xmath129 if @xmath93 is the radius of a hole which holds the mass of one atom , then @xmath130 and @xmath131 it will take @xmath132 trillion years to evacuate a hole of radius @xmath133 light year . at that time , the velocity of the outwardly expanding wave would be @xmath134 x @xmath135 m / s . further calculations show that the mass of a typical galaxy ( our own , for instance ) is about @xmath136 x @xmath137 kg / sun x @xmath66 suns / galaxy @xmath138 x @xmath139 kg . in the original one - atom - per - cubic - meter universe , a sphere would need to have a radius of @xmath140 x @xmath141 m @xmath142 x @xmath143 light years to contain that much mass . solving @xmath144 for @xmath61 shows that it would take @xmath145 in order to evacuate a hole of size large enough to have initially contained the mass of a typical galaxy . the velocity of the outward - expanding wave at that time would be @xmath146 x @xmath147 m / s , still less than @xmath148 of the speed of light . in analogy with section 3.1 , we remark that the equation @xmath149 can be solved for @xmath61 as a function of @xmath11 and @xmath93 . for a given value of @xmath11 , we see that @xmath150 as @xmath151 . this reconfirms the conclusion we made in section 3.1 of the infinite age of the universe we are considering here . later , we show how this model can be applied to develop a cosmological model for our universe . inow we turn to an investigation of the problem of the behavior of a spherical hole in an infinite , uniform universe from the point of view of general relativity . we will have occasion to use some of the previously - derived results and will show that the newtonian results are compatible with general relativity . ithe standard line element of riemannian geometry is given as @xmath152 , where the einstein summation convention is used , and @xmath63 and @xmath153 take on values 1 through 4 . the @xmath154 are the elements of the metric tensor . because the problem we are considering exhibits spherical symmetry , the coordinates @xmath155 will be identified with the standard spherical coordinates of special relativity : @xmath156 where @xmath157 is the speed of light in a vacuum . depending on context , the origin will be taken to be either the center of an attracting mass or , when dealing with a hole in a uniform universe , the center of the hole . a standard solution of einstein s equations , which is now well known , was first found by schwarzschild in 1916 . for a particle of mass @xmath0 in an otherwise empty space , the static line element can be reduced to @xcite @xmath158 where @xmath159 and @xmath160 are functions of @xmath11 only . the relativistic field equations are ( * ? ? ? ( 9.79 ) ) @xmath161 @xmath162 is the contracted riemann - christoffel tensor , @xmath28 is a scalar contracted from @xmath162 , @xmath163 is the energy - momentum tensor , @xmath164 is the cosmological constant introduced by einstein , and @xmath165 is a constant which is related to the constant of gravitation in newton s theory . @xmath165 is usually evaluated when poisson s equation for the potential is used as a first approximation . this is not done here for two reasons : ( 1 ) a different potential [ that for holes ; see equ . ( 8b ) ] is used , and ( 2 ) the value of @xmath165 is later derived independently . later , the gravitational force acting on a particle of mass @xmath0 in a uniform universe with a spherical hole whose center is at the origin will be considered . the results of the newtonian theory are presented here for purposes of comparison with the results of general relativity . earlier , we found the outward newtonian force at any point to be [ see equs . ( 1 ) and ( 2 ) ] @xmath166 this represents a conservative force field with corresponding potential @xmath167 while this may not be the standard way of defining @xmath168 , the ultimate use to which this is put is as a guide to solve the field equations . the conclusion of that venture in no way depends upon this newtonian potential . moreover , the similarity below between this newtonian potential for holes and the schwarzschild solution is certainly more than coincidental . from equation ( 8b ) , we finally note that , either inside or outside the hole , direct differentiation yields @xmath169 this is the negative of the value found for the potential of an attracting mass ( * ? ? ? * equ . ( 9.3 ) ) . now consider an isolated mass at the origin . the equations ( 7 ) become @xcite @xmath170 @xmath171 + \l , \eqno ( 9b)\ ] ] @xmath172 where @xmath173 are the only possible nonzero elements of the energy - momentum tensor . only perfect fluids of density @xmath174 and pressure @xmath175 will be considered , so that the energy - momentum tensor reduces to @xmath176 as will be shown later , the cosmological constant , @xmath164 , enters into the equations through the influence of distant matter . in the schwarzschild model , special relativity boundary conditions at infinity are usually applied to obtain the value zero for @xmath164 . a nonzero value for @xmath164 is included in these calculations only for consideration in later models where distant matter can not be ignored . in the empty space surrounding the mass , the pressure and density are zero , and so @xmath177 . at this point , standard calculations on equations ( 9 ) yield the solution for the metric @xmath178 where @xmath0 is the mass of the attracting object . thus , the schwarzschild line element ( 6 ) becomes @xmath179 following tolman ( * ? ? ? * sec 83 ) , we can write the geodesic equations for a moving particle as @xmath180 where the term in braces is the christoffel symbol of the second kind . again , standard calculations yield the geodesics for a particle or for a photon . for a particle , we find that ( @xmath181 is a constant , @xmath182 ) @xmath183 or , for a radially - traveling particle , @xmath184 the acceleration in ( 14b ) is negative for positive @xmath0 and @xmath185 . thus , a particle is accelerated inward for an attracting body . for a photon , we can also calculate the geodesics . standard asymptotic analysis shows that the photon is deflected as it passes the attracting mass . to first order in @xmath186 , where @xmath28 is the distance of closest approach to the origin , the angle @xmath187 satisfies @xmath188 for an incoming photon , and @xmath189 for an outgoing photon . the photon has been deflected from the straight line path a total of @xmath190 toward the source @xmath0 , which agrees well with observations made during solar eclipses . the conclusion to be drawn from equations ( 14b ) and ( 15c ) is that , for negative @xmath0 and @xmath191 , particles are accelerated away from the mass ( @xmath192 , see @xcite ) , and photons are deflected away from the mass ( @xmath193 ) . particles and photons are deflected away from the origin due to an outward directed newtonian force or due to an outward - curving geometry . in other words , mass deficiency repels matter and energy . iconsider a spatially infinite universe with matter uniformly distributed with density @xmath194 . by spherical symmetry and homogeneity , the line element has the form [ equ . ( 6 ) ] @xmath195 where @xmath159 and @xmath160 are constants independent of @xmath11 and @xmath61 . einstein s field equations for a perfect fluid at constant pressure @xmath196 and constant density @xmath194 reduce to [ see equs.(9 ) and ( 10 ) ] @xmath197 @xmath198 @xmath199 so that ( see ( * ? ? ? after ( 4.206 ) ) ) @xmath200 @xmath201 @xmath202 @xmath203 equation ( 18c ) shows that the cosmological constant @xmath164 is nonzero . this shows the importance of the boundary conditions at infinity . a zero value for the cosmological constant is derived using special relativity ( no mass ) boundary conditions at infinity . a nonzero value is found by assuming uniformly distributed mass at infinity . @xmath164 must be positive in order that @xmath194 be positive . the pressure @xmath204 will then be negative . this is extremely counterintuitive . however , as we pointed out earlier , the introduction of a hole into an infinite , homogeneous , newtonian universe causes matter to move away from the hole . this explains the negative value of the pressure , since pressure measures the tendency of matter to fill a hole . this concept represents a significant change in cosmological thought . an infinite , uniform , static universe is now allowed , along with negative pressure , although such a universe is unstable ( see secs . 4.4 and 5 ) . see @xcite , for example , for a contrasting view . hawking @xcite states the contrasting case even more strongly : we now know that the universal attractive nature of gravity is inconsistent with a static infinite universe . " all three of @xmath204 , @xmath194 , and @xmath164 are extremely small in magnitude . later , the value of @xmath205 is found . using the values @xmath206 x @xmath8 m@xmath207 kg s@xmath208 x @xmath48 kg / m@xmath9 , and @xmath209 x @xmath210 m / s , then @xmath211 x @xmath212 m@xmath213 . the value of density is an estimate based upon observations of luminous matter and should provide a lower bound for @xmath164 . the presence of even vast amounts of unobserved matter would increase its value but not change the conclusions drawn . see section 6 for a more accurate determination of @xmath164 . in order to determine @xmath160 from ( 16 ) , consider a radially - traveling photon ( @xmath214 ) such that @xmath215 represents the speed of light in a medium of density @xmath194 , and @xmath216 is the index of refraction . @xmath217 is a reduction factor for the speed of light traveling through the medium , and @xmath218 . for a fixed medium , this will be constant . thus the speed of light in the medium is @xmath219 . as @xmath194 decreases to zero , @xmath217 increases to one . for the low density problem we envision here , @xmath220 . therefore , we will assume @xmath221 and not concern ourselves with the difference . it should also be recognized from the above equations that the cosmological constant @xmath164 is proportional to the density . this interpretation will provide a justification for the appearance of that constant in the field equations for cosmological purposes . it also gives a cosmological meaning to that constant and will show the effect of distant matter on the behavior of particles . finally , in a sparse universe of discrete atoms , there will be no kinetic interactions between atoms , and it is permissible to take @xmath222 , with @xmath223 . inow consider the same homogeneous universe with a spherical hole of radius @xmath1 . take the origin at the center of the hole . the line element will now have the form @xcite @xmath224 where @xmath159 and @xmath160 are functions of @xmath11 and @xmath61 . the universe is no longer static , but it can still be considered to be spherically symmetric . physically , think of arriving at this state by superimposing on a homogeneous universe a negative mass @xmath0 spread out over a sphere of radius @xmath1 , or , equivalently , deleting a sphere of radius @xmath1 and mass @xmath0 from the uniform universe . mathematically , think of linearizing the field equations and superimposing the schwarzschild exterior solution with negative mass , @xmath225 , onto the flat geometry of a uniform universe given in section 4.2 , @xmath226 . the field equations are ( see @xcite ; compare with equ . ( 9 ) ; note that dots and primes represent partial derivatives with respect to @xmath61 and @xmath11 , respectively ) @xmath227 = e^{-\l}\left({{\l ' } \o r } - { 1\o{r^2}}\right ) + { 1\o{r^2 } } - \l , \eqno ( 21a)\ ] ] @xmath228 = -e^{-\l}\left({{\nu ' } \o r } + { 1\o{r^2}}\right ) + { 1\o{r^2 } } - \l , \eqno ( 21b)\ ] ] @xmath229 = c\left[\left(\rho + { p\o{c^2}}\right)r^2sin^2\t \left({{d\phi}\o{ds}}\right)^2 + { p\o{c^2}}\right ] = \ ] ] @xmath230 + { { e^{-\nu}}\o{c^2}}\left({{\ddot{\l}}\o 2 } + { { \dot{\l}^2}\o 4 } - { { \dot{\l } \dot{\nu}}\o 4}\right ) - \l , \eqno ( 21c)\ ] ] @xmath231 = e^{-\l}{{\dot{\l}}\o { rc } } , \eqno ( 21d)\ ] ] where @xmath3 and @xmath175 represent density and pressure . these equations can be solved exactly in the interior of the hole where the energy - momentum tensor is zero . equations ( 21a ) and ( 21b ) imply @xmath232 . choose @xmath233 . also , equation ( 21d ) implies @xmath234 . theorem @xcite then requires @xmath235 since the equations ( 21 ) reduce to the same form as equations ( 9 ) [ see equ . ( 11 ) ] . regularity at @xmath236 requires @xmath237 . therefore , @xmath238 in light of equation ( 15c ) , with @xmath237 , the @xmath164 term has no influence on photons which pass through the hole in a straight euclidean line . however , with @xmath237 and @xmath239 , equation ( 14b ) implies that particles are accelerated outward . by continuity , this outward acceleration must extend at least part way into the matter - filled region outside the hole . we wonder whether the singularity at @xmath240 in equation ( 22b ) provides an upper limit to the expansion of holes in a uniform universe . in the newtonian theory , inside the hole , we have [ equs . ( 8) ] @xmath241 , and @xmath242 , indicating that @xmath243 , and @xmath244 . this is @xmath245 times the value usually used for c , which is derived for the potential of an attracting point mass . the equation @xmath246 is derived in @xcite or @xcite for a time independent metric , which exists in the interior of the hole . see @xcite where it is suggested how to calculate @xmath164 experimentally , thus yielding a reliable value for @xmath194 . before trying to solve the field equations ( 21 ) outside of the hole , where matter is in motion and the metric is not static , let us make a connection between this and the newtonian theory . there is the approximation given above @xmath247 where @xmath187 is the newtonian potential . we assume that this is approximately valid in the region exterior to the hole even though the metric is not static there . the force acting on a particle of mass @xmath0 at any point outside the hole is [ see equ . ( 4a ) ] @xmath248 . \eqno ( 24a)\ ] ] since the bracketed term is positive ( @xmath249 ) , the force is outward - directed . thus , as time increases , the hole expands outward and individual particles move away from the origin . the corresponding potential is @xmath250 . \eqno ( 24b)\ ] ] note that @xmath187 decreases toward @xmath251 as @xmath11 increases ( @xmath252 points in the same direction as @xmath253 ) , so that large @xmath11 values correspond to large negative potentials . using this value for @xmath187 , we see that @xmath254 from equation ( 23 ) should be of the form @xmath255}\o{c^2 } } \eqno ( 25a)\ ] ] @xmath256\left({g\o{c^2r}}\right ) - 2\left[\left({4\o 3}\right)\pi \rho _ 0r_1 ^ 3\right]\left({g\o{c^2r}}\right ) . \eqno ( 25b)\ ] ] earlier , we showed that , for a particle which started at @xmath114 , the position in the newtonian theory at any later time @xmath61 is @xmath257 . in addition , if @xmath123 then @xmath258 , but @xmath259 as @xmath260 . it is reasonable to conclude that @xmath261 where @xmath262 , so that the last term in equation ( 25b ) can be replaced with a general function @xmath263 which would be expected to lie asymptotically between @xmath264 and @xmath265 as @xmath266 . we identify in the first coefficient of @xmath264 in equation ( 25b ) the term @xmath267 which is the mass @xmath268 which was initially removed from an infinite homogeneous universe of constant density @xmath194 to make a hole of radius @xmath1 centered at the origin . letting @xmath269 ( * ? ? ? ( 8.28 ) , ( 8.33 ) ) or ( * ? ? ? ( 5.118 ) ) , this term is exactly @xmath270 which corresponds to the term in the schwarzschild exterior solution with negative mass [ see equ . ( 11 ) ] . the coefficient of @xmath271 is exactly the @xmath164 value calculated above . thus , @xmath272 which differs from the schwarzschild solution only in the @xmath273 term . it should also be noticed that the schwarzschild solution , except for the @xmath273 term , corresponds exactly with the newtonian potential for holes . in particular , the @xmath270 term corresponds to the mass which was initially removed to form the hole of radius @xmath1 ( negative schwarzschild mass ) , and the @xmath274 term corresponds to the effect of negative gravity , that is , to mass at infinity . the geodesic equations ( 13 ) now become more complicated , so the analysis of the motion of photons and particles is made more complex . thus , at this point , only an educated guess as to the motion of photons and particles outside the hole can be made . because of the outward newtonian force , we claim that particles move outward in the relativistic theory as well . this conclusion also follows from equation ( 14b ) if we neglect the @xmath273 term . in fact , as was remarked earlier , this conclusion follows from continuity at least near the edge of the hole . also , in comparison with equation ( 15c ) , we claim that the trajectories of photons are bent outward in the matter - filled region exterior to the hole because of the negative @xmath0 [ negative schwarzschild mass see discussion after equ . ( 15c ) ] referred to after equation ( 25b ) ( again neglecting @xmath273 ) . thus , the linear theory ( superposition of solutions from section 4.2 and from this section , neglecting @xmath273 ) gives the conclusion that a hole repels matter and energy . in the newtonian theory , we showed that the outward acceleration was due to the effect of mass at infinity " , that is , only that mass outside of a sphere of arbitrarily large radius . from equation ( 14b ) with @xmath237 , it can be seen that the outward acceleration acting on particles is due to the @xmath164 term . thus the cosmological constant @xmath164 is associated with the distant mass , which shows that mach s principle @xcite , at least insofar as it deals with radial forces , is apparently incorporated into einstein s theory through the cosmological constant . this conclusion contrasts with conventional thinking @xcite . inow , consider the equations ( 21 ) outside of the hole , where matter is present . first , it seems likely that zero is a lower bound for the sum @xmath275 [ see equ . ( 18d ) ] . as matter moves away from the hole , the hole ( where @xmath276 ) will expand . in addition , as matter builds up at the edge of the hole and beyond , @xmath3 will increase there , but @xmath175 will also increase due to the increased kinetic activity of the matter there . we will take this to be an additional requirement : @xmath277 next , notice that subtraction of equation ( 21b ) from equation ( 21a ) results in @xmath278 = e^{-\l}\le({{\l ' + \nu ' } \o r}\ri ) . \eqno ( 26)\ ] ] since @xmath279 ( sec . 4.3 ) and @xmath280 , then @xmath281 . compare with ( * ? ? ? ( 9.121 ) ) . let us require further that @xmath282 and @xmath283 . also note from equation ( 21d ) that @xmath284 . now the geodesic equations ( 13 ) will be considered . these reduce to ( for @xmath285 ) @xmath286 @xmath287 @xmath288 @xmath289 @xmath290 one solution of equation ( 27b ) is @xmath291 ( that is , motion takes place entirely in that plane ) , and the corresponding solution of equation ( 27c ) is @xmath292 , where @xmath181 is a constant . using these , equations ( 27a ) and ( 27d ) are simplified to @xmath293 = 0 , \eqno ( 28a)\ ] ] @xmath294 = 0 . \eqno ( 28b)\ ] ] one interesting observation about equation ( 28a ) is that it is a completely general description of the motion of particles or photons in a non - static universe . as such , it also describes motion in a different model where we have a concentration of matter ( rather than a hole ) at the origin in an otherwise uniform universe . physical considerations , as well as superposition of the regular schwarzschild solution for an attracting mass onto the flat geometry of section 4.2 , indicate that matter is accelerated inward and the trajectories of photons are curved inward toward the excess concentration of matter . in that model , @xmath295 and @xmath296 . the conclusion for that model is that @xmath297 for particles and photons . in our model , @xmath298 and @xmath299 , and we expect on exactly the same physical grounds that @xmath300 that is , particles and photons are repelled away from the hole , or are attracted toward the mass in the spherical region s ( see fig . 4 ; note that coordinates in the figure are spherical ) which is symmetrically - placed on the opposite side of the particle or photon from the hole . first , consider motion of a particle . assume that it starts from rest ( @xmath301 ) and travels radially ( @xmath302 ) . equation ( 28a ) can then be written as @xmath303 . \eqno ( 29)\ ] ] by previous assumptions that @xmath304 and @xmath305 , this represents a positive initial acceleration . therefore , @xmath306 initially increases from zero . we conclude that @xmath306 and @xmath307 are both positive on some initial interval @xmath308 . that is , particles are initially accelerated outward away from the hole . physical considerations indicate that this situation will continue forever . on the other hand , if @xmath307 ever turned negative , then @xmath306 would start to decrease , and the troublesome term ( the first one on the right ) in equation ( 29 ) would decrease in significance . in fact , @xmath306 could never become zero with @xmath297 , since that would contradict equation ( 29 ) , so we conclude that @xmath306 is always positive . we infer that @xmath307 is also always positive . now consider the trajectory of a photon , with @xmath309 but @xmath310 . let @xmath311 be a parameter to replace @xmath312 , increasing along the geodesic in the direction of motion . the geodesic equation ( 28a ) has the same form as that for the motion of a particle , with the presence of the @xmath313 term . at the point where @xmath314 ( the point of closest approach to the origin ) , @xmath315 by equation ( 28a ) . by the same reasoning as before , we deduce that @xmath316 from that time forward and infer that @xmath317 for all time . perhaps it is not possible to actually prove mathematically that @xmath307 is positive for particles and photons without a more detailed description of the metric functions @xmath159 and @xmath160 . the main conclusion we wish to draw is that particles and photons are accelerated outward , away from the hole . this result is independent of the actual structure of the metric ( within the restrictions placed on it above ) . it only depends on @xmath281 which , in turn , depends on @xmath279 , a result derived in section 4.3 when dealing with the potential of a hole . we thus see that holes and gravitating bodies have exactly the opposite effect on matter and energy . iwe have shown , on the basis of the newtonian theory , that a spherical hole in an otherwise infinite and uniform universe tends to repel matter . using einstein s theory of general relativity above , we have come to the same conclusion for both matter and photons , the quantum packets of radiation . along the way , we have developed some interesting formul and concepts and have considered einstein s equations for a nonstatic line element and the corresponding geodesics . in addition , we have given a physical explanation of the cosmological constant and have shown the possibility of an infinite , but unstable , universe in the general theory of relativity . the consequences of these new discoveries in the field of cosmology are staggering . below , we develop an entirely new cosmology based upon holes and their repulsive characteristics in an infinite universe . the mathematical results given here form the basis for that development . inow , we consider the implications of this theory to cosmology . we assume that our universe started out in an infinite , uniform configuration infinitely long ago . then we develop some of the logical consequences of such an assumption and show that the resulting universe is surprisingly like our own universe today . in particular , the expansion of the universe can be explained in terms of an accelerating expansion due to negative gravity rather that an explosive but decelerating expansion due to the big bang . of course , any cosmology , especially one which reportedly refutes the big bang theory , must be able to explain several astronomical observations such as the microwave background radiation , the hubble expansion , and the relative abundance of hydrogen , helium , and other elements in the universe , among others . this is done also . in section 5.1 , the history of the universe is developed , starting with an infinite , uniform state , and it is shown how structure could evolve in the universe to its present state . in section 5.2 , the gravitational redshift and some of its consequences for the evolution of the universe are discussed . section 5.3 gives a comparison between this new cosmology and the big bang . iwe showed mathematically , using newtonian mechanics , that there exists an outward gravitational force acting at each point in an infinite universe with a spherical hole whose center is at the origin . at any point at a distance @xmath11 from the center of the hole , the outward force on a mass @xmath0 is given by equation ( 1 ) . then , we came to a similar conclusion using einstein s theory of general relativity . using this idea that an outward force would cause holes to expand , r. n. lewis @xcite showed how structure could form in a universe which started out in a uniform state . the idea is that two spherical holes in an otherwise infinite and uniform universe would migrate toward one another , they would eventually touch and form a three - dimensional figure - eight , and they would eventually merge into one larger sphere . the mass in between the two holes would be pushed aside . if three holes were in close proximity , they too would approach each other and eventually merge , pushing aside the mass between them . four holes in close proximity to each other , however , retain the possibility that they trap some of the mass in between them as they move toward one another , thus forming a larger concentration of matter which would then pull together under its own positive gravitational attraction . we assume the existence of many holes throughout the universe , without specifying how they came into being . perhaps there were always slight imperfections in the uniformity which , as we earlier pointed out , would gradually grow in size . another possibility for the introduction of holes is some quantum effect which would destroy the total uniformity of the universe . whatever the cause , we assume that it repeated itself throughout the universe to form enough holes so that the ensuing evolution described below could occur . these holes , or any regions of lower than average mass density , would expand outward , just as lumps , or regions of higher than average mass density , would attract other mass toward itself . in other words , the infinite , uniform universe is unstable . we can imagine many expanding holes existing throughout the universe . r. n. lewis @xcite referred to this configuration as the swiss - cheese universe . later , as two holes came together , a sheet of matter would form between them , and the sheets of matter , looking at the entire conglomeration , would look like a foamy mixture , named the foam universe . still later , as the holes continue to expand and encroach upon each other s territory , the matter in the sheets would migrate toward the edge of the sheet , forming filaments connecting matter concentrations . this network configuration is referred to as the network universe . even later still , the filaments would split and be drawn gravitationally toward a concentration of matter . this arrangement is referred to as the caltrop universe because the appearance of one of these concentrations of matter is similar to a medieval caltrop . when astronomers peer outward toward other galaxies , they notice just this type of structure . galactic clusters seem to be connected to each other by filaments of galaxies , and , in turn , they appear clustered , not randomly , but in sheets , just the structure predicted above . galaxies and clusters are , in turn , separated from each other by gigantic voids which occur naturally in the above - described theory . it therefore appears that this view of the universe is better able to explain the observed structure of the universe than is the big bang theory , whose proponents are still arguing about how the universe could have differentiated into stars , galaxies , and galactic clusters in the relatively short time since the big bang . what is necessary for the above evolution to have occurred as stated ? the only requirements are an infinite universe , matter distributed somewhat uniformly throughout it , and the existence of randomly - scattered holes . the matter that was once distributed uniformly has been pushed or pulled together into galaxies , surrounded by emptyness , that we observe when looking out into the universe today . assuming that these galaxies are distributed somewhat uniformly throughout our universe , as most cosmologists believe today , then the outward force of negative gravity must also exist at this time , tending to push the universe apart . we can imagine that the universe we observe ( perhaps 10 billion light years or more in radius ) has an average mass density slightly less that that at infinity . one possible mechanism for this to happen is the following . assuming that there were events that occurred to form each of the holes initially all over the universe , it is conceivable that the effect of those events sufficiently far away from us havent been felt yet on earth , so that we on earth observe the average mass density in our neighborhood to be less than that farther out . our neighborhood ( radius unspecified ) of the universe and the galaxies in it , therefore , effectively reside in a large hole , with the force of negative gravity pushing outward of course , under this scenario of locality , this same comment would apply to any observer situated at any point of the universe . therefore , the same lower - than - average mass density and the same expansion must be observed everywhere . ithe outward force due to negative gravity at any point was given above in equation ( 1 ) . in addition to that outward force , there is an additional inward force due to the positive gravitational force of all mass closer to the center of the hole than the mass [ equ.(2 ) ] . the total force was given earlier as equation ( 8a ) . we also gave the corresponding potential as equation ( 8b ) . it is well known that light escaping from a gravitating body undergoes a shift in frequency due to the gravitational field @xcite . the shift is given by @xmath318 where @xmath319 is the frequency of the emitted wave , @xmath320 is its change in frequency , and @xmath321 is the change in potential . in calculating @xmath321 , there is a contribution from the emitting galaxy and one from the observing galaxy ( our own ) , each dependent on the gravitational mass of the respective galaxies . for a galaxy of roughly the same size as our own , these contributions will cancel . on the other hand , a much more massive galaxy may exhibit a relatively large gravitational redshift . there is also a doppler - type effect due to relative velocity , which is currently believed to be the primary cause of observed frequency shift . assuming that the visible universe resides in an expanding hole ( see section 5.1 ) , there will also be a contribution from the potential of the hole due to negative gravity . this newtonian potential [ see equ . ( 8b ) ] is proportional to @xmath265 for large @xmath11 , and could overwhelm the contributions from the other gravitational effects . it could do the same relative to doppler effects as well . it is apparent that whatever frequency shift is observed , at least four causes must be taken into account in the analysis : frequency shift caused by ( 1 ) gravitational effect of emitting galaxy , ( 2 ) gravitational effect of absorbing galaxy , ( 3 ) gravitational effect of negative gravity , and ( 4 ) doppler effect due to relative motion . however , we can conclude that the effects of ( 3 ) and ( 4 ) above reinforce each other , so that the observed redshift is not entirely due to relative motion . in other words , the estimated velocity of recession from observations of redshifted light is exaggerated , especially for galaxies which are relatively far away . in the past , an observed redshift value has been interpreted as indicating a certain distance and velocity of recession ( hubble s law ) . however , if ( negative ) gravitational effects as well as relative motion are taken into account , the observed redshift would indicate a slower ( less doppler redshift ) galaxy which is closer to us . the observed redshift should not be attributed entirely to doppler effects , nor entirely to gravitational effects . the negative gravitational redshift becomes increasingly important the larger that @xmath11 becomes . we note that , for @xmath322 light years , equation ( 30 ) yields a gravitational redshift given by @xmath323 x @xmath324 , which is significant when dealing with galaxies at that distance . in observing distant galaxies , this must be taken into account . for example , the great wall @xcite , distances to which are predicted on the basis of observed redshift values , is at a distance where negative gravitational effects are significant . more importantly , however , these negative gravitational effects will affect the microwave background radiation ( see below ) . because of the above comments , the entire big bang theory needs to be reevaluated in light of negative gravity . an entirely new cosmology based upon expansion due to negative gravity has been developed . r. n. lewis @xcite shows one approach to this . in fact , his explanation gives a simple account for the present structure of the universe which the big bang theory can not explain adequately without internal contradictions . a new cosmology was developed in which large velocities ( see section 3.2 ) are achieved due entirely to the effects of negative gravity over much larger time intervals ( many orders of magnitude ) than are currently believed to be required for expansion to our present state . we also showed that hubble s law holds in this case . for this cosmology , all that is required is that the visible universe , as well as possibly much more , resides in a region of slightly smaller mass density than that at infinity . this would allow negative gravity , as well as all of its consequences , to function . the observed abundances of helium , deuterium , and other heavier elements in the universe can be explained by supernova activity during the longer time intervals . it is currently believed that supernova activity since the big bang was not sufficient to produce anywhere near the amount of heavier elements that are observed in the universe . the amount of helium present is estimated to be from @xmath325 to @xmath326 of the mass of the universe , with a preference for a figure of @xmath327 . it is thought that perhaps one - tenth of that amount could have been produced in supernovas in the @xmath325 billion years since the big bang . in a universe which has been around forever and which may have been nuclearly active for @xmath328 billion years , the observed abundances of helium and other heavier elements could easily have been produced . the uniform microwave background radiation could be accounted for by the radiation of all galaxies . our sun radiates at close to blackbody levels . we expect the same of other stars and galaxies . averaged out over regions of the sky , we expect the radiation coming from other galaxies to be uniform and of blackbody type . indeed , data from the cosmic background explorer @xcite support this conclusion . it is generally agreed that the main nonuniformity in the background radiation is due to the fact that we are in motion relative to the source of that radiation . most cosmologists believe that the motion is due to the big bang . the motion could be due to a general expansion caused by negative gravity . we also expect the radiation from other galaxies to be redshifted because of relative motion . from the discussion above , it is apparent that radiation is also redshifted because of the negative gravitational potential . this redshifted radiation is observed in the microwave region . it is interesting to note that the singularity implicit in equation ( 22b ) ( which gives the metric tensor : @xmath329 ) , @xmath330 yields a redshift from equation ( 30 ) of @xmath331 this value being independent of @xmath332 or @xmath3 . if the microwave background radiation is actually due to an integration of the radiation of all galaxies , it would indicate a redshift of @xmath333 , corresponding to a temperature of 2.73 k and stellar surface temperatures on the order of 5000 k ( possibly less in the cooler , earlier universe ) . thus , it seems likely that the origin of the microwave background radiation must be @xmath334 to @xmath335 times more distant than @xmath336 , if we are to attribute the redshift to gravitational , rather than doppler , effects . more likely , it is due to a combination of the two effects . the singularity needs to be investigated further . however , it must be remembered that our universe is not exactly as modeled , so that the singularity shown above for a uniform universe with a spherical hole may not be a singularity for an uneven universe such as our own . the value @xmath337 found above is larger than most observed galactic redshifts . however , some recent observations of quasar redshifts are on the order of @xmath338 . these can be explained by doppler effects or positive gravitational effects due to incredibly high concentrations of mass , possibly in combination with the negative gravitational redshift . our sun s gravitational redshift is about @xmath323 x @xmath339 . a redshift of @xmath340 could result from a star or quasar with mass - to - radius radio of @xmath323 x @xmath143 times that of our sun . weinberg ( * ? ? ? ( 11.6.20 ) ) gives an upper bound of @xmath341 for gravitational redshift in a universe with zero cosmological constant . as can be seen from the equation before weinberg s equation ( 11.6.20 ) , using our equation ( 11 ) for his @xmath342 , a positive @xmath343 mediates a higher value of @xmath344 . of course this view of cosmology denies the great cosmological principle of isotropy and homogeneity and the principle of covariance by giving preference to centers of holes , at least at the present time . if we travel backward in time , the universe as has been presented here should appear more homogeneous and colder , so that these principles would have been more valid at earlier times . but , of course , as we look outward toward the heavens , we see that the universe is not isotropic nor homogeneous , so this should not be of concern . olbers paradox has an interesting explanation in this new theory of cosmology . wesson @xcite gives an explanation based upon the finite age of the universe and on inflation of the universe . this new theory does not require inflation or the big bang , but we can still invoke wesson s argument of the finite age of the universe by stating that , at earlier times , the universe was less radiant ( less hot ) . furthermore , another interesting explanation lies in the interpretation of the light ray geodesics . these bend outward , so that radiation from far enough away never reaches us ! one final comment deals with whether the universe is open or closed . this and the structure of the universe seem to be two of the most important open questions in the study of cosmology . in any cosmology based upon negative gravity , the universe is not only open , but its expansion is accelerating outward . there is not any great mass pulling it inward . in fact , the mass at infinity " is pulling it apart . iwhy should we believe in a cosmology based upon negative gravity rather than one based upon the big bang ? the big bang theory was developed to explain the observed redshift of light coming from distant galaxies . this redshift , in turn , was explained by attributing it to recessional motion , mainly because no other explanation for the observed redshift was known . this motion was then attributed to a general expansion of the universe . it was argued that this expansion could only be explained by extrapolating back to a time when an explosion ( big bang ) caused the motion we apparently observe today . no other mechanism for expansion was known . the subsequent discovery of the microwave background radiation fit in with this theory very nicely . this radiation was thought to be the remnant radiation left over from the big bang . it was further deduced that the expansion must be decelerating due to the inward pull of the finite mass distribution . the theory was further refined to include other recent observations , including the presence and concentration of heavy elements . the theory still has trouble explaining the observed structure of the universe , including the presence of large voids and the clustering of galaxies . recent observations from the cosmic background explorer @xcite show the early universe " to be almost perfectly uniform and homogeneous , not leaving enough time for matter to have differentiated and formed into the universe that we observe today . the theory of the development of the universe presented here , which is based on negative gravity in an infinite universe , does not have this shortcoming . it predicts the gross structure of the universe on the basis of mass being pushed or pulled together by the combined negative gravitational effects of adjacent holes @xcite . in addition , it presents a viable alternate explanation of the observed redshift of galactic radiation and of the microwave background radiation . the observed redshift of galactic light can be explained as due partly to recessional motion of expansion and partly to gravitational redshift , both ( expansion and gravitational redshift ) caused by the negative gravitational potential of mass at infinity " . it is possible to separate out the contribution to the redshift @xmath344 due to negative gravity and that due to recessional motion , assuming that some form of hubble s law is still valid . in fact , earlier , we showed that @xmath345 ( velocity proportional to distance ) for such a universe , which shows that hubble s law ( with constant @xmath121 ) is valid , without the necessity of extrapolating back in time to a singularity . in addition , the microwave background radiation can also be associated with a redshift of radiation due to the negative gravitational potential . finally , the abundances of elements other than hydrogen can be explained by supernova activity over the incredibly long life of the luminous universe ( at least an order of magnitude larger than is currently believed ) . this new theory predicts an expansion which originated rather benignly in a cold , ( relatively ) motionless , infinite universe and which has accelerated outward and will continue to accelerate outward forever . the purpose of this section is to use some of the equations derived in previous sections to deduce accurate values for the average mass density of the universe , @xmath194 , hubble s constant , @xmath46 , and the cosmological constant , @xmath164 . these calculations are based upon fairly accurate data , including redshift and distance data to specific galaxies . the computations can be done for any such galaxies , but we concentrate on one , m87 , for a variety of reasons . section 6.1 details these reasons and the specific data we use . section 6.2 includes the calculations of the different constants . finally , in section 6.3 , a summary and conclusions as well as justification of the methods developed here are presented . iin order to perform the calculations of the values of the particular constants , specific astronomical data are needed . there are several very good reasons why the galaxy m87 would yield reliable data on which the conclusions we reach are based . second , it is close enough to us that its distance from us can be fairly accurately measured by standard ( not redshift ) means . this measured distance is @xmath347 million light years @xcite . we will use both figures in our calculations . any error in measuring this distance will show up later in our calculation of the values of the constants . m87 is also close enough that small @xmath344 approximations should be valid . third , it lies near the center of the virgo cluster of galaxies , and is relatively motionless with respect to that cluster . therefore , such motions caused by local gravitational effects ( such as rotation about the center ) are not an added complication . we assume that the entire redshift , @xmath344 , is due to universal expansion ( see above for more details ; this expansion is not due to the big bang ) , and that local gravitationally - induced motions are insignificant . m87 is far enough away that this seems to be a reasonable assumption . fourth , m87 lies in the general direction toward which our galaxy and galactic cluster are moving . the observation of our direction of motion is based upon relative microwave background radiation measurements . for this reason , the calculations based upon negative gravity should be relatively accurate on the assumption that the virgo cluster is farther out than we are along a radial direction from the center of expansion . one drawback of using m87 is that it is a huge galaxy relative to our own , roughly @xmath348 times as massive @xcite . as such , it may have associated with it a significant gravitational redshift compared with ours , which may have to be accounted for in the calculations . ignoring this possible gravitational redshift , as we do here , means that the results for @xmath349 and @xmath350 later in this paper are upper bounds and may need to be reduced , thus changing the other calculated values . one problem associated with the use of data from only one galaxy is that the galaxy we choose may not be typical . a more comprehensive study could be done using a statistical analysis of data from many different galaxies , with proper precautions being taken to account for , in particular , the phenomenon referred to in the fourth category above . however , the fact that our calculation below gives a value of the hubble constant in the middle of the experimentally - determined range should convince the reader that m87 is a reasonable choice and that this approach is a sound one . ifirst , using the notation @xmath344 for redshift , we identify three causes of redshift : ( 1 ) redshift due to recessional motion , @xmath349 , ( 2 ) redshift due to negative gravity , @xmath350 , and ( 3 ) gravitational redshift , @xmath351 . for a given value of @xmath344 , the wavelength of the radiation has been shifted by a factor of @xmath352 . therefore , @xmath353 . we assume that @xmath351 is negligible , so that the above equation can be replaced by @xmath354 or in these equations , @xmath157 is the speed of light in a vacuum , @xmath359 is the velocity of recession , @xmath11 is distance , @xmath194 is mass density , and @xmath6 is the gravitational constant . therefore , since the @xmath11 values are line - of - sight distances and are probably greater than the difference in distance from the center of the main expansion , the @xmath11 values should be reduced by an indeterminant amount , thus resulting in a larger value for @xmath194 . the approximate value of @xmath194 used earlier was @xmath47 x @xmath48 kg / m@xmath9 , roughly one - tenth the value calculated above . rubin @xcite finds that the average value of @xmath194 , at least in our visible universe , is @xmath369 x @xmath48 kg / m@xmath9 . this figure is based upon gravitational calculations and is about @xmath370 times larger than observations of luminous matter lead us to believe . thus , rubin s calculations indicate that a large percentage of the matter in the universe is nonluminous . if these are accurate figures , they indicate that the average mass density at infinity is roughly @xmath371 times what it is in our neighborhood , which is in line with the assumption made above and referred to in section 5 . we note that the range of values we derive by this method for hubble s constant is right in the middle of the range of values currently used for that value . it corresponds to the value @xmath373 km / s / mpc . this correlation of values must lend credence to this line of reasoning . it is extremely unlikely that we would end up with this degree of accuracy , compared with experimentally - determined values , if the underlying theory were not correct . iin this section , we have used certain astronomical data , certain well - established formul , and several new formul for calculating the values of the average mass density of the universe , @xmath194 , the hubble constant , @xmath46 , and the cosmological constant , @xmath164 . these values are found to be in the ranges : the fact that @xmath46 lies exactly in the middle of the experimentally - determined range of values for that constant must mean several things . first and most importantly , it signifies that the underlying theory of negative gravity must be correct . otherwise , it would be exceedingly improbable that the calculated value would be anywhere near the experimental value . in fact , we use formula ( 37 ) above which is based upon a gravitational potential which is on the order of @xmath378 times larger than the one normally used in astronomical calculations . with the values of @xmath11 used here , the difference is about @xmath379 orders of magnitude . if the underlying theory were not correct , the calculated values should be different to a corresponding degree . second , it shows that our use of the galaxy m87 was a reasonable choice . third , it serves to pin down the value of the hubble constant even closer . fourth , the values given for the cosmological constant and for the average mass density of the universe should be relatively accurate . it should be pointed out that this is the first attempt to actually assign a nonzero value to the cosmological constant and to calculate the average mass density in an infinite universe . finally , let us observe that , in the above calculations , there were several places where it was indicated that certain values might need to be reduced for various reasons , while , in other places , just the opposite was indicated . again , since the calculated value of the hubble constant falls in the experimentally - determined range , it is reasonable to conclude that the above mentioned changes approximately balance each other .
metal - organic frameworks ( mofs ) are organic - inorganic hybrid solids in which a usually low - dimensional inorganic subnetwork is connected with adequately functionalized organic molecules to form a 3d framework . the fascinating structures and properties of these materials have been the focus of intense research efforts over the last 15 years or so.@xcite at present , a wide variety of new compounds continues to be discovered , because of the huge combinatorial possibilities offered by the amalgamation of organic and inorganic chemistry . this new class of compounds has even proved able to produce interesting properties for industrial applications.@xcite one of the main reasons for this success is their porous character.@xcite various applications have been envisaged for mofs , principally in the field of gas storage and separation,@xcite but also for catalysis,@xcite luminescence,@xcite or magnetorefrigeration.@xcite mofs containing magnetic ions furthermore represent a goldmine for low - dimensional and frustrated magnetism.@xcite first , when the organic moiety of these materials is composed of magnetically inert molecules such as sufficiently long alkane chains , it can act as a simple spacer that prevents the establishment of electronically - driven magnetic interactions between e.g. spin chains or spin layers . in other words , such systems have a strongly low - dimensional character regarding through - bond magnetic interactions and , consequently , they only become magnetically ordered at low or very - low temperatures , when the one - dimensional ( 1d ) or two - dimensional ( 2d ) correlation length diverges and promotes sizable classical dipolar interactions between correlated spin blocks . second , in addition to the intrinsic low - dimensionality , magnetic mofs also show a large diversity in the type of topologies formed by the inter - connection of the magnetic ions ( m ) through the possible superexchange pathways ( m - o - m pathways in most cases ) . despite of the clear interest of these materials in the field of magnetism , hybrid frameworks are relatively little studied by neutron scattering techniques,@xcite although these methods are commonly used to investigate the magnetic structures and excitations of purely inorganic compounds . this is partly due to the difficulty in synthesizing deuterated compounds when complex organic ligands are involved , and partly due to the weak magnetic contributions to the total neutron scattering from these materials . recently , we reported a new cobalt^ii^ hydroxy - sebacate framework of formula co^ii^@xmath0(oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5.@xcite the compound contains co^ii^-based magnetic chains which are very well separated from each other ( 1 to 2 nm ) . we revealed the coexistence of magnetic long - range order ( lro ) and slow dynamics of the magnetization below @xmath7 k. this new compound is an excellent candidate for a detailed magnetic study for at least two reasons . first , the magnetic chains are much better isolated from each other than in most of the spin - chain compounds . second , the chains themselves present a rather complex topology of magnetic interactions between spin carriers which are well known for their large single - ion anisotropy . spin - chains attract much attention because they provide genuine examples for testing models that are less tractable in higher dimensions.@xcite for example , spin chains with nearest - neighbor antiferromagnetic interactions display a large variety of ground states with non - classical magnetic phenomena.@xcite although ideal 1d spin systems do not undergo lro at finite temperature due to strong quantum fluctuations,@xcite slight perturbations such as , for instance , weak interchain interactions can make the quantum critical state unstable to three - dimensional ( 3d ) lro.@xcite another striking feature of 1d magnetic systems is the so - called single - chain magnet ( scm ) behavior observed in a relatively large number of metallo - organic compounds.@xcite these systems , though not long - range ordered , present bulk magnetic behavior at low temperature that results from slow intrachain dynamics arising from the combination of a large magnetic anisotropy with ferromagnetic interactions . this was predicted in the pioneering work of r. j. glauber in 1963.@xcite in this article we provide deeper insight on the magnetism in the ordered phase of co^ii^@xmath0(oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5 . in section ii we recall the necessary information concerning the crystal structure and macroscopic magnetic behavior of the system at @xmath8 k. experimental details are given in section iii . our new results are presented and discussed in section iv . firstly , the zero - field magnetic structure is determined from neutron powder diffraction data . the magnetic field - dependence of this spin arrangement is qualitatively evaluated . secondly , we present new magnetic measurements for temperatures down to 70 mk of both magnetization and ac susceptibility , and deduce the characteristic energies of the observed dynamics . finally , in section v , we discuss the magnetic structure in relation with the magnetic couplings and anisotropy in the compound . we further propose a scenario to describe the dynamics . finally , in section vi we summarize the present work . ( oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5.@xcite co^ii^1o@xmath4 and co^ii^2o@xmath9 polyhedra are represented in yellow and red , respectively . ( i ) projection of the 3d metal - organic framework perpendicular to the magnetic chain axis . ( ii ) atomic structure of the 1d inorganic subnetwork.,width=264 ] . ( i ) j1 , j3 and j4 correspond to corner - sharing pathways having the angle values given on the figure , while j2 corresponds to an edge - sharing pathway with co^ii^1-o - co^ii^2 angles of about 91^o^ and 96^o^. ( ii ) spin topology viewed along the [ 101 ] crystallographic direction , i.e. by applying a 45^o^ rotation around the chain axis from the orientation of figure [ fig.2](i).,width=302 ] co^ii^@xmath0(oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5 adopts a _ _ p__2@xmath2__/c _ _ monoclinic structure in which magnetic chains develop along the * b-axis . isoreticular compounds , i.e. having a similar framework organization though eventually based on different organic ligands , have been reported with manganese^ii^ and iron^ii^ ions @xcite . figure [ fig.1 ] sheds light on the main structural characteristic of this compound : the strongly 1d character due to large distances between the inorganic chains . the distance between the centers of neighboring chains varies from 11 to 21 , with the shortest path ( i.e. in the [ 001 ] direction , see figure [ fig.1](i ) ) not involving chemical connections . from the point of view of the magnetism , the combination of these large interchain distances with either the absence of interchain chemical connection or its presence through magnetically inert molecules ( alkane chain with 9 consecutive c - c single bonds ) precludes any role of interchain interactions other than classical dipolar forces . the 1d inorganic subnetwork of this material contains two crystallographically distinct magnetic ions , co^ii^1 and co^ii^2 , that have distorted octahedral and distorted trigonal bipyramidal oxygen environments , respectively . both sites have intrinsically asymmetric crystal field environments as the oxygen atoms originate from carboxylate functions and hydroxide groups . the chains can be initially viewed as being constituted of a central ` zig - zag ' chain of corner - sharing co^ii^1o@xmath4 octahedra . in addition , co^ii^2o@xmath9 trigonal bipyramids share oxygen atoms with three consecutive co^ii^1o@xmath4 octahedra of the main chain . both magnetic sites are in the general 4__e _ _ position which , within a single inorganic chain , splits into two equivalent positions related by the 2-fold screw axis and labeled as ` a ' and ` b ' on figure [ fig.1](ii ) . four intrachain superexchange pathways can be distinguished and are emphasized on figure [ fig.2](i ) . first , yellow co^ii^1o@xmath4 octahedra are connected to each other along the chain axis by sharing corners through a large superexchange angle of about 136^o^ ( j1 ) . second , each co^ii^2o@xmath9 bipyramid ( red ) connects with two co^ii^1o@xmath4 octahedra by corner - sharing through smaller superexchange angles ( @xmath10102^o^ for j3 and @xmath10107^o^ for j4 ) ; and with one co^ii^1o@xmath4 octahedron by edge - sharing ( j2 ) with the nearly orthogonal angles ( @xmath1091^o^ and @xmath1097^o^ ) . the resulting spin topology is represented on figure [ fig.2](ii ) . in addition to these four m - o - m pathways , weaker intrachain magnetic ` super - superexchange ' pathways exist through the carboxylic functions of the sebacate ligand ( m - o - c - o - m pathways ) , leading to further terms expected to contribute to the full hamiltonian . the first magnetic characterizations of co^ii^@xmath0(oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5 have been previously reported.@xcite the compound undergoes magnetic long - range ordering below t@xmath12=5.4 k , as evidenced by heat capacity ( c@xmath13 ) measurements and by the appearance of magnetic scattering below t@xmath12 as observed by neutron diffraction . the presence of a broad bump in the zero - field c@xmath13(t ) curve , between the @xmath14-peak at t@xmath12 and 2 to 3 times t@xmath12 , was attributed to the effects of magnetic correlations within the chains . a complex behavior of the magnetic susceptibility was observed above the lro transition . the non - monotonic shape of the product @xmath15 could be described by introducing two effective energy parameters ( 20 and -30 k ) which account for both anistropy and ( ferromagnetic and antiferromagnetic ) exchange interactions . ac susceptibility measurements showed the existence of dynamical processes at t @xmath16 t@xmath12 . ac susceptibility data were taken on a physical properties measurement system ( quantum design , usa ) for frequencies of the ac magnetic field between 10 and 6500 hz . the analysis revealed the existence of a thermally activated behavior within the 3 - 4.5 k temperature range , with an energy barrier of about 67 k and a characteristic time of the order of 1.4@xmath17 s. the observed slow dynamics was attributed to the motion of thermally activated domain walls within the magnetic chains . finally , ac susceptibility measurements performed under various bias dc magnetic fields revealed a strong displacement of the main peak towards higher temperatures , for both in - phase and out - of - phase components . this latter observation was suggested to evidence the existence of field - induced magnetic order for temperatures above the zero - field transition temperature t@xmath12 . the two powder samples of co^ii^@xmath0(oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5 used in this work were synthesized hydrothermally according to the method reported previously.@xcite chemicals ( sebacic acid and naoh ) were not deuterated , obtained from commercial suppliers , and used as received . high - purity distilled water was used for the reaction medium . long - exposure laboratory x - ray powder diffraction revealed the samples to be single phase . neutron scattering experiments were conducted using the high - flux d20 two - axis diffractometer at the ill ( institut laue langevin , grenoble , france ) . the neutron wavelength was @xmath18 . all experiments were performed with a 6 mm diameter vanadium sample container and a cryomagnet providing temperatures down to 1.8 k and magnetic fields up to 6 t. zero - field data were first recorded by simply loading the 2 g powder into the sample holder . two long duration neutron diffraction patterns were recorded at 1.8 k and 20 k ( 8 hours for each ) . for measurements with an applied magnetic field , the sample was immersed in deuterated isopropanol in order to avoid reorientation of the powder particles . isopropanol freezes around t @xmath10 115 k in a glassy state , giving rise to a weak and broad bump in the data around @xmath19 . after initial rapid cooling of the sample in the remnant field of the magnet , the temperature was never raised above the glass transition temperature of isopropanol for the remainder of the experiment . it was therefore guaranteed that the powder grains remained fixed and that the same powder average was measured in each run . a diffraction pattern was recorded under these conditions , in zero - field at 10 k , and used for subtraction of the nuclear scattering . data were then recorded under zero - field and under two different applied magnetic fields ( 0.5 and 2 t ) . for each of these three field values , diffraction data were collected for different temperatures across the transition , from 1.8 k to higher temperatures . after these measurements , another zero - field 10 k pattern was recorded in order to confirm the absence of reorientation of the powder particles . the neutron patterns were analyzed by rietveld refinements using the fullprof program . @xcite the shape of the diffraction peaks was fitted using a pseudo - voigt function . besides the six peak profile parameters , the fit of the paramagnetic pattern recorded at 20 k comprised zero shift , cell parameters , preferred orientation of the powder particles , and scale factor . the agreement indices r@xmath20 and r@xmath21 are the factors defined in ref . . the difference resulting from the two long - duration patterns recorded at 1.8 and 20 k was used for the refinement of the magnetic structure . all parameters were fixed to their previously refined values ( 20 k ) , except the lattice parameters which were taken from a le bail decomposition@xcite of the 1.8 k pattern . the agreement factor noted r@xmath22 is the bragg agreement factor@xcite for the fit of the magnetic scattering observed on the difference pattern . the representational analysis techniques described by bertaut@xcite were used to determine the possible magnetic structures compatible with both the magnetic wavevector and the symmetry of the crystal structure . the irreducible representations of the propagation vector group were determined with the program basireps@xcite . the possible solutions were first tested by refining the mixing coefficients of the basis vectors , as implemented in the fullprof program.@xcite the final rietveld refinement of the magnetic scattering was performed in the ` spherical coordinates procedure ' of the software . the three parameters ( r , @xmath23 , @xmath24 ) per co^ii^ characterizing their atomic magnetic moments correspond to the value of the magnetic moment @xmath25 ( in bohr magnetons ) , the polar @xmath23 angle with z axis and the azimuth @xmath24 angle between the x axis and the projection of r onto the ( x , y ) plane . since this mode works only if z is perpendicular to the ( x , y ) plane , the present monoclinic space group was described in the laue class 1 1 2__/m _ _ for the magnetic refinement . two powder samples ( sample 1 - mass @xmath26 mg and sample 2 - mass @xmath27 mg ) were measured using low temperature superconducting quantum interference device ( squid ) magnetometers equipped with a miniature dilution refrigerator developed at the institut nel - cnrs grenoble @xcite . sample 1 was obtained from the 2 g sample used for neutron scattering experiments . the samples were prepared in a copper pouch and mixed with apiezon grease ( for thermalization and to avoid reorientation of the powder under magnetic field ) . they were attached to a copper tress suspended from the mixing chamber of the dilution fridge . @xmath28 hysteresis loops were measured for both samples in the ` high ' field magnetometer which reaches 8 t. ac and relaxation measurements were performed in the ` low ' field magnetometer , which reaches 3500 oe , for sample 1 . the magnitude of the alternating field was @xmath29 oe . the rietveld refinement of the 20 k neutron pattern was obtained by using the atomic coordinates determined from x - ray single - crystal diffraction at 100 k ( see figure [ rietveld_nuclear])@xcite . the good agreement between the calculated and recorded neutron patterns ( r@xmath20 = 0.115 and r@xmath21 = 0.127 ) suggests the absence of significant modifications of the crystal structure below 100 k. the refined unit cell parameters are : @xmath30 , @xmath31 , @xmath32 , and @xmath33 ^o^. the magnetic scattering was extracted from the data recorded at 1.8 k by subtracting the 20 k pattern . the resulting ` difference pattern ' shows weak magnetic bragg peaks both on top of some nuclear peaks of the ( _ h_,0,0 ) type and on the ( 0,0,_l _ ) and ( _ h_,0,_l _ ) peaks with odd _ l _ indices . these two latter families of reflections , absent in the high - temperature pattern , are forbidden by the _ c_-glide of the nuclear space group . these indications for such a non - bravais lattice ( breaking of the nuclear point - group symmetries ) imply a commensurate magnetic structure with a propagation vector * _ k_*=(0,0,0 ) . to simplify the analysis , group theory was used in order to predict the magnetic arrangements compatible with the nuclear symmetry , thus reducing the number of independent parameters in the refinement . in the landau theory , magnetic fluctuations in the paramagnetic state while approaching a second - order phase transition reflect the crystal symmetries . at the transition , one magnetic mode is stabilized . the magnetic modes compatible with the nuclear space group _ g _ (= _ _ p__2@xmath2__/c _ _ ) are given by the irreducible representations of the little group _ g@xmath34 _ that leave _ k _ * invariant , i.e. the irreducible representations of _ g _ for * * _ k_*=(0,0,0 ) . the application of the group theory , using the irreducible representations tabulated by kovalev@xcite and available in the basireps program,@xcite shows that the group representation can be reduced to a sum of four irreducible representations of order 1 : @xmath35 . [ cols="<,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] the four possible magnetic arrangements resulting from this analysis are summarized in table [ irreps ] . in principle , @xmath36 and @xmath37 can be excluded because they do not contain a ferromagnetic coupling while magnetic scattering is detected on the top of allowed nuclear peaks . moreover , testing the four models by refining the mixing coefficients of the basis vectors of the two magnetic moments indicates @xmath38 to most likely describe the magnetic structure . the three other models provide unreasonable magnetic moment values and/or much larger agreement factors . imposing the irreducible representation @xmath38 , the refinement of the six mixing coefficients converges and gives reasonable magnetic moment values . in particular , the magnetic moment of co^ii^1 is oriented along the chain axis ( b-axis ) , implying that the co^ii^1 sublattice forms a simple antiferromagnetic chain . in a second step , the refinement was realized by describing the magnetic moments by their spherical coordinates ( see section [ exp ] ) , in order to improve the stability of the refinement procedure . this procedure confirmed the orientation of the co^ii^1 moment to lie along the b-axis . thus , we constrained the magnetic moment of co^ii^1 with @xmath39^o^ and @xmath40 = 0@xmath41 and 180@xmath41 for the positions related by the screw axis , in accordance with @xmath38 . the magnetic moment of co^ii^2 , for its part , was refined with its three spherical coordinates by imposing the couplings relevant to @xmath38 : ( @xmath42 , @xmath43 , @xmath44 ) for co@xmath45 ions in positions 1 and 3 , and ( @xmath42 , @xmath43 , @xmath46 ) for co@xmath45 in positions 2 and 4 . using these four parameters ( @xmath47 for co@xmath48 and @xmath42 , @xmath43 , @xmath44 for co@xmath45 ) , the refinement procedure converges satisfactorily ( r@xmath49 , @xmath50=1.49 , see figure [ rietveld_magnet ] ) . starting from this model , lifting the additional constraints applied to the magnetic moment of co^ii^1 ( @xmath51 ) does not lead to a significantly different magnetic structure . in the final model , the refined @xmath43 value of about @xmath52 corresponds to the monoclinic angle of the unit cell , indicating that the co@xmath45 magnetic moment is roughly in the ( b,c * ) plane . consequently , @xmath44 ( @xmath53 ) defines both the rotation of the co@xmath45 magnetic moments in the ( * b,c * ) plane relative to the chain axis , and their canting angle between the positions related by the screw axis ( @xmath54 ) . model ( see text for details).,width=283 ] correspond to a setting with x = * a * , y = * c * , z = * b * ( laue class 1 1 2__/m _ _ , see section [ exp]).,width=321 ] this refinement gives magnetic moment values of 2.19(7 ) and 2.68(8 ) for co@xmath48 and co@xmath45 respectively ( see figure [ magn_struct ] and table [ table_magmom ] ) . the co@xmath48 sublattice is fully compensated ( @xmath51 ) , while co@xmath45 has both compensated ( @xmath55 ) and uncompensated ( @xmath56 ) components . the global ferromagnetic component for this material is thus of about @xmath57.co@xmath58mol@xmath59 . this value , after averaging over the three directions ( @xmath60.co@xmath58mol@xmath59 ) , is in good agreement with the size of the magnetization jump at very low field ( @xmath61.co@xmath58mol@xmath59 , see section [ mh ] and ref . ) . the magnetic structure is presented on figure [ magn_struct ] , and is further discussed in section [ discussion ] . * 7c & _ _ m__@xmath62 & _ _ m__@xmath63 & _ _ m__@xmath64 & * m & @xmath23 & @xmath24 + co@xmath581 - 1 & 0 & 2.19(7 ) & 0 & 2.19(7 ) & 0 & 0 + co@xmath581 - 2 & 0 & -2.19(7 ) & 0 & 2.19(7 ) & 180 & 0 + co@xmath581 - 3 & 0 & 2.19(7 ) & 0 & 2.19(7 ) & 0 & 0 + co@xmath581 - 4 & 0 & -2.19(7 ) & 0 & 2.19(7 ) & 180 & 0 + co@xmath582 - 1 & 0 & -2.13(5 ) & -1.64(8 ) & 2.68(8 ) & -36(2 ) & 101(3 ) + co@xmath582 - 2 & 0 & 2.13(5 ) & -1.64(8 ) & 2.68(8 ) & 216(2 ) & 101(3 ) + co@xmath582 - 3 & 0 & -2.13(5 ) & -1.64(8 ) & 2.68(8 ) & -36(2 ) & 101(3 ) + co@xmath582 - 4 & 0 & 2.13(5 ) & -1.64(8 ) & 2.68(8 ) & 216(2 ) & 101(3 ) + * the stability of the magnetic structure against the application of an external magnetic field was verified by measuring the magnetic scattering at 1.8 k under fields of 0.5 and 2 t ( see section [ exp ] ) . despite the complex experimental setup ( cryomagnet and dispersion of the compound in deuterated isopropanol ) , the high hydrogen content of the compound , and the diluted character of the magnetism , weak magnetic elastic scattering could be observed ( see figure [ patterns_3h ] ) . although the bad statistics preclude refinement of these data , the absence of drastic changes in the 1.8 k magnetic structure is evidenced up to 2 t. this result is in agreement with the shape of the isothermal magnetization at 1.8 k ( see section [ mh ] and ref . ) , i.e. a fast increase at very low field due to the ferromagnetic component of the magnetization , followed by a mild increase due to competition between the applied field and the local anisotropy of co@xmath58 ions . data were also recorded as a function of temperature for 0 , 0.5 and 2 t. figure [ 3ref_h ] summarizes the obtained data through the thermal evolution of three magnetic bragg reflections . we can first look at the ( 10@xmath65 ) reflection , which is characteristic of the interchain correlations ( anti-c reflection , see figure [ magn_struct ] ) . in zero - field ( see figure [ 3ref_h ] ) , the ( 10@xmath65 ) intensity increases below @xmath7 k , the transition temperature determined from macroscopic measurements @xcite . when a magnetic field is aplied , the intensity of this reflection remains finite above @xmath66 . this indicates the persistence of interchain correlations in presence of an applied field well above @xmath66 : up to 10 and 13 k for applied field values of 0.5 and 2 t respectively . this is not surprising since the interchain magnetic arrangement is ferromagnetic - like , so that a magnetic field is expected to stabilize these correlations . it is worth noting that these results are in good agreement i ) with the previously reported phase diagram established from the bias dc field dependence of the ac susceptibility maxima,@xcite and ii ) with the position of the maxima in the derivative of the temperature dependence of the magnetization : 9.8 k for 0.5 t and 13.9 k for 2 t ( data not shown).@xcite second , we consider the ( 001 ) and ( 100 ) reflections . these reflections are relevant for observing the evolution of the canting within the chains . for instance , considering @xmath67 , ( 001 ) is maximal for @xmath68 while ( 100 ) is completely absent , and vice versa for @xmath69 . the temperature dependence of these reflections depends strongly upon the magnetic field , especially at the onset of the interchain correlations when field - induced magnetic scattering is detected . this indicates significant rearrangements of the spin structure within the chains at temperatures for which magnetic field induces long - range ordering , and points toward a rich @xmath70 phase diagram . interestingly , at low temperature , the respective intensities of these reflections are little affected by the magnetic field up to 2 t. this suggests that the applied magnetic field only provokes a small canting with respect to the zero field magnetic structure . ) reflections recorded upon warming under zero - field ( left ) , 0.5 t ( middle ) , and 2 t ( right ) . the plotted magnetic neutron scattering is obtained after subtraction of a pattern recorded in zero - field at 10 k.,width=321 ] @xmath28 hysteresis loops start to open below 2 k as shown in figures [ fig_hyst1 ] and [ fig_hyst2 ] . the maximum coercive field is about 0.5 t at low temperature . these hysteresis loops are strongly affected by the dynamics existing in the chains , resulting in relaxation at a given field during the extraction measurements . this can be seen clearly in the data at 800 mk or 1 k , for example . vs. @xmath71 for sample 1 at several temperatures between 100 mk and 1.5 k , measured with field increments of 500 oe every 110 s for @xmath72 k , and of 50 oe every 80 s for @xmath73 k. , width=302 ] below 1 k , magnetic avalanches ( that is to say , abrupt reversals of the magnetization ) were measured in sample 1 , at fields between 0.4 and 0.5 t ( see figure [ fig_hyst1 ] ) . these avalanches are typical of samples with slow relaxation processes at very low temperature ( see for example ref . ) and have been observed in other ordered spin - chains @xcite . the avalanche mechanism can be described as follows : due to slow dynamics ( large energy barrier , pinned domain - walls , ) , when the sample is driven out of thermodynamic equilibrium , by an applied field , the magnetization does not follow the field . at a given field ( the resonant tunneling field or the nucleation field for example ) , the reversal of the magnetization becomes easier , and spins can start to flip . then , because the system is at very low temperature , the specific heat is very low and the energy @xmath74 released in this reversal heats the neighboring spins , which in turn can flip . this process results in an increase of the internal temperature of the sample , which itself induces a faster relaxation , and provokes a global reversal of the magnetization termed a magnetic avalanche . in some cases avalanches can be suppressed by slowing the field sweeping rate in the hysteresis loop , or by improving the thermal contact between the sample and the mixing chamber of the dilution fridge which provides the cooling power . the latter was done in the present compound with sample 2 , which was better thermalized than sample 1 ( smaller size and more apiezon grease ) . indeed , magnetic avalanches could be suppressed in this sample ( see figure [ fig_hyst2 ] ) . interestingly , around @xmath75 t , there is however a change of slope in the magnetization curve , showing that the magnetization starts to respond to the applied magnetic field . this indicates that @xmath76 corresponds to an energy scale of the system , and that the previously observed avalanches trigger at a field which is characteristic of a faster relaxation of the system . however , because the measurements were made on a powder and the magnetic structure is very complex , the origin of this energy scale is not readily determinable . vs. @xmath71 for sample 2 at several temperatures between 100 mk and 1.5 k , measured with field increments of 1000 oe every 110 s for @xmath72 k , and of 500 oe every 110 s for @xmath73 k. , width=302 ] as shown in ref . , and observed here in the @xmath28 hysteresis loops , co^ii^@xmath0(oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5 presents slow dynamics in the ordered regime , characterized by an energy barrier of about 60 k which is attributed to domain - wall dynamics . in order to get a better understanding of the dynamics we have conducted complementary experiments at temperatures lower than 2 k. first , we have measured the ac susceptibility from 2.11 hz down to the lowest frequency reachable in our magnetometer , 1 mhz . in this frequency range , the ac susceptibility still shows a frequency dependence , which we characterize by tracking the maximum in the imaginary part of the susceptibility @xmath77 ( see inset of figure [ fig_tau ] ) . expecting that , at the maxima , the measurement frequency @xmath78 matches the relaxation time @xmath79 of the system , we have plotted @xmath80 as a function of the inverse of the temperature of the maximum @xmath81 , together with the higher frequency results from ref . ( see figure [ fig_tau ] ) . these data can be fitted with an arrhenius law @xmath82 with @xmath83 s and @xmath84 k in the whole frequency range , that is to say , over 7 decades . this shows that the mechanism behind the dynamics is robust in the ordered regime , and remains thermally activated down to at least 2.1 k. the existence of a distribution around the central barrier energy e@xmath85 was previously suggested by the shape of the cole - cole plots ( @xmath77 vs. @xmath86).@xcite the parameter extracted from the fit of these plots to a generalized debye model@xcite ( @xmath87 for t = 4 k , see ref . ) , can be used to estimate the full width at half maximum of this distribution ( @xmath88 k ) , by following the phenomelogical formalism developped by mydosh for the study of spin glasses.@xcite vs. @xmath89 in a semi - logarithmic scale from ac susceptibility measurements . red points are from mpms ( qd ) measurements ( 10 hz@xmath90 6.5 khz ) @xcite , and blue squares from measurements in dilution fridge magnetometers ( lt ) ( 1 mhz @xmath91 hz ) . the line is a fit to the arrhenius law with @xmath83 s and @xmath84 k. inset : imaginary part @xmath77 of the susceptibility at low temperature . , width=302 ] to probe the dynamics at lower temperatures , wherein we expect a slower response , we have performed measurements of the relaxation of the magnetization as a function of time . the protocol is the following : ( i ) the sample is heated to about 2 k ( temperature at which there is no hysteresis in the magnetization ) @xcite , ( ii ) a field @xmath92 oe is applied , ( iii ) the sample is cooled towards the target temperature , ( iv ) the field is removed and the magnetization is measured as a function of time . we note that , according to the neutron scattering experiments under applied magnetic field ( see section [ mag_struc ] ) , a field of 3500 oe does not seem to destabilize the magnetic structure . this implies that the induced magnetization is due to the reversal of magnetic domains or to canting phenomena . the relaxation curves obtained in zero field are shown in figure [ relax ] for temperatures between 0.9 and 1.8 k. the relaxation is not purely exponential , and can be fitted with stretched exponential functions . this confirms the existence of a distribution of relaxation times . to investigate more precisely the barrier distribution , we followed the approach of ref . by plotting the measured magnetization as a function of a single variable , @xmath93 . all curves can be scaled using the parameter @xmath94 s ( see figure [ fig_mtlnt ] ) , thus confirming the picture of a distribution of barriers crossed by thermal activation . a barrier distribution @xmath95 can then be deduced from the derivative of the magnetization with respect to @xmath96 , @xmath97 , which is plotted on the inset of figure [ fig_mtlnt ] . vs time @xmath98 at several temperatures between 0.9 and 1.8 k. the sample is heated at about 2 k and cooled to the target temperature in an applied field of -3500 oe . at @xmath99 , the field is removed . , width=283 ] vs the energy @xmath100 , for 0.9 k@xmath101 k. @xmath102 s is chosen so that all the magnetization curves collapse . inset : derivative @xmath97 vs @xmath96 showing the barrier distribution ( see text ) . the line is a gaussian law centered on @xmath103 k with @xmath104=16.5 . , width=302 ] the obtained barrier distribution is almost symmetric , but rather large : the full width at half maximum @xmath105 evaluated by the fit to a gaussian function is about 10 k. surprisingly , the barrier distribution is centered on @xmath106 k , that is to say about half of the energy barrier @xmath107 obtained from ac susceptibility measurements ( in zero field ) . in addition , the value of @xmath108 is three orders of magnitude larger than the one deduced from ac measurements . these results indicate the existence of different relaxation processes for the two types of measurement . this is in strong contrast with other spin - chains on which the same analysis was made @xcite , and where the obtained barrier energy was similar from ac susceptibility and relaxation of the magnetization . vs the energy @xmath100 with @xmath109 s , obtained for relaxation curves measured in applied fields of 500 , 1000 , 1500 , 2000 and 2500 oe between 0.8 and 1.2 k. the curve at @xmath110 ( inset of figure [ fig_mtlnt ] ) is shown for comparison . the inset gives the central barrier energy @xmath107 vs applied field @xmath71 . the line is a fit for @xmath111 to the equation : @xmath112.,width=298 ] we have also performed measurements of the magnetization relaxation in non - zero applied fields , from 500 to 2500 oe . the protocol described above was used for temperatures between 0.8 and 1.2 k. note that for fields larger than 2000 oe , and at high enough temperatures , the relaxation is so fast that avalanches occur . this phenomenon , which involves self - heating of the sample , prohibits access to the full barrier distribution at these fields . nevertheless , the same scaling procedure was applied and led to the same @xmath108 ( @xmath113 s ) . furthermore , the barrier distribution could be obtained ( see figure [ fig_dmdevsh ] ) . two main features are clear from this figure . first , the distribution width is much smaller in applied field than in zero field . in the presence of an applied magnetic field , the full width at half - height @xmath105 varies between 3.5 and 4.5 k , a value to be compared with the 10 k obtained in zero applied field . second , the mean energy barrier decreases when the field increases . the dependence is linear for non - zero applied fields ( see inset of figure [ fig_dmdevsh ] ) . this is consistent with the two - level system picture used for the analysis , in which the energy barrier would be reduced by the zeeman energy : @xmath114 , where @xmath115 is the energy barrier in zero field , and @xmath116 is the magnetic moment parallel to the applied field . the obtained @xmath115 value is smaller than the experimental @xmath117 ( @xmath118 k against @xmath106 k - see inset of figure [ fig_dmdevsh ] ) , but within the barrier distribution measured at zero field . the slope @xmath116 is equal to 30 @xmath119 . this might indicate that the energy barrier does not concern single spin flips but rather blocks of spins of effective moment @xmath116 . in that sense , the larger distribution width , as well as the difference between @xmath117 and @xmath115 , might be explained by a more complex domain structure in zero field than in finite applied field for which a domain selection occurs ( as shown by magnetization measurements ) . the complex magnetic structure and the diversity of the exchange paths in this cobaltous chain make the interpretation of the above results challenging . nonetheless , the observed slow dynamics has a large signal , and is very clean and reproducible , suggesting that there should be a well - defined origin of this phenomenon . in the following , we examine the exchange paths and the relevant energy scales in the system , and present a possible scenario explaning the observations described in the previous section . the magnetic structure results from the magnetic interactions and the local anisotropy of the co@xmath120 ions . it is interesting to compare the anisotropy expected in the two types of environments adopted by the co@xmath120 ions . as mentioned in section [ struct ] , the co@xmath1201 coordination consists of distorted octahedra , whereas the co@xmath1202 ion is 5-coordinated in a trigonal bipyramid . in the former , the magnetic moment is almost in the basal plane , like already observed in other magnetic structures involving co@xmath120 ions in similar environments.@xcite . the j1 exchange coupling ( see below ) could be responsible for the proposed orientation , thus implying that this coupling is stronger than the in - plane anisotropy of the co@xmath121 ions . nevertheless , our neutron data can not exclude that the magnetic moments lie exactly in the basal plane . in the latter , co@xmath1202 , the anisotropy can be stronger @xcite than in the octahedral coordination : it has been observed to be even larger when the co@xmath120 ion is out of the basal plane , which is the case here ( see figure [ fig_coordination ] ) , and to be of order 25 k @xcite . 1 site , right : co@xmath1202 site . both cations contain one hydroxide group among their ligand atoms.,width=283 ] we examine now the exchange paths that appear to be relevant for the understanding of the intrachain magnetic structure . their estimation is deduced from both the goodenough - kanamori rules@xcite , and the values reported in cobalt - based magnetic systems with similar exchange networks @xcite . they are summarized in both table [ tablej ] and figure [ fig.2 ] . the strongest magnetic coupling j1 is antiferromagnetic and induces the collinear antiferromagnetic arrangement of the co@xmath48 sublattice . the j2 ( weak ferromagnetic ) and j3 ( weak antiferromagnetic ) interactions determine the relative orientation between the co@xmath1201 and co@xmath1202 sublattices . the strong canting of the co@xmath122 ion with respect to the co@xmath1201 ion demonstrates the importance of the single - ion anisotropy of this ion . in that sense , given the anisotropy of the ions , the experimentally deduced magnetic structure satisfies the three interactions j1 , j2 and j3 . the j4 coupling , which is expected to be weakly antiferromagnetic , is frustrated in the obtained magnetic structure , thus suggesting a smaller coupling compared with j2 and j3 . * 5c exchange & exchange & sharing & exchange & estimated + constant & path&type&type & value + j1 & co1a - oh - co1b & corner & af & [ -20,-40 ] k + j2 & co1-o - co2 & edge & f & [ 10 , 30 ] k + j3 & co1-oh - co2 & corner & af & [ -15,-25 ] k + j4 & co1-o - co2 & corner & af & [ -15,-25 ] k + given the structural characteristics of co^ii^@xmath0(oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5 presented in section [ struct ] , the 3d magnetic lro in this material can only be explained by interchain dipolar interactions . the energy of the dipolar interaction between two spins @xmath123 and @xmath124 separated by a distance @xmath125 is @xmath126 \label{dipolar}.\ ] ] in the present case , the interchain dipolar interaction is typically a few millikelvins , and implies a ratio of about 10@xmath127 between the intrachain and interchain interactions . in such a system with a strong 1d character , the critical temperature describing lro originates from the combination of the strong intrachain correlations , resulting in spin blocks exhibiting a super - spin ( c.f . uncompensated component of the magnetic structure ) , and the interchain interactions between these super - spins @xcite . in the present material , this picture is most readily applied along the crystallographic * c-axis which corresponds to the shortest interchain distance . in that direction the dipolar interaction stabilizes an interchain ferromagnetic arrangement between the intrachain super - spins . however , the complexity of the magnetic structure makes this description much more complicated for the @xmath128 direction . the reproduction of the magnetic motif observed in this direction is not favourable when considering the orientation of the super - spins emerging from the ferromagnetic component . the dipolar interactions between individual spins have thus to be considered along a-axis . preliminary calculations of the dipolar energy suggest that the obtained orientation is more energetically favourable with respect to the dipolar interaction than its 180@xmath129 counterpart . in lro spin chains , very strong correlations exist within the chains , while the spin chains are weakly coupled to each other . the result is that when the temperature is larger than the interchain interactions ( of a few millikelvins in our case ) , the dynamics are governed by intrachain dynamics , which can be described in terms of domain - walls motion . this was observed in several ordered spin chains @xcite : dependent on the ratio between the local anisotropy and exchange interactions , the dynamics can be described in terms of single - chain magnets or motion of extended domain - walls . in co^ii^@xmath0(oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5 , the presence of two co@xmath120 sites with a different single - ion anisotropy , combined with the existence of several exchange couplings , makes the description in terms of domain - walls complicated . to get a better understanding , we return to the striking result of the analysis of the dynamics : the ac susceptibility ( i.e zero - field ) dynamics is accounted for by an energy barrier of @xmath84 k over a large range of temperature , whereas the relaxation of the magnetization ( i.e. from a finite magnetization state ) can be described by a barrier of half value @xmath106 k. to go further , we will have to keep in mind that the magnetic structure is only canted by the field applied during measurements of the magnetization relaxation ( see section [ neut_field ] ) . the zero - field dynamics measured by ac - susceptibility can be ascribed to the motion of domain - walls along the chains . due to the strong anisotropy , we can expect that these domain - walls are relatively narrow . then the observed energy barrier would be due to the cost in energy ( exchange and anisotropy ) to move the wall at the domain boundary . the ranges of the exchange constants and anisotropy ( see section [ discussion1 ] ) is consistent with the obtained energy barrier . we recall that ( i ) magnetization relaxation measurements were done after a field of -3500 oe was applied , and that ( ii ) the sample being a powder , the field acts on the three directions of the crystal . moreover , as already mentioned , a fast increase of the magnetization is observed when a field is applied . this can be attributed to a selection of domains with respect to the resulting spin component in the ferromagnetic - like direction , the @xmath130 direction . in the other directions , or at larger fields , the magnetization increases smoothly by spin canting . because of the strong anisotropy of the co@xmath1202 ions , this induced magnetization can be assigned to the co@xmath1201 ions . this canting is effective only along the directions a * and * c * , since the co@xmath1201 ions are oriented along the * b-axis . when the field is removed , the return to equilibrium , that is to say , the relaxation of the magnetization , will consist of two processes : the reorientation of the co@xmath1201 ions and the creation of domains to recover the zero magnetization . it is only the second process which needs to overcome a barrier by nucleation of domain - walls inside the chains . the obtained @xmath117 value should be associated with this nucleation . the question that immediately arises is why @xmath117 is half of @xmath107 . a possible reason could be that the nucleation occurs at the boundary of a chain ( or near a defect ) , where it is more energetically favourable , thus creating only one domain - wall . on the contrary , the dynamics probed by ac susceptibility concern existing domains where the two boundaries of the domains are moving in response to the ac - field ( because the magnetization has to remain zero ) . the obtained characteristic relaxation time @xmath108 is three orders of magnitude larger in magnetization measurements than in ac susceptibility measurements , suggesting that the nucleation process is much slower . to get a deeper insight in these dynamics , and especially the field induced dynamics , further measurements on single crystals are needed . these would allow us to determine if an anisotropy exists in the dynamics ( which should be the case in our scenario ) , and to provide a more quantitative analysis of these observations . in conclusion , we have determined the magnetic structure of co^ii^@xmath0(oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5 . it is made of spin - chains involving two sublattices : one is purely antiferromagnetic , the second gives rise to a ferromagnetic component along the c-axis . due to the magnetic propagation vector being * _ k_*=(0,0,0 ) , the magnetic motif is reproduced in the neighboring unit cells forming the three dimensional lattice . the magnetic structure at low temperature is not modified upon the application of a magnetic field , at least up to 2 t. only canting within the chains is observed . furthermore , the applied field reinforces the interchain correlations resulting in a higher ordering temperature which increases up to 13 k at 2 t ( to be compared to the transition temperature @xmath7 k ) . at very low temperature , in the ordered phase , slow dynamics are observed in a large temperature / frequency range which are ascribed to dynamics of domains inside the chains . interestingly , these dynamics are extremely sensitive to the presence of a magnetic field . finally , we notice that the nature of the present material , comprising inorganic chains interconnected through organic linkers , is reminiscent of other metal - organic frameworks that have shown high intrinsic flexibility in response to high pressures @xcite or to the absorption of guest molecules @xcite . this point is particularly interesting as it might be possible to tune the organization of the chains relative to each other and thus the magnetic ground state . for this purpose , powder x - ray diffraction experiments under back pressure of co@xmath1 ( up to 100 bars ) and h@xmath1 ( up to 200 bars ) were performed . however , the crystal structure of co^ii^@xmath0(oh)@xmath1(c@xmath2@xmath3h@xmath2@xmath4o@xmath0)@xmath5did not seem to be affected . experiments under higher pressure , using a diamond anvil cell , will be the scope of future experiments . c. paulsen is warmly acknowledged for the use of his magnetometers . the authors thank the institut laue langevin , grenoble , france for provision of the research facilities . t. hansen and a. daramsy are acknowledged for their assistance during the d20 experiments . thanks mesr for his ph.d . funding and j. s. white for reading of the article . we acknowledge the support of the european community - research infrastructures under the fp7 capacities specific programme , microkelvin project number 228464 . a. k. cheetham , c. n. r. rao , and r. k. feller , chem ( cambridge ) 2006 * , 4780 . j. r. long , and o. m. yaghi , chem . soc . rev . * 38 * , 1213 ( 2009 ) . g. frey , chem . rev . * 37 * , 191 ( 2008 ) . a. u. czaja , n. trukhan , and u. mller , chem . soc . rev . * 38 * , 1284 ( 2009 ) . h. li , m. eddaoudi , m. okeeffe , and o. m. yaghi , nature , bf 402 , 276 ( 1999 ) . g. frey , c. mellot - draznieks , c. serre , f. milange , j. dutour , s. surbl , and i. margiolaki , science * 5743 * , 2040 ( 2005 ) . l. j. murray , m. dinc , and j. r. long , chem . soc . rev . * 38 * , 1294 ( 2009 ) . li , r. j. kuppler , and h - c . zhou , chem . soc . rev . * 38 * , 1477 ( 2009 ) . j. y. lee , o. k. farha , j. roberts , k. a. scheidt , s. t. nguyen , and j. t. hupp , chem . soc . rev . * 38 * , 1450 ( 2009 ) . m. d. allendorf , c. a. bauer , r. k. bhakta , and r. j. t. houk , chem . soc . rev . * 38 * , 1330 ( 2009 ) . r. sibille , t. mazet , b. malaman , and m. franois , chem . j. * 18 * , 12970 ( 2012 ) . m. kurmoo , chem . soc . rev . * 38 * , 1353 ( 2009 ) . e. ressouche , j. x. boucherle , b. gillon , p. rey , and j. schweizer , j. am . soc . * 115 * , 3610 ( 1993 ) . r. feyerherm , a. loose , m. a. lawandy , and j. li , j. phys . solids * 63 * , 71 ( 2002 ) . r. feyerherm , a. loose , m. a. lawandy , and j. li , appl . a : mater . sci . process . * 74 * , s778 ( 2002 ) . j. l. manson , h. n. bordallo , j. w. lynn , q. huang , r. feyerherm , a. loose , l. chapon , and d. n. argyriou , appl . a : mater . sci . process . * 74 * , s722 ( 2002 ) . r. feyerherm , a. loose , p. rabu , and m. drillon , solid state sci . * 5 * , 321 ( 2003 ) . a. zheludev , b. grenier , e. ressouche , l. p. regnault , z. honda , and k. katsumata , phys . b * 71 * , 104418 ( 2005 ) . e. lhotel , v. simonet , e. ressouche , b. canals , d. b. amabilino , c. sporer , d. luneau , j. veciana , and c. paulsen , phys . b * 75 * , 104429 ( 2007 ) . r. a. mole , j. a. stride , t. unruh , and p. t. wood , j. phys . matter * 21 * , 076003 ( 2009 ) . j. saines , j. r. hester , and a. k. cheetham , phys . b * 82 * , 144435 ( 2010 ) . a. mesbah , r. sibille , t. mazet , b. malaman , s. lebegue , and m. franois , j. mater . chem . * 20 * , 9386 ( 2010 ) . a. mesbah , b. malaman , t. mazet , r. sibille , and m. franois , crystengcomm . * 12 * , 3126 ( 2010 ) . r. a. mole , j. a. stride , p. f. henry , m. hoelzel , a. senyshyn , a. alberola , c. j. gmez garcia , p. r. raithby , and p. t. wood , inorg . 50 * , 2246 ( 2011 ) . p. j. saines , h. h. m. yeung , j. r. hester , a. r. lennie , and a. k. cheetham , dalton trans . * 40 * , 6401 ( 2011 ) . o. fabelo , l. candillas - delgado , i. puente orench , j. a. rodriguez - velamaza , j. campo , and j. rodriguez - carvajal , inorg . chem . * 50 * , 7129 ( 2011 ) . o. fabelo , l. candillas - delgado , j. pasan , p. diz - gallifa , c. ruiz - perez , f. lloret , m. julve , i. puente orench , j. campo , and j. rodriguez - carvajal , inorg . chem . * 52 * , 12818 ( 2013 ) . g. nnert , h - j . koo , c. v. colin , e. m. bauer , c. bellitto , c. ritter , g. righini , and m - h . whangbo , inorg . chem . * 52 * , 753 ( 2013 ) . r. sibille , a. mesbah , t. mazet , b. malaman , s. capelli , and m. franois , j. sol . state chem . * 186 * , 134 ( 2012 ) . r. sibille , t. mazet , b. malaman , t. gaudisson , and m. franois , inorg . chem . * 51 * , 2885 ( 2012 ) . r. georges , j. j. borrs - almenar , e. coronado , curly , and m. drillon , _ one - dimensional magnetism : an overview of the models _ , in magnetism : molecules to materials i : models and experiments ( eds j. s. miller and m. drillon ) , wiley - vch verlag gmbh & co. kgaa , weinheim , frg , 2002 , p. 1 . _ order and disorder in quantum spin systems _ , edited by t. tonegawa , s. miyashita , and h. takayama , suppl . 145 ( progress of theoretical physics , japan , 2002 ) . f. d. m. haldane , phys . * 45 * , 1358 ( 1980 ) . f. d. m. haldane , phys . b * 25 * , 4925 ( 1982 ) . h. a. bethe , z. phys . * 71 * , 205 ( 1931 ) . g. gomez - santos , phys . b * 41 * , 6788 ( 1990 ) . y. uchiyama , y. sasago , i. tsukada , k. uchinokura , a. zheludev , t. hayashi , n. miura , and p. bni , phys . lett . * 83 * , 632 ( 1999 ) . a. caneschi , d. gatteschi , n. lalioti , c. sangregorio , r. sessoli , g. venturi , a. vindigni , a. rettori , m. g. pini , and m. a. novak , angew . , int . ed . * 40 * , 1760 ( 2001 ) . c. coulon , h. miyasaka , and r. clrac , struct . bonding ( berlin ) * 122 * , 163 ( 2006 ) . d. gatteschi , and a. vindingni , _ single - chain magnets _ , in molecular magnets : physics and applications ( eds j. bartolom , l. fernando , and j. f. fernandez ) , springer , 2014 , p. 191 . j. saines , p. jain , and a. k. cheetham , chem . * 2 * , 1929 ( 2011 ) . r. sibille . _ organic - inorganic hybrid solids based on dicarboxylate molecules and 3d or 4f elements : relations between structure and magnetic properties , dimensionality effects _ , thse . vandoeuvre - ls - nancy : universit de lorraine , * 2012. http://docnum.univ-lorraine.fr/public/ddoc_t_2012_0163_sibille.pdf r. j. glauber , j. math . phys . 4 * , 294 ( 1963 ) . j. rodriguez - carvajal , physica b * 192 * , 55 ( 1993 ) . l. b. mccusker , r. b. von dreele , d. e. cox , d. loer , and p. scardi , j. appl . crystallogr . * 32 * , 36 ( 1999 ) . a. le bail , h. duroy , and j. l. fourquet , mater . * 23 * , 447 ( 1988 ) . e. f. bertaut , in _ magnetism _ , edited by g. t. rado , and h. a. shul , academic : newyork , 1963 , vol . iii , chap . j. rodriguez - carvajal , _ basireps : a program for calculating irreducible representation of little groups and basis functions of polar and axial vector properties . _ part of the fullprof suite of programs , 2007 available at http://www.ill.eu/sites/fullprof ; c. ritter , solid state phenom . * 170 * , 263 ( 2011 ) . c. paulsen , in _ introduction to physical techniques in molecular magnetism : structural and macroscopic techniques - yesa 1999 _ , edited by f. palacio , e. ressouche , and j. schweizer , servicio de publicaciones de la universidad de zaragoza , zaragoza , 2001 , p. 1 . o. v. kovalev , _ representation of the crystallographic space groups , irreducible representations , induced representations , and corepresentations _ , 2nd ed . ( gordon and breach science , amsterdam , 1993 ) . j .- g . park and c. paulsen , physics * 6 * , 55 ( 2013 ) . e. lhotel , d. b. amabilino , c. sporer , d. luneau , j. veciana and c. paulsen , phys . b * 77 * , 064416 ( 2008 ) . e. lhotel , e. n. khatsko and c. paulsen , phys . b * 74 * , 020402 ( r ) ( 2006 ) . m. hagiwara , j. magn . magn . mater . * 177 - 181 * , 89 ( 1998 ) . d. huser , a. j. van duyneveldt , g. j. nieuwenhuys , and j. a. mydosh , j. phys . c * 19 * , 3697 ( 1986 ) . e. lhotel , v. simonet , j. ortloff , b. canals , c. paulsen , e. suard , t. hansen , d. j. price , p. t. wood , a. k. powell , and r. ballou , eur . j. b * 86 * , 248 ( 2013 ) . ideally , the sample should be heated over @xmath7 k , but the working conditions of the dilution fridge prevents from heating to such a high temperature . j. j. prjean and j. souletie , j. physique * 41 * , 1335 ( 1980 ) . j. b. goodenough , _ magnetism and chemical bond _ ( john wiley and sons publ . , new york,1963 ) . e. j. kinast , c. a. dos santos , d. schmitt , o. isnard , m. a. gusmao and j. b. m. da cunha , j. alloys comp . * 491 * , 41 ( 2010 ) . m. c. viola , m. j. martinez - lope , j. a. alonso , j. l. martinez , j. m. de paoli , s. pagola , j. c. pedregosa , m. t. fernandez - diaz , and r. e. carbonio , chem . mat . * 15 * , 1655 ( 2003 ) . m. e. foglio , m. c. dos santos , g. e. barberis , j. m. rojo , j. l. mesa , l. lezama , and t. rojo , j. phys . condens . matter * 14 * , 2025 ( 2002 ) . l. robertson , m. duttine , m. gaudon , and a. demourgues , chem . mat . * 23 * , 2419 ( 2011 ) . t. jurca , a. farghal , p. h. lin , i. korobkov , m. murugesu , and d. s. richeson , j. am . soc . * 133 * , 15814 ( 2011 ) . m. drillon , and p. panissod , j. magn . mater . * 188 * , 93 ( 1998 ) . s. ostrovsky , w. haase , m. drillon , and p. panissod , phys . b * 64 * , 134418 ( 2001 ) . c. coulon , r. clrac , w. wernsdorfer , t. colin , and h. miyasaka , phys . rev . lett . * 102 * , 167204 ( 2009 ) . a. u. ortiz , a. boutin , a. h. fuchs , and f. x. coudert , phys . lett . * 109 * , 195502 ( 2012 ) . k. j. gagnon , c. m. beavers , and a. clearfield , j. am . soc . * 135 * , 1252 ( 2013 ) . s. h. lapidus , g. j. halder , p. j. chupas , and k. w. chapman , j. am . soc . * 135 * , 7621 ( 2013 ) . c. serre , f. millange , c. thouvenot , m. nogues , g. marsolier , d. lour , and g. frey , j. am . soc . * 124 * , 13519 ( 2002 ) .
massive data from online social media and online social networking ( osn ) enables accurate analysis of the human behavior and of the interaction times between individuals through technology such as in word - of - mouth marketing , opportunistic networking and viral spreading . empirical measurements contradict the commonly assumed exponential distribution of poissonian inter - event times @xcite in spreading processes or epidemic models . multiple publications @xcite report that _ the interactivity time _ , defined as the duration between two consecutive tasks like sending emails , accessing web pages , instant messaging and phone calls , follow power - law distributions . these findings recently led to non - markovian analyses , addressed in the work of cator _ et al . _ @xcite , iribarren and moro @xcite , van mieghem and van de bovenkamp @xcite and schweizer _ et al . given that the inter - activity distributions are heavy tailed , word - of - mouth spreading , viral infections or dynamics of memes are expected to endure or survive longer , compared to basic markovian models @xcite . barabasi @xcite infers that heavy - tailed distributions may arise from a priority queue , where individuals execute tasks of which the majority can be completed in short time , but some tasks wait long due to a perceived priority . barabasi s priority queue model fits the distribution of durations between events quite well , leading to a power - law distribution with an exponent @xmath0 around 1 . a power - law random variable @xmath1 has the probability density function @xmath2 where @xmath3 and @xmath4 is the lower bound for @xmath5 . the probability density function ( pdf ) of a lognormal random variable @xmath5 for @xmath6 is @xmath7 } { \sigma t\sqrt{2\pi } } \label{pdf_lognormal}\ ] ] where @xmath8 are called the parameters of the lognormal pdf , that are the mean and variance of @xmath9 as shown in appendix [ sec_properties_lognormal ] . when we assume that inter - event durations are power - law distributed , we encounter the following issues : 1 . in many cases , only a part of the data ( the tail larger than @xmath10 ) is modeled by a power - law . the lower bound @xmath10 in ( [ pdf_powerlaw ] ) does not correspond to the physical minimum of the random variable @xmath5 , but @xmath10 is fitted from the data by ignoring smaller values that do not obey the power - law . often , these smaller values may have a large probability to occur , so that their neglect is difficult to justify . in other words , only a part of the process ( above @xmath10 ) is modeled by a power - law ( [ pdf_powerlaw ] ) , while the other part ( below @xmath10 ) is not . most processes or measurements possess both a lower as well as an upper bound . apart from the lower bound @xmath10 , an upper bound @xmath11 is often invoked , at which a cut - off is observed : the power - law behavior is confined to the range @xmath12 , although @xmath13 and @xmath14 . however , it is often unclear whether the process in the deep tail still obeys a power - law distribution or some other , much faster decreasing distribution . the upper bound @xmath11 is usually empirically determined , rather than based on the physical maximum of @xmath5 . as long as @xmath15=c\int_{\kappa}^{\infty}t^{-\gamma}dt=\left ( \frac{\kappa}{\tau}\right ) ^{1-\gamma}\ ] ] is small ( with respect to the measurement precision ) , the upper bound @xmath11 is justified , else other validation arguments are needed . 3 . we demonstrate here that the binning of data ( either by the data - provider or by the researcher ) alters the shape of a lognormal distribution into an apparent power - law . particularly in relation to human activities or behaviors , we question in this paper the widely assumed power - law distribution . we present measurements of inter - event durations from digg.com and reddit.com , two online social news aggregators @xcite , and from the enron data set , a collection of emails sent by employees of the company enron , and argue that a lognormal distribution is a valid candidate for the distribution of human inter - event durations in section [ sec_measurements ] . the problem of fitting a lognormal is explained in section [ sec_fittinglognormal ] , followed by previously reported lognormally distributed data sets in section [ sec_reported_parameters ] . existing theoretical models are compared in section [ sec_debate ] , a plausible interpretation for lognormal human behavior is proposed in section [ sec_plausible_explanation_lognormal_interactivity_times ] and section [ sec_conclusion ] concludes . mathematical properties of the lognormal distribution are deferred to the appendix [ sec_properties_lognormal ] . appendix [ sec_likelihood_testing ] presents results of likelihood tests for the observed distributions . all events in an osn are based on users activities , such as posts , friend - requests and comments . a complete data set including all activities of users from digg.com for a duration of 4 years , described in tang _ _ @xcite and doerr _ et al . _ @xcite , allows us to analyze the time frame in which users of digg.com add their friends . _ @xcite found that reaction times in a retweet network from twitter and digg are close to a lognormal distribution with parameters @xmath16 and @xmath17 . the process of adding friends shows bursts of activity also observed and analyzed in email communication @xcite . in these publications the question arises , whether the observed distribution of inter - activity times is described by a power - law , lognormal or a cascading poisson process . @xcite describe that circadian and weekly activity cycles in human behavior are the factors that lead to heavy tails . they proposed a cascading poisson process , consisting of a _ nonhomogeneous _ poisson process that reflects periodicity and a _ homogeneous _ poisson process describing active intervals , which is shown to model the interactivity distributions of e - mail communication . because the network of digg.com is directed ( like in twitter.com or other osns ) , a user can be followed by other users to become their friend , while a user can not add followers . this means that the process of adding friends is solely based on the user himself , whereas obtaining followers depends on the activities of other users . the random variable @xmath18 denotes the time between the addition of two friends and , similarly , @xmath19 is the time between receiving two followers . figures [ fig_interfriendaddtime_seconds ] and [ fig_binsize1s_followers ] depict the distribution of durations between adding friends @xmath18 and of receiving followers @xmath19 for 7.4 million friendship relations in digg . fitting the data , i.e. all realizations of @xmath18 , by the state of the art technique by clauset _ _ @xcite to a power - law ( [ pdf_powerlaw ] ) results in an exponent @xmath20 for @xmath21 , whereas all realizations of @xmath19 are fitted best by a lognormal ( [ pdf_lognormal ] ) with parameters @xmath22 and @xmath23 . for @xmath19 an estimated p - value of 0.0 indicates that a power - law fit provides not the best solution , whereas the p - value for @xmath18 of @xmath24 indicates a reasonable fit for a power - law distribution . the reason that the distribution of @xmath19 does not fit the lognormal distribution over the whole range lies in the nature of digg.com . @xcite showed that only a few users were active over a long period , while just a fraction of them was actually submitting stories . only 2% of all registered users succeeded to have their submissions promoted to the frontpage @xcite . since the username of a submitter appears next to the story , these users receive a lot of followers during the relatively small period that their story was listed on the front page . therefore , @xmath25 is large for small @xmath26 in fig . [ fig_binsize1s_followers ] . in summary , even ignoring the small time values , the random variable @xmath18 and @xmath19 possess different distributions . the process that generates @xmath18 is more likely described by a variant of barabasi s priority queue model , leading to power - law behavior . since the random variable @xmath19 is generated by the collective dynamics of different individuals ( that are likely weakly dependent ) , central limit law arguments may point towards the lognormal distribution @xcite . similar properties occur in sending and receiving emails . obviously , a user can only send an email when he is online , whereas emails arrive at a user s inbox at his email server at all times . we analyzed the durations between receiving and sending emails in the enron data set , which contains emails of all employees of enron during roughly 6 years , starting in january 1998 until february 2004 . similar distributions arise as shown later in figs . [ fig_enron_sent_diff ] and [ fig_enron_received_diff ] . a third data set from reddit.com , an osn in which users mainly submit , comment or vote on bookmarks . the pdf of the duration @xmath27 between consecutive comments and submissions in reddit are shown in fig . [ fig_reddit_comment_and_submission ] . the tails of the distributions in fig . [ fig_reddit_comment_and_submission ] seem nicely fitted by a power - law distribution with exponents @xmath28 . however , as mentioned in the introduction , the increasing regime in @xmath29 for small values of @xmath30 , the peak of @xmath31 nor the concave form of the pdf can be modeled by a power - law distribution . the two main approaches are based on the pdf and the edf ( empirical distribution function ) , after a logarithmic transformation of the data . figure [ fig_ccdf_seconds_normal_followers ] depicts the data fitted to the distribution function of a normal distribution ( [ distribution_lognormal ] ) and shows that the parameters of the edf are about the same as those in fig . [ fig_binsize1s_followers ] ( pdf approach ) . ( same data as in fig . [ fig_binsize1s_followers ] ) . ] we generate @xcite @xmath32 realizations of a lognormal random variable @xmath5 , which we denote by the set @xmath33 @xmath34 . all @xmath32 realizations lie in the interval @xmath35 $ ] , where the minimum value is @xmath36 and the maximum value is @xmath37 . the binning operation @xcite divides the entire data interval @xmath38 $ ] into @xmath39 sub - intervals of length @xmath40 and the @xmath41-th subinterval @xmath42 $ ] for @xmath43 is associated with a bin @xmath44 , that contains the number of realizations of @xmath5 or data points of set @xmath45 lying within the @xmath41-th subinterval,@xmath46 clearly , for a given set @xmath45 , increasing the binsize @xmath47 decreases the number of bins @xmath39 . the effect of binning on the pdf @xmath48 is depicted in fig . [ fig_different_binning_m10_s2_lognormnoise ] , where @xmath49 realizations from a lognormal random variable with parameters @xmath50 and @xmath51 are drawn . as demonstrated in appendix [ sec_properties_lognormal ] , the scaled random variable @xmath52 has parameter @xmath53 , but @xmath54 , implying that binning only changes the parameter @xmath55 , but leaves the parameter @xmath56 invariant ! by binning ( scaling ) the distribution with different binsizes , the up - going regime , where @xmath48 increases with @xmath30 , disappears and the observable part of the distribution evolves towards a straight line on a log - log plot . binning the data with larger binsizes decreases the parameter @xmath55 , even to the extent that @xmath55 may become negative . if @xmath57 decreases , the maximum of the lognormal at @xmath58 tends to zero ( infinitely far to the left on a log - log scale ) . therefore , just the decreasing part of the quadratic shape of ( [ pdf_lognormal ] ) will be visible in a log - log plot . binned per minute . ] binning the data in different time units , say per minute instead of per second , scales the distribution by a factor of 60 ( since 60 seconds equals 1 minute ) , which will shift the parameter @xmath59 of the cdf towards the left to @xmath60 . however , a remarkable property of the lognormal distribution is that the parameter @xmath56 will not change after linear scaling as shown in the appendix [ sec_properties_lognormal ] . binned per hour . ] an important consequence of a binning operation is illustrated in the pdfs in fig . [ fig_pdf_minutes_lognormal_followers ] and fig . [ fig_pdf_hours_lognormal_followers ] , that show the pdf of @xmath19 binned per minute and per hour , respectively . conforming to the theory , the parameter @xmath55 decreases by a factor of about @xmath61 , while the parameter @xmath56 keeps its value , but the shape of the distribution changes . the distribution , binned per hour , is shown in fig . [ fig_pdf_hours_lognormal_followers ] , where the distribution can be confounded with a power - law distribution ( with exponential cut - off ) . moreover , this misinterpretation may be justified , because the exponent @xmath62 of a power - law fitted to the hourly binned data is close to the exponent @xmath20 found for the @xmath18 distribution in fig . [ fig_interfriendaddtime_seconds ] . [ fig_pdf_hours_lognormal_followers_exponential ] shows the complementary cumulative distribution function of @xmath19 binned per hour , fitted using the method of clauset _ et al . _ @xcite . the resulting power - law exponent in fig . [ fig_pdf_hours_lognormal_followers_exponential ] is @xmath63 . the visualization in fig . [ fig_pdf_hours_lognormal_followers_exponential ] is misleading because the original distribution is , as shown earlier , a lognormal distribution . where data is binned per hour . ] for completeness we like to refer to table [ tab : estimatedparametersusingthemethodofclausetetal ] in appendix [ sec_likelihood_testing ] , where we list fitted parameters , p values and the result of log - likelihood tests , conducted using the method of virkar and clauset @xcite . as shown in the table , the larger the used binsizes , the higher the p values indicating that a power - law fits the data . for small binsizes ( low p values ) the results are not conclusive or rather in favor of a lognormal distribution . to further demonstrate the effect , figs . [ fig_pdf_minutes_lognormal_followers ] , [ fig_pdf_hours_lognormal_followers ] and [ fig_pdf_hours_lognormal_followers_exponential ] depict distributions in which the data was artificially binned by minutes and hours . in reality , such binning operation might occur if the data provider ( the osn or web service ) returns values per hour or even larger scales . in other words , data measured in larger time steps over a certain total time range thus erases the possibility to distinguish between a lognormal and power - law distribution . the sending time of an email in the enron data set until may 2001 was stored per minute , whereas afterwards ( june 2001 - february 2004 ) the sending time is stored with an accuracy of a second . the resulting pdf of time differences between sent emails is plotted in fig . [ fig_enron_sent_diff ] , where the black and red dots represent the measurements per second and per minute , respectively . the parameters of the fitted lognormal ( @xmath64 ) and power - law distribution @xmath65 ) correspond to those of digg.com in fig . [ fig_binsize1s_followers ] and fig . [ fig_interfriendaddtime_seconds ] . the pdf of inter - arrival times of received emails , split in observations per minute ( red ) and per second ( black ) in fig . [ fig_enron_received_diff ] , can be compared to the durations of added followers , where the parameters are again in the same range as shown before . we found that the edf of the data ( not shown ) agrees with the pdfs in fig . [ fig_enron_sent_diff ] and fig . [ fig_enron_received_diff ] , underlining the likeliness that a lognormal random variable models the data better than a power - law . table [ tab : powerlawexponentsfoundinhumandynamics ] lists inter - arrival times and the fitted exponents found in publications in which the aspect of binning was ignored . rather small power - law exponents @xmath0 between @xmath66 and @xmath67 are found in multiple data sets of human activity . [ c]lll data - set & exponent ( @xmath0 ) & reference + time between sending emails & 1 & barabasi ( 2005 ) @xcite + time between sending emails & 1.2 & eckmann _ et al . _ ( 2004 ) @xcite + time between clicks in website usage & 1 , 1.25 & gonalves and ramasco ( 2008 ) @xcite + time between similar actions in web data & 1.1 , 1.2 , 1.8 & radicchi ( 2008 ) @xcite + time between messages in instant - messaging & 1.53 & leskovec and horvitz ( 2008 ) @xcite + time between phone calls & 0.9 & candia _ et al . _ ( 2008 ) @xcite + time between phone calls & 0.7 & karsai _ et al . _ ( 2012 ) @xcite + time between sending emails & 1 & karsai _ et al . _ ( 2012 ) @xcite + time between sending short messages & 0.7 & karsai _ et al . _ ( 2012 ) @xcite lognormal distributions were reported for similar human activity as shown in table [ tab : addlabel ] , which extends the collection of limpert _ et al . the oldest analysis was conducted by boag @xcite in 1949 . based on the scaling invariance of @xmath56 as demonstrated in appendix [ sec_properties_lognormal ] , the parameter @xmath56 can be compared over different measurements , whereas the parameter @xmath55 can not , since @xmath55 depends on the units in which the lognormal random variable is measured . interestingly , all @xmath56 s in table [ tab : addlabel ] lie within an amazingly small range of @xmath68 , which shows that the parameter @xmath56 only varies over about one order of magnitude in different measured phenomena . if @xmath56 is rather small as in our measurements where @xmath69 , the first term in ( [ loglogform ] ) dominates and a quadratic function appears in a log - log plot . consequently , these rather small values of @xmath56 and those in table [ tab : addlabel ] contradict the common deductions made from ( [ loglogform ] ) in appendix [ sec_properties_lognormal ] , namely that only for large values of @xmath56 power - law and lognormal distributions are indistinguishable . equations ( [ pdf_lognormal_power_law_like_form ] ) and ( [ power_law_exponent_lognormal ] ) in appendix [ sec_properties_lognormal ] explain why the lognormal distribution may seem linear in a certain regime . [ c]llll @xmath55 & @xmath56 & process & reference + 5.547 & 2.126 & email forwarding & iribarren and moro @xcite + @xmath70 & @xmath71 & email forwarding & stouffer _ et al . _ @xcite + @xmath72 hour & & & + @xmath73 days & & email forwarding & stouffer _ et al . _ @xcite + 2.47 & 0.38 & infection times & nishiura @xcite + 2.47 & 0.36 & latency periods of diseases & limpert @xcite + 14 days & 1.14 & latency periods of diseases & sartwell @xcite + 100 days & 1.24 & latency periods of diseases & sartwell @xcite + 2.3 hours & 1.48 & latency periods of diseases & sartwell @xcite + 2.4 days & 1.47 & latency periods of diseases & sartwell @xcite + 12.6 days & 1.50 & latency periods of diseases & sartwell @xcite + 21.4 days & 2.11 & latency periods of diseases & sartwell @xcite + 9.6 months & 2.5 & survival times after cancer diagnosis & boag @xcite + 15.9 monts & 2.8 & survival times after cancer diagnosis & feinleib and macmahon @xcite + 17.2 months & 3.21 & survival times after cancer diagnosis & feinleib and macmahon @xcite + 14.5 months & 3.02 & survival times after cancer diagnosis & boag @xcite + 60 years & 1.16 & age of onset of alzheimer & horner @xcite + 4 days & & incubation periods ( viral infections ) & lessler _ et al . _ @xcite + 3 to 5 & @xmath71 & task completion & linden @xcite + 0.5 & 1.4 & strike duration & lawrence @xcite + & & time of individual activities & mohana _ et al . _ @xcite + 0.43 & 1.634 & call duration & spedalieri _ et al . _ @xcite + 3.5 & 0.70 & message holding time & barcelo and jordn @xcite + 7.439 & 0.846 & transmission holding time & barcelo and jordn @xcite + 3.29 & 0.890 & channel holding time & barcelo and jordn @xcite + 3.3 & 0.89 & channel holding time & barcelo and jordn @xcite + @xmath74 & @xmath75 & & + @xmath76 & @xmath77 & call holding time & bolotin @xcite + & & citations & eom and fortunato @xcite + & & citations & redner @xcite + 1 to 2 & @xmath78 & citations & stringer _ _ et al.__@xcite + & 1.095 & citations & radicchi _ _ et al.__@xcite + @xmath79 & @xmath80 & & + @xmath81 & @xmath82 & retweeting behavior & doerr _ _ et al.__@xcite + 5.29 & 0.42 & distribution of votes on pages of digg.com & van mieghem _ @xcite + & & & mitchell @xcite discusses the fascinating and abundant appearance of power - law distributions and mentions processes such as self - criticality , that are related to phase transitions , as the main producers of power - laws , but she remarks that the precise nature as well as the deviations from a power - law are still open to debate . table [ tab : addlabel ] illustrates that the research on lognormal distributions has a long history . unfortunately , also the processes that produce ( almost ) lognormal behavior are not well understood . mitzenmacher @xcite overviews processes leading to power - law and lognormal distributions , emphasizing that minor changes in the process will lead to either a power - law or a lognormal distribution . mitzenmacher mentions the work of gabaix @xcite , who analyzed the size distribution of cities in the united states . interestingly , gabaix @xcite found that the city size distribution follows a zipf distribution , which is a power - law distribution ( [ pdf_powerlaw ] ) with an exponent @xmath83 . gabaix argues that cities can not become infinitely small , a fact that imposes a lower bound to their size . when modeling the size of cities as a markov chain with a fixed number of cities , which grow stochastically as proposed by gibrat @xcite , then the steady state of the markov chain will follow zipf s distribution with an exponent @xmath83 . if there is no lower bound on the city size , so that cities can be arbitrarily small , then the distribution degenerates to a lognormal distribution . gibrat @xcite , whose work is often associated with the law of proportionate effect @xcite , argues in his model that transition probabilities or the variance of transition probabilities in a growth process are independent from the size . gibrat , who estimated the distribution of city sizes , concluded that the law of proportionate growth leads to a distribution that is lognormal . however , simon @xcite showed that gibrat s law of the proportionate effect may lead to other heavy - tailed distributions as well . champernowne @xcite and cordoba @xcite analyzed the income distribution of england and wales with markov theory . again , their crucial assumption is that incomes have a lower bound . eeckhout @xcite analyzed the distribution of city sizes by using accurate data from the us census in 2000 and found that the heavy tail obeys zipf s law . but , the entire distribution is better described by a lognormal than a pareto distribution . by comparing the census data from 1990 and 2000 , eeckhout shows that the growth of a city is independent of its size . the parameters of the lognormal distribution found by eeckhout are @xmath84 and @xmath85 . the difference with the above mentioned markov chain approach lies in the fact that eeckhout @xcite modeled the process by a multiplicative process , which leads to a lognormal distribution . this multiplicative process , proposed by kapteyn @xcite in 1903 for the first time and later coined the law of proportionate effect by gibrat @xcite , is based on the central limit theorem applied to a multiplicative process , which leads to lognormal distributed sizes @xcite . gibrat s growth process is defined as @xmath86 where the size @xmath87 of an item at state @xmath30 depends upon the previous size @xmath88 times a positive , random factor @xmath89 . by taking the logarithm of both sides in ( [ gibrat_governing_eq ] ) and denoting @xmath90 , we obtain , after iteration , @xmath91 by the central limit theorem @xcite , @xmath92 , we arrive at , for large @xmath30,@xmath93\rightarrow\frac{1}{\sqrt{2\pi}}\int_{-\infty}^{\frac{y - t\mu}{\sigma\sqrt{t}}}e^{-u^{2}/2}\;du\ ] ] from which approximately @xmath94 where @xmath55 denotes the mean and @xmath95 the variance of the sequence @xmath96 . for large @xmath30 , a power - law distribution ( [ pdf_powerlaw ] ) tends to zero as @xmath97 , while a lognormal distribution ( [ pdf_lognormal ] ) tends considerably faster as @xmath98 \right ) $ ] to zero , illustrating that the deep tails are significantly different . malevergne @xcite addressed this fact by the concept of a slowly varying function , in particular , for @xmath99 and @xmath100 , the power - law distribution ( [ pdf_powerlaw ] ) features @xmath101 the lognormal distribution ( [ pdf_lognormal ] ) on the other hand is not slowly varying . in the limit @xmath99 , a lognormal distribution will , for @xmath102 , always tend to zero : @xmath103 this different tail behavior questions whether exponential bin sizes should be used in a log - log plot , because these may hide the rapid decrease in the tail . perhaps , a plausible explanation of the appearance of lognormal human interactivity times is that the logarithm @xmath104 of a human interactivity time @xmath27 is gaussian or normally distributed ( see appendix [ sec_properties_lognormal ] ) , most likely as a consequence of the central limit theorem . roughly speaking , the central limit theorem applies for a large number of weakly dependent random variables , where none of them is dominant . thus , rather than concentrating on the time @xmath27 , it seems more natural to focus on the random variable @xmath104 as the decisive quantity . the logarithm of a human related measure often occurs : for example , in our hearing system , the intensity of sound is logarithmically experienced . but more importantly , the large variability in human performance also seems logarithmically distributed @xcite and the measured interactivity times are strongly related to the large differences in human performance and/or behavior . on the other hand , if @xmath105 is an exponential random variable with mean @xmath106 for @xmath107 ( else @xmath108 ) and @xmath109 , then @xmath110 = \pr\left [ y\leq\log t\right ] = 1-e^{-\left ( \gamma-1\right ) \left ( \log t-\log\tau\right ) } = 1-\left ( \frac{t}{\tau } \right ) ^{1-\gamma}\ ] ] so that @xmath111 = \frac{\gamma-1 } { \tau^{1-\gamma}}t^{-\gamma}$ ] is a perfect power - law probability density function as in ( [ pdf_powerlaw ] ) . however , if @xmath104 @xmath112 were exponentially distributed with mean @xmath106 , the memoryless property of the exponential distribution @xcite would indicate that @xmath113 = \pr\left [ \log t\geq t\right]\ ] ] in other words , given that the logarithm of a human interactivity time is larger than @xmath114 time units , the probability that @xmath104 exceeds @xmath115 time units is equal to the probability that @xmath104 exceeds @xmath30 time units , for any @xmath114 and @xmath30 larger than @xmath116 , precisely as if the log - interactivity time @xmath114 never has been spent or waited , which is quite counterintuitive for a duration between consecutive activities . alternatively , with @xmath117 and @xmath118 , the memoryless property of @xmath104 implies scale - freeness in @xmath27 : @xmath119 = \pr\left [ \log t\geq t\right]\ ] ] which shows , ignoring the condition @xmath120 , the independence on the scale @xmath121 . but , we have shown in section [ sec_effect_binning ] that rescaling the human interactivity time ( by different bin sizes ) definitely alters the distribution . from these arguments , we can infer that a power - law time @xmath27 is less defendable than a lognormally distributed @xmath27 . in this paper , the interactivity durations of individuals , between creating friendship relations , writing emails , commenting and voting on online content are analyzed . we found that the distribution of durations to add friends follows a power - law with an exponent of @xmath122 , whereas the durations to acquire followers are well described by a lognormal with @xmath123 and @xmath124 . due to the small probability of executing two tasks in a small time interval ( typically ignored in fitting a power law ) , we claim that a lognormal distribution covers the entire activity time range better than a power distribution . in addition , we show that binning of lognormally distributed data can seriously affect the perception : the parameter @xmath55 shifts towards smaller values , but the parameter @xmath56 of a lognormal distribution does not change after a binning or scaling operation . in the extreme case , only the heavy tail of the lognormal distribution ( [ pdf_lognormal ] ) remains , which follows a power - law distribution ( [ pdf_powerlaw ] ) with an exponent of @xmath0 close to 1 . as explained in the appendix , there exists an interval in which the lognormal distribution is indistinguishable from a power - law distributions with power - law exponent @xmath125 , for small @xmath126 . similar observations and concerns , discussed for a long time in the literature of city size and income distributions and reviewed in section [ sec_debate ] , supports that a lognormal distribution is modeling the whole data range better than power - laws . finally , section [ sec_plausible_explanation_lognormal_interactivity_times ] argues that the logarithm of a quantity associated with human activities rather than the quantity itself is the better descriptor , because human performance seems to fit a lognormal . 99 p. van mieghem . . cambridge university press , 2014 . barabsi . . , 435:207211 , may 2005 . eckmann , e. moses , and d. sergi . entropy of dialogues creates coherent structures in e - mail traffic . , 101(40):1433314337 , 2004 . b. gonalves and j. j. ramasco . . , 78.2026123 , 2008 . f. radicchi . . , 80.026118 , aug 2009 . j. leskovec and e. horvitz . . in _ proceeding of the 17th international conference on world wide web _ , pages 915924 . acm , 2008 . j. candia , m. c. gonzlez , p. wang , t. schoenharl , g. madey , and a .- barabsi . uncovering individual and collective human dynamics from mobile phone records . , 41(22):224015 , 2008 . m. karsai , k. kaski , a .- barabsi , and j. kertsz . . , 2:397 , may 2012 . e. cator , r. van de bovenkamp , and p. van mieghem . susceptible - infected - susceptible epidemics on networks with general infection and cure times . , 87:062816 , jun 2013 . j. l. iribarren and e. moro . . , 103(3):038702 , jul 2009 . p. van mieghem and r. van de bovenkamp . . , 110:108701 , mar 2013 . i. scholtes , n. wider , r. pfitzner , a. garas , c. j. tessone , and f. schweitzer . causality - driven slow - down and speed - up of diffusion in non - markovian temporal networks . september 2014 . m. karsai , m. kivel , r. k. pan , k. kaski , j. kertsz , a. l. barabsi , and j. saramki . . , 83(2):025102 , s. tang , n. blenn , c. doerr , and p. van mieghem . . , 13(5):11631175 , 2011 . c. doerr , n. blenn , s. tang , and p. van mieghem . . in _ computer communications _ , volume 35 , pages 796809 , 2012 . p. van mieghem . human psychology of common appraisal : the reddit score . , 13(6):14041406 , 2011 . r. d. malmgren , d. b. stouffer , a. e. motter and l. a. n. amaral . , 105(47):18153 , 2008 . r. d. malmgren , d. b. stouffer , a. s. l. o. campanharo and l. a. n. amaral on universality in human correspondence activity . , 325(5948):16961700 , 2009 . c. doerr , n. blenn , and p. van mieghem . . , 8(5):e64349 , 05 2013 . a. clauset , c. r. shalizi , and m. e. j. newman . . , 51(4):661703 , e. limpert , w. a. stahel , and m. abbt . log - normal distributions across the sciences : keys and clues . , 51(5):341352 , 2001 . j. w. boag . . , 11(1 ) , 1949 . d. b. stouffer , r. d. malmgren , and l. a. n. amaral . . arxiv : physics/0510216 , 2005 . d. b. stouffer , r. d. malmgren , and l. a. n. amaral . log - normal statistics in e - mail communication patterns , arxiv : physics/0605027 2006 . h. nishiura . . , 2007 p. e. sartwell . . , 51:310318 , 1950 . m. feinleib and b. macmahon . variation in the duration of survival of patients with the chronic leukemias . , 15 , 1960 . r. d. horner . , 126(3):40914 , 1987 . j. lessler , n. g. reich , r. brookmeyer , t. m. perl , k. e. nelson , and d. a. cummings . , 9(5):291300 , may 2009 . w. j. van der linden . . , 31(2 ) , 2006 . r. j. lawrence . . , 147(3 ) , 1984 . s. mohan , m. gopalakrishnan , h. balasubramanian , and a. chandrashekar . . , 58(6):827831 , 2006 . a. spedalieri , i. martn - escalona , and f. barcel . in _ wireless communications and networking conference _ , pages 23812386 . ieee , 2005 . f. barcel and j. jordn . . electronics letters , vol.34 no.2 _ , pages 146147 , 1998 . v. a. bolotin . . , 12:433438 , 1994 . eom and s. fortunato . . , 6(9):e24926 , s. redner . . , 58(6):4954 , 2005 . m. stringer , m. sales - pardo , and l. a. n. amaral . effectiveness of journal ranking schemes as a tool for locating information . , 3(2):e1683 , 2008 . p. van mieghem , n. blenn , and c. doerr . . , 83:251261 , 2011 . 10.1140/epjb / e2011 - 20124 - 0 . m. mitchell , _ complexity : a guidede tour _ , oxford university press , 2009 . m. mitzenmacher . . , 1(2 ) , 2003 . x. gabaix . . , 114:114739 , 1999 . r. gibrat . . librairie du recueil sirey , 1931 , french . h. a. simon . . , 42(34):425440 , 1955 . d. g. champernowne . . , 318351 , 1953 . j. c. cordoba . . , 49(4):14631468 , 2008 . j. eeckhout . . , 94(5):14291451 , september 2004 . j. c. kapteyn . . , 19(3):205206 , 1918 . y. malevergne , v. pisarenko , and d. sornette . . swiss finance institute research paper series 09 - 40 , swiss finance institute , 2009 . w. shockley , on the statistics of individual variations of productivity in research laboratories , proceedings of the institute of radio engineers ( ire ) , vol . 279 - 290 , 1957 . y. virkar and a. clauset . power - law distributions in binned empirical data . , 8(1):89119 , 03 2014 . f. black and m. s. scholes . . , 81(3):63754 , may - june 1973 . d. garcia , p. mavrodiev , and f. schweitzer . social resilience in online communities : the autopsy of friendster . in _ proceedings of the first acm conference on online social networks _ , acm , february 2013 . h. kesten . . , cxxxi:207248 , 1973 . f. radicchi , s. fortunato , and c. castellano . . , 105(45 ) , 2008 . a lognormal random variable @xcite is defined as @xmath109 where @xmath127 is a gaussian or normal random variable . hence , @xmath128 . the distribution function @xmath129=\pr[y\leq\log t]$ ] is @xmath130 du=\frac{1}{2}\left ( 1+\text{erf}% \left ( \frac{t-\mu}{\sigma\sqrt{2}}\right ) \right ) \label{distribution_lognormal}%\ ] ] where erf(@xmath26 ) is the error function . the probability density function ( pdf ) of a lognormal random variable @xmath5 follows from the definition @xmath131 $ ] , for @xmath6 , as ( [ pdf_lognormal ] ) , where @xmath8 are called the parameters of the lognormal pdf , while the mean and variance are @xcite @xmath132 = e^{\mu}e^{\frac{\sigma^{2}}{2}}\ ] ] and @xmath133 = e^{2\mu}e^{\sigma^{2}}\left ( e^{\sigma^{2 } } -1\right)\ ] ] the limit @xmath134 reduces to a dirac delta function at @xmath135 , thus @xmath136 . given the mean and variance , the parameters of the lognormal are found as @xmath137 } { \left ( e\left [ x\right ] \right ) ^{2}}\right ) \label{parameter_sigma^2}%\ ] ] and @xmath138 -\frac{\sigma^{2}}{2 } \label{parameter_mu}%\ ] ] although @xmath139 \geq0 $ ] , we remark that the parameter @xmath55 can be negative . moreover , ( [ parameter_sigma^2 ] ) and ( [ parameter_mu ] ) show that the scaled lognormal random variable @xmath52 , where @xmath140 is a positive real number , has mean @xmath141 and @xmath142 . hence , _ scaling _ by a factor @xmath140 does not change the parameter @xmath56 , which has interesting consequences for binning and measured data : the unit ( e.g. second versus hours ) in which the random variable is measured does not alter the parameter @xmath56 , only the parameter @xmath55 . the change of the argument @xmath143 in @xmath48 leads to @xmath144 } { \sigma\sqrt{2\pi } } \label{pdf_lognormal_e^u}%\ ] ] illustrating that the scaled lognormal pdf @xmath145 is a gaussian pdf @xmath146 with mean @xmath147 . the maximum of @xmath48 occurs at @xmath148 and equals @xmath149 , which follows directly from ( [ pdf_lognormal_e^u ] ) . moreover , we find easier from ( [ pdf_lognormal_e^u ] ) than from ( [ pdf_lognormal ] ) that @xmath150 and that @xmath151 . this means that any lognormal starts at @xmath152 from zero , increases up to the maximum at @xmath153 after which it decreases towards zero at @xmath154 . thus , the lognormal is bell - shaped , but , in contrast to the gaussian , the lognormal pdf is not symmetric around its maximum at @xmath148 and can be seriously skewed . the expression for the lognormal pdf in ( [ pdf_lognormal ] ) can be rewritten @xcite in a power - law - like form as @xmath155 where the exponent @xmath156 equals@xmath157 which illustrates that a lognormal random variable behaves as a power - law random variable , provided the last fraction in ( [ power_law_exponent_lognormal ] ) is negligibly small , say @xmath158 . the latter happens when @xmath159 . thus , when @xmath160 $ ] or in terms of the lower bound @xmath161 and upper bound @xmath162 defined in section [ sec_introduction ] , the pdf ( [ pdf_lognormal ] ) of a lognormal random variable is almost indistinguishable from the pdf ( [ pdf_powerlaw ] ) of a power - law random variable with power exponent @xmath125 . the @xmath30-interval @xmath163 $ ] exceeds the maximum @xmath58 of the lognormal pdf and is clearly longer when @xmath56 is larger ( as well as the tolerated accuracy @xmath158 increases ) . _ @xcite mention that , if @xmath164 , then the exponent @xmath156 in ( [ power_law_exponent_lognormal ] ) varies less than 0.3 units over a range of three orders of magnitude . taking the logarithm on both sides of ( [ pdf_lognormal ] ) results in @xmath165 if @xmath56 is large , then the second term @xmath166 in ( [ loglogform ] ) dominates , which leads to a straight line with in a log - log plot resembling a power - law with exponent @xmath83 . on the other hand , if @xmath56 is small or @xmath167 , the first , quadratic term in ( [ loglogform ] ) dominates . we apply the method of clauset _ et al . _ @xcite and virkar and clauset @xcite to our data to distinguish between power - law and lognormal distributions . @xcite approached the problem of estimating the exponent @xmath0 in ( [ pdf_powerlaw ] ) by testing different distributions . in their data , lognormals and power - laws were not clearly distinguishable either : for some tested data - sets , a lognormal distribution actually achieved a higher p - values ( goodness of fit ) than power - laws , but log ratio tests suggested that other tested distributions are closer to power - laws . by using the technique in @xcite to fit the distribution of @xmath18 , a power - law with exponent @xmath20 was found with a reasonable @xmath168-value of @xmath24 . the distribution of @xmath19 is most likely not a power - law because the @xmath168-value of @xmath169 . table [ tab : estimatedparametersusingthemethodofclausetetal ] lists the parameters for all used datasets using clauset and virkar s fitting technique @xcite with the according log - likelihood ( log - lh ) and p values . a positive log - likelihood ratio indicates that the power - law is favored over the lognormal . [ c]l\|cll\|ll\|ll\|r distribution & & power - law & & lognormal & & log - lh & & # of data - points + ( binned in ) & exponent ( @xmath0 ) & p & xmin & @xmath55 & @xmath56 & p & lr & + & & & & & & & & + adding friends @xmath18 ( seconds ) & 1.53 & 0.23 & 59 & 0.54 & 2.26 & 0.0015 & 418.67 & 7,156,722 + adding follower @xmath19 ( seconds ) & 2.03 & 0.0 & 12 & 10.45 & 2.75 & 0 & -1327.5 & 6,734,405 + adding follower @xmath19 ( hours ) & 1.81 & 0.37 & 116 & 6.43 & 3.1 & 0 & -297.66 & 41,184 + sending emails ( seconds ) & 2.33 & 0.0 & 86 & 9.53 & 3.24 & 0 & -373.7441 & 68,346,901 + sending emails ( minutes ) & 2.03 & 0.21 & 1 & 6.9 & 3.32 & 0 & -391.13 & 62,031,701 + receiving emails ( seconds ) & 2.01 & 0.0 & 1 & 10.31 & 2.73 & 0 & -958.67 & 69,528,701 + receiving emails ( minutes ) & 2.21 & 0.05 & 182 & 6.23 & 3.54 & 0 & -136.31 & 15,365,001 + commenting on reddit.com ( seconds ) & 1.08 & 0.0 & 112 & 10.2 & 3.13 & 0 & -0.0024 & 180,156,539 + commenting on reddit.com ( hours ) & 1.91 & 0.01 & 64 & 0.88 & 2.41 & 0 & -20.73 & 50,044 + & & & & & & & &
one of the hallmarks of biological systems is their tremendous specificity in binding reactions between receptors and ligands . on the molecular level , a prominent example is antigen - antibody recognition , which allows our immune system to react to pathogens in a highly specific way . although traditionally much attention has been devoted to the biochemical aspects of receptor - ligand binding , physical concepts are equally important in this context . in particular , a physical transport process is required to bring receptor and ligand to sufficient proximity for binding . a helpful concept is the notion that transport has to lead to the formation of an _ encounter complex _ , which then can react to form the final receptor - ligand complex @xcite . in the language of stochastic dynamics , the formation of the encounter complex is a first passage problem which can be treated with appropriate tools from statistical physics . in many situations , the transport process is simple diffusion . however , more complex situations also exist , like the setup in affinity chips , where ligands are transported by hydrodynamic flow into a reaction chamber loaded with receptors @xcite . in cell adhesion , the physical transport processes required for specific bond formation tend to be even more complex , because here receptors and ligands are attached to surfaces and their movement is determined by the dynamics of the objects they are attached to . one important example in this context are white bloods cells , which circulate the body with the blood flow and whose receptor - mediated binding to ligand - coated walls is usually studied experimentally in flow chambers @xcite . in order to fight pathogens in the surrounding tissue , white blood cells have to extravasate from the blood vessels . initial binding is provided by transmembrane receptors from the selectin family binding to carbohydrate ligands on the vessel walls . here , the probability to form an encounter complex is determined by the translational and rotational movement of the cell as determined by hydrodynamic , thermal and other external forces . similar situations also arise in microbiology , when bacteria adhere to the intestinal wall @xcite , in malaria infection , when infected red blood cells adhere to the vessel walls @xcite , in the initial stages of pregnancy , when the developing embryo adheres to the uterus @xcite , and in biotechnology , e.g. , when sorting cells on microfluidic chips @xcite . in this paper , we address this situation theoretically by combining methods from hydrodynamics and stochastic dynamics . in we show the situation which is theoretically analyzed in the following . a spherical particle with radius @xmath0 moves with hydrodynamic flow in positive @xmath1-direction at a height @xmath2 above a wall . the simplest possible flow pattern is linear shear flow with shear rate @xmath3 . for the usual dimensions in flow chamber experiments with white blood cells , this is the relevant flow profile . in the absence of external forces , there is no reason for the particle to drift towards the wall and the formation of an encounter complex has to rely completely on thermal diffusion . in many situations of interest , however , there are forces pushing the particle towards the wall , e.g. , gravitational or electric forces . in physiological blood flow , cell density is high and the driving force for encounter is provided mainly be hydrodynamic or contact interactions with other cells . for the sake of computational simplicity and for conceptual clarity , here we consider the simplest case of a force driving the particle onto the wall , namely a constant gravitational force directed in negative @xmath2-direction . therefore , we introduce a mass density difference @xmath4 between the particle and the surrounding fluid . again this is the relevant situation in flow chamber experiments , which are usually done with a diluted solution of cells , thus ruling out a dominant role for cell - cell interactions . receptors are modeled as patches on the particle surface , while ligands are modeled as patches on the boundary wall . the formation of an encounter complex is then identified with the first approach of any pair of receptor and ligand patches which is smaller than a prescribed capture radius @xmath5 . the underlying stochastic process is rather complex due to position - dependent mobilities resulting from the hydrodynamic equations . in order to solve the corresponding mean first passage time problem , here we use computer simulations of the appropriate langevin equation . a short report of some of our results has been given before @xcite . we start in by introducing the relevant concepts from hydrodynamics at small reynolds numbers , in particular the friction and mobility matrices resulting from the stokes equation for a rigid particle in linear shear flow above a wall . in we combine these results with concepts from stochastic dynamics in order to arrive at a langevin equation describing particle motion subject to hydrodynamic , gravitational and thermal forces . due to the position - dependent mobility functions , we deal with multiplicative noise , that is special care is needed to derive and interpret the noise terms . in our numerical scheme is applied to a sphere falling in shear flow . the comparison of the measured stationary height distribution function with the exact solution provides a favorable test for our numerical treatment . in we show that for the case of homogeneous coverage of sphere and wall the mean first passage time to contact can be solved exactly , again in excellent agreement with our numerical procedure . in we explain why the choice for the initial height is not essential . in the next two sections , we present and explain our simulations results , first for movement restricted to two dimensions in and then for the full three dimensional case in . we finally conclude in by discussing the biological and biotechnological relevance of our results . due to their small sizes , the hydrodynamics of cells is in the low reynolds number regime . using a typical cell size @xmath6 m and a typical velocity v = mm / s ( that is the flow velocity at a distance @xmath7 to a wall with linear shear flow of rate @xmath8 100 hz ) , the reynolds number is @xmath9 , where @xmath10 g/@xmath11 and @xmath12 pa s are density and viscosity of water , respectively . therefore , we essentially deal with the stokes equation for incompressible fluids : @xmath13 where @xmath14 is the fluid velocity field , @xmath15 is the pressure field and @xmath16 is the force density on the fluid by the particle . here , we use the induced force picture , i.e. , the fluid equations of motion are extended to the interior of the particle and the particle is replaced by an appropriate force density @xmath16 acting on the fluid @xcite . the unperturbed flow field has to satisfy the homogeneous version of as well as no - slip boundary conditions at the wall . in this paper , we use the simplest possible example , namely linear shear flow , @xmath17 . the effective flow field in the region occupied by the rigid sphere reads @xmath18 where @xmath19 are the translational and rotational velocities of the sphere , respectively . @xmath20 is the position of its center , @xmath0 the sphere radius and @xmath21 the theta step - function . the particle is subject to forces and torques which follow from the force density as @xmath22 because we consider a rigid object , higher moments of the force density are not required in our context . for the unperturbed flow at the mid - point of the sphere , we make the following definitions : @xmath23 where the vector @xmath24 is called _ vorticity _ and the tensor @xmath25 _ rate of strain tensor_. because we restrict ourselves to linear shear flow , all higher moments of the unperturbed flow vanish . due to the linearity of the stokes equation , a linear relationship exists between the force density @xmath16 and the driving flow , which is the difference between real and unperturbed flows @xcite . specified for the first moments of the force density , it leads to the relation @xmath26 where the _ shear force _ @xmath27 with @xmath28 . it results from the perturbation of the flow by the presence of the wall and vanishes for free flow . the two matrices @xmath29 and @xmath30 are conveniently written as @xmath31 where the @xmath32 are the symmetric _ friction matrices _ and the superscripts t , r and d stand for _ translational _ , _ rotational _ and _ dipolar _ , respectively . in order to obtain the translational and rotational velocities of the sphere as a function of the hydrodynamic forces and torques , we have to invert : @xmath33\ .\end{aligned}\ ] ] the symmetric matrix @xmath34 is called _ mobility matrix_. it is convenient to define the mobility tensors through @xmath35 in order to calculate the friction and mobility tensors for the special case of a sphere in linear shear flow above a wall , we follow the procedure from ref . @xcite . the friction tensors @xmath32 introduced in and the mobility tensors @xmath36 introduced in are expressed in terms of scalar functions together with irreducible tensors formed form the kronecker symbol @xmath37 , the levi - civita symbol @xmath38 and the normal vector @xmath39 . the scalar friction and mobility functions are not known analytically , but can be obtained to high accuracy by the following numerical scheme . one introduces the variable @xmath40 , where @xmath0 is the radius of the sphere and @xmath2 is its height above the wall . thus , @xmath41 can take values from the interval @xmath42 $ ] . in the limit @xmath43 , that is far away from the wall , one can expand the friction functions in powers of @xmath41 . in the limit @xmath44 , that is close to the wall , analytical results can be obtained with lubrication theory . in order to cover the whole interval , the two limit solutions are matched using a pad summation scheme @xcite . more details of this implementation are given in appendix [ appendix : mobility ] . in contrast to the tabulated finite element results from ref . @xcite , this implementation gives correct results for any possible configuration . the motion of a particle subject to thermal , hydrodynamic and direct external forces like gravity is called _ stokesian dynamics _ @xcite . in this section we derive the corresponding stochastic differential equation ( _ langevin equation _ ) . the langevin equation will allow us to base our statistical treatment on the repeated simulation of individual trajectories . because we are interested in the over - damped ( stokes ) limit , we can neglect inertia in newton s second law : @xmath45 where @xmath46 , @xmath47 and @xmath48 are hydrodynamic , direct and thermal forces acting on the sphere . an analogous balance exists for the torques . for the following , forces and torques as described above are united in one symbol . for example , from now on the symbol @xmath49 denotes @xmath50 , a six - dimensional vector comprising force @xmath49 and torque @xmath51 , and @xmath52 denotes the six - dimensional particle translational / rotational velocity vector . + in the absence of brownian forces , @xmath53 and @xmath54 . inserting this into then gives @xmath55 and the particle trajectory can be found with a simple euler algorithm as @xmath56 . in the presence of brownian motion , the situation is more complex , because thermal noise leads to terms of the order @xmath57 and special care has to be taken to include all terms up to order @xmath58 . due to the fluctuation - dissipation theorem , for our problem gaussian white noise reads @xmath59 here , the subscript @xmath41 corresponds to the fact that the thermal force @xmath60 is a random process . the left part of states that the forces that the fluid exerts on the particles are equally distributed in all directions so there is no net drift due to thermal fluctuations . the right part of states that forces at different times are not correlated , which is a good approximation because the diffusive forces act on a much faster time scale than the hydrodynamic forces . because the mobility matrix @xmath61 is position - dependent , we deal with so - called _ multiplicative noise_. since the @xmath62-correlation in can be considered to be the limit of a process with an intrinsic time scale for thermal relaxation , which is much faster than the time scale of hydrodynamic movement , the stratonovich interpretation of the stochastic process is appropriate @xcite . this means that for each time step , the mobility functions have to be evaluated at @xmath63 ( rather than at @xmath64 as in the it interpretation ) . the stratonovich interpretation also implies that the rules for integration and coordinate transformation are the same as for the riemann integral in non - stochastic calculus . the presence of the thermal noise converts the position function @xmath64 into a random process @xmath65 . multiplicative noise can result in additional drift terms . we therefore write the langevin equation as @xmath66 where in comparison to the deterministic equation we have added both the gaussian white noise @xmath67 ( to be interpreted in the stratonovich sense ) and some drift term @xmath68 . the drift term @xmath68 can be derived by requiring to be equivalent to the appropriate smoluchowski equation . the details of these calculations are given in appendix [ appendix : gradient_term ] . the result is @xmath69 for additive noise , that is for position - independent mobility functions , the additional drift term would vanish . in the case of position - dependent mobility matrices , the noise term @xmath67 alone would lead to a drift of the particle towards regions of lower mobility ( that is towards the wall , where mobility vanishes due to the no - slip boundary condition ) . this drift , however , is exactly compensated by the additional term @xmath68 . for the following , it is useful to non - dimensionalize . for length , the natural scale is sphere radius @xmath0 . for time , we use @xmath70 , which is the time needed to diffuse the distance @xmath0 . for force , we use @xmath71 , the stokes force at velocity @xmath72 , that is in linear shear flow a distance @xmath0 away from the wall . the scalar friction and mobility functions appearing in @xmath61 , @xmath30 and @xmath29 , also become dimensionless as explained in appendix [ appendix : mobility ] . the langevin equation now reads @xmath73 where the _ pclet number _ @xmath74 measures the relative importance of deterministic to brownian motion . in the limit @xmath75 the particle only exhibits diffusive motion and in the limit @xmath76 it is no longer subjected to diffusion . the second dimensionless parameter @xmath77 measures the relative importance of direct forces / torques versus the shear force / torque . measuring the time in units of the diffusive time scale is appropriate for pclet numbers of order ten or less . for simulations with larger pclet numbers it is more suitable to scale time with the inverse shear rate @xmath78 this has the effect of dividing by @xmath79 . in order to solve numerically , it has to be discretized with respect to time . the appropriate euler algorithm can be derived by first rewriting in the it - version , which adds another drift term to the equation . as explained in appendix [ appendix : euler ] , the two drift terms together lead to the result @xmath80 its discretized version is simply @xmath81\delta t + { { \mathbf g}}(\delta t ) + \mathcal{o}(\delta t^2)\ .\ ] ] this final result has been derived before in a different way by brady and bossis @xcite . for vanishing shear flow , it also agrees with the classical result by ermak and mccammon @xcite . in appendix [ appendix : euler ] , we describe the algorithms used to implement , in particular the algorithm to generate the thermal forces @xmath82 . as explained in the introduction , we consider a sphere whose density is slightly larger than that of the fluid . due to this density difference @xmath4 a constant drift towards the wall exists . as we will see later , this drift ensures that on average the sphere will bind to the wall in finite time . the two independent parameters defined in for this model system are @xmath79 and @xmath83 , with the earth acceleration constant @xmath84 m / s@xmath85 . for later considerations , it is convenient to introduce also the parameter @xmath86 , which we call the _ pclet number in @xmath2-direction_. @xmath79 and @xmath87 represent the strengths of the hydrodynamic and gravitational forces in respect to the thermal force , respectively . out of the three parameters @xmath79 , @xmath88 and @xmath87 , only two are independent , because @xmath89 . we first consider the path of a sphere falling in shear flow after it has been dropped at some initial height . illustrates the effect of the pclet number by showing some representative simulation trajectories . for @xmath90 the motion of the sphere is purely deterministic and only governed by the parameter @xmath88 . in the diffusive limit @xmath91 , the sphere makes a pure random walk ( except for the drift in @xmath2-direction due to the gravitational force ) . as the mobility matrix does only depend on the height of the sphere above the wall ( cf . appendix [ appendix : mobility ] ) , the motion in the @xmath2-direction is independent of the position in the @xmath92-plane and the orientation of the sphere . therefore , it can be treated separately . the probability density @xmath93 for the sphere to be at height @xmath2 at time @xmath41 is the solution to a one dimensional smoluchowski equation @xmath94 this equation can not be solved analytically as the mobility function @xmath95 is not known in closed form . in we show numerical solutions obtained by simulating the equivalent langevin equation . one clearly sees that first the @xmath62-function at @xmath96 is broadened due to diffusion and then develops into a stationary solution which has its maximum at the wall . this stationary solution has a simple analytical form which follows from by integrating @xmath97 : @xmath98 thus , the stationary solution is simply the barometric formula , as it should be for thermodynamic reasons . we also find that the first two moments ( mean and variance ) are the same : @xmath99 in the limit of vanishing gravitational force ( @xmath100 ) , the probability distribution becomes flat and the probability of finding the sphere does not peak at the wall anymore . if the sphere and the wall are homogeneously covered with receptors and ligands , respectively , an encounter complex is established whenever the sphere comes sufficiently close to the wall . the mean time which elapses after the sphere is set free at some initial position until an encounter complex is established is then identical with the mean first passage time ( mfpt ) for a sphere dropped at initial height @xmath101 to reach the height @xmath102 . note again that the motion in @xmath2-direction is independent of the values of the other coordinates . for a particle diffusing in an interval @xmath103 $ ] , with @xmath102 being an absorbing boundary and @xmath104 a reflective boundary , the mfpt @xmath105 to reach @xmath102 when started at @xmath106 $ ] is the solution to the following ordinary differential equation @xcite @xmath107 in our case , @xmath108 . the drift term is @xmath109 and the diffusive term @xmath110 , where @xmath111 is a scalar mobility function as explained in appendix [ appendix : mobility ] . the general solution to is @xcite @xmath112 this , can be reduced up to an integral over @xmath113 : @xmath114 thus the dependence of @xmath115 on @xmath87 , the only parameter in this problem , is obtained exactly . it is important to note that the compact form for the mfpt in is a result of the constant vertical force . for a more general vertical potential force @xmath116 with a potential @xmath117 , can be reduced to @xmath118 this equation shows that the potential must satisfy the condition @xmath119 for the mfpt to be finite . this holds true , e.g. , for the gravitational force studied here or for the interaction of a charged object with an oppositely charged wall , but not , e.g. , for a lennard - jones potential . the integral over the scalar mobility function @xmath120 can easily be calculated numerically as @xmath120 behaves well in the full range of @xmath2 . in fact @xmath121 can be approximated by its leading term from the lubrication analysis , i.e. , @xmath122 . we then find @xmath123.\end{aligned}\ ] ] a numerical analysis shows that the approximation deviates only by a few percent from the exact solution . thus , @xmath115 is logarithmically divergent if the absorbing point is close to the wall , @xmath124 , and linearly divergent if the starting point is at infinite height , @xmath125 . for a sphere homogeneously covered with receptors each having a capture radius @xmath5 , the mean time for forming an encounter complex is @xmath126 . this time will serve as a useful limiting result in some of the considerations presented in the next sections . the exactly known result provides also a good test for the algorithm we implemented . in a the mfpt obtained from simulation experiments and from quadrature of are compared . the two results agree very well ( see appendix [ appendix : error ] for a discussion of the statistical and systematic errors of the simulation results ) . in b we show the numerically obtained distribution of first passage times . one clearly sees that the larger @xmath87 , the stronger they peak around the mean . we conclude the case of homogenous coverage by noting that in order to obtain dimensionalized results , one has to multiply the mfpt by the diffusive time scale @xmath127 . this result does not depend on shear rate @xmath3 because vertical and horizontal motion are decoupled and rotational motion is not relevant here . however , it depends on viscosity @xmath128 , which sets the time scale for vertical motion . if one switched off thermal fluctuations , the falling time would be exactly the same as the mfpt from , but this is a special result for constant force and not true in general . if one removed the wall , the translational symmetry in @xmath2-direction would not be broken and the mfpt would be @xmath129 , that is the logarithmic term in would be missing . we now turn to spatially resolved receptor coverage , that is we consider a sphere which is covered by @xmath130 equidistantly spaced receptor patches . for the moment being , the wall is still considered to be homogeneously covered with ligands . the mfpt @xmath131 now will depend on the initial position @xmath132 and the initial orientation @xmath133 as well as on the absorbing boundary @xmath134 in diffusion space . the latter is given by the special receptor and ligand geometry . in an experimental setup with linear shear flow it is possible to measure only particles which have been initially at a certain height . this is due to the fact that their average velocity as obtained from the solution of the stokes equation depends on their height in a unique way @xcite . however , it is almost impossible to prepare a certain initial orientation @xmath133 or @xmath92-position relative to the ligands . therefore , the quantity of interest to us will be a mfpt which is averaged over all possible initial orientations @xmath133 and all initial positions @xmath92 , which will be denoted as @xmath135 . the dependence of @xmath136 on the initial height for @xmath137 can be derived exactly . for homogeneous ligand coverage the quantity of interest is @xmath138 where @xmath134 is the absorbing hyper - surface in @xmath139-space and @xmath140 a normalization constant . absorption is only possible if @xmath141 , thus if we look at some intermediate height @xmath142 , then @xmath143 where @xmath144 is the conditional probability to pass the height @xmath145 with the orientation @xmath146 when starting with the initial orientation @xmath133 at @xmath101 . @xmath147 is independent of the initial orientation and can be calculated by means of . now averaging over the initial orientation gives @xmath148t(\vec{\theta}_m , z_m \| c)\\ \nonumber & = t(z_0 \| z_m ) + { \frac{1}{v_{\vec{\theta}}}}\int d^3 \vec{\theta}_m\ ; t(\vec{\theta}_m , z_m \| c ) = t(z_0 \| z_m ) + { \langle t(\vec{\theta}_m , z_m \| c ) \rangle}_{\vec{\theta}_m}. \end{aligned}\ ] ] thus , if the orientation - averaged mfpt is known for some initial height @xmath137 , then the mfpt for any other initial height @xmath149 can be calculated by means of equations and . in , this result is verified by simulations for the two - dimensional case , that is the sphere can only move in the @xmath150-plane and rotate only around the @xmath151-axis , compare . due to the decomposition , the initial height is not essential . in the following , we therefore will always use the value @xmath152 , that is the sphere has to fall by one radius until it hits the substrate for the first time . we now study the effect of shear rate for heterogeneous receptor distribution if the sphere is restricted to move only in two dimensions . then , the receptor patches can be equidistantly distributed over the circumference as illustrated in . each receptor patch has a capture height of @xmath5 and a width of @xmath153 . the 2d receptor density is then @xmath154 . orientation is now represented by a single angle @xmath155 . the absorbing boundary @xmath134 is illustrated in . for each receptor patch , binding can occur over a range @xmath156 , which consists of two parts . the inner part is valid already for @xmath157 and reflects the overlap due to a finite @xmath5 . the outer part is results from a finite @xmath158 . together this leads to @xmath159 . the receptor patches establish a periodicity with period @xmath160 . as the number of receptor patches grows , this period decreases and one finally achieves overlap . then , encounter becomes possible for all values of @xmath155 , that is we are back to the case of homogeneous receptor coverage . in our case of non - homogeneous coverage , the mfpt depends on @xmath161 and @xmath101 . for the following simulations @xmath162 , @xmath163 and @xmath152 is chosen unless other values are explicitly mentioned . a shows the mfpt as a function of the pclet number @xmath79 . note that in the log - log plot , an apparent plateau appears at small value of @xmath79 , although in a linear plot there would be monotonous decay . three regimes can be distinguished . for @xmath164 ( _ diffusive limit _ ) the transport by the imposed shear flow is negligible and only diffusive transport is present . for very large values of @xmath79 , @xmath165 plateaus at the value given by independent of @xmath130 . in this limit the time for rotation to any certain orientation is negligible compared to the mean time to fall down close to the wall , therefore , the result for rotational symmetry is recovered . between these two limits the mfpt decreases monotonically with increasing @xmath79 . b shows the data from a plotted as a function of the receptor density @xmath166 . the larger @xmath79 the less pronounced is the dependence of @xmath165 on @xmath130 . for @xmath164 , however , @xmath165 strongly depends on @xmath130 . the latter relation is better illustrated in c. there , at @xmath164 , @xmath165 is shown for a wide range of @xmath130 . the simulations were done for fixed patch size @xmath158 but for four different values of the capture radius @xmath5 ( cf . ) . for @xmath167 , @xmath165 reaches the value given by . as described by , @xmath165 is the smaller the larger @xmath5 is . an increase in the number of receptor patches @xmath130 leads to a strong decrease for the mfpt , however , no special scaling behavior can be observed . it is remarkable that the limiting value for the case of homogeneous receptor coverage is already reached for @xmath168 . the larger the capture radius @xmath5 the more pronounced is this effect . this can be understood by observing that the effective patch size as given by the angle @xmath169 ( see ) is monotonically increasing with increasing @xmath5 . we next try to qualitatively understand the effect of shear rate for the simulation results shown in a. in general , it is very hard to separate the effects of diffusion and convection . the time for binding at @xmath164 is determined purely by diffusion effects and will be denoted by @xmath170 . as shear flow increases , the rotation of the sphere is increasingly dominated by convection . we now derive a convection time @xmath171 which competes with the diffusion time @xmath170 at large pclet number . for very large pclet number , we expect the mfpt to be the sum of the homogeneous result from plus this additional time @xmath171 . an important question then is at which @xmath79 the convection time @xmath171 become smaller than the diffusion time @xmath170 . on order to estimate @xmath171 , we note that the main effect of increased shear rate is faster rotation in the direction of flow . once a receptor has rotated by an angle @xmath160 such that it opposes a ligand on the substrate , there is some probability @xmath172 that the sphere is at the correct height that an encounter can occur . if no encounter occurs with the complementary probability @xmath173 , the sphere has to rotate about another angle @xmath174 until the next receptor points downwards . supposing the time , @xmath175 , to rotate about the angle @xmath174 is large enough that there is no correlation between the height of the sphere before and after the rotation , then , an encounter occurs again with probability @xmath172 ( therefore this analysis also does not hold at very large @xmath79 ) . thus , the mean time @xmath171 for encounter is @xmath176 where the series has been summed up by means of the geometric formula . in the last term we assumed that the probability @xmath172 for the proper height is small due to a small capture distance @xmath5 . it follows from the stationary probability distribution @xmath177 given by : @xmath178 the time @xmath179 to rotate about half of the angle @xmath174 is approximately @xmath180 . therefore , we get @xmath181 in this analysis , the convection time @xmath171 scales inversely with the number of receptor patches @xmath130 and the pclet number @xmath79 . as @xmath79 increases , @xmath171 gets smaller than @xmath170 and then dominates the overall outcome . comparing to the simulation data for @xmath164 shows that this crossover occurs in the range @xmath182 and that the corresponding value of @xmath79 increases with increasing receptor number @xmath130 , exactly as observed in the simulation data over the full range of @xmath79 . however , the exact scaling of this data is not @xmath183 for large @xmath79 as predicted by . in practice , the decay is somehow slower due to correlations between the height of the sphere at two successive instances of a receptor pointing downwards , which we have neglected in our analysis . we briefly comment on the effect of the downward driving force , that is @xmath87 . above , we have found that in two cases , homogeneous coverage from and convection - dominated rotation from , the mfpt scales inversely with @xmath87 . this scaling behavior is indeed found in the simulations , except that for very large values of @xmath87 , the mfpt approximates a constant value ( data not shown ) . the reason is that the larger @xmath87 , the smaller the mean time to fall below the height @xmath184 . as indicated by , then the sphere stays below this height until an encounter occurs . this implies that in this limit , the mfpt depends only on rotational motion and the falling motion is irrelevant . we now introduce spatially resolved ligands into the 2d - model . a shows the model definition : the ligand patches are considered to have the same radius @xmath185 as the receptor patches and they are located at a distance @xmath186 from each other . this results in a one - dimensional ligand density given by @xmath187 . the mean first passage time will now also depend on the initial @xmath1-position , @xmath188 , where @xmath134 is the hypersurface in @xmath189 space where a receptor patch touches a ligand patch . but similarly as in the above section in regard to initial orientation , the dependence on the initial @xmath1-position is of minor interest and therefore , we will discuss the mfpt averaged over the initial position and orientation , denoted by @xmath190 . b shows that by varying the pclet number we can identify the same three regimes for all ligand - densities as before . for @xmath75 in the limit of pure diffusive transport , @xmath190 approaches a finite value , depending on @xmath191 and @xmath192 . with increasing @xmath79 , @xmath193 decreases monotonically and finally for @xmath194 reaches the value of the mfpt in the limit of homogeneous receptor and ligand coverage . in contrast to above , however , in this limit the shear flow not only restores rotational invariance of the sphere , but in addition also translational invariance of the substrate . a provides more details for @xmath190 as a function of @xmath192 in the diffusive limit ( @xmath164 ) . we find that in the range @xmath195 the mfpt is almost not affected by ligand concentration : as long as the ligand patches are sufficiently close to each other , a receptor patch touching the wall will most probably find a ligand before diffusing away again . the situation changes completely with small ligand density . for @xmath196 the averaged mean first passage time @xmath190 scales with the ligand density @xmath192 as @xmath197 . this can be understood by calculating the position - averaged mfpt @xmath198 for a particle diffusing in an interval @xmath199 $ ] with diffusion constant @xmath200 , which gives @xmath201 . this suggests that the quadratic scaling with @xmath186 results from the diffusive motion between adjacent ligand patches . b summarizes our results for the dependence of the 2d mfpt @xmath190 on ligand density @xmath192 and receptor density @xmath191 in the diffusive limit . clearly there exists a large plateau around the value for the case of homogeneous coverage @xmath202 . this implies that if ligands and receptors patches are not too strongly diluted , the mean encounter time is still close to the optimal value given by . on the other hand if the number of receptor and/or ligand patches is highly reduced the mean encounter time is strongly increased . we finally turn to the full 3d - situation , that is the sphere may diffuse about all three axes as described by and . receptors are located in spherical patches which are randomly distributed over the sphere . each receptor patch has a radius @xmath158 and a height ( capture length ) @xmath5 . that is the appropriate generalization of the situation shown in for the 2d - case . thus , for @xmath130 receptor patches the receptor density is @xmath203 ( for @xmath204 ) . in contrast to the preceeding sections where the receptor patches could be regularly distributed over the circumference , this is no longer possible on the surface of a sphere . therefore , we distribute the patches randomly over the sphere with equal probability for each position , with a hard disk overlap algorithm making sure that no two patches overlap @xcite . one has to bear in mind that then for small @xmath130 two different distributions may have slightly different binding properties . this effect becomes weaker for larger @xmath130 , therefore in the following we will only use @xmath205 . the quantity we measure in our simulations is now @xmath206 in the case of homogeneous ligand coverage and @xmath207 in the case that the ligands are located in spherical patches on a 2d - lattice . thus , we average the mfpt over the initial orientations and positions as explained above . in order to explore the dependence of @xmath206 on @xmath130 and @xmath79 we first simulated the receptor ligand encounter in the case of homogeneous ligand coverage @xmath208 . in order to average over the initial positions we started each run with a randomly chosen initial orientation . after 100 runs we generated a new distribution , thus averaging out also the effect of different receptor distributions . in order to achieve reasonable statistics , we typically used 100,000 runs . our results are shown in a. again we find three different regimes as a function of the pclet number @xmath79 . this proves that qualitatively the basic results of the 2d - treatment remain valid in 3d . however , in detail there are important differences . in contrast to the 2d results presented above , @xmath206 in the limit @xmath209 is no longer given by if @xmath130 is small . that is due to the fact that for @xmath76 the receptor patches effectively behave as ring - like structures . the rotation of such a ring about the @xmath1- or @xmath151-axis is not affected by @xmath79 and thus still depends on diffusion . for large @xmath130 the rings cover the whole sphere and for @xmath76 @xmath206 is again given by . in b we plot the @xmath75 limit of @xmath206 as a function of the number of receptor patches @xmath130 , for different values of the capture radius @xmath5 . the fitted straight line for @xmath210 shows that @xmath206 approximately behaves like @xmath211 . neglecting effects of curvature , the average distance between two receptors patches is @xmath212 and the mean time to diffuse that distance is @xmath213 . this provides a simple explanation for the observed scaling behavior . for high @xmath130 , the mfpt reaches a plateau value , given by . this plateau value depends on @xmath5 and is the smaller the larger @xmath5 . also the crossover from the asymptotic behavior at small @xmath130 to the plateau at large @xmath130 is shifted with increasing capture radius @xmath5 towards smaller @xmath130 . in c we show the effect of a finite ligand density @xmath192 at @xmath164 . for the simulations we distributed the ligands in circular patches of radius @xmath214 on a quadratic lattice with lattice constant @xmath186 , thus , resulting in a ligand density @xmath215 . in our implementation , the intersection between the receptor patch and the wall is approximated by an appropriate circle , because it is easy to check if this circle overlaps with the ligand patch . the fits given in c show that for small @xmath192 , the mfpt scales as @xmath216 . because the curves for different @xmath130 appear to be rather similar , in the inset we plot the ratio of different pairs of these curves . as this results in approximately constant plateaus , we conclude that the scaling with ligand density is hardly effected by @xmath130 . as in 2d , the inverse scaling with ligand density can be understood in simple terms by noting that the mfpt to diffusional capture scales like @xmath217 . at a coverage around @xmath218 , saturation occurs as it did for receptor coverage . we finally discuss the influence of the receptor geometry described by the parameters @xmath5 and @xmath158 . because @xmath79 changes the mfpt in a monotonous way , it is sufficient to study the diffusive limit @xmath219 . a and b show @xmath206 as a function of @xmath158 for @xmath220 and @xmath221 , respectively . in order to obtain smooth curves , in this case only one receptor distribution was used for all runs . we find that the curves can be fitted well to the function @xmath222 where the second term is the homogeneous result from . this means that even for vanishing receptor size @xmath223 the mfpt remains finite . this makes sense because above we have shown that the effective patch size is determined both by @xmath158 and @xmath5 . in detail , showed that capture occurs over the solid angle @xmath224 with @xmath225 . for small @xmath5 and @xmath158 , this allows us to define an effective patch size @xmath226 where we have used @xmath227 . suppose now that the sphere diffuses over the time @xmath228 until a receptor patch points downwards , then it may encounter a ligand with a probability @xmath172 that is given by the normalized area of one effective receptor patch : @xmath229 if no encounter occurs , the sphere has to diffuse again a time @xmath228 until the next encounter can occur . this leads to the mean encounter time @xmath230 . putting everything together gives with @xmath231 and @xmath232 . if checked against our simulation results , we indeed find that the fit parameter @xmath104 is an increasing function of @xmath5 , but varies only slightly with @xmath130 . the fit parameter @xmath233 scales approximately as @xmath183 and varies with @xmath5 , also consistent with the above analysis . in c @xmath206 is plotted as a function of @xmath158 for several values of @xmath5 and @xmath234 . one clearly sees that increasing @xmath158 has a much smaller impact on @xmath206 than a comparable increase in @xmath5 , which is qualitatively well described by the preceeding analysis . in a and b the receptor density is varied over almost four orders of magnitude by changing @xmath158 , but the largest measured decrease for @xmath206 is only by a factor four . in contrast , an increase of the receptor density by one order of magnitude due to ten - fold more receptor patches leads to a decrease of @xmath206 by almost also one order of magnitude . however , this is only true as long as @xmath130 is not too large , as for large @xmath130 @xmath206 saturates at the limiting value of homogeneous receptor coverage ( cf . the crossover from the @xmath235 behavior to the saturation should take place when the average distance between two receptor patches @xmath236 becomes comparable to the size of one receptor patch . this corresponds to @xmath237 or @xmath238 . this estimate predicts that the crossover takes place between several tens to several hundreds of receptor - patches , depending on @xmath5 , in agreement with the data shown in b. in this paper we have calculated the mean first passage times ( mfpt ) for initial encounter between spatially resolved receptors on a brownian particle in linear shear flow and spatially resolved ligands on the boundary wall . our main results were obtained by repeated simulations of the discretized langevin equation . each data point shown corresponds to at least 100,000 simulation runs . it is important to note that these simulations are very time consuming because we resolve objects of the size of @xmath239 , that is for @xmath36m - sized particles we resolve the nm - scale . in general , we found that the mfpt was always monotonically decreased when the pclet number was increased . that means that a particle which is covered with receptors in a way that it binds well to ligands already in the diffusive limit is even better suited to initiate binding at finite shear rate . in our simulations we modeled the receptor geometry using three parameters : the number of receptor - patches @xmath130 , the radius of the receptor patches @xmath158 , and the capture radius @xmath5 . the efficiency of binding is mainly increased by @xmath130 , but only up to a saturation value of the order of hundred . an increase of @xmath158 leads only to a weak enhancement of binding efficiency . the influence of @xmath5 to the mfpt is threefold : i ) it reduces the mean falling time , ii ) it increases the effective patch size , and iii ) according to the stationary probability distribution for the @xmath2-direction , it becomes more probable for the sphere to be within the encounter zone when @xmath5 is increasing . an additional but more indirect effect of receptor protrusions is that the further the cell is away from the wall , the faster it can rotate ( even in the diffusive limit ) due to the larger mobility . as shown by rotations play a role only within binding range , i.e. , for @xmath141 . therefore , a large @xmath5 lets the cell also benefit from faster rotations . summarizing our findings in regard to receptor geometry we conclude that the most efficient design for particle capture under flow is to cover the particle with hundreds of receptor patches ( @xmath130 above threshold ) , each with a rather small area ( small @xmath158 ) , but formed as a protrusion ( large @xmath5 ) . indeed , this strategy seems to be used by white blood cells , which have evolved intriguing mechanisms both on the molecular and cellular scale in order to adhere effectively to the endothelium under the conditions of hydrodynamic flow . the typical size of white blood cells is @xmath240 m and they are covered with a few hundreds of protrusions ( _ microvilli _ ) with the receptors ( most notably l - selectin ) localized to the microvilli tips @xcite . in general , the microvilli of white blood cells are much more complex than the parameter @xmath5 in our model : they are rather long ( typical length 350 nm , that is @xmath241 ) and have their own physical properties ( e.g. , very flexible in the transverse direction and viscous in the longitudinal direction ) @xcite . nevertheless , it is striking that elevation of the receptors above the main cell surface seems to be a major design principle for white blood cells . in fact , the same strategy appears to be used also by malaria - infected red blood cells , which are known to develop a dense coverage with elevated receptor patches ( _ knobs _ ) on the cell surface @xcite . a typical value for the cell radius is 3.5 @xmath36 m @xcite . the knobs have a typical height of 20 nm , a radius of about 90 nm and a distance of 200 nm ( for red blood cells infected by single parasites ) @xcite . this dense and elevated coverage suggests that like the white blood cells , the malaria - infected red blood cells also function in the regime of homogeneous coverage . in order to discuss the motion of white blood cells in more detail , it is instructive to consider the parameters for a typical flow chamber experiment . in aqueous solution and at room temperature , @xmath10 g/@xmath11 , @xmath12 pa s , and @xmath242 . then , the dimensionless parameters determining cell motion become @xmath243 where @xmath0 is given in @xmath244 , @xmath4 in units of g/@xmath11 and the shear rate @xmath245 in units of 1 per seconds ; @xmath58 is the diffusive time scale . for leukocytes in flow chambers we typically have @xmath246 , @xmath247 and @xmath248 , thus , for the two pclet numbers we get @xmath249 and @xmath250 , respectively then , @xmath251 , that is the effect of hydrodynamic deterministic motion will be very strong . the experimental time scale is given by the time for transversing the field of view , which is about 3 s at a shear rate of 100 hz and length of 670 @xmath36 m . the diffusive time scale @xmath58 for leukocytes is about 600 s ( 10 min ) , which reflects their large size and shows that diffusive motion is by far not sufficient to initiate binding . binding becomes more favorable in the presence of convection . for a start height of one radius above the wall ( @xmath152 ) , our calculations give a mfpt of about 5 s , that is much less than the diffusive time . however , this is still much larger than the experimental time scale . this proves that only those cells have a chance to bind that flow very close to the wall , exactly as observed experimentally . in vivo , white blood cells therefore depend also on other mechanisms driving them onto the substrate , including contact and hydrodynamic interactions with other cells . these effects have been studied in detail before . for example , munn and coworkers have shown that adhesion of leukocytes close the the vessel wall in post - capillary venules is enhanced by red blood cells passing them @xcite . king and hammer have shown , using an algorithm capable of simulating several cells , that already adherent leukocytes can recruit other leukocytes via hydrodynamic interactions @xcite . the results presented here , when specified to leukocytes , show that indeed these mechanisms are crucial for effective leukocyte capture under flow . our results also suggest that leukocytes are sufficiently large that thermal fluctuations are not dominant . this changes when studying smaller particles , e.g. , receptor - covered spheres with @xmath252 m , whose binding also has been investigated with flow chambers @xcite . shows that the pclet numbers scale strongly with particle radius @xmath0 , therefore , these beads are subject to much stronger thermal fluctuations than leukocytes . in ref . @xcite it has been verified that indeed in equilibrium such particles obey the barometric distribution from . in ref . @xcite it was found that the adhesion probability @xmath253 is proportional to the ligand - density , @xmath254 . with @xmath255 it follows that @xmath256 as found by our simulations in the limit of low ligand densities . throughout this paper we have considered the generic case of a constant downward acting force due to a density difference between the sphere and the surrounding fluid . in future work it might be interesting to examine also other forces which can easily be done in the framework presented here . as the addition formula for falling and rotational mfpt was not derived under the assumption of a specific force , it is also true for non - constant forces . for general potential forces the falling time has then to be replaced by . also the rotational mfpt is influenced by a vertical force via the stationary height distribution . neglecting gravitational force and considering only short - ranged forces like van der waals or electrostatic forces would result in infinite mfpts for the setup of the halfspace . this problem , however , can be solved by using an additional wall acting as an upper boundary @xcite . in this paper we assumed a rigid brownian particle . for cells , elastic deformations might be relevant . for free flow , a simple scaling estimate shows that the critical value for the shear rate leading to substantial deviations from the spherical shape is @xmath257 @xcite , where @xmath258 pa and @xmath259 nm are young modulus and thickness of the cellular envelope , respectively . the fact that the young modulus e appears here indicates that cells tend to passively deform less than vesicles , whose elasticity is characterized rather by the bending rigidity @xcite . the scaling estimate leads to a critical shear rate of @xmath260 hz , which is above the value of a few @xmath261 hz ( corresponding to @xmath262 for white blood cells ) which often provides an upper limit in flow chamber experiments . similar but more complicated scaling arguments can be made for lubrication forces which arise when the cell approaches the wall @xcite . to fully understand the rate of association between a receptor - covered particle in shear flow and a ligand - covered wall , our analysis should be completed by the implementation of an adhesion scenario , which in general should also include molecular determinants like residence times and receptor flexibility . if one assumes that a bond between two encountering molecules is formed with a certain rate , then , the mfpt for encounter as reported here should be a good approximation for the mean adhesion time in the limit of zero shear rate , because in this limit the duration of each encounter should be sufficiently long for the formation of an adhesion contact . then , the proper knowledge of the mfpt could also be used to design a cell sorting experiment . suppose one has a mixture of different cells each bearing some receptors and the wall is covered with one kind of ligand . then , the cells are flowed into the chamber and flow is stopped . certainly , only cells that bear receptors which fit to the ligands can attach to the wall . if the flow is then turned on again , the attached cells will be separated from the other cells . if the no - flow period is much shorter than the mfpt , only a few cells can attach . if the no - flow period is much longer than the mfpt , attached cells might already start to spread and are therefore difficult to remove . only if the no - flow period is of the order of mfpt one gets an appreciable number of weakly attached cells . in this sense our theoretical analysis might be essential for appropriate biotechnological applications . we thank reinhard lipowsky for general support . this work was supported by the german research foundation ( dfg ) through the emmy noether program and by the center for modelling and simulation in the biosciences ( bioms ) at heidelberg . for the numerical implementation of the friction and mobility tensors for a sphere in linear shear flow above a wall we use the results from refs . @xcite . this implementation procedure has been described and tested in detail in ref . @xcite . in this appendix , we briefly summarize it for the sake of completeness . writing the friction tensors in terms of irreducible tensors formed from @xmath263 defines the scalar friction functions . in the case that the normal vector to the wall is @xmath39 , these tensors read [ frictionfunctions ] @xmath264 this defines the scalar friction functions @xmath265 . the scalar friction functions @xmath266 depend only on the inverse distance of the sphere from the wall , that is the dimensionless variable @xmath40 , which takes values from the interval @xmath42 $ ] . the friction functions can be expanded in powers of @xmath41 . the numerically obtained first 20 coefficients of such a series expansion of the dimensionless scalar friction functions [ frictionfunctions2 ] @xmath267 are tabulated in ref . @xcite . for the other three dimensionless scalar friction functions @xmath268 the first 32 coefficients of a series expansion in powers of @xmath41 are tabulated in ref . @xcite . for small values of @xmath41 the series expansion converges quite well and only a few coefficients are needed to obtain accurate results . however , for @xmath269 , i.e. , close to the wall , the friction functions are better described in a lubrication expansion , which reads @xmath270 the coefficients @xmath271 for the eight friction functions defined above can be found in ref . @xcite . in order to match the two limit cases , the the asymptotic expansion of the @xmath269 limit is subtracted from the friction functions @xmath272 leading to a new series expansion : @xmath273 this series is truncated at @xmath274 and the coefficients @xmath275 are calculated from the coefficients @xmath276 . next the coefficients @xmath275 ( @xmath277 ) are not used to calculate the taylor sum , but rather to calculate the pad approximant to this function . the pad approximant is given as @xmath278 where the coefficients @xmath279 are the solution to @xmath280 finally the numerically implemented friction functions become @xmath281 for the calculation of the coefficients @xmath279 of the pad approximant we use the algorithm provided by the numerical recipes @xcite . having implemented the scalar friction functions , the implementation of the mobility tensors proceeds by substituting @xmath282 in the above decomposition of the friction tensors . this defines the scalar mobility functions @xmath283 . using the dimensionless scalar mobility functions can be calculated from the scalar friction functions : [ mobilityfunctions2 ] @xmath284 in we use our implementation to plot the eight dimensionless mobility functions . the limit of an unbounded flow corresponds to @xmath285 and results in @xmath286 where @xmath287 is the unity matrix . ( [ frictionmatrix2 ] ) reduces to @xmath288 which are the well - known stokes laws for the friction force and torque exerted on a sphere moving in a fluid with relative velocity @xmath289 . for the linear shear flow considered here , @xmath290 and @xmath291 . the probability distribution @xmath292 of a brownian particle subject to external force / torque @xmath49 satisfies a continuity equation @xmath293 . the probability flux @xmath294 contains a diffusive and a convective part @xcite : @xmath295 where @xmath296 and @xmath61 are diffusion and mobility matrices , respectively , and @xmath49 is external force . in equilibrium , the flux has to vanish and the probability distribution has to become the boltzmann distribution . this leads to the einstein relation @xmath297 , which is a special case of the fluctuation - dissipation theorem . using and the einstein relation in the continuity equation leads to the smoluchowski equation @xcite : @xmath298 we now will derive the equivalent langevin equation . in the case of constant mobility ( _ additive noise _ ) , e.g. , @xmath299 , the appropriate langevin equation is given by @xmath300 where @xmath301 is a gaussian white noise term and the stratonovich interpretation is used as explained in the main text . however if @xmath61 depends on @xmath302 ( _ multiplicative noise _ ) , an additional drift term occurs in the langevin equation @xmath303 the following derivation of the drift term @xmath68 proceeds in two steps @xcite . first we perform a coordinate transformation which makes the noise additive . in the case of additive noise the langevin equation ( [ langevin - constant ] ) and the fokker - planck equation ( [ smoluchowski ] ) are equivalent . then starting from the fokker - planck equation in the new coordinates we perform the transformation back to the old coordinates . requiring the transformed fokker - planck equation to be of the same form as in , determines the drift term @xmath68 . as we use the stratonovich interpretation for the noise process the usual rules for differentiation and integration apply and we can perform the following coordinate transformation @xmath304 with some regular matrix @xmath305 . the langevin equation for the transformed coordinates then reads @xmath306 from the requirement that @xmath307 , that is @xmath308 we can fix @xmath305 to be the inverse of a matrix @xmath309 with @xmath310 as @xmath61 is a symmetric positive definite matrix , it is always possible to find a matrix @xmath309 with @xmath311 . defining @xmath312 the new langevin equation for the primed coordinates and with additive noise reads @xmath313 the corresponding probability distribution @xmath314 is the solution of the smoluchowski equation @xmath315 next we transform ( [ smol-02 ] ) back to the unprimed coordinates . the preservation of probability requires that @xmath316 where @xmath317 is the jacobian of the coordinate transformation @xcite : @xmath318 inserting ( [ proppreserv ] ) into ( [ smol-02 ] ) gives @xmath319 dividing by @xmath317 we obtain for the first term on the right hand side of ( [ smol - mid ] ) @xmath320 here we made use of the identities @xmath321 again using the identity ( [ trafo-03 ] ) the second term of the right hand side of ( [ smol - mid ] ) can be evaluated to be @xmath322 adding both terms and inserting the definitions ( [ def - b ] ) and ( [ def - fnew ] ) we have @xmath323 comparing this with the required result ( [ smoluchowski ] ) we can read off @xmath68 @xmath324 finally shifting @xmath325 we obtain the langevin equation as given by combined with . in order to solve numerically we use an euler algorithm . as the physical situation requires to use the stratonovich interpretation of the noise term @xmath301 , the displacement @xmath326 of a particle from time @xmath41 to time @xmath327 depends on the position of the particle at time @xmath328 , which is not known at time @xmath41 . as usual , this problem is solved by rewriting the langevin equation in the it - version . then the noise term can be evaluated at time @xmath41 and as a compensation an additional drift term @xmath329 is added to @xcite . because @xmath330 , we arrive at . in this equation , the random displacements @xmath331 must satisfy @xmath332 following ref . @xcite , @xmath333 is calculated from a weighted sum of normal deviate random numbers @xmath334 satisfying @xmath335 . this sum is given by @xmath336 where the weighting factors are the elements of the matrix @xmath309 defined in ( [ def - b ] ) . they can recursively be calculated according to @xmath337 in the case of a sphere above a wall we obtain the following dimensionless weighting factors ( cf . @xcite ) @xmath338 as pointed out in ref . @xcite , using the euler method , instead of normal deviate random variables any uncorrelated random variable @xmath339 can be chosen , as long as they fulfill the required relation for the first moments @xmath340 . thus , it is much faster to generate the random numbers according to @xmath341 , with @xmath342 being uncorrelated random variables uniformly distributed in @xmath42 $ ] . for the calculation of the random numbers we use the pseudo random number generator ran3 from the numerical recipes @xcite . calculating the new configuration after each time - step using ( [ langevin - euler ] ) is straightforward for the spatial degrees of freedom . for the update of the orientation of the sphere we use a coordinate system spanned by three orthonormal basis - vectors @xmath343 . the origin of this coordinate system shall be identical with the center of mass of the sphere and the relative orientation of this system and of the sphere are kept fixed . given then an orientation update form ( [ langevin - euler ] ) @xmath344 , we decompose each of the basis vectors @xmath345 into a component parallel to @xmath133 denoted by @xmath346 and a component perpendicular to @xmath133 denoted by @xmath347 ( the index @xmath348 is dropped for the sake of simplicity ) . these components are given by @xmath349 then the orientation update affects only @xmath347 and the updated @xmath350 is given by ( with @xmath351 ) @xmath352 applying the euler algorithm to a mean first passage time problem gives rise to two sorts of errors . first there exists the statistical error , which is proportional to @xmath353 , where @xmath354 is the number of iterations the algorithm is applied . the extent of the statistical error of the measured mean value can be calculated during the simulation . for the measurements performed in sections [ sec:2d ] and [ sec:3d ] typically @xmath355 iterations where chosen resulting in statistical errors in the range of @xmath356 . error - bars in these sections refer to the statistical error . the systematic error for the mean first passage time calculated by use of an euler algorithm scales with @xmath357 , although the error of the particle position is only of the order of @xmath58 @xcite . thus to decrease the systematic error by a factor of 10 one must increase the numerical cost by a factor of 100 . one way to obtain accurate results at moderate numerical cost is to measure the mean first passage time for various intermediate numerical time steps . fitting these results to @xmath358 allows the extrapolation to @xmath359 . shows an example where this procedure was applied to the case of homogeneous coverage as considered in . the resulting mean first passage time then deviates by @xmath360 from the value obtained from quadrature of . this is the same accuracy as we have for the implemented mobility functions themselves ( cf . appendix [ appendix : mobility ] ) .
the recent discovery @xcite of a new particle with properties consistent with a standard model ( sm ) higgs boson is clearly the most significant news from the large hadron collider ( lhc ) . this discovery was supported by evidence for a higgs boson found by the cdf and d0 collaborations at the tevatron @xcite and completes our picture of the sm . however , the sm leaves many fundamental questions open perhaps the most pressing issue being that the sm does not explain the value of the electroweak scale , the higgs mass , itself . clearly , a prime goal after the discovery is to thoroughly test the sm nature of this higgs - like signal . with the measurements in various channels , a comprehensive study of the properties of the higgs - like state becomes possible and has the potential for revealing whether or not the higgs sector is as simple as envisioned in the sm . in a recent study @xcite , we analyzed the extent to which the results from the lhc and tevatron , as published by the end of 2012 , constrain the couplings of the higgs boson - like state at @xmath0 . to this end we assumed that only sm particles appear in the higgs decays , but tree - level higgs couplings to the up - quarks , down - quarks and vector bosons , relative to the sm are free parameters . moreover , we considered the case that new particles appearing in loops modify the effective higgs couplings to gluons and/or photons . we found that the sm expectation is more than @xmath7 away from fits in which : a ) there is some non - sm contribution to the @xmath8 coupling of the higgs ; or b ) the sign of the top quark coupling to the higgs is opposite that of the sm . in both these cases good fits with @xmath9-values @xmath10 can be achieved . since option b ) is difficult to realize in realistic models , it would seem that new physics contributions to the effective couplings of the higgs are preferred . in this paper we carry our work a step further and investigate the extent to which current data constrain invisible decays ( @xmath11 , where @xmath12 is the lightest susy particle ) or undetected decays ( such as @xmath13 , where @xmath14 is a light cp - odd , perhaps singlet scalar ) of the higgs - like particle . with a small @xmath15 of about @xmath16 . for the global fits , there is no difference between invisible and yet undetected decays , and we will use the term invisible " for both , genuine invisible and merely undetected . when talking about specific signatures , like monojets + missing energy , it is however important to make the distinction in this case we will use `` invisible '' in the strict sense . ] our parametrization and fitting procedure is the same as in @xcite , where we introduced scaling factors @xmath17 relative to sm couplings . we treat the couplings to up - type and down - type fermions , @xmath18 and @xmath19 , as independent parameters ( but we only consider the case @xmath20 , and we assume that the @xmath21 are family universal ) . moreover , we assume a custodial symmetry in employing a single @xmath22 . in addition to the tree - level couplings given above , the @xmath23 has couplings to @xmath24 and @xmath8 that are first induced at one loop and are completely computable in terms of @xmath18 , @xmath19 and @xmath1 if only loops containing sm particles are present . we define @xmath25 and @xmath26 to be the ratio of these couplings so computed to the sm ( @xmath27 ) values . however , in some of our fits we will also allow for additional loop contributions @xmath28 and @xmath29 from new particles ; in this case @xmath30 and @xmath31 . limits on `` invisible '' higgs decays from global fits to lhc data were obtained previously in @xcite and within the mssm in @xcite . since then , much more data has become available . we therefore find it worthwhile to re - investigate the status of invisible decays . moreover , we consider more general deviations from sm couplings than the previous works . the possibility of probing directly the ( genuine ) invisible branching fraction of the higgs in various channels at the lhc has been investigated already some time ago @xcite . it was revisited more recently in light of the 125 gev higgs signal at the lhc in @xcite . the channels considered were monojets or 2 jets plus missing energy . associated production , leading to high-@xmath32 @xmath33 s plus missing energy . however , stronger limits were obtained from monojet searches . ] these make it possible to constrain the product of the higgs production cross section in the gluon gluon fusion ( ggf ) , or vector boson fusion ( vbf ) and vh associated production modes , times the invisible branching fraction . specifically , they put constraints on @xmath34 and @xmath35 . we will not include these direct limits in our fits but will comment on the impact of these probes on the valid parameter space . probing yet undetected @xmath13 decays ( with the pseudoscalars further decaying in various channels ) was discussed in @xcite results depend strongly on the manner in which the @xmath14 decays . the results of our global fits are presented in section 2 . in section 3 , we discuss further probes of invisible or undetected decays . in the case of higgs decays into dm particles , a strong interplay with direct dm searches arises . this is discussed in section 4 . a summary and conclusions are given in section 5 . distributions for the branching ratio of invisible higgs decays . the full , dashed , and dotted lines correspond , respectively , to the cases of 1 ) sm couplings , 2 ) arbitrary @xmath28 and @xmath29 , and 3 ) deviations of @xmath36 from unity . in addition , we show as dash - dotted line the variant of case 3 ) with @xmath37 and @xmath38 . [ fig : onedim ] ] as mentioned , we consider several cases in our analysis . we start the discussion with the simplest case , case 1 ) , of an sm higgs boson augmented by invisible / undetected decays , we require that all couplings be equal to the sm values ( @xmath39 ) but allow for ( yet ) invisible decays of the higgs state in addition to the normal decays to sm particles . we first recall that , for the most part , the observed rates in many production - decay channels are not far from the sm - predictions and , in particular , are not suppressed . in addition , there are a few final states that have a somewhat enhanced rate relative to the sm , most notably the @xmath8 final state . since invisible decays reduce the branching ratio to the ( visible ) sm final states , it is to be expected that @xmath40 is strongly limited . this is confirmed in our fit . we find that @xmath41 ( @xmath42 ) at @xmath7 ( @xmath43 ) . the @xmath44 distribution as function of the invisible rate for case 1 ) is shown as the solid line in fig . [ fig : onedim ] . the @xmath44 at the minimum is @xmath45 , the same as for the sm @xcite . with 20 degrees of freedom ( d.o.f . ) this corresponds to a @xmath9-value of @xmath46 , as compared to a @xmath9-value of @xmath47 for the pure sm ( 21 d.o.f . ) . let us now turn to case 2 ) , where we allow for extra loop contributions to the @xmath24 and @xmath8 couplings of the higgs boson , denoted @xmath28 and @xmath29 , respectively . such loops would involve particles beyond those present in the sm . if these bsm particles receive mass from the higgs mechanism , then they will give contributions to @xmath28 and @xmath29 that approach a constant value as their mass increases in size they do not decouple . allowing for arbitrary @xmath28 and @xmath29 the minimum @xmath44 decreases from the sm fit value of @xmath48 to @xmath49 with the best values of @xmath50 and @xmath51 for a non - zero value of the invisible rate , @xmath52 . the one - dimensional ( 1d ) distribution of @xmath53 as a function of @xmath40 , after profiling over @xmath28 and @xmath29 , is plotted as the dashed line in fig . [ fig : onedim ] . the 95% cl upper limit on the invisible rate increases dramatically to @xmath54 . larger @xmath40 can be accommodated by increasingly large values of @xmath28 so that the overall production rates in sm final states from @xmath24 fusion processes remain the same . rates in sm final states for vector boson fusion ( vbf ) induced processes will decline somewhat and this is the primary reason for the increase in @xmath44 with increasing @xmath40 . the @xmath55 , @xmath7 and @xmath43 ranges in the @xmath28 versus @xmath29 plane are shown in the top plot in fig . [ fig : delta - cg - cp ] . it is interesting to compare to the case of arbitrary @xmath28 and @xmath29 without invisible decays , for which the @xmath55 and @xmath7 ranges are shown as black and gray contours in fig . [ fig : delta - cg - cp ] . allowing for invisible decays , we find @xmath56 , @xmath57 ( @xmath58 ) as compared to @xmath59 , @xmath60 ( @xmath61 ) when invisible decays are absent . contours of @xmath40 versus @xmath28 and versus @xmath29 are shown in the bottom row of fig . [ fig : delta - cg - cp ] . we see that , as mentioned , larger @xmath40 can be accommodated by increasing @xmath28 . on the other hand , larger negative @xmath28 , as for example induced by light stops with large mixing , disfavors invisible decays . finally , @xmath40 is rather insensitive to @xmath29 . contours obtained by allowing for additional loop contributions @xmath62 and @xmath63 to the effective couplings of the higgs to gluons and photons . the red , orange and yellow areas ( black , grey and light grey areas in greyscale print ) are the 68% , 95% and 99.7% cl regions , respectively . the black and grey ellipses in the top plot show the 68% and 95% cl contours when invisible decays are absent . we find @xmath64 , @xmath65 and @xmath66 ( where the errors correspond to 1d profiling ) . at the best fit point , marked as a white star , we find @xmath67 ( for 18 d.o.f . ) . [ fig : delta - cg - cp ] , title="fig : " ] + contours obtained by allowing for additional loop contributions @xmath62 and @xmath63 to the effective couplings of the higgs to gluons and photons . the red , orange and yellow areas ( black , grey and light grey areas in greyscale print ) are the 68% , 95% and 99.7% cl regions , respectively . the black and grey ellipses in the top plot show the 68% and 95% cl contours when invisible decays are absent . we find @xmath64 , @xmath65 and @xmath66 ( where the errors correspond to 1d profiling ) . at the best fit point , marked as a white star , we find @xmath67 ( for 18 d.o.f . ) . [ fig : delta - cg - cp ] , title="fig : " ] contours obtained by allowing for additional loop contributions @xmath62 and @xmath63 to the effective couplings of the higgs to gluons and photons . the red , orange and yellow areas ( black , grey and light grey areas in greyscale print ) are the 68% , 95% and 99.7% cl regions , respectively . the black and grey ellipses in the top plot show the 68% and 95% cl contours when invisible decays are absent . we find @xmath64 , @xmath65 and @xmath66 ( where the errors correspond to 1d profiling ) . at the best fit point , marked as a white star , we find @xmath67 ( for 18 d.o.f . ) . [ fig : delta - cg - cp ] , title="fig : " ] in case 3 ) , we allow for arbitrary @xmath36 while varying @xmath40 . we do not allow for extra contributions from bsm particles to the @xmath24 and @xmath8 higgs couplings , we assume @xmath68 and @xmath69 . it is interesting to consider several sub - cases . for general @xmath36 ( limited however to the range @xmath70 ) , we find @xmath71 , @xmath72 and @xmath73 . the best fit point has @xmath74 for 17 d.o.f . and is basically the same as without invisible decays . a large @xmath40 can yield a small @xmath75 when simultaneously increasing both @xmath76 ( so as to increase the @xmath24 production rates ) and @xmath1 ( so as to increase the vbf production rates ) in order to compensate for the decreasing branching ratios to sm particle final states . ( we also need to increase @xmath77 in order to have enough decays into @xmath78 and @xmath79 . ) as in the fits of @xcite , the minimum @xmath44 is achieved for negative @xmath18 , something that is very difficult in the context of most theoretical models , see the comment at the end of this section . in addition , the large values of @xmath80 required for a good fit with large @xmath40 imply an enhancement of the isospin @xmath81 cross section , which may be achieved if the higgs state at @xmath82 125 gev mixes with higher su(2 ) representations @xcite . ( within extensions of the higgs sector by an arbitrary number of su(2 ) doublets and/or singlets only , one generally obtains @xmath83 . ) allowing for large negative @xmath18 and large positive @xmath1 ( up to @xmath84 , @xmath85 ) , @xmath40 can only be constrained to @xmath86 at 95% cl , see fig . [ fig : cu - cd - cv ] . allowing for deviations of @xmath18 , @xmath19 , @xmath1 from 1 , but without extra loop contributions , @xmath87 . same color code as in fig . [ fig : delta - cg - cp ] . allowing up to 100% deviations in @xmath18 and @xmath1 , the @xmath7 ( @xmath43 ) limit is @xmath88 ( 0.92 ) . [ fig : cu - cd - cv ] , title="fig : " ] allowing for deviations of @xmath18 , @xmath19 , @xmath1 from 1 , but without extra loop contributions , @xmath87 . same color code as in fig . [ fig : delta - cg - cp ] . allowing up to 100% deviations in @xmath18 and @xmath1 , the @xmath7 ( @xmath43 ) limit is @xmath88 ( 0.92 ) . [ fig : cu - cd - cv ] , title="fig : " ] but restricting either @xmath89 ( left ) or @xmath90 ( right ) . [ fig : cu - cd - cv-2 ] , title="fig : " ] but restricting either @xmath89 ( left ) or @xmath90 ( right ) . [ fig : cu - cd - cv-2 ] , title="fig : " ] restricting the fit to @xmath89 , as relevant for a higgs sector consisting of doublets+singlets , the minimum @xmath44 remains at @xmath91 but the @xmath53 increases rapidly with the invisible rate . we find @xmath92 in this case , and @xmath93 at 95% cl . obviously , imposing @xmath94 not only greatly restricts the production cross sections that can be achieved in vector boson fusion but also restricts the two - photon partial width of the higgs and therefore decreases the value of @xmath40 that can allow reasonable consistency with the experimental observations . @xmath40 versus @xmath18 for @xmath89 is shown in the left plot in fig . [ fig : cu - cd - cv-2 ] . another interesting case is to require @xmath90 while allowing @xmath80 . in this case , the minimum @xmath44 point is characterized by @xmath95 , @xmath96 , @xmath97 , and the fit worsens to @xmath98 . further , @xmath40 can only be constrained to @xmath99 at 95% cl . for illustration , see the right - hand - side plot in fig . [ fig : cu - cd - cv-2 ] . perhaps most interesting from the theoretical point of view is the case in which @xmath37 and @xmath89 is required . with these constraints , @xmath100 ( not far from the sm fit value of @xmath48 ) and the 95% cl limit on @xmath40 is @xmath101 . the @xmath53 distribution for this case is shown as dash - dotted line in fig . [ fig : onedim ] . for @xmath40 versus @xmath18 and @xmath1 , see fig . [ fig : brinv - cu - cd - pos - cvle1 ] . contours obtained by requiring @xmath89 and @xmath37 . same color code as in previous figures . [ fig : brinv - cu - cd - pos - cvle1 ] , title="fig : " ] contours obtained by requiring @xmath89 and @xmath37 . same color code as in previous figures . [ fig : brinv - cu - cd - pos - cvle1 ] , title="fig : " ] allowing for deviations of @xmath18 , @xmath19 , @xmath1 from 1 . in the top row of plots , @xmath38 but @xmath102 may have either sign . in the bottom row , @xmath89 and @xmath37 . color code _ etc . _ as in previous figures . [ fig : cu - cd - cv - par ] , title="fig : " ] allowing for deviations of @xmath18 , @xmath19 , @xmath1 from 1 . in the top row of plots , @xmath38 but @xmath102 may have either sign . in the bottom row , @xmath89 and @xmath37 . color code _ etc . _ as in previous figures . [ fig : cu - cd - cv - par ] , title="fig : " ] allowing for deviations of @xmath18 , @xmath19 , @xmath1 from 1 . in the top row of plots , @xmath38 but @xmath102 may have either sign . in the bottom row , @xmath89 and @xmath37 . color code _ etc . _ as in previous figures . [ fig : cu - cd - cv - par ] , title="fig : " ] + allowing for deviations of @xmath18 , @xmath19 , @xmath1 from 1 . in the top row of plots , @xmath38 but @xmath102 may have either sign . in the bottom row , @xmath89 and @xmath37 . color code _ etc . _ as in previous figures . [ fig : cu - cd - cv - par ] , title="fig : " ] allowing for deviations of @xmath18 , @xmath19 , @xmath1 from 1 . in the top row of plots , @xmath38 but @xmath102 may have either sign . in the bottom row , @xmath89 and @xmath37 . color code _ etc . _ as in previous figures . [ fig : cu - cd - cv - par ] , title="fig : " ] allowing for deviations of @xmath18 , @xmath19 , @xmath1 from 1 . in the top row of plots , @xmath38 but @xmath102 may have either sign . in the bottom row , @xmath89 and @xmath37 . color code _ etc . _ as in previous figures . [ fig : cu - cd - cv - par ] , title="fig : " ] the relevant 2d correlations between parameters , illustrating the discussion above , are shown in fig . [ fig : cu - cd - cv - par ] for : the fit requiring @xmath38 but allowing arbitrary signs for @xmath102 ( top row ) ; and the fit requiring both @xmath89 and @xmath37 ( bottom row ) . in order to see the impact of invisible decays on the coupling fits , we have superimposed the @xmath55 and @xmath7 regions from @xcite obtained for @xmath103 . let us end this section with a comment on @xmath104 . a negative sign of @xmath105while maintaining a positive sign of @xmath106is actually not easy to achieve . if the top quark and higgs bosons are considered as fundamental fields , it would require that the top quark mass is induced dominantly by the vev of at least one additional higgs boson which is not the higgs boson considered here , and leads typically to various consistency problems as discussed , , in @xcite . truly invisible higgs decays can be probed at the lhc in monojet searches in either the ggf mode where a gluon is radiated from the initial state , or in vbf when one of the jets is missed . invisible decays can also be probed in @xmath107 associated production with @xmath108 . in @xcite , sensitivity to the monojet searches is phrased in terms of limits on @xmath109 a 95% cl upper limit of @xmath110 was obtained using the cms monojet analysis at @xmath111 tev and @xmath112 fb@xmath113 @xcite . the relative contributions of the gluon and vector boson fusion production mechanisms were assumed to be the same as in the sm after the analysis cuts . of course , this need not apply if @xmath36 are allowed to vary . ( assuming only one production channel , the 95% cl upper limits are @xmath114 and @xmath115 . ) the projected limit for @xmath116 tev and @xmath117 fb@xmath113 is @xmath118 . versus @xmath119 , on the left for the fit of @xmath28 and @xmath29 , on the right for the fit of @xmath36 with @xmath38 . the red , orange and yellow regions are 68% , 95% and @xmath120% cl regions , respectively . the dark ( light ) blue bands show the constraints from the monojet search , @xmath121 ( @xmath122 ) at 95% cl , with the bands obtained by varying @xmath29 ( left plot ) or @xmath1 ( right plot ) within their fitted @xmath7 ranges . [ fig : monojet ] , title="fig : " ] versus @xmath119 , on the left for the fit of @xmath28 and @xmath29 , on the right for the fit of @xmath36 with @xmath38 . the red , orange and yellow regions are 68% , 95% and @xmath120% cl regions , respectively . the dark ( light ) blue bands show the constraints from the monojet search , @xmath121 ( @xmath122 ) at 95% cl , with the bands obtained by varying @xmath29 ( left plot ) or @xmath1 ( right plot ) within their fitted @xmath7 ranges . [ fig : monojet ] , title="fig : " ] since the signals are also proportional to the possibly non - standard higgs production cross section @xmath123 in the ggf mode or @xmath124 in the vbf mode , @xmath34 and @xmath125 . upper limits on @xmath126 will thus constrain @xmath40 as function of @xmath119 , as shown in fig . [ fig : monojet ] for case 2 ) on the left and case 3 ) with @xmath89 on the right . these plots should be compared to @xmath40 versus @xmath28 in fig . [ fig : delta - cg - cp ] and @xmath40 versus @xmath18 in fig . [ fig : cu - cd - cv-2 ] . the dark ( light ) blue bands indicate @xmath127 ( @xmath122 ) , with the band obtained by varying @xmath29 or @xmath1 within @xmath7 . as can be seen , the monojet searches are already quite complementary in constraining invisible higgs decays when there is a large increase in the production cross section . another analysis @xcite considered searching for invisible higgs in the 2 jets and missing @xmath128 channel showing that a @xmath129 signal could be observed at 8 tev for @xmath130 fb@xmath113 for an sm production cross section provided @xmath131 , while the lhc at 14 tev could probe @xmath132 with @xmath133 fb@xmath113 . since the 2 jets @xmath134 channel is dominated by vbf production ( 86% after analysis cuts ) , this can be useful to constrain the cases with @xmath135 . for example , @xmath136 @xmath137 could be probed at 8 tev for @xmath138 and @xmath139 @xmath140 , thus covering a large fraction of the currently allowed parameter space at large values of @xmath141 in fig . 3 and fig . 4 ( right panel ) . let us finally comment on decays that may in principle be detectable . the ability of the lhc to probe for higgs decays into light pseudoscalars , @xmath13 , depends on the decays of the @xmath14 s . the most likely @xmath14 decays are @xmath142 , dominant for @xmath143 , and @xmath144 , dominant for @xmath145 . a review , with detailed referencing , of the possibilities for the lhc in various production modes in the cases of these decays is given in @xcite . in two - higgs - doublet models @xmath146 ( where @xmath147 is a light quark , @xmath148 or @xmath149 ) can be significant if @xmath150 or @xmath151 . lhc sensitivity in this case has been examined for @xmath107 production in @xcite . in all the different @xmath14 decay scenarios pretty much full lhc luminosity , @xmath152 at @xmath153 is required to place strong limits ( @xmath154 at 95% cl ) on @xmath13 decays . another potentially interesting decay channel , that may have escaped observation , is @xmath155 , with a soft photon . this may arise for instance in @xmath156 decays followed by @xmath157 , where @xmath158 denotes a gravitino @xcite or a goldstino @xcite . assuming that the invisible particle which the higgs potentially decays into is the dark matter of the universe , the lhc bounds on @xmath40 can be turned into bounds on the dm scattering off nucleons , mediated by higgs exchange , _ cf . _ @xcite . these bounds are often much stronger than the current limits from xenon100 for @xmath159 gev ( _ i.e. _ @xmath160 ) . both the invisible width of the higgs and the spin - independent cross - section for scattering on protons depend on the square of the higgs dm dm coupling @xmath161 . if the dm is a majorana fermion , @xmath162 , the invisible width arising from @xmath163 decays is given by @xmath164 where @xmath165 and @xmath166 is defined by @xmath167 . in case of the dm being a real scalar , @xmath168 , we have @xmath169 and @xmath170 the spin - independent cross - section for scattering on a nucleon , considering only the higgs exchange diagram , can then be directly related to the invisible width of the higgs : @xmath171 ^ 2 \ , \label{eq : sigsi}\ ] ] with @xmath172 for a majorana fermion and @xmath173 for a real scalar ; @xmath174 is the reduced mass and @xmath175 are the quark ( gluon ) coefficients in the nucleon . we take the values @xmath176 , @xmath177 , and @xmath178 from an average of recent lattice results @xcite . the gluon and heavy quark ( @xmath179 ) coefficients are related to those of light quarks , and @xmath180 at leading order . since the contribution of heavy quarks to the scattering amplitude originates from their contribution to the @xmath181 coupling , we write the effect of @xmath28 , the last term in eq . ( [ eq : sigsi ] ) , in terms of an additional top quark contributing to the @xmath181 coupling ; numerically @xmath182 with only the sm top - quark contribution taken into account for computing @xmath183 . for the numerical evaluation of @xmath184 , we use micromegas @xcite in which the relation between the heavy quark coefficients and the light ones are modified by qcd corrections . this amounts to taking @xmath185 as a function of the mass of the dm particle , for @xmath186 ( from bottom to top ) for the case of a majorana @xmath162 ( left panel ) or a real scalar @xmath168 ( right panel ) when @xmath187 and @xmath188 , an sm higgs plus invisible decays . the red dashed curves show the xenon100 exclusion limit . [ brinv - sigmasi ] , title="fig : " ] as a function of the mass of the dm particle , for @xmath186 ( from bottom to top ) for the case of a majorana @xmath162 ( left panel ) or a real scalar @xmath168 ( right panel ) when @xmath187 and @xmath188 , an sm higgs plus invisible decays . the red dashed curves show the xenon100 exclusion limit . [ brinv - sigmasi ] , title="fig : " ] the results for @xmath184 versus the dm mass and for different @xmath40 are displayed in fig . [ brinv - sigmasi ] for a majorana fermion ( left panel ) and a real scalar ( right panel ) assuming @xmath187 . as can be seen , for a majorana fermion the current xenon100 limits @xcite exclude , for example , @xmath189 when @xmath190 . for scalar dm , the cross sections are larger , and xenon100 excludes @xmath189 for @xmath191 gev . these limits become much stronger when @xmath105 and/or @xmath192 are large provided they have the same sign . further , these limits become stronger when we include a non - zero value of @xmath193 . for example , for @xmath194 we find that @xmath184 increases by a factor 1.8 as compared to the case @xmath195 for any given value of @xmath40 . this increase is due in part to the new contribution in eq . ( [ eq : sigsi ] ) and in part because a larger coupling of the dm to the higgs is necessary to maintain a constant value of @xmath40 . note that imposing universality of quark couplings to the higgs has an impact on our predictions for @xmath184 since all quark flavors contribute to this observable , whereas universality plays basically no role for higgs decays as only the third generation is important . when @xmath196 and @xmath197 , corresponding to the best fit for case 3 ) , there is a destructive interference between the @xmath198-type and @xmath199-type quark contributions such that @xmath184 is much below the current limit . note however that a negative sign of @xmath105 would require that @xmath106 is induced dominantly by the vev of a higgs boson which is not the higgs boson considered here ; if such a higgs boson also couples to dark matter it could then contribute significantly to the si cross section . without a complete model for the higgs sector it is therefore difficult to make generic predictions in this case . when the dm candidate is a dirac fermion and one assumes the same amount of matter and anti - matter in the early universe , the results for @xmath184 are simply a factor @xmath200 lower then those obtained in the majorana case . however if this fermion also couples to the @xmath33 , this gives an additional positive contribution to @xmath184 , thus leading to stronger constraints from direct detection experiments . similar arguments hold for the case of a complex scalar , as compared to a real scalar . assuming that the 125 gev state observed at the lhc is , indeed , a higgs boson a very important question is whether or not it has decays to non - sm particles that escape undetected . truly invisible decays include , for instance , @xmath201 lsp+lsp ( the lsp being the lightest supersymmetric particle in r - parity conserving supersymmetry and a dm candidate ) while undetected , but not intrinsically invisible , decays are typified by @xmath13 where @xmath14 is a light pseudoscalar of an extended higgs sector . in these and many other cases , `` invisible '' @xmath23 decays provide a portal to bsm physics that might prove hard to detect in any other way . in this paper , we have assessed the extent to which currently available data constrain invisible ( or undetected ) @xmath23 decays . by performing fits to all public data from the lhc and the tevatron experiments , we have shown that the 95% cl limits for @xmath40 obtained depend very much upon the higgs sector model . assuming a higgs boson with sm couplings , @xmath202 is allowed at 95% cl . allowing the @xmath18 , @xmath19 and @xmath1 reduced coupling factors ( that are defined as the coefficients multiplying the up - quark , down - quark and vector boson couplings relative to the sm values ) to deviate from unity , much larger invisible decay rates are possible . in the most theoretically motivated case where @xmath90 , @xmath203 and @xmath94 the latter being required for any two - doublet or two - doublet plus singlets model we find @xmath204 at 95% cl . limits on invisible ( or yet undetected ) higgs decays also depend strongly on whether or not there are bsm particles that provide extra loop contributions to the @xmath24 and @xmath8 couplings of the @xmath23 . in the simplest case of an sm - like higgs with bsm loop contributions to its @xmath24 and @xmath8 couplings , the invisible decay rate can go up to about 60% at 95% cl . in the absence of invisible decays , the best fits to the lhc data suggest significant deviations of @xmath18 and/or @xmath205 from unity . we have shown that including the additional possibility of invisible / undetected @xmath23 decays makes even larger deviations accompanied by large @xmath40 values consistent with lhc observations . if @xmath40 really were as large as the @xmath7 limits we derived , one might hope that direct detection of invisible and/or hard - to - detect @xmath23 decays would be possible . estimates suggest that @xmath206 at @xmath153 will typically be required . finally , we have also shown that if @xmath207 is due to @xmath23 decays to a pair of dm particles , there are significant constraints on the size of @xmath40 from the non - observation of spin - independent dm scattering on nucleons , the most important such limits currently being those from the xenon100 experiment . these constraints are much stronger for scalar dm than for majorana or dirac fermions . overall , our results suggest a continued competition between limits on @xmath184 and those on @xmath40 as direct detection experiments achieve improved sensitivity and increasingly accurate measurements of the properties of the @xmath23 become available with future lhc running . in short , precision measurements of the properties of the @xmath23 could well continue to provide the strongest constraints on a number of types of bsm physics , including the existence of light ( mass @xmath208 ) weakly interacting massive or hard - to - detect bsm particles . while this paper was being refereed , atlas and cms presented major updates of their higgs results based on @xmath209 fb@xmath113 of data in most channels . taking these new results into account , we find @xmath210 at 95% cl for a higgs boson with sm couplings , and @xmath211 at 95% cl when @xmath90 , @xmath203 and @xmath94 are left free to vary @xcite . gb thanks a. pukhov for useful discussions . this work was supported in part by us doe grant de - fg03 - 91er40674 and by in2p3 under contract pics fr usa no . 5872 . gb and sk acknowledge morever partial support from the french anr dmastrolhc . ue acknowledges partial support from the french anr lfv - cpv - lhc , anr str - cosmo and the european union fp7 itn invisibles ( marie curie actions , pitn - ga-2011 - 289442 ) . this work originated from discussions started at the aspen center for physics ( acp ) which is supported by the national science foundation grant no . phy-1066293 ; gb , ue , jfg and sk thank the acp for hospitality and an inspiring working atmosphere . gb , jfg and sk also thank the galileo galilei institute for theoretical physics ( ggi florence ) for hospitality and the infn for partial support . finally , jfg acknowledges the hospitality of the kavli institute for theoretical physics which is supported by the national science foundation under grant no . nsf phy11 - 25915 . 10 collaboration , g. aad _ et al . _ , `` observation of a new particle in the search for the standard model higgs boson with the atlas detector at the lhc , '' _ phys.lett . _ * b716 * ( 2012 ) 129 , http://www.arxiv.org/abs/1207.7214[1207.7214 ] . collaboration , s. chatrchyan _ et al . _ , `` observation of a new boson at a mass of 125 gev with the cms experiment at the lhc , '' _ phys.lett . _ * b716 * ( 2012 ) 3061 , http://www.arxiv.org/abs/1207.7235[1207.7235 ] . , `` updated combination of cdf and d0 searches for standard model higgs boson production with up to 10.0 fb@xmath113 of data , '' http://www.arxiv.org/abs/1207.0449[1207.0449 ] . g. belanger , b. dumont , u. ellwanger , j. gunion , and s. kraml , `` higgs couplings at the end of 2012 , '' http://www.arxiv.org/abs/1212.5244[1212.5244 ] . j. r. espinosa , m. muhlleitner , c. grojean , and m. trott , `` probing for invisible higgs decays with global fits , '' _ jhep _ * 1209 * ( 2012 ) 126 , http://www.arxiv.org/abs/1205.6790[1205.6790 ] . d. carmi , a. falkowski , e. kuflik , t. volansky , and j. zupan , `` higgs after the discovery : a status report , '' _ jhep _ * 1210 * ( 2012 ) 196 , http://www.arxiv.org/abs/1207.1718[1207.1718 ] . p. p. giardino , k. kannike , m. raidal , and a. strumia , `` is the resonance at 125 gev the higgs boson ? _ * b718 * ( 2012 ) 469474 , http://www.arxiv.org/abs/1207.1347[1207.1347 ] . p. giardino , k. kannike , m. raidal , and a. strumia , `` reconstructing higgs boson properties from the lhc and tevatron data , '' _ jhep _ * 1206 * ( 2012 ) 117 , http://www.arxiv.org/abs/1203.4254[1203.4254 ] . n. desai , b. mukhopadhyaya , and s. niyogi , `` constraints on invisible higgs decay in mssm in the light of diphoton rates from the lhc , '' http://www.arxiv.org/abs/1202.5190[1202.5190 ] . j. gunion , `` detecting an invisibly decaying higgs boson at a hadron supercollider , '' _ phys.rev.lett . _ * 72 * ( 1994 ) 199202 , http://www.arxiv.org/abs/hep-ph/9309216[hep-ph/9309216 ] . d. choudhury and d. roy , `` signatures of an invisibly decaying higgs particle at lhc , '' _ _ * b322 * ( 1994 ) 368373 , http://www.arxiv.org/abs/hep-ph/9312347[hep-ph/9312347 ] . o. j. eboli and d. zeppenfeld , `` observing an invisible higgs boson , '' _ phys.lett . _ * b495 * ( 2000 ) 147154 , http://www.arxiv.org/abs/hep-ph/0009158[hep-ph/0009158 ] . r. godbole , m. guchait , k. mazumdar , s. moretti , and d. roy , `` search for ` invisible ' higgs signals at lhc via associated production with gauge bosons , '' _ phys.lett . _ * b571 * ( 2003 ) 184192 , http://www.arxiv.org/abs/hep-ph/0304137[hep-ph/0304137 ] . h. davoudiasl , t. han , and h. e. logan , `` discovering an invisibly decaying higgs at hadron colliders , '' _ phys.rev . _ * d71 * ( 2005 ) 115007 , http://www.arxiv.org/abs/hep-ph/0412269[hep-ph/0412269 ] . r. e. shrock and m. suzuki , `` invisible decays of higgs bosons , '' _ _ * b110 * ( 1982 ) 250 . y. bai , p. draper and j. shelton , `` measuring the invisible higgs width at the 7 and 8 tev lhc , '' _ jhep _ * 1207 * ( 2012 ) 192 [ arxiv:1112.4496 [ hep - ph ] ] . a. djouadi , a. falkowski , y. mambrini , and j. quevillon , `` direct detection of higgs - portal dark matter at the lhc , '' http://www.arxiv.org/abs/1205.3169[1205.3169 ] . d. ghosh , r. godbole , m. guchait , k. mohan , and d. sengupta , `` looking for an invisible higgs signal at the lhc , '' http://www.arxiv.org/abs/1211.7015[1211.7015 ] . m. almarashi and s. moretti , `` reinforcing the no - lose theorem for nmssm higgs discovery at the lhc , '' _ _ * d84 * ( 2011 ) 035009 , http://www.arxiv.org/abs/1106.1599[1106.1599 ] . c. englert , m. spannowsky , and c. wymant , `` partially ( in)visible higgs decays at the lhc , '' _ _ * b718 * ( 2012 ) 538544 , http://www.arxiv.org/abs/1209.0494[1209.0494 ] . h. e. logan and m. -a . roy , `` higgs couplings in a model with triplets , '' phys . d * 82 * ( 2010 ) 115011 [ arxiv:1008.4869 [ hep - ph ] ] . a. falkowski , s. rychkov , and a. urbano , `` what if the higgs couplings to w and z bosons are larger than in the standard model ? , '' _ jhep _ * 1204 * ( 2012 ) 073 , http://www.arxiv.org/abs/1202.1532[1202.1532 ] . d. choudhury , r. islam , a. kundu and b. mukhopadhyaya , `` anomalous higgs couplings as a window to new physics , '' arxiv:1212.4652 [ hep - ph ] . collaboration , s. chatrchyan _ et al . _ , `` search for dark matter and large extra dimensions in monojet events in @xmath212 collisions at @xmath111 tev , '' _ jhep _ * 1209 * ( 2012 ) 094 , http://www.arxiv.org/abs/1206.5663[1206.5663 ] . c. petersson , a. romagnoni , and r. torre , `` higgs decay with monophoton + met signature from low scale supersymmetry breaking , '' _ jhep _ * 1210 * ( 2012 ) 016 , http://www.arxiv.org/abs/1203.4563[1203.4563 ] . e. dudas , c. petersson , and p. tziveloglou , `` low scale supersymmetry breaking and its lhc signatures , '' http://www.arxiv.org/abs/1211.5609[1211.5609 ] . c. p. burgess , m. pospelov and t. ter veldhuis , `` the minimal model of nonbaryonic dark matter : a singlet scalar , '' nucl . b * 619 * ( 2001 ) 709 [ hep - ph/0011335 ] . s. kanemura , s. matsumoto , t. nabeshima and n. okada , `` can wimp dark matter overcome the nightmare scenario ? , '' phys . d * 82 * ( 2010 ) 055026 [ arxiv:1005.5651 [ hep - ph ] ] . x. -g . he and j. tandean , `` hidden higgs boson at the lhc and light dark matter searches , '' phys . d * 84 * ( 2011 ) 075018 [ arxiv:1109.1277 [ hep - ph ] ] . y. mambrini , `` higgs searches and singlet scalar dark matter : combined constraints from xenon 100 and the lhc , '' phys . d * 84 * ( 2011 ) 115017 [ arxiv:1108.0671 [ hep - ph ] ] . p. j. fox , r. harnik , j. kopp and y. tsai , `` missing energy signatures of dark matter at the lhc , '' phys . d * 85 * ( 2012 ) 056011 [ arxiv:1109.4398 [ hep - ph ] ] . a. djouadi , o. lebedev , y. mambrini and j. quevillon , `` implications of lhc searches for higgs portal dark matter , '' phys . b * 709 * ( 2012 ) 65 [ arxiv:1112.3299 [ hep - ph ] ] . p. junnarkar and a. walker - loud , `` the scalar strange content of the nucleon from lattice qcd , '' http://www.arxiv.org/abs/1301.1114[1301.1114 ] . g. belanger , f. boudjema , a. pukhov and a. semenov , `` micromegas3.1 : a program for calculating dark matter observables , '' arxiv:1305.0237 [ hep - ph ] . g. belanger , f. boudjema , a. pukhov , and a. semenov , `` dark matter direct detection rate in a generic model with micromegas 2.2 , '' _ comput.phys.commun . _ * 180 * ( 2009 ) 747767 , http://www.arxiv.org/abs/0803.2360[0803.2360 ] . collaboration , e. aprile _ et al . _ , `` dark matter results from 225 live days of xenon100 data , '' _ phys.rev.lett . _ * 109 * ( 2012 ) 181301 , http://www.arxiv.org/abs/1207.5988[1207.5988 ] . g. belanger , b. dumont , u. ellwanger , j. gunion , and s. kraml , in preparation .
there are numerous situations in which additive ( main effects ) models are insufficient for predicting an outcome of interest . in medical diagnosis , the co - occurrence of two symptoms may lead a doctor to be confident that a patient has a certain disease whereas the presence of either symptom without the other would provide only a moderate indication of that disease . this situation corresponds to a positive ( i.e. , synergistic ) interaction between symptom variables . on the other hand , suppose both symptoms convey redundant information to the doctor about the patient so that knowing both provides no more information about the disease status than either one on its own . this situation is again not additive , but this time there is a negative interaction between symptoms . fitting regression models with interactions is challenging when one has even a moderate number , @xmath0 , of measured variables , since there are @xmath1 interactions of order @xmath2 . for this paper , we focus on the case of pairwise ( @xmath3 ) interaction models , although the ideas we develop generalize naturally to higher - order interaction models . we consider a regression model for an outcome variable @xmath4 and predictors @xmath5 , with pairwise interactions between these predictors . in particular our model has the form @xmath6 where @xmath7 . regardless of whether the predictors are continuous or discrete , we will refer to the additive part as the `` main effect '' terms and the quadratic part as the `` interaction '' terms . our goal is to estimate @xmath8 and @xmath9 , where @xmath10 and @xmath11 . the factor of one half before the interaction summation is a consequence of our notational decision to deal with a symmetric matrix @xmath12 of interactions rather than a vector of length @xmath13 . we take @xmath11 throughout this paper because it simplifies notation , but everything carries over if we remove this restriction . indeed , we provide this as an option in the ` hiernet ` ( pronounced `` hair net '' ) package . we observe a training sample , @xmath14 , and our goal is to select a subset of the @xmath15 main effect and interaction variables that is predictive of the response , and to estimate the values for the nonzero parameters of the model . it is a well - established practice among statisticians fitting ( [ eq : model ] ) to only allow an interaction into the model if the corresponding main effects are also in the model . such restrictions are known under various names , including `` heredity , '' `` marginality , '' and being `` hierarchically well - formulated '' [ @xcite ] . there are two types of restrictions , which we will call _ strong _ and _ weak hierarchy _ : @xmath16 some statisticians argue that models violating strong hierarchy are not sensible . for example , according to @xcite , _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ `` [ t]here is usually no reason to postulate a special position for the origin , so that the linear terms must be included with the cross - term . '' _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ to see that violating strong hierarchy amounts to `` postulating a special position for the origin , '' consider writing an interaction model as @xmath17 . first of all , we would only take @xmath18 if we have special reason to believe that the regression surface must go through the origin . likewise , taking @xmath19 but @xmath20 would only be appropriate if we actually believe that @xmath21 s effect on @xmath4 should only be present specifically when @xmath22 is nonzero . in most situations , we do not think that the variable @xmath22 that we measured is any more special than @xmath23 . yet if our model with @xmath22 violates strong hierarchy , then our model with @xmath23 ( for any @xmath24 ) is strongly hierarchical . this argument suggests that violations to hierarchy occur in special situations whereas hierarchy is the default . another argument in favor of hierarchy has to do with statistical power . in the words of @xcite : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ `` [ l]arge component main effects are more likely to lead to appreciable interactions than small components . also , the interactions corresponding to larger main effects may be in some sense of more practical importance . '' _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in other words , rather than looking at all possible interactions , it may be useful to focus our search on those interactions that have large main effects . indeed , the method we propose in this paper makes direct use of this principle . as a final argument for hierarchy , it is useful to distinguish between two notions of sparsity , which we will call _ parameter sparsity _ and _ practical sparsity_. parameter sparsity is what most statisticians mean by `` sparsity '' : the number of nonzero coefficients in the model . practical sparsity is what someone actually collecting data cares about : the number of variables one needs to _ measure _ to make predictions at a future time . the hierarchy restriction favors models that `` reuse '' measured variables whereas a nonhierarchical model does not . the top left panel of figure [ fig : olive ] gives a small example where this difference is manifest . in fact , a simple calculation shows that this difference can be quite substantial : we can have a hierarchical and a nonhierarchical interaction model with the same parameter sparsity but with the nonhierarchical method having a practical sparsity of @xmath25 whereas the hierarchical method s practical sparsity is just @xmath2 . for the strong hierarchical lasso . filled nodes correspond to nonzero main effects , and edges correspond to nonzero interactions . ] while taking these arguments to the extreme leads to the use of strong hierarchy exclusively , we develop the case of weak hierarchy in parallel throughout this paper . weak hierarchy , as the name suggests , can be thought of as a compromise between strong hierarchy and imposing no such structure and appears as a principle in certain statistical methods such as classification and regression trees [ @xcite ] and multivariate additive regression splines [ @xcite ] . the lasso [ @xcite ] is a method that performs both model selection and estimation . it penalizes the squared loss of the data with an @xmath26-norm penalty on the parameter vector . this penalty has the property of producing estimates of the parameter vector that are sparse ( corresponding to model selection ) . given a design matrix @xmath27 and response vector @xmath28 , the lasso is the solution to the convex optimization problem , @xmath29 where @xmath30 is the vector of ones . the penalty parameter , @xmath31 , controls the relative importance of fitting to the training data ( sum - of - squares term ) and of sparsity ( @xmath26 penalty term ) . a natural extension of the lasso to our interaction model ( [ eq : model ] ) would be to take @xmath32 $ ] and @xmath33 , where the columns of @xmath34 correspond to elementwise products of the columns of @xmath35 . we will refer to this method as the _ all - pairs lasso _ since it is simply the lasso applied to a data matrix which includes all pairs of interactions ( as well as all main effects ) . it is common with the lasso to standardize the predictors so that they are on the same scale . in this paper , we standardize @xmath35 so that its columns have mean 0 and standard deviation 1 ; we then form @xmath36 from these standardized predictors and , finally , center the resulting columns of @xmath36 . by centering @xmath37 and @xmath38 , we may take @xmath39 . the lasso s @xmath26 penalty is neutral to the pattern of sparsity , allowing any sparsity pattern to emerge . the notions of strong and weak hierarchy introduced in section [ sec : strong - weak - hier ] represent situations in which we want to exclude certain sparsity patterns . there has been a growing literature focusing on methods that produce _ structured sparsity _ [ @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ] . these methods make use of the _ group lasso _ penalty ( and generalizations thereof ) which , given a predetermined grouping of the parameters , induces entire groups of parameters to be set to zero [ @xcite ] . given a set of groups of variables , @xmath40 , these methods generalize the @xmath26 penalty by @xmath41 where @xmath42 , @xmath43 is @xmath44 projected onto the coordinates in @xmath45 , and @xmath46 is a nonnegative weight . hierarchical structured sparsity is obtained by choosing @xmath40 to have nested groups . for example , @xcite consider the penalty @xmath47 likewise , the framework of @xcite if specialized to this paper s focus would lead to a penalty of the form @xmath48 for some @xmath49 and @xmath50 . in fact , @xcite suggest a penalty for generalized additive models with interactions that reduces to ( [ eq : bach ] ) in the linear model case , with @xmath51 and @xmath52 independent of @xmath53 . here , we propose a lasso - like procedure that produces sparse estimates of @xmath54 and @xmath12 while satisfying the strong or weak hierarchy constraint . in contrast to much of the structured sparsity literature which is based on group lasso penalties , our approach , presented in section [ sec : our - proposed - method ] , involves adding a set of convex constraints to the lasso . although we find this form of constraint more naturally interpretable , we show ( remark [ re3 ] ) that this problem can be equivalently expressed in a form that relates it to penalties from the structured sparsity literature such as ( [ eq : bach ] ) . a key advantage of our specific choice of penalty structure is that it admits a simple interpretation of the effect of the hierarchy demand . unlike other hierarchical sparsity methods , which do not pay much attention to the particular choice of norms ( as long as @xmath42 ) , our formulation is carefully tailored to allow it to be related directly back to the lasso , permitting one to understand specifically how hierarchy alters the solution ( section [ sec : effect - constraint ] ) . this feature of our estimator gives it a transparency that exposes the effects ( both positive and negative ! ) of the hierarchy constraint . furthermore , our characterization suggests that the demand for hierarchy is analogous to the demand for sparsity a form of `` regularization . '' we develop an unbiased estimator of the degrees of freedom of our method ( section [ sec : degrees - freedom ] ) and an interpretable upper bound on this quantity , which also points to hierarchy as regularization . in particular , we show that we do not `` spend '' in degrees of freedom for main effects that are forced into the model by the hierarchy constraint . another difference from much of the structured sparsity literature , which aims to develop a broad treatment of structured and hierarchical sparsity methods , is that our focus is narrowed to the problem of interaction models . our restricted scope allows us to address specifically the performance of such a tool to this important problem . in section [ sec : related - work ] , we review previous work on the problem of hierarchical interaction model fitting and selection . these methods fall into three categories : multi - step procedures , which are defined by an algorithm [ @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ] ; bayesian approaches , which specify the hierarchy requirement through a prior [ @xcite ] ; and , most related to this paper s proposal , regularized regression methods , which are defined by an optimization problem [ @xcite , @xcite , @xcite , @xcite , @xcite ] . in section [ sec : empirical - study ] , we study via simulation the statistical implications of imposing hierarchy on an interactions - based estimator under various scenarios ( in both the lasso and stepwise frameworks ) . in section [ sec : algorithmics ] we present an efficient algorithm for computing our estimator . real data examples are used to illustrate a distinction we draw between `` parameter sparsity '' and `` practical sparsity '' and to discuss hierarchy s role in promoting the latter . in section [ sec : sparsity - lasso ] , we introduced the _ all - pairs lasso _ , which can be written as @xmath55 where @xmath56 and @xmath57 is the loss function , typically@xmath58 , but may also include a ridge penalty on the coefficients as discussed later or may be substituted for the binomial negative log - likelihood . the one - half factors in front of terms involving @xmath12 are merely a consequence of the notational choice to represent @xmath12 as a symmetric matrix ( with @xmath11 for @xmath59 ) . in this paper , we propose a modification of the all - pairs lasso that produces models that are guaranteed to be hierarchical . as motivation for our proposal , consider building hierarchy into the optimization problem as a constraint , @xmath60 \\[-8pt ] \nonumber & & \qquad\mbox{s.t.}\qquad \theta= \theta^t , \vert\theta_j \vert_1\le\|\beta_j\| \qquad\mbox{for } j=1,\ldots , p,\end{aligned}\ ] ] where @xmath61 denotes the @xmath53th row ( and column , by symmetry ) of @xmath12 . notice that if @xmath62 , then @xmath63 and @xmath64 and thus @xmath65 and @xmath66 . while the added constraints enforce strong hierarchy , they are not convex , which makes ( [ eq : nonconvex ] ) undesirable as a method . in this paper , we propose a straightforward convex relaxation of ( [ eq : nonconvex ] ) , which we call the _ strong hierarchical lasso _ , @xmath67 \\[-8pt ] \nonumber & & \qquad\mbox{s.t . } \qquad\theta= \theta^t , \left.\begin{aligned } \vert\theta_j\vert_1&\le{\beta^+}_j+ { \beta^-}_j \\ { \beta^+}_j&\ge0 , { \beta^-}_j\ge0 \end{aligned } \right\}\mbox { for } j=1,\ldots , p , \nonumber\end{aligned}\ ] ] where we have replaced the optimization variable @xmath8 by two vectors @xmath68 . after solving the above problem , our fitted model is of the form @xmath69 . while we might informally think of @xmath70 and @xmath71 as positive and negative parts of a vector @xmath72 , that is , that @xmath73 , this is not actually the case since at a solution we can have both @xmath74 and @xmath75 indeed , if we were to add the constraints @xmath76 for @xmath59 to ( [ eq : hiernet ] ) , then these would be positive and negative parts and so @xmath77 , giving us precisely problem ( [ eq : nonconvex ] ) . this observation establishes that ( [ eq : hiernet ] ) is a convex relaxation of ( [ eq : nonconvex ] ) . the hierarchy constraints can be seen as an embedding into our method of david cox s `` principle '' that `` large component main effects are more likely to lead to appreciable interactions than small components . '' the constraint @xmath78 budgets the total amount of interactions involving variable @xmath79 according to the relative importance of @xmath79 as a main effect . one additional advantage of the convex relaxation is that the constraint is less restrictive . if the best fitting model would have @xmath80 large but @xmath81 only moderate , this can be accommodated by making @xmath82 and @xmath83 both large . [ rem1 ] another possibility for the hierarchy constraint that we have considered is @xmath84 ; however , we have found that this can lead to an overabundance of interactions relative to main effects . [ rem2 ] it is desirable to include in the loss function @xmath85 an _ elastic net _ term , @xmath86 , to ensure uniqueness of the solution [ @xcite ] . we think of @xmath87 as a fixed tiny fraction of @xmath88 , such as @xmath89 , rather than as an additional tuning parameter . such a modification does not complicate the algorithm , but simplifies the study of the estimator . in all numerical examples and in the ` hiernet ` package , we use this elastic net modification . [ re3 ] we prove in section 2 of the supplementary materials [ @xcite ] that ( [ eq : hiernet ] ) may equivalently be written as @xmath90 this reparameterization of the problem shows its similarities to the group lasso based methods . in place of the more standard penalty @xmath91 of ( [ eq : bach ] ) , we use @xmath92 . in section [ sec : effect - constraint ] , we show that this unusual choice of penalty admits a particularly simple interpretation for the effect of imposing hierarchy . in section [ sec : strong - weak - hier ] , we also introduced the notion of weak hierarchy . by simply removing the symmetry constraint on @xmath12 , we get what we call the _ weak hierarchical lasso _ , @xmath93 \\[-8pt ] \nonumber & & \qquad\mbox{s.t.}\qquad \left.\begin{aligned } \vert\theta_j \vert_1&\le{\beta^+}_j+{\beta^-}_j \\ { \beta^+}_j&\ge0 , { \beta^-}_j\ge0 \\ \end{aligned } \right\}\mbox { for } j=1,\ldots , p.\end{aligned}\ ] ] even though at a solution to this problem , @xmath94 is not symmetric , we should think of the interaction coefficient as @xmath95 since this is what multiplies the interaction term @xmath96 when computing @xmath97 . [ rem4 ] we can build further on the connection between ( [ eq : bach ] ) and ( [ eq : hiernet ] ) discussed in remark [ re3 ] . our removal of the symmetry constraint in ( [ eq : weak - hiernet ] ) is analogous to the technique of duplicating columns of the design matrix used in the overlap group lasso [ @xcite ] . a favorable property that distinguishes our method from previous approaches discussed in section [ sec : related - work ] is the relative transparency of the role that the hierarchy constraint plays in our estimator . this aspect is developed in section [ sec : effect - constraint ] . although our primary focus in this paper is on the gaussian setting of ( [ eq : model ] ) , our proposal extends straightforwardly to other situations , such as the logistic regression setting in which the response is binary . in this case , we simply have @xmath98 be the appropriate negative log - likelihood , @xmath99 , where @xmath100^{-1}$ ] . in section 3 of the supplementary materials [ @xcite ] , we show that solving this problem requires only a minor modification to our primary algorithm . it should also be noted that our estimator ( and the algorithms developed to compute it ) is designed for both the @xmath101 and @xmath102 setting . as a preliminary example , consider predicting whether a sample of olive oil comes from southern apulia based on measurements of the concentration of @xmath103 fatty acids [ @xcite ] . the dataset consists of @xmath104 samples , and we average our results over 100 random equal - sized train - test splits . we compare three methods : ( a ) a standard lasso with main effects only ( mel ) , ( b ) the all - pairs lasso ( apl ) , and ( c ) the strong hierarchical lasso ( hl ) . the top left panel of figure [ fig : olive ] shows an interesting difference between hl and apl . we see that , on average , at a parameter sparsity level of five , the hl model uses four of the measured variables whereas apl uses six . using the hierarchical model to classify a future olive oil , we only need to measure four rather than six of the fatty acids . the top right panel of figure [ fig : olive ] shows the predictive performance ( versus the practical sparsity ) of the three methods . it appears that hl enjoys the `` best of both worlds , '' matching the good performance of mel for low practical sparsity levels ( since it tends to pick out the main effects first ) and the good performance of apl at high practical sparsity levels ( since it can incorporate predictive interactions ) . finally , the bottom panel of the figure provides a visual display of a sequence of hl s solutions ( by varying @xmath88 ) . nonzero main effects are shown as filled nodes , and edges indicate nonzero interactions . since all edges are incident to filled nodes , we see that strong hierarchy holds . in the next section , we present several properties of our estimator that shed light on the effect of adding the convex hierarchy constraint to the lasso . among these properties is an unbiased estimate of the degrees of freedom of our estimator . we view this degrees - of - freedom result as valuable primarily for the sake of understanding the effect of hierarchy . while such an estimate could be used for parameter selection , we prefer cross validation to select @xmath88 since this is more directly tied to the goal of prediction . a key advantage of formulating an estimator as a solution to a convex problem is that it can be completely characterized by a set of optimality conditions , known as the karush tucker ( kkt ) conditions . these conditions are useful for understanding the effect that the hierarchy constraint in ( [ eq : hiernet ] ) and ( [ eq : weak - hiernet ] ) has on our solutions . in this section , we will study the simplest case , taking @xmath98 to be the quadratic loss function with no elastic net penalty . we let @xmath105 denote partial residuals ( where @xmath106 denotes elementwise multiplication , @xmath107 the vector of fitted values and @xmath108 the @xmath53th predictor ) , and we assume that @xmath109 . for linear regression , the kkt conditions are known as the normal equations and can be written as @xmath110 the all - pairs lasso solution satisfies @xmath111}{\vert x_jx_k\vert^2},\ ] ] where @xmath112 denotes the soft - thresholding operator defined by @xmath113 . written this way , we see that the lasso is similar to linear regression , but all coefficients are shrunken toward 0 , with some coefficients ( those for which @xmath114 ) set to zero . it is instructive to examine the corresponding statements for the strong and weak hierarchical lasso methods.=-1 [ prop : constraint ] the coefficients of the strong and weak hierarchical lassos with @xmath115 and taking @xmath98 to be the quadratic loss ( with no elastic net penalty ) satisfy : strong : @xmath116}{\vert x_jx_k\vert^2};\end{aligned}\ ] ] weak : @xmath117}{\vert x_jx_k\vert^2}\end{aligned}\ ] ] for some @xmath118 , @xmath59 with @xmath119 when @xmath120 ( and likewise for @xmath121 ) . see section 1 of the supplementary materials [ @xcite ] . the @xmath122 appearing in the above two properties are optimal dual variables corresponding to the @xmath53th hierarchy constraint for the strong and weak hierarchical lasso problems , respectively . when @xmath123 , we have @xmath119 ( or @xmath124 ) by complementary slackness . comparing these expressions to those of the all - pairs lasso gives insight into the effect of the constraint . property [ prop : constraint ] reveals that the overall form of the all - pairs lasso and hierarchical lasso methods is identical . the difference is that the hierarchy constraint leads to a reduction in the shrinkage of certain main effects and an increase in the shrinkage of certain interactions . in particular , we see that when the hierarchy constraints are loose at the solution , that is , @xmath120 , the weak hierarchical lasso s optimality conditions become identical to the all - pairs lasso ( since @xmath124 ) for all coefficients involving @xmath108 . for the strong hierarchical lasso , when both the @xmath53th and @xmath2th constraints are loose , the optimality conditions match those of the all - pairs lasso for the coefficients of @xmath108 , @xmath125 and @xmath126 . the methods differ when constraints are active , that is , when @xmath127 , which allows @xmath128 ( or @xmath121 ) to be nonzero . intuitively , this case corresponds to the situation in which hierarchy would not have held `` naturally '' ( i.e. , without the constraint ) , and the corresponding dual variable plays the role of reducing @xmath129 in @xmath26-norm and increasing @xmath130 until the constraint is satisfied . the way in which the weak and strong hierarchical lasso methods perform this shrinkage is different , but both are identical to the all - pairs lasso when all constraints are loose . in section [ sec : our - proposed - method ] , we showed that adding the constraint @xmath131 would guarantee that hierarchy holds . however , we have not yet shown that the same is true of the convex relaxation s constraint , @xmath132 . in particular , while @xmath133 , we could still have @xmath134 . this would correspond to a model in which @xmath135 is used in the model , but @xmath79 is not . intuitively , we would expect that if @xmath74 , then @xmath136 is analogous to getting an exact zero in linear regression ( i.e. , a zero probability event ) . in this section , we establish that this is in fact the case . in particular , we study ( [ eq : hiernet ] ) and ( [ eq : weak - hiernet ] ) where @xmath98 includes an elastic net term . the importance of this modification is that it ensures uniqueness , simplifying the analysis . as noted in remark [ rem2 ] , we think of @xmath137 as a small , fixed proportion of @xmath88 rather than as a separate tuning parameter . [ prop : strong - hier ] suppose @xmath37 is absolutely continuous with respect to the lebesgue measure on @xmath138 . if @xmath139 solves ( [ eq : hiernet ] ) , where @xmath98 is the quadratic loss with an @xmath87 ridge penalty , then strong hierarchy holds with probability 1 , that is , @xmath140 see appendix [ app : proof - hier ] . to understand how dropping the symmetry constraint leads to the `` or '' statement required of weak hierarchy , note that @xmath135 is in the weak hierarchical lasso model if and only if @xmath141 . this holds only if @xmath62 or @xmath142 . [ prop : weak - hier ] suppose @xmath37 is absolutely continuous with respect to the lebesgue measure on @xmath138 . if @xmath139 solves ( [ eq : weak - hiernet ] ) , where @xmath98 is the quadratic loss with an @xmath87 ridge penalty , then weak hierarchy holds with probability 1 , that is , @xmath143 see appendix [ app : proof - hier ] . in classical statistics , the degrees of freedom of a procedure refer to the dimension of the space over which its fitted values can vary . it is useful in that it provides a measure of how much `` fitting '' the procedure is doing . this notion can be generalized to adaptive procedures such as the lasso [ @xcite , @xcite , @xcite , @xcite ] . see ( ryan ) @xcite for a thorough discussion . if given data @xmath28 , a procedure @xmath144 produces fitted values @xmath145 , the degrees of freedom of the procedure @xmath144 is defined to be @xmath146 [ prop : df ] suppose @xmath147 . an unbiased estimate of the degrees of freedom of the strong hierarchical lasso , with quadratic loss and no ridge penalty , is given by @xmath148 where @xmath149 with @xmath36 containing the interactions , and @xmath150 is a projection matrix which depends on the sign pattern of @xmath139 and on the set of hierarchy constraints that are tight . see appendix [ app : df ] . figure [ fig : df ] provides a numerical evaluation of how well @xmath151 estimates @xmath152 . we fix @xmath153 and @xmath9 , and we generate @xmath154 monte carlo replicates @xmath155 . for each replicate , we fit the strong hierarchical lasso along a grid of @xmath88 values to get @xmath156 and @xmath157 . from these values , we compute monte carlo estimates of @xmath152 from the definition in ( [ eq : df - def ] ) and of @xmath158 $ ] . estimates @xmath152 . monte carlo estimates of @xmath158 $ ] ( @xmath37-axis ) versus monte carlo estimates of @xmath152 ( @xmath159-axis ) for a sequence of @xmath88 values ( circular ) are shown . one - standard - error bars are drawn and are hardly visible . our bound on the unbiased estimate is plotted with diamonds . ] while @xmath160 can be calculated from the data and is therefore useful as an unbiased way of calibrating the amount of fitting the strong hierarchical lasso is doing , this expression is difficult to interpret . however , it turns out that we can _ bound _ @xmath160 by a quantity that does make more sense : [ prop : df - bound ] let @xmath161 , @xmath162 and @xmath163 . then , @xmath164 holds almost surely , where @xmath165 . see appendix [ app : df ] . by contrast , for the all - pairs lasso in the case that @xmath166 and the design matrix is full rank , we have @xmath167 $ ] [ @xcite ] . in other words , the strong hierarchical lasso does not `` pay '' ( in terms of fitting ) for those main effects , @xmath168 , that are forced into the model by the hierarchy constraint to accommodate a strong interaction . notice that we do pay for a nonzero main effect if _ both _ @xmath169 and @xmath170 are nonzero . this makes sense since the constraint could be satisfied with just one of these variables nonzero , but in this case it is advantageous to the fit to make both nonzero . in figure [ fig : df ] , we find that this bound is in expectation visually indistinguishable from @xmath171 $ ] . there has been considerable interest in fitting interaction models in statistics and related fields . we focus here on an overview of methods that aim at forming predictive models that satisfy the hierarchical interactions restriction . many statistics textbooks discuss a simple stepwise procedure in which one iteratively considers adding or removing the `` best '' variable ( whether it be main effect or interaction ) ; they add that one should only consider including an interaction if its main effects are in the model [ e.g. , see backward elimination in @xcite , section 6.1.3 ] . in doing so , they are enforcing the strong hierarchy restriction . such procedures are ubiquitous [ @xcite , @xcite ] as are more recent versions [ @xcite , @xcite , @xcite , @xcite ] . another approach is to perform model selection first without considering hierarchy and then to include any lower - order terms necessary to satisfy hierarchy as a post - processing step [ @xcite ] . finally , @xcite and @xcite consider modifying the lars algorithm [ @xcite ] so that hierarchy is enforced . another set of procedures for building hierarchical interaction models comes from a bayesian viewpoint . @xcite adapts the _ stochastic search variable selection _ ( ssvs ) approach of @xcite to produce strong or weak hierarchical interaction models . ssvs makes use of a hierarchical normal mixture model to perform variable selection in regression . every variable has a latent binary variable indicating whether it is `` active . '' conditional on this latent variable , each coefficient is a 0-mean normal with variance determined by the latent importance of the coefficient . the original ssvs paper chooses a prior in which the importance of each variable is an independent bernoulli . @xcite introduces dependence into the prior so that @xmath172 is important only if @xmath173 and/or @xmath174 is important as well . @xcite formulate a nonconvex optimization problem to get sparse hierarchical interaction models . they write @xmath175 , where @xmath54 are the main effect coefficients and then apply @xmath26 penalties on @xmath54 and @xmath176 . notice that @xmath177 implies @xmath178 and @xmath179 . the nonconvexity arises in writing @xmath172 as the product of optimization variables . most similar to this paper s proposal is a series of methods which formulate convex optimization problems to give sparse hierarchical interaction models . @xcite modify the nonnegative garrote [ @xcite ] by adding linear inequality constraints to enforce hierarchy . in this sense , our method can be seen as the adaption of their approach to the lasso . finally , as discussed in section [ sec : sparsity - lasso ] , another set of convex methods makes use of the group lasso penalty [ @xcite ] . @xcite [ and , relatedly , @xcite ] describe composite absolute penalties ( cap ) , a very broad class of penalties that can achieve group and hierarchical sparsity . to achieve `` hierarchical selection , '' they put forward the principle that a penalty of the form @xmath180 , with @xmath181 , induces @xmath182 to be zero only when @xmath183 is zero as well . for hierarchical interaction models , they suggest a penalty of the form @xmath184 $ ] . this framework has been developed in the structured sparsity literature [ e.g. , @xcite ] . @xcite introduce vanish , which uses this nested - group principle to achieve hierarchical sparsity in the context of nonlinear interactions . their penalty in the setting of ( [ eq : model ] ) is @xmath185 $ ] . as noted in remark [ re3 ] , our proposal is closer to cap and vanish than it may first appear . our problem can be rewritten to have a penalty of the form @xmath186.$ ] in this sense , the penalty is in the spirit of cap and related methods although it does not quite fall into the class of cap ( since ours involves a sum of norms of norms ) . it is most similar to vanish in that it combines all of @xmath61 into the term involving @xmath173 . our main interest in this section is to study the advantages and disadvantages of restricting one s interaction models to those that honor hierarchy . clearly , the effectiveness of such a strategy depends on the true model generating the data . we take @xmath187 and @xmath188 ( 435 two - way interactions ) and consider four scenarios : truth is hierarchical : @xmath189 ; truth is anti - hierarchical : @xmath190 ; truth only has interactions : @xmath191 for all @xmath53 ; truth only has main effects : @xmath192 for all @xmath193 . in cases ( i ) , ( ii ) , ( iv ) , we set 10 elements of @xmath54 to be nonzero ( with random sign ) , and , in cases ( i ) , ( ii ) , ( iii ) , we set 20 elements of the submatrix of @xmath10 to be nonzero . the signal - to - noise ratio ( snr ) for the main effects part of the signal is about 1.5 whereas the snr for the interactions part is about 1 . we study the effectiveness of the hierarchy constraint in the context of both the lasso and forward stepwise regression . forward stepwise regression refers to a greedy strategy for generating a sequence of linear regression models in which we start with an intercept - only model and then at each step add the variable that leads to the greatest decrease in the residual sum of squares . we choose forward stepwise as a basis of comparison since it has a simple modification that we think may be the hierarchical interactions approach most commonly used by statisticians . the modification is to restrict the set of interactions that could be added at a given step to only those between main effect variables currently in the model . a backward stepwise version of this approach is suggested in @xcite . we compare six methods , corresponding to each cell of the following table : [ cols="<,^,^,^",options="header " , ] we will refer to this in the proofs that follow . including the elastic net penalty , @xmath194 , is equivalent to replacing @xmath38 and @xmath37 in ( [ eq : linear ] ) by @xmath195 suppose we solve ( [ eq : linear ] ) with the above design matrix . by lemma [ lem : project ] , @xmath196 for some @xmath197 satisfying @xmath198>0\ ] ] for all @xmath199 . let @xmath200 be the linear operator that selects the part of a vector corresponding to the variable @xmath201 . now , @xmath202 implies that @xmath203 and @xmath204 . we showed earlier that we can not have @xmath205 and @xmath206 . this means that for any @xmath193 , there must be an @xmath207 for which @xmath208 corresponds to @xmath209 or @xmath210 . thus , @xmath211 means that @xmath212 or @xmath213 for each @xmath193 . this implies that @xmath214 and thus @xmath215 . we show now that @xmath216 in terms of our above notation , this is @xmath217 where @xmath218 is the vector with all zeros except for @xmath219 . let @xmath220 , and consider the set @xmath221^t\bigl(e_j^+-e_j^-\bigr)=0 , \\ & & \hspace{32pt}\bigl[({\widetilde{x}}_\varepsilon p)^+\bigl(z-\bigl(p{\widetilde{x}}_\varepsilon^t \bigr)^+w\bigr)\bigr]^te_j^+>0 , \\ & & \hspace*{77pt}\bigl[({\widetilde{x}}_\varepsilon p)^+\bigl(z-\bigl(p{\widetilde{x}}_\varepsilon^t \bigr)^+w\bigr)\bigr]^te_j^->0\bigr\}.\end{aligned}\ ] ] in light of ( [ eq : sets ] ) , fixing @xmath222 automatically specifies @xmath223 . the outer union is restricted to those subsets @xmath222 of @xmath224 elements that would have @xmath225 and @xmath226 . the event in ( [ eq : zeroprob ] ) is contained in @xmath227 since it corresponds to the case in which @xmath222 is @xmath228 . we begin by showing that @xmath228 is in this restricted union with probability one . we have already argued that at a solution we must have @xmath229 for all @xmath193 . now , @xmath74 and @xmath75 together imply that @xmath230 since otherwise we could lower the objective by reducing @xmath169 and @xmath170 without leaving the feasible set . therefore , it would be sufficient to show that @xmath231 . we do so by observing that @xmath232 is a finite union of zero probability sets . we begin by establishing that @xmath233 . to do so , we write @xmath234 for some @xmath235 . now , @xmath236 , so we can write @xmath237 , where @xmath238 and thus @xmath239 . since @xmath240 , it follows that @xmath241 . write @xmath242 for @xmath243 , and observe that @xmath244 $ ] so that @xmath245 \\ & = & \varepsilon\bigl[u_1^tu_2-u_1^{\theta^+t}u_2^{\theta^-}-u_1^{\theta ^-t}u_2^{\theta^+ } \bigr ] \\ & = & -\varepsilon\biggl[0+\sum_{jk}\bigl[u_1^{\theta^+ } \bigr]_{jk}\bigl[u_2^{\theta ^-}\bigr]_{jk}+ \bigl[u_1^{\theta^-}\bigr]_{jk}\bigl[u_2^{\theta^+ } \bigr]_{jk}\biggr].\end{aligned}\ ] ] now , for each @xmath193 we must have @xmath246 since @xmath247 . thus , for each @xmath193 , there is a r4 or r5 row in @xmath248 and thus @xmath249 implies that @xmath250_{jk}=0 $ ] or @xmath251_{jk}=0 $ ] for each @xmath193 and likewise @xmath252_{jk}=0 $ ] or @xmath253_{jk}=0 $ ] . therefore , @xmath254 . now , @xmath255 now , @xmath256 since @xmath257 and @xmath258 and @xmath259 has full column rank . thus , @xmath260 . this completes the first part of the proof . next , we show that @xmath261 as long as @xmath262\times e_j^\pm>0 $ ] . now , since @xmath263 , the only row of @xmath264 that has @xmath265 is the r1 row ; but clearly @xmath266 . thus , @xmath267 and @xmath268 . it follows that @xmath269 putting these two parts of the proof together establishes that@xmath270 . thus , @xmath227 is a finite union of lebesgue measure 0 sets . this shows that @xmath231 as long as @xmath37 is absolutely continuous with respect to the lebesgue measure on @xmath138 . an argument nearly identical to that of the previous section establishes that @xmath271 with probability one . thus , if @xmath134 and @xmath272 , then both @xmath273 and @xmath274 . it follows then that @xmath275 . this establishes weak hierarchy . the fit in terms of the active set is given by @xmath276 where @xmath277 . of course , @xmath278 , and we can solve the kkt conditions to get the rest of the optimal dual variables in terms of the active set @xmath279+c,\ ] ] where @xmath280 satisfies @xmath281+c\ge0.$ ] note that @xmath282 and thus @xmath228 and @xmath283 depend on @xmath37 even though we do not write this explicitly . we will continue writing @xmath284 to mean specifically @xmath285 . for @xmath286 in a neighborhood of @xmath37 , we might guess that @xmath287 and @xmath288 , where @xmath289+c , \\ g\bigl(y'\bigr)_{{\mathcal{a}}({\hat\phi } ) } & = & 0.\end{aligned}\ ] ] to verify this guess , we need to check that the pair @xmath290 satisfies the optimality conditions at @xmath286 , @xmath291 first of all , @xmath292 holds since @xmath293 and @xmath294 . likewise , @xmath295 . now @xmath296 , so by continuity of @xmath297 , we have @xmath298 for all @xmath286 in a small enough neighborhood , @xmath299 , of @xmath37 . this establishes that @xmath300 . now @xmath301 , so complementary slackness holds . to see that the first optimality condition holds , we can simply plug @xmath302 into the left - hand side . all that remains is to show that @xmath303 . if we knew that @xmath304 , then by continuity of @xmath305 we could argue that over a small enough neighborhood , @xmath306 , @xmath307 . however , it could be the case that @xmath308 for some @xmath207 , that is , @xmath309 . nonetheless , one can show that there is a set @xmath310 of measure 0 for which @xmath311 implies that @xmath312 and @xmath313 for all @xmath286 in a neighborhood of @xmath37 . lemma 9 of ( ryan ) @xcite proves this result for a nearly identical situation . the fit @xmath314 is a piecewise affine function of @xmath37 . using stein s formula for the degrees of freedom [ as described in ryan , @xcite ] , we get that @xmath315=e\bigl[{\operatorname{tr}}\bigl\{({\widetilde{x}}p ) ( { \widetilde{x}}p)^+\bigr\ } \bigr]=e\bigl[\operatorname{rank}({\widetilde{x}}p)\bigr],\ ] ] where @xmath277 . we bound this by an estimate that is more interpretable : @xmath316 . clearly , r2r7 are linearly independent rows . thus , the rank of @xmath317 is at least @xmath318 . now , an r1 row is linearly independent of r2r8 precisely when @xmath319 has @xmath320 . to see this , note that if @xmath321 , then r1 is certainly linearly independent of r2r8 and likewise for @xmath322 ; however if @xmath323 , then @xmath324 for all @xmath325 , and therefore this row of r1 lies in the span of r3r7 . thus , this means there are @xmath326 additional linearly independent rows . finally , we consider r8 . clearly , r8 lies in the span of r4r5 for @xmath324 since @xmath327 at a solution . but if @xmath328 , then it is linearly independent of r1r8 . therefore , r8 adds @xmath329 to the rank where we have used that @xmath330 at a solution ( since @xmath331 ) and recalling that @xmath332 for the rows of r8 . in summary , we have shown that the row - rank is @xmath333 recalling that there are @xmath224 columns , we get that @xmath334 the lagrangian of ( [ eq : prox ] ) is given by @xmath335 where @xmath336 is the dual variable corresponding to the hierarchy constraints and @xmath337 are the dual variables corresponding to the nonnegativity constraints . for notational convenience , we have written @xmath338 for @xmath61 , and we have dropped the subscripts on @xmath339 . the kkt conditions are @xmath340 where @xmath341 is a subgradient of the absolute value function evaluated at @xmath342 . the three conditions involving @xmath337 implies that @xmath343_+$ ] . the stationarity condition involving @xmath344 implies that @xmath345 . now , define @xmath346_+-[\tilde\beta^-+t\alpha]_+.$ ] the remaining kkt conditions involve @xmath347 alone : @xmath348 observing that @xmath297 is nonincreasing in @xmath336 and piecewise linear suggests finding @xmath347 as done in algorithm [ alg : onerow ] . _ inputs : _ @xmath349 . 1 . find @xmath350 define @xmath351_+-[\tilde\beta_j^-+t\alpha]_+.$ ] 1 . if @xmath352 , take @xmath353 and go to step 2 . form knot the set @xmath354 and let @xmath355 . 3 . evaluate @xmath356 for @xmath357 . 4 . if @xmath358 for some @xmath357 , take @xmath359 and go to step 2 . find adjacent knots , @xmath360 , such that @xmath361 . take @xmath362/(p_2-p_1).\ ] ] 2 . return @xmath363 $ ] and @xmath364_+$ ] . we thank will fithian , max grazier - gsell , brad klingenberg , balasubramanian narasimhan and ryan tibshirani for useful conversations and two referees and an associate editor for helpful comments .
the essential point of stochastic resonance @xcite is the following : if a stochastic resonator is subjected to an external influence , commonly referred to as the signal , its response exhibits signal features , which are most pronounced at a certain level of noise present in the system . the fascinating aspect of stochastic resonance is therefore that , counter - intuitively , an increasing noise level does not steadily deteriorate the transmission of the signal through the resonator . on the contrary , noise may rather be used to optimise the transmission process @xcite . the implications of these effects are intensively investigated . in biological systems , for example , stochastic resonance apparently plays a role in the neural transmission of information @xcite . from a technical point of view one might possibly see the emergence of a novel type of detector which incorporates an optimal amount of noise to perform best . in this regard superconducting quantum interference devices ( squids ) have been studied @xcite . they may be made to detect weak magnetic fields reasonably well without the usual costly shielding from environmental noise . in recent years it was shown that the performance of a single stochastic resonator can be enhanced , if it is embedded into an ensemble of other stochastic resonators which are properly coupled @xcite@xcite@xcite@xcite@xcite : compared to being operated isolated the response of the resonator to the signal increases within the coupled ensemble . however , if the coupling becomes too strong this response was found to deteriorate again . thus in an ensemble of stochastic resonators the coupling strength turns out to be a second design parameter : apart from the noise level it can be tuned to achieve an optimal performance of the embedded resonator . this effect was called _ array - enhanced stochastic resonance _ @xcite . it will possibly find technical exploitation and might also be relevant to biological systems , for example , coupled neurons . the aim of the present paper is a further investigation of the phenomena related to array - enhanced stochastic resonance . to this end we study a simple prototype system of coupled two - state stochastic resonators under periodic modulation . the model we propose allows to analytically calculate the weak - signal limits of two prominent stochastic - resonance quantities , spectral power amplification ( spa ) and signal - to - noise ratio ( snr ) , respectively . following the concept of array - enhanced stochastic resonance both quantities refer to the response of a resonator as part of the coupled ensemble . the impact of the coupling on both response measures will be studied . there is a close link of the present analysis to the work presented in @xcite and @xcite , which to our knowledge are the most comprehensive in the field . however , compared to @xcite , where the spa of a system of globally coupled bistable elements was studied analytically , too , the present approach offers a more refined description , since it allows to avoid a mean - field approximation . compared to @xcite , on the other hand , where the snr of a chain of locally coupled elements was investigated in a ( rather extensive ) simulation , the snr can now be obtained analytically . for a quick review of the results the reader is referred to the 3rd and 4th passage of the summary section . from a conceptual point of view the model proposed here can be seen as an extension of the two - state model developed by mcnamara and wiesenfeld @xcite to study stochastic resonance in noisy bistable systems . instead of considering individual two - state elements as in @xcite these elements are now arranged in a chain . a simple next - neighbour interaction is introduced which brings the model close to the system studied in @xcite . the interaction is chosen in a way that the resulting evolution of the elements is given by glauber s stochastic ising model @xcite . a detailed description of this approach is given in the next section which also provides the necessary background . we note that the connection to the glauber model is made for mathematical convenience . we are not concerned with the observation of stochastic resonance in ising systems which was the central theme in @xcite and @xcite . our purpose will also lead to the assumption of arrhenius - type transition rates of uncoupled unmodulated elements which is unusual in the context of ising systems . first results of the present analysis had been published in @xcite . a further interesting approach to arrays of stochastic resonators is c entered around a response of a more collective nature . here stochastic resonance is studied in the summed output of @xmath0 resonators which need not necessarily be coupled @xcite@xcite@xcite . recently , it was shown analytically that the snr of this output approaches the input snr , if a sufficiently large number of uncoupled resonators is used @xcite . whether in some way coupling may still be beneficial for the collective response of the present model will be the subject of a future study . only the summed output of a very large number of resonators shall be briefly discussed here . it will be shown that in this case the snr always deteriorates under coupling . this indicates that the coupling - induced improvement of the performance of stochastic resonators associated with array - enhanced stochastic resonance is a local rather than a global effect . an early theory of stochastic resonance in noisy bistable systems with one ( generally continuous ) variable was worked out by mcnamara and wiesenfeld @xcite . they studied the effect of a periodic modulation of these systems in a way which is independent of the precise dynamics involved . to this end the behaviour of the bistable system was approximated by a random telegraph process . for simplicity this process was taken to be symmetric , randomly switching between two states @xmath1 . for convenience we shall consider here @xmath2 . the probabilities @xmath3 to find the process in state @xmath4 satisfy @xmath5 and their time - evolution is governed by @xmath6 where @xmath7 denotes the rate of the transition @xmath8 . these rates must be extracted from the precise dynamics at hand . to build a theory of stochastic resonance they were assumed to be approximately given by @xcite @xmath9\ ] ] with @xmath10 being the small parameter of the theory . obviously , at this level the periodic modulation of the bistable system with ( angular ) frequency @xmath11 and phase @xmath12 is taken into account by the cosine term , only . this simple model allows to calculate the long - time limit of the power spectrum of the random telegraph process averaged over a uniformly distributed phase @xmath12 . the spectrum consists of a continuous part and a delta function at modulation frequency @xmath11 . the continuous part of the spectrum is the lorentzian of the unperturbed process times a frequency - dependent prefactor . the latter is close to 1 and governs the modulation - induced transfer of broad - band power to the delta peak in the spectrum . from this spectrum mcnamara and wiesenfeld obtained their central theoretical result with respect to stochastic resonance : an exact analytical expression for the signal - to - noise ratio ( snr ) of the two - state process . this snr is defined as the ratio of the weight of the delta function to the continuous part of the spectrum at modulation frequency . in view of the present work we shall neglect the signal - induced suppression of the continuous part of the spectrum and consider the linear response approximation of the mcnamara - wiesenfeld result with respect to the small parameter @xmath10 . doing so one finds for the snr @xmath13 if the approximation ( [ mwrate ] ) is performed on a particular noisy bistable system subject to a periodic modulation , the parameters @xmath14 and @xmath10 become functions of noise strength and modulation amplitude , respectively . the dependence of the snr on the noise intensity can then be studied and the occurrence of stochastic resonance may be established for that particular system . as an example the overdamped double - well system was given in @xcite . the model equation reads @xmath15 where @xmath16 is gaussian white noise with @xmath17 and @xmath18 . using a modified kramers formula for the transition rates valid for sufficiently low modulation frequencies mcnamara and wiesenfeld found @xmath19 { } \delta = \frac{a}{d}.\ ] ] the central idea of the present work is to pass from the single bistable element to a set of coupled bistable elements by extending the two - state model of mcnamara and wiesenfeld . to this end we consider a chain of two - state elements which for convenience is taken to be of infinite length . if the elements interact the simple gain - loss balance ( [ mwmaster ] ) has to be modified to describe the evolution of the set of probabilities @xmath20 to find the chain in a particular configuration @xmath21 at time @xmath22 . introducing a formal operator @xmath23 defined for any function @xmath24 by @xmath25 the new gain - loss balance reads @xmath26 with the initial condition @xmath27 the @xmath20 are interpreted as the transition probabilities @xmath28 of a markovian process with infinitely many discrete components . the statistics of the process is fully determined by ( [ master ] ) . to introduce interactions we assume that the transition rates ( [ mwrate ] ) of an element depend on the states of its next neighbours . a simple choice for this coupling controlled by a parameter @xmath29 is @xmath30,\ ] ] where @xmath31 is the mcnamara - wiesenfeld rate ( [ mwrate ] ) . with positive @xmath29 neighbouring elements prefer to be in the same state , whereas they tend to be in opposite states , if @xmath29 is negative . both tendencies grow with growing coupling strength @xmath32 . to avoid negative transition rates @xmath33 has to be required . we note that the assumed type of coupling is thus able to model ferromagnetic - type interactions for which a coupling - induced improvement of the performance of stochastic resonators was found in @xcite and @xcite . for positive @xmath29 the present model is particularly close to the model simulated in @xcite , since both are linear arrays of next - neighbour - coupled elements . for convenience we shall often use the terms ferromagnetic or antiferromagnetic coupling instead of coupling with positive or negative @xmath29 . of course , other types of transition rates ( [ transrate ] ) could be considered as well . the advantage of the present choice is that most of the relevant stochastic properties of the resulting model are already known . they were studied by glauber who introduced this model as a stochastic form of the ising model @xcite . ( we note that in the glauber model the term @xmath34 in ( [ transrate ] ) is replaced by a general time - dependent parameter @xmath35 . ) as in the mcnamara - wiesenfeld model it is again possible to find an analytical expression for the snr in leading order of the modulation parameter @xmath10 which is the subject of the next section . the respective analysis is essentially an exploitation of glauber s work . it may be possible to reduce the dynamics of some coupled systems to the two - state model given by ( [ master ] ) and ( [ transrate ] ) and hence to directly employ the snr formula derived in the next section however , this approach is certainly more involved than the rate expansion ( [ mwrate ] ) needed to make use of the mcnamara - wiesenfeld snr . we shall not pursue this more general aspect of the model in further detail . instead we are interested in devising a simple prototype system which allows to study the impact of the coupling on the stochastic - resonance effect . to this end we are looking for simple assumptions on the dependence of the model parameter @xmath14 , @xmath29 , and @xmath10 on some noise intensity and signal amplitude , respectively . we proceed in two steps : first , we retain the glauber s original relation of the present model to the ising model given by the hamiltonian @xmath36 this relation is defined by an assumption on the probability @xmath37 to find the @xmath38-th element in state @xmath39 , if all other elements are fixed : in an adiabatic limit the @xmath37 of both models are to be identical . this implies that @xmath40 holds which allows to establish a relation between the parameters of both models . one finds @xmath41 and @xmath42 , respectively @xcite . the role of the noise intensity is thus played by the temperature @xmath43 . to skip unnecessary parameters we set @xmath44 and @xmath45 . in the present paper we assume that @xmath46 , which shall be referred to as the signal , is given by @xmath47 for weak amplitudes @xmath48 this results in @xmath49 . ( in section [ srsnr ] we shall also consider periodic signals ( [ forcing ] ) with an additional noisy background . ) second , a plausible assumption has to be made on the temperature dependence of the time - scale parameter @xmath50 which is not effected by requiring ( [ detailedbal ] ) . it is assumed that it retains the qualitative dependence ( [ mwparam ] ) found by mcnamara and wiesenfeld for the double - well system , i.e. , @xmath51 with properly scaled temperature @xmath43 . this dependence is typical for many rate processes . an appropriate temperature dependence of @xmath14 is also necessary to turn the elements into stochastic resonators for vanishing coupling . for simplicity @xmath52 is considered . the prototype system we are looking for is thus specified by ( [ master ] ) and ( [ transrate ] ) together with @xmath53 { } \gamma = \tanh \left(\frac{2j}{t } \right ) , \makebox[0.3 cm ] { } \delta = \frac{h_0}{t}.\ ] ] we emphasise that this system is not an approximation of the ising system ( [ hamiltonian ] ) in the same way as the double - well system ( [ doublewell ] ) is approximated by ( [ mwmaster]),([mwrate ] ) , and ( [ mwparam ] ) , respectively : the ising system and the stochastic glauber model are based on totally different dynamical concepts . nevertheless the devised prototype system appears to capture in a simple way the essence of the effect of coupling and forcing on coupled continuous - variable systems like the double well . to motivate this , let @xmath54 denote the height of the barrier to be surmounted in order to escape from state @xmath4 . in a very simple approximation the effect of next - neighbour coupling and forcing on the barrier height could be modelled as @xmath55 @xmath56 comparing this to ( [ hamiltonian ] ) one finds that within this simple picture the effect of coupling and forcing on the barrier heights is the same as their effect on the energy levels of the ising model . ( we note that for coupled double - well systems the chosen approximation would only hold for weak coupling . for strong coupling the systems may no longer be bistable and the entire concept of a simple - minded two - state approximation breaks down . ) within an adiabatic limit and with some noise strength @xmath43 the transition rate @xmath57 of the barrier system would roughly read @xmath58 from here rates of the form ( [ transrate ] ) together with parameters ( [ params ] ) can be obtained in two steps . first , the factor related to @xmath59 is kept as the parameter @xmath14 with a respective choice of @xmath59 . second , the factor related to @xmath60 , which describes the effect of forcing and coupling , is replaced by the respective term of ( [ transrate ] ) , which is easier to handle . a connection between both types of rates is again made by requiring that they lead to identical stationary distributions in the sense of ( [ detailedbal ] ) . a particular close relation between the glauber - dynamics model and the barrier system should thus arise within the regime of quasistatic response . nevertheless we shall not restrict the following investigation of the simple prototype system to the quasistatic regime . whether or not the suggested model provides insights which are also relevant to more realistic settings and possibly helpful in the design of coupled stochastic resonators remains to be seen . the aim of this section is to calculate the snr of the response of a two - state element embedded into the chain of coupled elements given by ( [ master ] ) and ( [ transrate ] ) . this calculation will be done in terms of the parameters @xmath14 , @xmath29 , and @xmath10 and is essentially based on glauber s work . we shall only take account of terms in leading order of the modulation amplitude @xmath10 . the resulting snr will then be an extension of the simplified mcnamara - wiesenfeld result ( [ mwsnr ] ) . at the end of the section the general result will be applied to the prototype system specified by ( [ params ] ) . in addition to the snr we shall also give an analytical expression for the so - called spectral power amplification ( spa ) . following glauber we start by calculating the averages @xmath61 . from ( [ master ] ) one derives @xmath62 . it results in @xmath63 \nonumber \\ & & + \left[1 - \frac{\gamma}{2 } ( r_{k-1,k } + r_{k , k+1})\right ] \delta \cos ( \omega t + \phi)\end{aligned}\ ] ] with @xmath64 . a closed set of equations for the @xmath65 is found by linearising ( [ mean ] ) in @xmath10 . in this case the @xmath66 can be taken from the unperturbed model ( @xmath67 ) . after long times these @xmath66 read @xmath68 with @xmath69 @xcite . the resulting long - time - limit set of equations can be further simplified . since all elements are forced uniformly , a particular element can not be distinguished from any other , once the initial distribution has been forgotten . hence , all elements will have identical statistics and the indices in ( [ mean ] ) can be skipped . together with the previous assumption one obtains @xmath70 obviously , the long - time dynamics of the average @xmath71 is identical to the long - time dynamics of the average of an uncoupled element with rescaled relaxation rate @xmath72 and rescaled modulation term @xmath73 . the resulting long - time limit of the averaged state simply reads @xmath74 @xmath75 { } \tan \psi = - \frac{\omega}{\alpha(1-\gamma)},\ ] ] where @xmath76 is the response to static signals @xmath77 : @xmath78 formally this static response might even exceed @xmath79 . we conclude that in this case @xmath10 is not sufficiently small and the linearisation of ( [ mean ] ) is no longer justified . later on it will be shown that this does not amount to a considerable restriction . the power spectrum of an element is determined from the average @xmath80 . like in @xcite its @xmath22-dependence is removed by averaging @xmath81 over an uniformly distributed initial phase @xmath12 of the modulation term in ( [ transrate ] ) . introducing the correlation function @xmath82 [ \sigma_k(t+\tau)-\langle \sigma_k(t+\tau ) \rangle ] \rangle $ ] one finds @xmath83 the second term on the r.h.s . contributes a delta function with weight @xmath84 at signal frequency @xmath11 to the ( one - sided ) power spectrum as defined below . the first term , on the other hand , forms the continuous part of the spectrum . for the purpose of calculating the snr in leading order of the modulation amplitude @xmath10 it is sufficient to approximate this continuous part by the power spectrum of the unperturbed model . this implies that we neglect any possible signal - induced transformation of the continuous part of the spectrum as we did to obtain the simplified mcnamara - wiesenfeld result ( [ mwsnr ] ) . for the unperturbed model the long - time correlation function @xmath85 is not effected by the average @xmath81 . it also no longer depends on the index of the element . according to glauber @xmath85 reads @xmath86 with modified bessel functions @xmath87 . using the relation @xcite @xmath88 which holds for @xmath89 and @xmath90 the power spectrum @xmath91 is calculated . for the one - sided spectrum defined by @xmath92 at @xmath93 we eventually obtain @xmath94 with @xmath95 and @xmath96 . now the snr of the response of a two - state element embedded into the chain can be calculated , which is the central result of this section . it reads @xmath97 for vanishing coupling it reduces to the simplified mcnamara - wiesenfeld result ( [ mwsnr ] ) . in addition to the effect of the coupling on the snr we also wish to study its impact on the delta peak in the power spectrum . in this regard the spectral power amplification ( spa ) is a convenient measure ( cf . e.g. jung in @xcite ) . it is defined as the ratio of power contained in the signal peaks of output to input spectrum , i.e. , the ratio of the weights of the respective delta functions . because of its dependence on the input spectrum the spa depends on the precise dynamics to be modelled by the two - state chain . for the prototype system specified by ( [ params ] ) the spa reads @xmath98 , or explicitly @xmath99 @xmath100 where @xmath101 is the spa of static signals . obviously , the spa does no longer depend on the signal amplitude @xmath102 . for convenience we also remove the @xmath102-dependence of the snr by considering @xmath103 for the prototype system the parameter @xmath104 involved in the unperturbed spectrum ( [ spectrum ] ) simplifies to @xmath105 @xcite . both quantities , spa ( [ spa ] ) and rescaled snr ( [ snr ] ) , only depend on three parameters : temperature @xmath43 , coupling strength @xmath106 , and signal frequency @xmath11 . restrictions arise from @xmath48 , which lead to a simple expression for @xmath10 , as well as from the linearisation of ( [ mean ] ) , which results in an upper bound on @xmath10 as discussed in connection with ( [ meanstatic ] ) . together one finds @xmath107 it implies that finite @xmath102 place a lower bound on @xmath43 and an upper bound on @xmath106 . however , these restrictions are rather weak : @xmath102 can be made arbitrarily small because its size is immaterial within the present weak - signal approximation . we are now in a position to study the impact of the coupling on the response measures spa and snr of a single resonator embedded into the chain . this will be the subject of the following two sections . the spa ( [ spa ] ) has an unique maximum over temperature @xmath43 and coupling parameter @xmath106 for any time - dependent signal ( [ forcing ] ) . the maximum spa is obtained for a frequency - dependent value @xmath108 which is always positive . in other words : a properly tuned ferromagnetic - type coupling yields the best spa performance of the two - state resonator element which is embedded into the chain . for our simple model this maximum can be studied analytically . the partial derivatives of the spa with respect to @xmath43 and @xmath106 are found to vanish at pairs @xmath109 given by @xmath110 and @xmath111 respectively . for @xmath108 to be real and finite , @xmath112 has to hold . with this restriction ( [ simple2 ] ) has a unique solution @xmath113 for any given @xmath114 . the corresponding @xmath108 is found via ( [ simple ] ) . within the given range of @xmath115 one can easily check on ( [ simple ] ) that @xmath108 is indeed always positive . an inspection of the second derivatives of the spa finally reveals that there is a maximum at @xmath109 . its peak value is implicitly given by @xmath116 from ( [ simple ] ) , ( [ simple2 ] ) , and ( [ spamax ] ) the following features of the maximum can be derived which are shown in fig.1 : tuning the signal frequency @xmath11 from very large to vanishing small values the temperature @xmath115 falls from @xmath117 to @xmath118 . at the same time , the coupling strength @xmath108 as well as the peak height @xmath119 increase from vanishing small to very large values . in addition , an analysis of the curvatures @xmath120 and @xmath121 , both expressed in terms of @xmath115 , shows that the sharpness of the maximum grows as its height increases . this indicates that to achieve optimal performance the system parameters @xmath43 and @xmath106 have to be tuned with increasing accuracy as the signal frequency decreases . fig.1 also includes a comparison to the spa of an uncoupled element . at given @xmath114 this spa has a maximum located at a temperature @xmath122 , implicitly given by @xmath123 on the interval @xmath124 , whereby its peak value is found to be @xmath125 . it can be shown from ( [ simple2 ] ) and ( [ simple3 ] ) that @xmath115 always exceeds @xmath122 . the fact that @xmath108 was found to be never zero also implies that @xmath119 always exceeds @xmath126 . however , looking at fig.1 it is obvious that at frequencies @xmath127 both compared quantities of the coupled element approach those of the uncoupled one and @xmath108 approaches zero . the coupling - induced increase in the spa is thus vanishingly small at sufficiently high frequencies of the signal . this comparison allows to distinguish between two different types of spa behaviour : at low frequencies the spa is enhanced under ferromagnetic coupling whereas at high frequencies it is basicly not , although a tiny increase still occurs . both situations are illustrated at selected frequencies in fig.2 and fig.3 , respectively . the low frequency spa exactly reproduces the qualitative effect found analytically by jung et al . @xcite in a system of globally interacting elements . the high frequency behaviour , on the other hand , where the spa does practically not increase under coupling , has to our knowledge not been reported before . in both cases fig.2 and fig.3 clearly demonstrate that at fixed coupling parameter @xmath106 stochastic resonance occurs : the spa has a maximum over temperature @xmath43 which , unfortunately , can not be established analytically . fig.2(bottom ) shows that while ferromagnetic coupling improves the spa at any fixed temperature @xmath43 this improvement is lost , if the coupling becomes too strong . there is thus an optimal coupling strength @xmath128 for every given temperature , which can even be exactly calculated at any @xmath114 . one finds @xmath129 which is always positive . since the spa does not have a further extremum over @xmath106 , it implies that antiferromagnetic coupling always decreases the spa . moreover , @xmath128 increases with growing temperature t and decreases as the signal frequency grows . in fig.2 the spa at the optimal coupling strength is included ( dash - dotted curves ) . in fig.3(bottom ) , where the high - frequency spa is illustrated , the coupling - induced improvement of the spa is hardly detectable and @xmath128 is almost zero . considering the spa of time - independent signals ( @xmath130 ) given by ( [ spa_static ] ) one finds that it has a maximum over temperature for antiferromagnetic coupling @xmath131 , only , whereby it decreases as the coupling strength @xmath132 grows . for @xmath133 the spa increases with increasing @xmath106 as well as with decreasing @xmath43 . it formally diverges for @xmath134 and @xmath135 , respectively . in both cases the weak - signal limit ( [ range ] ) breaks down . in general , the spa ( [ spa ] ) is given by its static value ( [ spa_static ] ) times a dynamical factor @xmath136 ( cf.([meanresult ] ) ) . here @xmath137 is the ratio @xmath138 of signal frequency to long - time relaxation rate , which also governs the phase shift @xmath139 in ( [ meanresult ] ) . with growing @xmath137 the elements gradually loose their ability to follow the signal : the spa weakens and the phase shift grows . this effect occurs , for example , if the signal frequency @xmath11 increases . subsequently , the spa decreases with growing @xmath11 . the impact of @xmath137 is also responsible for the occurrence of stochastic resonance and optimal coupling in the spa . here a decrease of the long - time relaxation rate @xmath72 plays the crucial role : the elements dynamics slows down as @xmath106 increases or as @xmath43 decreases . thus this slow - down occurs whenever the static spa grows . hence , the dynamical factor always counteracts the static spa as @xmath43 or @xmath106 is changed . eventually , the increase of the monotonous static spa is outperformed by the decrease of the dynamical factor which results into a maximum of the spa over @xmath43 and @xmath106 , respectively . in other words , stochastic resonance and optimal coupling occur . as illustrated in fig.2 the spa can be seen as a transition between two limits , the static spa given by @xmath140 ( @xmath141 , dashed curves ) and @xmath142 ( @xmath143 , dotted curves ) , respectively . since both limits intersect at @xmath144 , i.e. at @xmath145 , the plots nicely show the well - known approximate matching of time scales at the spa peak . this matching does not only occur over @xmath43 ( fig.2 , top ) , but over @xmath106 ( fig.2 , bottom ) , too . if the signal frequency is changed , the peaks of the spa shift whereby all curves share the static spa as a limit . this results into the qualitative frequency dependence found by mcnamara and wiesenfeld for the delta function part of the spectrum of the double - well system ( [ doublewell ] ) . roughly speaking , an improvement of the spa under ferromagnetic coupling only occurs at @xmath146 , where the impact of the static spa is not yet outperformed by the dynamical factor . this explains why there is no improvement at signal frequencies @xmath147 : since under ferromagnetic coupling @xmath148 holds , @xmath146 can not be fulfilled in this case . at @xmath149 ( fig.3 ) the spa is already within line width of the limit @xmath150 , where the impact of the dynamical factor prevails . this limit is basicly given by the modulation term in ( [ simplemean ] ) . its non - monotonous temperature dependence is known to be no longer associated with a matching of time scales @xcite . the static response , which is thus the origin of the coupling - induced enhancement of the spa , can be interpreted as follows . without coupling and for a given fixed signal the two - state elements prefer to be in the state of lowest energy ( [ hamiltonian ] ) or highest barrier ( [ deltau ] ) , respectively . since the signal is homogeneous , this state is the same for all elements . without a signal but under ferromagnetic coupling neighbouring elements tend to be in the same state , too , although they do not favour a particular state @xmath151 or @xmath152 . taking now signal and coupling together , both effects add up . the tendency to find the elements in the state favoured by the signal grows and hence @xmath76 and the static spa grow compared to the uncoupled case . with antiferromagnetic coupling , on the other hand , neighbouring elements prefer to be in opposite states . this counteracts the effect of the signal and leads to a decrease in the spa . finally , the decrease of the static spa with growing temperature directly follows from ( [ detailedbal ] ) . there the imbalance in the distribution of probability between the two states , which dependent on the view taken either differ in energy or barrier height , decreases as the system heats up : the mean response @xmath76 weakens due to increasing fluctuations . the investigation of the snr presented here basicly relies on numerical evaluations of ( [ snr ] ) . an explicit calculation of the power spectrum ( [ spectrum ] ) already yields a rather complicated expression which does not lend itself to a detailed analytical study . at finite signal frequencies @xmath114 the snr displays the same qualitative dependence on temperature @xmath43 and coupling parameter @xmath106 as shown for the spa in fig.2 and fig.3 , respectively . at low frequencies the snr is enhanced under ferromagnetic coupling whereas at high frequencies it is not . due to this close similarity we omitted the respective plots for the snr . we note , however , that this similarity is by no means a trivial result , since the spectrum ( [ spectrum ] ) is itself a non - monotonous function of @xmath43 and @xmath106 . the low - frequency snr qualitatively reproduces the snr behaviour found in a chain of next - neighbour - coupled overdamped double - well systems simulated in @xcite . it is the first analytical confirmation of this behaviour . we note that this correspondence occurs although the simulation in @xcite was performed with strong forcing while the present model is studied within a weak - signal limit . as for the spa , the high - frequency behaviour of the snr has to our knowledge not been reported before . if the elements are not coupled , their snr ( [ mwsnr ] ) does not depend on the signal frequency . this was already found in @xcite , if the signal - induced reduction of the continuous part of the spectrum is neglected , as ( [ mwsnr ] ) clearly shows . under coupling this frequency independence of the snr for ferromagnetic coupling numerical evaluations of ( [ snr ] ) predict a decrease of the snr with growing signal frequency . only in the limit of low and high frequencies the snr approaches constant values . in both cases the snr formula ( [ snr ] ) simplifies significantly . one finds @xmath153 @xmath154 where @xmath155 is the simplified mcnamara - wiesenfeld snr ( [ mwsnr ] ) of uncoupled elements . @xmath156 is the static snr while @xmath157 represents the leading order term of its high - frequency expansion . fig.4 illustrates the typical qualitative behaviour of the snr as a function of the signal frequency . at various frequencies it shows a set of snr curves over temperature @xmath43 and coupling parameter @xmath106 , respectively . the dashed curves represent @xmath156 and @xmath157 , respectively . the apparent fact that the snr decreases with growing signal frequency implies that its static value can not be exceeded at any other signal frequency . then it follows immediately from ( [ snrlim1 ] ) that the coupling - induced enhancement of the snr possesses an upper limit : it can not be better than a factor four . this is a rigorous result for the general chain model and not limited to the prototype system , where a special choice of the parameters was made . due to the simplicity of the chain model this result may well reflect a limit of more general nature for the improvement of a weak - signal snr . fig.5 shows the static snr over temperature @xmath43 . it differs qualitatively from the static spa which according to ( [ spa_static ] ) has a monotonous behaviour . turning briefly to antiferromagnetic coupling one finds that the continuous part of the spectrum is insensitive to the sign of the coupling parameter @xmath106 . now even the static snr decreases under coupling . as shown in fig.4(bottom ) it is found below the high - frequency snr curve , i.e. , here the snr increases with growing signal frequency . since at medium frequencies again a transition occurs between both curves similar to the transition in fig.4(top ) , a second maximum in the snr over @xmath43 may emerge in this situation . the investigation of the impact of the coupling on the snr can also be extended to the more natural situation , where the signal itself is embedded into noise . the question to address is , whether the improvement of the snr , which was established for independent internal noise sources , can still be found with additional coherent external noise . ( this problem does not occur for the spa , where only the height of the signal peak is of interest . ) to study this case it is assumed that the input spectrum consists of the previous signal peak described by @xmath158 and a noise part @xmath159 . within the weak - signal limit it was shown that the signal peak of the input spectrum leads to a respective peak in the output spectrum at signal frequency , only : no additional peaks at multiples of that frequency occur . therefore , any additional contribution to the continuous part of the output spectrum at signal frequency can only arise from @xmath160 . in analogy to ( [ meanresult ] ) one finds that this additional contribution is given by @xmath161 . together with ( [ snr ] ) the resulting snr @xmath162 reads , again in units of @xmath163 , @xmath164 it can be expressed in terms of the snr @xmath165 ( cf.([snr ] ) ) and the input snr @xmath166 written in units of @xmath163 , respectively . one finds @xmath167 @xmath162 is thus a steadily growing function of the previously studied snr @xmath165 , whereby it can not exceed the input snr @xmath168 . ( the latter is a meanwhile well - known result of linear response theory @xcite which is in fact the limit we are taking here . ) since @xmath168 is constant at fixed signal frequency @xmath11 , a coupling - induced increase of @xmath165 will lead to an increase in @xmath162 , too . the improvement of the snr under coupling is thus preserved with external coherent noise . from ( [ noisysnr2 ] ) one can easily show that the maximum enhancement of @xmath162 compared to the uncoupled chain reaches the previously found factor four for vanishing input noise , only . in general , this factor is smaller approaching 1 , if the input noise is so strong that the input snr and hence @xmath162 go to zero . at the end we shall make a brief excursion to the collective response of the chain , which was mentioned in the introduction . we consider the following sum which involves the states of @xmath0 elements : @xmath169 it is not difficult to show that the spa of this new quantity is simply @xmath170 times the previously studied spa ( [ spa ] ) . for the snr , however , this new situation is completely different . previously the continuous part of the spectrum was determined from the auto - correlation function @xmath171 of an element alone ( cf . ( [ correl1 ] ) ) . now this spectrum will involve cross - correlation contributions @xmath172 of different elements @xmath173 , too . the new contributions to the continuous part of the spectrum change the qualitative behaviour of the snr under coupling . we shall demonstrate this in the limit @xmath174 . for this case glauber calculated the unperturbed spectrum @xcite which reads in its one - sided version @xmath175 inserting this spectrum into ( [ snr ] ) , multiplied by @xmath170 due to the mentioned increase of the spa , one finds for the new snr per element @xmath176 which , of course , again reduces to the simplified mcnamara - wiesenfeld snr ( [ mwsnr ] ) for vanishing coupling . clearly , the snr is now a decreasing function of the coupling parameter and the sign of the latter is no longer important . hence , there is a drastic difference in the snr of single and collective response with respect to coupling . of course , we have so far merely investigated the limits of ( [ global0 ] ) , @xmath177 and @xmath174 , respectively . we expect that the snr of the collective response of only a few elements still increases under coupling @xcite . a detailed analysis of the present model with respect to these collective effects will be the subject of a further investigation . in this paper we extended the two - state model of stochastic resonance introduced by mcnamara and wiesenfeld to a chain of infinitely many coupled two - state elements which are periodically modulated . the interaction of the elements was chosen in a way that the chain evolves according to glauber s stochastic ising model . in analogy to the work of mcnamara and wiesenfeld and based glauber s results analytical expressions for the signal - to - noise ratio ( snr ) and the spectral power amplification ( spa ) have been obtained in the limit of weak modulations . here both quantities refer to the response of a single element as part of the chain . instead of approximating the dynamics of a particular coupled system by the chain model , we used the latter to build a prototype system which hopefully captures the essential features of an entire class of coupled stochastic resonators . to this end additional assumptions were made on the dependence of the model parameters on some noise intensity and signal amplitude , respectively . the prototype system was used to study the effect of the coupling on the response of a single resonator embedded into the chain . the results show that array - enhanced stochastic resonance occurs for ferromagnetic - type coupling in spa and snr . the qualitative features of the effects reproduce those previously found in coupled stochastic resonators . the simplicity of the chain model allowed for a detailed analytical investigation of the spa . for the snr the observed effects have been confirmed analytically for the first time . in addition , it was found that the improvement of the single - resonator response compared to the response of the uncoupled resonator still occurs , if the signal is embedded into noise . for the snr this improvement was shown to be limited by a factor four which is reached for vanishing input noise , only . a brief excursion into the collective response of @xmath0 resonators , on the other hand , disclosed that coupling can not improve the snr , if @xmath0 is very large . a closer look at the mechanisms behind the effects revealed that in the present model an improvement of the stochastic resonators under coupling is associated with the regime of quasistatic response . since the model studied here possesses an upper bound to the time scale of its dynamics , the desired improvement is essentially restricted to sufficiently slow signals . for the spa this improvement is based on a stronger tendency of the two - state elements to align in parallel , if signal and ferromagnetic coupling act together , compared to this tendency caused by the signal alone . the reason why the improvement of the spa is lost , if the coupling is too strong , was found to be the slow - down of the system dynamics under coupling : it simply prevents the resonators from responding quasistatically . for review - type articles on stochastic resonance see , e.g. : + p. jung , phys . * 234 * , 175 ( 1993 ) . + f. moss , d. pierson , d. ogorman , int . 4(6 ) , 1383 ( 1994 ) . + k. wiesenfeld , f. moss , nature 373 * , 33 ( 1995 ) . + a. bulsara and l. gammaitoni , physics today , march 1996 , 39 ( 1996 ) . + l. gammaitoni , p. hnggi , p. jung , and f. marchesoni , rev . * 70 * , 223 ( 1998 ) + a bibliography on stochastic resonance can be found at http://www.pg.infn.it / sr/. j. k. douglas , l. wilkens , e. pantazelou , and f. moss , nature * 365 * , 337 ( 1993 ) . + s. m. besrukov and i. vodyanoy , nature * 378 * , 362 ( 1995 ) . + j. e. levin and j. p. miller , nature * 380 * , 165 ( 1996 ) . + p. cordo , j. t. ingils , s. verschueren , j. j. collins , d. m. merfeld , s. rosenblum , s. buckley , and f. moss , nature * 383 * , 769 ( 1996 ) . + j. j. collins , t. t. imhoff , and p. grigg nature * 383 * , 770 ( 1996 ) . r. rouse , siyuan han , and j. e. lukens , appl . phys . lett . * 66 * , 108 ( 1995 ) . + a. d. hibbs , a. l. singsaas , e. w. jacobs , a. r. bulsara , j. j. bekkedahl , and f. moss , j. appl . * 77 * , 2582 ( 1995 ) . + a. d. hibbs and b. r. whitecotton , in _ applied nonlinear dynamics and stochastic systems near the millennium _ , ed . by j. b. kadtke and a. bulsara , ajp conference proceedings , vol . 411 , p.227 ( aip , woodbury , new york , 1997 ) . a. bulsara , g. schmera , phys . rev . * e47 * , 3734 ( 1993 ) . + a. bulsara , a. maren , and g. schmera , biol . cybern . * 70 * , 145 ( 1993 ) . + m. e. inchiosa , a. r. bulsara : phys . lett . * a200 * , 283 ( 1995 ) and phys . rev . * e52 * , 327 ( 1995 ) .
in this paper we introduce a general lagrangian framework to study the dissipative properties of two component systems composed of a high - loss and lossless components which can have gyroscopic properties . this framework covers any linear lagrangian system provided it has a finite number of degrees of freedom , a nonnegative hamiltonian , and losses accounted by the rayleigh dissipative function , ( * ? ? ? 10.11 , 10.12 ) , ( * ? ? ? 8 , 9 , 46 ) . such physical systems include , in particular , many different types of damped mechanical systems or electric networks . [ [ motivation . ] ] motivation . + + + + + + + + + + + we are looking to design and study two - component composite dielectric media consisting of a high - loss and low - loss component in which the lossy component has a useful property ( or functionality ) such as magnetism . we want to understand what is the trade - off between the losses and useful properties inherited by the composite from its components . our motivation comes from a major problem in the design of such structures where a component which carries a useful property , e.g. , magnetism , has prohibitively strong losses in the frequency range of interest . often this precludes the use of such a lossy component with otherwise excellent physically desirable properties . then the question stands : is it possible to design a composite system having a useful property at a level comparable to that of its lossy component but with significantly reduced losses over a broad frequency range ? an important and guiding example of a two - component dielectric medium composed of a high - loss and lossless components was constructed in @xcite . the example was a simple layered structure which had magnetic properties comparable with a natural bulk material but with 100 times lesser losses in a wide frequency range . this example demonstrated that it is possible to design a composite material / system which can have a desired property comparable with a naturally occurring bulk substance but with significantly reduced losses . in order to understand the general mechanism for this phenomenon , the authors in @xcite considered a general dynamical system as the model for such a medium where the high - loss component of the medium was represented as a significant fraction of the entire system . we have found that for such a system the losses of the entire structure become small provided that the lossy component is sufficiently lossy . the general mechanism of this phenomenon , which we proved in @xcite , is the _ modal dichotomy _ when the entire set of modes of the system splits into two distinct classes , _ high - loss _ and _ low - loss modes _ , based on their dissipative properties . the higher loss modes for a wide range of frequencies contribute very little to losses and this is how the entire structure can be low loss , whereas the useful property is still present . in that work the way we accounted for the presence of a useful property was by simply demanding that the lossy component was a significant fraction of the entire composite structure without explicitly correlating the useful property and the losses . in this paper , we continue our studies on the modal dichotomy and overdamping , which began in @xcite , but now we focus are attention on gyroscopic - dissipative systems . although our results on modal dichotomy apply to the full generality of the dynamical systems considered here , our results on overdamping in this paper will be restricted to systems without gyroscopy . consideration of overdamping in systems with gyroscopy will be considered in a future work . [ [ overview - of - results . ] ] overview of results . + + + + + + + + + + + + + + + + + + + + now in order to account in a general form for the physical properties of the two - component composite system with a high - loss and lossless components we introduce a lagrangian framework with dissipation . this framework can be a basis for studying the interplay of physical properties of interest and dissipation . these general physical properties include in particular gyroscopy ( or gyrotropy ) , which is intimately related to magnetic properties , different symmetries and other phenomena such as overdamping that are not present in typical dynamical systems . in the lagrangian setting , the dissipation is taken into account by means of the rayleigh dissipation function ( * ? ? ? 10.11 , 10.12 ) , ( * ? ? * sec . 8 , 9 , 46 ) . in the case of two - component composite with lossy and lossless components , the rayleigh function is assumed to affect only a fraction @xmath0 of the total degrees of freedom of the system . in a rough sense , the loss fraction @xmath1 signifies the fraction of the degrees of freedom that are effected by losses . we now give a concise qualitative description of the main results of this paper which are discussed next and organized under the following three topics : overdamping and selective overdamping in section [ sodan ] ; virial theorem for dissipative systems and equipartition of energy in section [ svir ] ; standard spectral theory vs. krein theory in sections [ smodel ] and [ sspan ] . * overdamping and selective overdamping. overdamping is a regime in dissipative lagrangian systems in which some of the systems eigenmodes are overdamped , i.e. , have exactly zero frequency or , in other words , are non - oscillatory when losses are sufficiently large . the phenomenon of overdamping has been studied thoroughly but only in the case when the entire system is overdamped @xcite , ( ? ? ? 9 ) , @xcite meaning all its modes are non - oscillatory . the lagrangian framework and the methods we develop in this paper are applicable to the more general case allowing for only a fraction of the modes to be susceptible to overdamping , a phenomenon we call _ selective overdamping_. our analysis of overdamping and this selective overdamping phenomenon is carried out in section [ sodan ] . our main results are theorems [ tcovd ] , [ tpovd ] , [ cpodr ] , [ thmod ] , and [ tseod ] . our interest in selective overdamping is threefold . first , as we prove in this paper the selective overdamping of the system is a rather _ universal phenomenon _ that can occur for a two - component system when the losses of the lossy component are sufficiently large . moreover , we show that _ the large losses in the lossy component _ cause not only the modal dichotomy , but also _ results in overdamping of all of the high - loss modes while a positive fraction of the low - loss modes remain oscillatory_. second , since mode excitation is efficient only when its frequency matches the frequency of the excitation force , if these high - loss modes go into the overdamping regime having exactly zero frequency these modes can not be excited efficiently and therefore their associated losses are essentially eliminated . this explains the absorption suppression for systems composed of lossy and lossless components . third , as we will show , as the losses in the lossy component increase the overdamped high - loss modes are more suppressed while all the low - loss oscillatory modes are more enhanced with increasingly high quality factor . _ this provides a mechanism for selective enhancement of these high quality factor , low - loss oscillatory modes and selective suppression of the high - loss non - oscillatory modes_. in fact , we show that _ the fraction of all overdamped modes is exactly the loss fraction _ @xmath1 _ _ which satisfies _ _ @xmath0 , and half of these overdamped modes are the high - loss modes . thus it is _ exactly the positive fraction _ @xmath2 _ _ of modes which are low - loss oscillatory modes__. in addition , _ these latter modes have extremely high quality factor which increases as the losses in the lossy component increase_. an example of this behavior using the electric circuit in section [ scircuits ] was shown numerically in @xcite . * virial theorem for dissipative systems and equipartition of energy. recall that the virial theorem from classical mechanics ( ? ? ? * ; 3.4 ) ( which we review in appendix [ apdxvirthm ] ) is about the oscillatory transfer of energy from one form to another and its equipartition . for conservative linear systems whose lagrangian is the difference between the kinetic and potential energy , e.g. , a spring - mass system or a parallel lc circuit , the virial theorem says that for any time - periodic state of the system , the time - averaged kinetic energy equals the time - averaged potential energy . this result is known as the equipartition of energy because of the fact that the total energy of the system , i.e. , the hamiltonian , is equal to the sum of the kinetic and potential energy . it is this result that we generalize in this paper for dissipative lagrangian systems and their damped oscillatory modes . in particular , we show that _ for any damped oscillatory mode of the system the kinetic energy equals the potential energy _ and so _ there is an equipartition of the system energy for these damped oscillatory modes into kinetic and potential energy_. our precise statements of these results can be found in section [ svir ] in theorem [ tvir ] and corollary [ tvir_eqp ] . we have also found that the transition to overdamping is characterized by a breakdown of the virial theorem . more specifically , as the losses of the lossy component increase some of the modes become overdamped with complete ceasing of any oscillations and with breaking of the equality between kinetic and potential energy . such a _ breaking of the equipartition of the energy can viewed as a dynamical characterization of the overdamped modes ! _ so we now have two different characterizations of overdamping : spectral and dynamical . the effectiveness of this dynamical characterization is demonstrated in our study of the selective overdamping phenomenon and in proving that a positive fraction of modes always remain oscillatory no matter how large the losses become in the lossy component of the composite system . * standard spectral theory vs. krein spectral theory. the study of the eigenmodes of a lagrangian system often relies on the fact the system evolution can be transformed into the hamiltonian form as the first - order linear differential equations . however , it is not emphasized enough that the hamiltonian setting does not lead directly to the standard spectral theory of self - adjoint or dissipative operators but that only it might be reduced to one in some important and well known cases by a proper transformation as , for instance , in the case of a simple oscillator . the general spectral theory of hamiltonian systems is known as the _ krein spectral theory _ @xcite , @xcite , @xcite . this theory is far more complex than the standard spectral theory and it is much harder to apply . we have found , as discussed in section [ smodel ] , a general transformation that reduces the evolution equation to the standard spectral theory under the condition of positivity of the hamiltonian . the standard spectral theory is complete , well understood , and allows for an elaborate and effective perturbation theory . we used that all in our analysis of the modal dichotomy and symmetries of the spectrum in section [ sspan ] as well as the selective overdamping phenomenon in section [ sodan ] . our main results on spectral symmetries are proposition [ pssym ] and corollary [ cssym ] . our principal results on the modal dichotomy can be found in theorems [ tmdic ] , [ cpodr ] and corollaries [ cpfsp ] , [ cmdic ] . in fact , it is important to point out that the key result of this paper that leads to the modal dichotomy which is essential in the analysis of overdamping is proposition [ pevbd ] which relates to certain eigenvalue bounds . moreover , these bounds and the dichotomy come from some general results that we derive in appendix [ apxebd ] on the perturbation theory for matrices with a large imaginary part . [ [ organization - of - the - paper . ] ] organization of the paper . + + + + + + + + + + + + + + + + + + + + + + + + + + the rest of the paper is organized as follows . section [ smodel ] sets up and discusses the lagrangian framework approach and model used in this paper to study the dissipative properties of two - component composite systems with a high - loss and a lossless component . in section [ scircuits ] we apply the developed approach to an electric circuit showing all key features of the method . sections [ svir][sodan ] are devoted to a precise formulation of all significant results in the form of theorems , propositions , and so on ( with section [ sproofs ] providing the proofs of these results ) . more specifically , section [ svir ] discusses our extension of the virial theorem and equipartition of energy from classical mechanics for conservative lagrangian systems to dissipative systems . section [ sspan ] is on the spectral analysis of the system eigenmodes including spectral symmetries , modal dichotomy , and perturbation theory in the high - loss regime . section [ sodan ] contains our analysis of the overdamping phenomena including selective overdamping . appendices [ apxsc ] and [ apxebd ] contain results we need that we believe will be of use more generally in studying dissipative dynamical systems , but especially for composite systems with high - loss and lossless components . in particular , appendix [ apxebd ] is on the perturbation theory for matrices with a large imaginary part . finally , appendices [ apenerg ] and [ apdxvirthm ] review the virial theorem from classical mechanics from the energetic point of view . the lagrangian and hamiltonian frameworks have numerous well known advantages in describing the evolution of physical systems . these advantages include , in particular , the universality of the mathematical structure , the flexibility in the choice of variables and the incorporation of the symmetries with the corresponding conservations laws . consequently , it seems quite natural and attractive to integrate dissipation into the lagrangian and hamiltonian frameworks . we are particular interested in integrating dissipation into systems with gyroscopic features such as dielectric media with magnetic components and electrical networks . we start with setting up the lagrangian and hamiltonian structures similarly to those in @xcite and @xcite . in particular , we use matrices to efficiently handle possibly large number of degrees of freedom . since we are interested in systems evolving linearly the lagrangian @xmath3 is assumed to be a quadratic function ( bilinear form ) of the coordinates @xmath4 _ { r=1}^{n}$ ] ( column vector ) and their time derivatives @xmath5 , that is @xmath6{l}\dot{q}\\ q \end{array } \right ] ^{\mathrm{t}}m_{\mathrm{l}}\left [ \begin{array } [ c]{l}\dot{q}\\ q \end{array } \right ] , \qquad m_{\mathrm{l}}=\left [ \begin{array } [ c]{ll}\alpha & \theta\\ \theta^{\mathrm{t } } & -\eta \end{array } \right ] , \label{dlag1}\ ] ] where @xmath7 denotes the matrix transposition operation , and @xmath8 and @xmath9 are @xmath10-matrices with real - valued entries . in addition to that , we assume @xmath8 to be symmetric matrices which are positive definite and positive semidefinite , respectively , and we assume @xmath9 is skew - symmetric , that is @xmath11 notice that only the skew - symmetric part of @xmath9 in ( [ dlag1 ] ) matters for the system dynamics because of the form of the euler - lagrange equations ( [ dlag4 ] ) and so without loss of generality we can assume as we have that the matrix @xmath9 is skew - symmetric . in fact , the symmetric part of the matrix @xmath9 if any alters the lagrangian by a complete time derivative and consequently can be left out as we have done . the lagrangian ( [ dlag1 ] ) can be written in the energetic form@xmath12 where @xmath13 are interpreted as the kinetic and potential energy , respectively . by hamilton s principle the dynamics of the system are governed by the euler - lagrange equations @xmath14 which are the following second - order ordinary differential equation ( odes ) @xmath15 the above second - order odes can be turned in the first - order odes with help of the hamiltonian function defined by the lagrangian through the legendre transformation @xmath16 hence,@xmath17 and , consequently @xmath18 = \frac{1}{2}\dot{q}^{t}\alpha\dot{q}+\frac{1}{2}q^{t}\eta q. \label{dlag7}\ ] ] in particular , the hamiltonian is the sum of the kinetic and potential energy , that is@xmath19 we remind that the hamiltonian function @xmath20 is interpreted as the system energy which is a conserved quantity , that is@xmath21 the function @xmath20 defined by ( [ dlag7 ] ) is a quadratic form associated with a matrix @xmath22 through the relation @xmath23{l}p\\ q \end{array } \right ] ^{\mathrm{t}}m_{\mathrm{h}}\left [ \begin{array } [ c]{l}p\\ q \end{array } \right ] , \label{dlag8}\ ] ] where @xmath22 is the @xmath24 matrix having the block form @xmath25{ll}\alpha^{-1 } & -\alpha^{-1}\theta\\ -\theta^{\mathrm{t}}\alpha^{-1 } & \theta^{\mathrm{t}}\alpha^{-1}\theta+\eta \end{array } \right ] = \left [ \begin{array } [ c]{ll}\mathbf{1 } & 0\\ -\theta^{\mathrm{t } } & \mathbf{1}\end{array } \right ] \left [ \begin{array } [ c]{ll}\alpha^{-1 } & 0\\ 0 & \eta \end{array } \right ] \left [ \begin{array } [ c]{ll}\mathbf{1 } & -\theta\\ 0 & \mathbf{1}\end{array } \right ] , \label{dlag7a}\ ] ] where @xmath26 is the identity matrix . notice that the representations ( [ dlag8 ] ) , ( [ dlag7a ] ) combined with the inequalities ( [ dlag2 ] ) imply@xmath27 the hamiltonian evolution equations then take the form@xmath28{l}p\\ q \end{array } \right ] = jm_{\mathrm{h}}\left [ \begin{array } [ c]{l}p\\ q \end{array } \right ] , \quad j=\left [ \begin{array } [ c]{ll}0 & -\mathbf{1}\\ \mathbf{1 } & 0 \end{array } \right ] \text { , } \label{dlag9}\ ] ] observe that the symplectic matrix @xmath29 satisfies the following relations@xmath30 the rayleigh dissipation function . * up to now we have dealt with a conservative system satisfying the energy conservation ( [ dlag7b ] ) . let us introduce now dissipative forces using rayleigh s method described in ( * ? ? ? 10.11 , 10.12 ) , ( * ? ? ? * sec . 8 , 9 , 46 ) . the _ rayleigh dissipation function _ @xmath31 is defined as a quadratic function of the generalized velocities , namely @xmath32 where the scalar @xmath33 is a dimensionless _ loss parameter _ which we introduce to scale the intensity of dissipation . in particular , @xmath34 is an @xmath10 symmetric matrix with real - valued entries and , most importantly , is positive semidefinite . the dissipation is then introduced through the following general euler - lagrange equations of motion with forces@xmath35 where @xmath36 are generalized dissipative forces and @xmath37 is an external force , yielding the following second - order odes @xmath38 now by a linear ( dissipative ) _ lagrangian system _ we mean a system whose state is described by a time - dependent @xmath39 taking values in the hilbert space @xmath40 with the standard inner product @xmath41 ( i.e. , @xmath42 , where @xmath43 denotes the conjugate transpose , i.e. , @xmath44 ) whose dynamics are governed by the odes ( [ radis2a ] ) . the energy balance equation for any such state , which follows from ( [ radis2a ] ) , is @xmath45 where @xmath46 as in ( [ dlag7c ] ) but now instead of ( [ dlag3_1 ] ) we have@xmath47 we continue to interpret @xmath48 , @xmath49 as the kinetic and potential energies , @xmath50 as the system energy , and @xmath51 as the dissipated power . the term @xmath52 is interpreted as the rate of work done by the force @xmath53 . the physical significance of the various energetic terms @xmath54 , where @xmath55 , is that for a complex - valued state @xmath56 with real - valued @xmath57 both @xmath58 and @xmath59 are also states representing physical solutions and @xmath60 . these latter two terms in the sum reduce to the previous definitions of the energetic term @xmath54 above in ( [ dlag3_1 ] ) , ( [ dlag7 ] ) , or ( [ radis1 ] ) for real - valued vector quantities . we now introduce an important quantity the _ loss fraction _ @xmath61 . it is defined as the ratio of the rank of the matrix @xmath34 to the total degrees of freedom @xmath62 of the system@xmath63 from our hypothesis @xmath64 it follows that @xmath65 . we consider the dissipative lagrangian system to be a model of a two - component composite with a lossy and a lossless component whenever the loss fraction condition @xmath66 is satisfied . we then associate the range of the operator @xmath34 , i.e. , @xmath67 , with the lossy component of the system and consider the lossy component to be highly lossy when @xmath68 , i.e. , the high - loss regime . our paper is focused on this case . * on the eigenmodes and the quality factor . * to study the dissipative properties of the dissipative lagrangian system ( [ radis2a ] ) in the high - loss regime @xmath68 , one can consider the eigenmodes of the system and their quality factor in this regime . an eigenmode of this system ( [ radis2a ] ) is defined as a nonzero solutions of the odes ( [ radis2a ] ) with no forcing , i.e. , @xmath69 , having the form @xmath70 , where @xmath71 and @xmath72 are called its frequency and damping factor , respectively . since the system is dissipative , i.e. , dissipates energy as interpreted from the energy balance equation ( [ radis4 ] ) , the damping factor must satisfy @xmath73 which implies that @xmath74 . thus , an eigenmode is a state of the dissipative lagrangian system which is in damped harmonic motion and the motion is oscillatory provided @xmath75 . an important quantity which characterizes the quality of such a damped oscillation is the quality factor @xmath76 ( i.e. , @xmath77-factor ) of the eigenmode which is defined as the reciprocal of the relative rate of energy dissipation per temporal cycle , i.e. , @xmath78 it follows immediately from this definition and ( [ radis4_1 ] ) that @xmath76 depends only on the complex frequency @xmath79 of the eigenmode , in particular,@xmath80 with the convention @xmath81 if @xmath82 . now the matrix hamilton form of the euler - lagrange equation involving rayleigh dissipative forces and external force ( [ radis2 ] ) reads@xmath83{ll}\beta r & 0\\ 0 & 0 \end{array } \right ] \right ) m_{\mathrm{h}}u+\left [ \begin{array } [ c]{l}f\\ 0 \end{array } \right ] , \quad u=\left [ \begin{array } [ c]{l}p\\ q \end{array } \right ] . \label{radis3}\ ] ] it is important to recognize that the evolution equation ( [ radis3 ] ) are not quite yet of the desired form which is the most suitable for the spectral analysis . advancing general ideas of the hamiltonian treatment of dissipative systems developed in @xcite we can factor the matrix @xmath22 as @xmath84 where the matrix @xmath85 is the block matrix@xmath86{ll}k_{\mathrm{p } } & 0\\ 0 & k_{\mathrm{q}}\end{array } \right ] \left [ \begin{array } [ c]{ll}\mathbf{1 } & -\theta\\ 0 & \mathbf{1}\end{array } \right ] = \left [ \begin{array } [ c]{ll}k_{\mathrm{p } } & -k_{\mathrm{p}}\theta\\ 0 & k_{\mathrm{q}}\end{array } \right ] , \label{radis6}\\ k_{\mathrm{p}}=\sqrt{\alpha}^{-1},\text { } k_{\mathrm{q}}=\sqrt{\eta } \label{radis6a1}\ ] ] which manifestly takes into account the gyroscopic term @xmath9 . here @xmath87 and @xmath88 denote the unique positive semidefinite square roots of the matrices @xmath89 and @xmath90 , respectively . in particular , it follows from the properties ( [ dlag2 ] ) and the proof of ( * ? ? ? * , theorem vi.9 ) that @xmath91 , @xmath92 are @xmath10 matrices with real - valued entries with the properties @xmath93 the introduction of the matrix @xmath85 according to @xcite is intimately related to the introduction of force variables@xmath94 consequently , recasting the evolution equation ( [ radis3 ] ) in terms of the force variables @xmath95 yields the desired canonical form , namely @xmath96 and@xmath97{cc}\omega_{\mathrm{p } } & -\mathrm{i}\phi^{\mathrm{t}}\\ \mathrm{i}\phi & 0 \end{array } \right ] , \text { \ } b = k\left [ \begin{array } [ c]{ll}r & 0\\ 0 & 0 \end{array } \right ] k^{\mathrm{t}}=\left [ \begin{array } [ c]{ll}\mathsf{\tilde{r } } & 0\\ 0 & 0 \end{array } \right ] , \text { \ } f=\left [ \begin{array } [ c]{l}k_{\mathrm{p}}f\\ 0 \end{array } \right ] , \label{radis8a}\\ \omega_{\mathrm{p}}=-\mathrm{i}2k_{\mathrm{p}}\theta k_{\mathrm{p}}^{\mathrm{t}},\text { \ \ } \phi = k_{\mathrm{q}}k_{\mathrm{p}}^{\mathrm{t}},\text { \ \ } \mathsf{\tilde{r}}=k_{\mathrm{p}}rk_{\mathrm{p}}^{\mathrm{t}}. \label{radis8b}\ ] ] observe that the @xmath24 matrices @xmath98 , @xmath99 in block form are hermitian and positive semidefinite , respectively , that is @xmath100 moreover , the matrix @xmath99 does not have full rank and its rank is that of @xmath34 , i.e. @xmath101 the _ system operator _ @xmath102 has the following important properties @xmath103 where the latter property comes from the fact that @xmath104 and @xmath105 . also , recall that for any square matrix @xmath106 , one can write @xmath107 , where @xmath108 and @xmath109 denote the real and imaginary parts of the matrix @xmath106 , respectively . now by the _ canonical system _ we mean the system whose state is described by a time - dependent @xmath110 taking values in the hilbert space @xmath111 with the standard inner product @xmath112 whose dynamics are governed by the odes ( [ radis8 ] ) . these odes will be referred to as the _ canonical evolution equation_. in this paper we will focus on the case when there is no forcing , i.e. , @xmath113 ( or , equivalently , @xmath69 ) . in this case , the canonical system represents a dissipative dynamical system ( since @xmath114 ) with the evolution governed by the semigroup @xmath115 . as we shall see in this paper and as mentioned in the introduction , there are some serious advantages in studying this dissipative dynamical system , i.e. , the canonical system , over the lagrangian or hamiltonian systems . in a previous work @xcite of ours , we studied canonical systems whose states are solutions of a linear evolution equation in the canonical form ( [ radis8 ] ) with system operator @xmath102@xmath116 @xmath117 , @xmath118 having exactly the properties ( [ radis10 ] ) in which the matrix @xmath99 did not have full rank . in that paper , the canonical system was a simplified version of an abstract model of an oscillator damped retarded friction that modeled a two - component composite system with a lossy and lossless component . we showed that the state @xmath110 of such a system satisfied the energy balance equation@xmath119 = -w_{\mathrm{dis}}\left [ v(t)\right ] + w\left [ v\left ( t\right ) \right ] , \label{radis11}\ ] ] in which the system energy @xmath120 $ ] , dissipated power @xmath121 $ ] , and @xmath122 $ ] the rate of work done by the force @xmath123 were given by @xmath124 = \frac{1}{2}\left ( v\left ( t\right ) , v\left ( t\right ) \right ) , \text { } w_{\mathrm{dis}}\left [ v(t)\right ] = \beta\left ( v\left ( t\right ) , bv\left ( t\right ) \right ) , \text { } w\left [ v\left ( t\right ) \right ] = \operatorname{re}\left ( v\left ( t\right ) , f\left ( t\right ) \right ) . \label{radis11a}\ ] ] an important result of this paper which follows immediately from the block form ( [ radis8a ] ) , ( [ radis8b ] ) and the relation between the variables @xmath95 , @xmath125 , @xmath126 from ( [ radis7 ] ) , ( [ radis3 ] ) , ( [ dlag6 ] ) ] is that a state @xmath77 of the lagrangian system , whose dynamics are governed by second - order odes ( [ radis2a ] ) , and the corresponding state @xmath95 of the canonical system , whose dynamics are governed by the canonical evolution equation ( [ radis8 ] ) , the energetics are equivalent in the sense @xmath127 = \mathcal{t}\left ( \dot{q},q\right ) + \mathcal{v}\left ( \dot{q},q\right ) , \text { \ } w_{\mathrm{dis}}\left [ v\right ] = 2\mathcal{r}\left ( \dot{q}\right ) , \text { \ } w\left [ v\right ] = \operatorname{re}\left ( \dot{q},f\right ) \text { , } \label{radis11b}\ ] ] where , as you will recall from the energy balance equation ( [ radis4 ] ) , @xmath128 was the system energy for the lagrangian system . * on the eigenmodes and the quality factor . * we will be interested in the eigenmodes of the canonical system ( [ radis8 ] ) and their quality factor . an eigenmode of the canonical system is defined as a nonzero solutions of the odes ( [ radis8 ] ) with no forcing , i.e. , @xmath113 , having the form @xmath129 . the quality factor @xmath130 $ ] of such an eigenmode is defined as in @xcite by @xmath131=2\pi\frac{\text{energy stored in system}}{\text{energy lost per cycle}}=\left\vert \operatorname{re}\zeta\right\vert \frac{u\left [ v\left ( t\right ) \right ] } { w_{\mathrm{dis}}\left [ v(t)\right ] } . \label{radis12}\ ] ] by the energy balance equation ( [ radis11 ] ) it follows that @xmath131=-\frac{1}{2}\frac{\left\vert \operatorname{re}\zeta\right\vert } { \operatorname{im}\zeta } , \label{radis12_1}\ ] ] with the convention @xmath130=+\infty$ ] if @xmath82 . now given an eigenmode @xmath132 of the lagrangian system [ radis2a ] ) , it follows that the corresponding state @xmath95 of the canonical system ( [ radis8 ] ) , related by ( [ radis7 ] ) , is an eigenmode of the form @xmath133 ( excluding the case @xmath134 , which can only occur if @xmath135 ) . importantly , the energetic equivalence ( [ radis11b ] ) holds and their quality factors ( [ radis4_3 ] ) , ( [ radis12_1 ] ) are equal , i.e.,@xmath136 = q_{\zeta}\text{. } \label{radis13}\ ] ] a more detailed discussion on the eigenmodes of the two systems and the relationship between them can be found in section [ sspan ] . one of the important applications of our methods described above is electric circuits and networks involving resistors representing losses . a general study of electric networks with losses can be carried out with the help of the lagrangian approach , and that systematic study has already been carried out in this paper . for lagrangian treatment of electric networks and circuits we refer to ( * ? ? ? 9 ) , ( * ? ? ? 2.5 ) , @xcite . we illustrate the idea and give a flavor of the efficiency of our methods by considering below a rather simple example of an electric circuit as in fig . [ figc1 ] with the assumptions@xmath137 this example has the essential features of two component systems incorporating high - loss and lossless components . , @xmath138 , @xmath139 , two inductances @xmath140 , @xmath141 , a resistor @xmath142 , and two sources @xmath143 , @xmath144 . this electric circuit example fits within the framework of our model . indeed , since resistors represent losses , this two component system consists of a lossy component and a lossless component the right and left circuits , respectively.,width=459,height=251 ] * the lagrangian system . * to derive evolution equations for the electric circuit in fig . [ figc1 ] we use a general method for constructing lagrangians for circuits , ( * ? ? ? 9 ) , that yields@xmath145 where @xmath48 and @xmath49 are respectively the kinetic and the potential energies , @xmath146 is the lagrangian , and @xmath31 is the rayleigh dissipative function . notice that @xmath147 and @xmath148 are the currents . the general euler - lagrange equations of motion with forces are , ( * ? ? ? 8),@xmath149 where @xmath77 are the charges and @xmath53 the sources@xmath150{c}q_{1}\\ q_{2}\end{array } \right ] , \quad f=\left [ \begin{array } [ c]{c}e_{1}\\ e_{2}\end{array } \right ] , \label{circ3}\ ] ] yielding from ( [ circ0])([circ3 ] ) the following second - order odes@xmath151 with the dimensionless loss parameter@xmath152 that scales the intensity of losses in the system , and@xmath153{cc}l_{1 } & 0\\ 0 & l_{2}\end{array } \right ] , \text { \ \ } r=\left [ \begin{array } [ c]{cc}0 & 0\\ 0 & \ell \end{array } \right ] , \text { \ \ } \eta=\left [ \begin{array } [ c]{cc}\frac{1}{c_{1}}+\frac{1}{c_{12 } } & -\frac{1}{c_{12}}\\ -\frac{1}{c_{12 } } & \frac{1}{c_{2}}+\frac{1}{c_{12}}\end{array } \right ] . \label{circ6}\ ] ] recall , the loss fraction @xmath1 defined in ( [ radis4a ] ) is the ratio of the rank of the matrix @xmath34 to the total degrees of freedom @xmath62 of the system which in this case is @xmath154 thus the lagrangian system ( [ circ4 ] ) has all the properties described in sections [ slagrangian ] , [ slagrmodeleigenmqfac ] and since the loss fraction condition ( [ radis4b ] ) , i.e. , @xmath0 , is satisfied then it is model of a two - component composite with a lossy and a lossless component . * the canonical system . * following the method in section [ scansys ] , we introduce the variables @xmath95 defined by ( [ dlag6 ] ) , ( [ radis3 ] ) , ( [ radis7 ] ) in terms of the matrix @xmath85 defined in ( [ radis6 ] ) , ( [ radis6a1 ] ) as@xmath155{ll}k_{\mathrm{p } } & 0\\ 0 & k_{\mathrm{q}}\end{array } \right ] , \text { \ \ } k_{\mathrm{p}}=\sqrt{\alpha}^{-1}=\left [ \begin{array } [ c]{ll}\frac{1}{\sqrt{l_{1 } } } & 0\\ 0 & \frac{1}{\sqrt{l_{2}}}\end{array } \right ] , \text { \ \ } k_{\mathrm{q}}=\sqrt{\eta}.\nonumber\end{gathered}\ ] ] as @xmath156 is a @xmath157 matrix in this case , then we can compute its square root explicitly using the formula@xmath158 where @xmath159 denotes the positive square root , which holds for any @xmath157 matrix @xmath160 with @xmath161 . consequently , in the lagrangian system ( [ circ4 ] ) in terms of the variable @xmath95 is the desired canonical evolution equation from ( [ radis8 ] ) , namely , @xmath162 and@xmath163{cc}0 & -\mathrm{i}\phi^{\mathrm{t}}\\ \mathrm{i}\phi & 0 \end{array } \right ] , \text { \ } b=\left [ \begin{array } [ c]{ll}\mathsf{\tilde{r } } & 0\\ 0 & 0 \end{array } \right ] , \text { \ } f=\left [ \begin{array } [ c]{l}k_{\mathrm{p}}f\\ 0 \end{array } \right ] , \label{circ11}\\ \phi = k_{\mathrm{q}}k_{\mathrm{p}}^{\mathrm{t}},\text { \ \ } \mathsf{\tilde{r}}=k_{\mathrm{p}}rk_{\mathrm{p}}^{\mathrm{t}}=\left [ \begin{array } [ c]{ll}0 & 0\\ 0 & \frac{\ell}{l_{2}}\end{array } \right ] .\nonumber\end{gathered}\ ] ] in this section we will introduce a new virial theorem for dissipative lagrangian systems in terms of the eigenmodes of the system . this result generalizes the virial theorem from classical mechanics ( * ? ? ? * ; 3.4 ) , as discussed in appendix [ apdxvirthm ] , for conservative lagrangian systems for these types of modes . recall that the virial theorem is about the oscillatory transfer of energy from one form to another and its equipartition . we can now state precisely and prove our generalization of the virial theorem to dissipative ( @xmath118 ) lagrangian systems ( [ radis2a ] ) and the equipartition of energy for the oscillatory eigenmodes . in fact , the proof in section [ sproofs ] shows that our theorem is true for more general system of odes in the form ( [ radis2a ] ) since essentially all that really matters is that the energy balance equation ( [ radis4 ] ) holds for the eigenmodes . [ virial theorem][tvir]if @xmath70 is an eigenmode of the lagrangian system ( [ radis2a ] ) then the following identity holds if @xmath75 : @xmath164 where @xmath165 and @xmath166 are the kinetic and potential energy , respectively , defined in ( [ radis4_0 ] ) . on the other hand , if @xmath167 then the identity ( [ tvirs1 ] ) no longer holds and either @xmath135 or the identity @xmath168 must hold . [ energy equipartition][tvir_eqp]for systems with @xmath169 , if @xmath70 is an eigenmode of the lagrangian system ( [ radis2a ] ) with @xmath75 , then the following identity holds@xmath170 in other words , for the oscillatory eigenmodes there is an equipartition of the system energy , i.e. , @xmath128 , between their kinetic energy @xmath48 and their potential energy @xmath49 . in this section we study time - harmonic solutions to the euler - lagrange equation ( [ radis2a ] ) which constitutes a subject of the spectral theory . an important subject of this section is the study of the relations between standard spectral theory and the quadratic pencil formulation as it arises naturally as the time fourier transformation of the euler - lagrange evolution equation . as introduced in section [ slagrmodeleigenmqfac ] , the eigenmodes of the lagrangian system are nonzero solutions of the odes ( [ radis2a ] ) with @xmath69 having the form @xmath70 . for a fixed @xmath33 , these modes correspond to the solutions of the quadratic eigenvalue problem ( qep)@xmath171 for the quadratic matrix pencil in @xmath79@xmath172 the set of eigenvalues ( spectrum ) of the pencil @xmath173 is the set@xmath174 which are exactly those values @xmath79 for which a solution to the qep ( [ pfsp1 ] ) exists . the spectral theory of polynomial operator pencils @xcite can be applied to study the eigenmodes but it has its disadvantages such as being more complicated than standard spectral theory . thus an alternative approach to the spectral theory is desirable . often the alternative approach is to use the hamiltonian system and consider its eigenmodes , that is , the nonzero solutions of the hamiltonian equations ( [ radis3 ] ) with @xmath69 having the form @xmath175 . these modes correspond to the solutions of the eigenvalue problem@xmath176 for the matrix@xmath177{ll}\beta r & 0\\ 0 & 0 \end{array } \right ] \right ) m_{\mathrm{h}}. \label{pfsp5}\ ] ] an advantage to this approach is the simple correspondence via ( [ dlag6 ] ) , ( [ radis3 ] ) between the set of modes of the two systems , namely , the eigenmodes of the hamiltonian system are solutions of ( [ radis3 ] ) having the block form @xmath178{l}p\\ q \end{array } \right ] e^{-\mathrm{i}\zeta t}$ ] in which @xmath70 an eigenmode of the lagrangian system and @xmath179 . in particular , this means the matrix @xmath180 and the pencil @xmath181 have the same eigenvalues and hence the same spectrum , i.e.,@xmath182 a major disadvantage of this approach is that @xmath183 is a non - self - adjoint matrix such that the standard theory of self - adjoint or dissipative operators does not apply without further transformation of the system , even in the absence of losses , i.e. , @xmath184 . instead , in this case it is the krein spectral theory @xcite , @xcite , @xcite that is often used . but this theory is far more complex than the standard spectral theory and much harder to apply . our approach to these spectral problems which overcomes the disadvantages in the pencil or krein spectral theory is to use the canonical system and consider its eigenmodes , that is , the nonzero solutions of the canonical evolution equation ( [ radis8 ] ) with @xmath113 having the form @xmath185 . these modes correspond to the solutions of the eigenvalue problem@xmath186 for the system operator @xmath187 with hermitian matrices @xmath98 , @xmath99 with @xmath188 . the advantage of this approach is that the standard spectral theory can be used since @xmath189 is a dissipative operator and when losses are absent @xmath190 is self - adjoint . this is a serious advantage since the spectral theory is significantly simpler and allows the usage of the deep and effective results from perturbation theory such as those developed in @xcite to study the modes of such canonical evolution equations having the general form ( [ radis8 ] ) . this is the approach we take in this paper to study spectral symmetries and the dissipative properties of the eigenmodes of the lagrangian system such as the modal dichotomy and overdamping phenomenon . * correspondence between spectral problems . * we now conclude this section by summarizing the correspondence between the two main spectral problems of this paper , namely , between the standard eigenvalue problem ( [ pfsp7 ] ) and the quadratic eigenvalue problem ( [ pfsp1 ] ) . we do this in the next corollary which uses the following proposition that tells us the characteristic matrix of the system operator @xmath191 can be factored in terms of the quadratic matrix pencil @xmath192 . first we introduce some notation that will be useful . the hilbert space @xmath111 with standard inner product @xmath112 can be decomposed as @xmath193 into the orthogonal subspaces @xmath194 , @xmath195 with orthogonal matrix projections@xmath196{cc}\mathbf{1 } & 0\\ 0 & 0 \end{array } \right ] , \text { \ \ } p_{\mathrm{q}}=\left [ \begin{array } [ c]{cc}0 & 0\\ 0 & \mathbf{1}\end{array } \right ] . \label{pfsp8a}\ ] ] in particular , the matrices @xmath98 , @xmath99 , and @xmath102 defined in ( [ radis8])([radis8b ] ) are block matrices already partitioned with respect to the decomposition @xmath197 and any vector @xmath198 can be represented uniquely in the block form@xmath199{c}\varphi\\ \psi \end{array } \right ] , \text { \ \ where } \varphi = p_{\mathrm{p}}w,\text { \ \ } \psi = p_{\mathrm{q}}w . \label{pfsp8b}\ ] ] then with respect to this decomposition we have the following results . [ ppfsp]if @xmath200 then @xmath201{cc}k_{\mathrm{p } } & \zeta^{-1}\mathrm{i}\phi^{\mathrm{t}}\\ 0 & \mathbf{1}\end{array } \right ] \left [ \begin{array } [ c]{cc}\zeta^{-1}\mathbf{1 } & 0\\ 0 & \zeta\mathbf{1}\end{array } \right ] \left [ \begin{array } [ c]{cc}c(\zeta,\beta ) & 0\\ 0 & \mathbf{1}\end{array } \right ] \left [ \begin{array } [ c]{cc}k_{\mathrm{p}}^{\mathrm{t } } & 0\\ -\zeta^{-1}\mathrm{i}\phi & \mathbf{1}\end{array } \right ] . \label{pfsp8}\ ] ] [ spectral equivalence][cpfsp]for any @xmath202 , @xmath203 in particular , the system operator @xmath102 and quadratic matrix pencil @xmath204 have the same spectrum , i.e. , @xmath205 moreover , if @xmath200 then the following statements are true : 1 . if @xmath206 and @xmath207 then @xmath199{c}-\mathrm{i}\zeta\sqrt{\alpha}q\\ \sqrt{\eta}q \end{array } \right ] , \text { \ \ where } c(\zeta,\beta)q=0,\text { \ \ } q\not = 0 . \label{cpfsp3}\ ] ] 2 . if @xmath208 and @xmath209 then @xmath210{c}-\mathrm{i}\zeta\sqrt{\alpha}q\\ \sqrt{\eta}q \end{array } \right ] \not = 0 . \label{cpfsp4}\ ] ] in section [ stvspen ] it was shown that the study of the eigenmodes of the lagrangian system ( [ radis2a ] ) reduces to the quadratic eigenvalue problem ( [ pfsp1 ] ) and this motivates a study of the spectrum @xmath211 of the quadratic matrix pencil @xmath204 in ( [ pfsp2 ] ) . corollary [ cpfsp ] makes it clear that we can instead study the eigenmodes of the canonical system ( [ radis8 ] ) and , in particular , the system operator @xmath102 spectrum satisfies @xmath212 . the purpose of this section is to give a detailed analysis of the set @xmath213 . recall , the system operator @xmath214 , @xmath118 from ( [ radis8 ] ) with @xmath24 matrices @xmath98 , @xmath99 has the fundamental properties ( [ radis10a ] ) , ( [ radis10b])@xmath215 where @xmath216 . these properties are particularly important in describing the spectrum @xmath213 of the system operator @xmath102 . for instance , the next proposition on the spectrum for nondissipative ( @xmath184 ) lagrangian systems ( [ radis2a ] ) follows immediately from these properties which would otherwise not be exactly obvious for gyroscopic systems ( i.e. , @xmath217 ) . [ real eigenfrequencies]for nondissipative lagrangian systems ( [ radis2a ] ) , that is when @xmath218 , all the eigenfrequencies @xmath79 are real , and consequently any eigenmode evolution is of the form @xmath219 with @xmath82 . if @xmath184 ( i.e. , no dissipation ) then @xmath220 is a hermitian matrix thus , the spectrum @xmath221 is a subset of @xmath222 . hence , if @xmath70 is an eigenmode of the lagrangian systems ( [ radis2a ] ) with @xmath184 then by corollary [ cpfsp ] we have @xmath223 and so @xmath82 . this completes the proof . in the next few sections we will give a deeper analysis of the spectral properties of @xmath102 including spectral symmetries in section [ sspsy ] and the modal dichotomy in sections [ sevmd ] , [ smdhlr ] . in order to do so we must first introduce some notation . in the hilbert space @xmath111 , denote by @xmath224 , @xmath225 the nonzero eigenvalues of @xmath99 ( counting multiplicities ) with the smallest denoted by @xmath226 in particular , the spectrum of @xmath99 is @xmath227 where @xmath228 . denote the largest eigenvalue of @xmath98 by @xmath229 . it follows from the fact that @xmath98 is a hermitian matrix which is skew - symmetric that @xmath230 where @xmath231 denotes the operator norm on square matrices . the next proposition describes the spectral symmetries of the system operator @xmath232 which follow from the property @xmath233 . [ spectral symmetry][pssym]the following statements are true : 1 . the characteristic polynomial of @xmath102 satisfies @xmath234 for every @xmath202 . in particular , the spectrum @xmath235 of the system operator @xmath232 has the symmetry@xmath236 2 . if @xmath237 is an eigenvector of the system operator @xmath232 with corresponding eigenvalue @xmath79 then @xmath238 is an eigenvector of @xmath232 with corresponding eigenvalue @xmath239 . 3 . if @xmath184 ( i.e. , no dissipation ) then @xmath240 @xmath241 for every @xmath202 . we will denote the discs centered at the eigenvalues of @xmath242 with radius @xmath229 by@xmath243 two subsets of the spectrum @xmath244 of the system operator @xmath245 @xmath117 which play a central role in our analysis are@xmath246 [ eigenvalue bounds][pevbd]the following statements are true : 1 . the eigenvalues of the system operator @xmath102 lie in the union of the closed discs whose centers are the eigenvalues of @xmath242 with radius @xmath229 , that is,@xmath247 2 . if @xmath207 and @xmath206 then@xmath248 3 . if @xmath79 is an eigenvalue of @xmath102 and @xmath249 then @xmath250 [ spectral clustering][cssym]the eigenvalues of the system operator @xmath187 , @xmath118 lie in the closed lower half of the complex plane , are symmetric with respect to the imaginary axis , and lie in the union of the closed discs whose centers are the eigenvalues of @xmath242 with radius @xmath229 . moreover , if @xmath184 ( i.e. , no dissipation ) then the eigenvalues of @xmath220 are real and symmetric with respect to the origin . [ modal dichotomy i][tmdic]if @xmath251 then @xmath252 furthermore , there exists unique invariant subspaces @xmath253 , @xmath254 of the system operator @xmath255 @xmath117 with the properties@xmath256 where @xmath111 . moreover , the dimensions of these subspaces satisfy@xmath257 [ high - loss susceptible subspace][dmdic]for the system operator @xmath258 @xmath117 with @xmath259 we will call its @xmath260-dimensional invariant subspace @xmath254 the high - loss susceptible subspace . we will call its @xmath261-dimensional invariant subspace @xmath262 the low - loss susceptible subspace . our reasoning for the definitions of these subspaces is clarified with the following corollary . [ high - loss subspace : dissipative properties][cmdic]if @xmath263 then@xmath264 furthermore , the quality factor ( [ radis12_1 ] ) of any eigenmode of the canonical system ( [ radis8 ] ) in the high - loss susceptible subspace @xmath254 satisfies @xmath265\leq\frac{1}{2}\frac{\omega_{\text{max}}}{\beta b_{\text{min}}-\omega_{\text{max}}}<\frac{1}{2}. \label{cmdic2}\ ] ] in particular , as the losses go to @xmath266 the damping factor and quality factor of any such eigenmode goes to @xmath267 and @xmath268 , respectively , that is,@xmath269=0 . \label{cmdic3}\ ] ] we conclude this section with the following remarks . the spectrum of @xmath270 restricted to the low - loss susceptible subspace @xmath262 , that is , the set @xmath271 in ( [ cmdic1 ] ) , is close to the real axis and actually coalesces to a finite set of real numbers as losses @xmath272 . this statement is made precise in the following section with theorem [ cpodr ] and the asymptotic expansions in ( [ pod14 ] ) . on the other hand , results on quality factor for the eigenmodes in @xmath273 is far more subtle than the results in corollary [ cmdic ] for the eigenmodes in the high - loss susceptible subspace @xmath254 . for gyroscopic - dissipative systems considered in this paper , corollary [ cmdic ] above and proposition [ pqaollm ] below give a partial description of the nature of the quality factor for large losses , i.e. , @xmath68 , in the lossy component of the composite system . when gyroscopy is absent , i.e. , @xmath169 , then a more complete analysis for quality factor can be carried out , which we have done in section [ sodan ] of this paper in connection to our studies on the overdamping phenomenon . consideration of quality factor and overdamping in dissipative systems with gyroscopy , however , requires a more subtle and detailed analysis that will be considered in a future work . we are interested in describing the spectrum @xmath274 of the system operator @xmath275 , @xmath118 in the high - loss regime , i.e. , @xmath68 . we do this in this section by giving an asymptotic characterization , as @xmath272 , of the modal dichotomy as described in theorem [ tmdic ] and corollary [ cmdic ] . in order to do so we need to give a spectral perturbation analysis of the matrix @xmath270 as @xmath272 . fortunately , this analysis has already been carried out in @xcite . we now introduce the necessary notion and describe the results . the hilbert space @xmath111 with standard inner product @xmath112 is decomposed into the direct sum of orthogonal invariant subspaces of the operator @xmath99 , namely @xmath276 where @xmath277 ( the range of @xmath99 ) is the loss subspace of dimension @xmath278 with orthogonal projection @xmath279 and its orthogonal complement , @xmath280 ( the nullspace of @xmath99 ) , is the no - loss subspace of dimension @xmath281 with orthogonal projection @xmath282 . the operators @xmath98 and @xmath99 with respect to the direct sum ( [ pod5 ] ) are the @xmath157 block operator matrices@xmath283{cc}\omega_{2 } & \theta\\ \theta^{\ast } & \omega_{1}\end{array } \right ] , \quad b=\left [ \begin{array } [ c]{cc}b_{2 } & 0\\ 0 & 0 \end{array } \right ] , \label{pod7}\ ] ] where @xmath284 and @xmath285 are restrictions of the operators @xmath98 and @xmath99 respectively to loss subspace @xmath286 whereas @xmath287 is the restriction of @xmath98 to complementary subspace @xmath288 . also , @xmath289 is the operator @xmath290 whose adjoint is given by @xmath291 . the perturbation analysis in the high - loss regime @xmath68 for the system operator @xmath292 described in ( * ? * a , theorem 5 & proposition 11 ) introduces an orthonormal basis @xmath293 diagonalizing the self - adjoint operators @xmath294 and @xmath295 from ( [ pod7 ] ) with@xmath296 where@xmath297 then for @xmath68 the system operator @xmath298 is diagonalizable with basis of eigenvectors @xmath299 satisfying@xmath300 which split into two distinct classes @xmath301 with the following properties . * the high - loss class : * the eigenvalues have poles at @xmath302 whereas their eigenvectors are analytic at @xmath302 , having the asymptotic expansions@xmath303 the vectors @xmath304 , @xmath305 form an orthonormal basis of the loss subspace @xmath286 and@xmath306 in particular , @xmath224 , @xmath225 are all the nonzero eigenvalues of @xmath99 ( counting multiplicities ) . * the low - loss class : * the eigenvalues and eigenvectors are analytic at @xmath302 , having the asymptotic expansions@xmath307 the vectors @xmath304 , @xmath308 form an orthonormal basis of the no - loss subspace @xmath309 and@xmath310 by ( * ? ? ? * a , proposition 7 ) we know that the asymptotic formulas for the real and imaginary parts of the complex eigenvalues @xmath311 as @xmath272 are given by @xmath312 observe that the expansions ( [ pod16 ] ) imply @xmath313 justifying the names high - loss and low - loss . the following theorem is the goal of this section . it characterizes the spectrum @xmath244 of the system operator @xmath187 and the modal dichotomy from theorem [ tmdic ] and corollary [ cmdic ] in the high - loss regime @xmath68 in terms of the high - loss and low - loss eigenvectors . [ modal dichotomy ii][cpodr]for @xmath33 sufficiently large , the modal dichotomy occurs as in theorem [ tmdic ] and corollary [ cmdic ] with the following equalities holding : @xmath314 and@xmath315 in particular , @xmath254 and @xmath253 , the high - loss and low - loss susceptible subspaces of @xmath270 , respectively , have as a basis the high - loss eigenvectors @xmath316 and the low - loss eigenvectors @xmath317 , respectively . in the next section we will study overdamping phenomena for lagrangian systems ( [ radis2a ] ) . an important role in our analysis will be played by the eigenmodes @xmath318 , @xmath319 of the canonical system ( [ radis8 ] ) which by the modal dichotomy split into the two distinct classes based on their dissipative properties:@xmath320 it should be emphasized here that the classification of these modes into high - loss and low - loss is based solely on the behavior of their damping factor in ( [ pod17 ] ) as losses become large , i.e. , as @xmath272 , and not necessarily on their quality factor which we will discuss at the end of this section . moreover , the damping factor is related to energy loss by the energy balance equation ( [ radis11 ] ) , ( [ radis11a ] ) satisfied by these modes which implies the dissipated power is@xmath321 & = -2\operatorname{im}\zeta_{j}\left ( \beta\right ) u[v_{j}\left ( t,\beta\right ) ] , \text { \ \ } 1\leq j\leq2n,\label{pod20a}\\ u[v_{j}\left ( t,\beta\right ) ] & = \frac{1}{2}\left ( w_{j}\left ( \beta\right ) , w_{j}\left ( \beta\right ) \right ) e^{2\operatorname{im}\zeta_{j}\left ( \beta\right ) t}\nonumber\end{aligned}\ ] ] where @xmath322 $ ] defined in ( [ radis11a ] ) was interpreted as the system energy . in particular , by ( [ pod17 ] ) , ( [ pod20a ] ) , and the fact @xmath323 it follows that their system energy for any _ fixed _ satisfies@xmath325=0,\text { \ \ } 1\leq j\leq n_{r};\label{pod20b}\\ \text{low - loss } & \text { : \ \ } \lim_{\beta\rightarrow\infty}u[v_{j}\left ( t,\beta\right ) ] = \frac{1}{2},\text { \ \ } n_{r}+1\leq j\leq2n.\nonumber\end{aligned}\ ] ] this combined with ( [ pod17 ] ) , ( [ pod20a ] ) is the justification for the names high - loss and low - loss modes . * asymptotic formulas for the quality factor . * the perturbation analysis in the high - loss regime @xmath68 for the quality factor @xmath326 $ ] , @xmath319 of the eigenmodes from ( [ pod20 ] ) as given by ( [ radis12 ] ) , ( [ radis12_1 ] ) has already been carried out in ( * ? ? ? * a , prop . we will now describe the results . the quality factor @xmath327 $ ] , @xmath328 for each high - loss eigenmode has a series expansion containing only odd powers of @xmath329 implying for @xmath68 it is a nonnegative decreasing function in the loss parameter @xmath33 which has the asymptotic formula as @xmath272@xmath330 = \frac { 1}{2}\frac{\left\vert \rho_{j}\right\vert } { b_{j}}\beta^{-1}+o\left ( \beta^{-3}\right ) , \ \ 1\leq j\leq n_{r}. \label{pod21}\ ] ] it follows from this discussion that as @xmath272 @xmath330 \searrow0,\ \ 1\leq j\leq n_{r } \label{pod21_1}\ ] ] ( here we will use the notation @xmath331 @xmath268 or @xmath332 to denote that a function of @xmath33 is decreasing to @xmath268 or increasing to @xmath267 , respectively , as @xmath272 ) . the quality factor @xmath327 $ ] , @xmath308 for each low - loss eigenmode has two possibilities : ( i ) @xmath327 = + \infty$ ] for @xmath68 ; ( ii ) @xmath327 $ ] has a series expansion containing only odd powers of @xmath329 implying either @xmath326 \searrow0 $ ] or @xmath333 \nearrow+\infty$ ] as @xmath272 . finding necessary and sufficient conditions for when the cases occur is a subtle problem which is still open in general , but is solved completely in the next section on overdamping when @xmath169 under the nondegeneracy condition @xmath334 . we will now consider the cases @xmath335 [ cf . ( [ pod22 ] ) , ( [ pod22_1 ] ) ] or @xmath336 ( cf . prop . [ pqaollm ] ) . in the former case ( which is the typical case see ( * ? ? ? * a , remark 9 ) ) , we have as @xmath272 the asymptotic formula@xmath337 = \frac{1}{2}\frac{\left\vert \rho_{j}\right\vert } { d_{j}}\beta+o\left ( \beta ^{-1}\right ) , \ \ \text{if } d_{j}\not = 0 \label{pod22}\ ] ] and so it follows in this case that as @xmath272 @xmath337 \searrow0\text { , if } \rho_{j}=0\text { and } q\left [ w_{j}\left ( \beta\right ) \right ] \nearrow+\infty,\text { if } \rho_{j}\not = 0\text{. } \label{pod22_1}\ ] ] * asymptotic oscillatory low - loss modes . * now the low - loss modes from ( [ pod20 ] ) with @xmath336 will play a key role in the study of the selective overdamping phenomenon in section [ spad ] . we call such modes the asymptotic oscillatory low - loss modes since if @xmath336 then the limiting function @xmath338 ( for fix @xmath339 ) is an oscillatory function in @xmath339 with period @xmath340 . thus as @xmath272 we denote this by @xmath341 from the quality factor discussion above for the low - loss modes the next proposition follows immediately , regardless of whether @xmath335 or not . [ q - factor : asymptotic oscillatory low - loss modes][pqaollm ] any low - loss eigenmodes from ( [ pod20 ] ) with @xmath336 has a quality factor @xmath327 $ ] with the property that either @xmath327 = + \infty$ ] for all @xmath68 or @xmath327 \nearrow+\infty$ ] as @xmath272 ( i.e. , is an unbounded increasing function of the loss parameter @xmath33 ) . the phenomenon of _ overdamping _ ( also called _ heavy damping _ ) is best known for a simple damped oscillator . namely , when the damping exceeds certain critical value all oscillations cease entirely , see , for instance , ( * ? ? 2 ) . in other words , if the damped oscillations are described by the exponential function @xmath342 with a complex constant @xmath79 then in the case of overdamping ( heavy damping ) @xmath343 . our interest to overdamping is motivated by the fact that if an eigenmode becomes overdamped it can not resonate at any finite frequency . consequently , the contribution of such a mode to losses at finite frequencies becomes minimal , and that provides a mechanism for the absorption suppression for systems composed of lossy and lossless components . the treatment of overdamping for systems with many degrees of freedom involves a number of subtleties particularly in our case when the both lossy and lossless degrees of freedom are present , i.e. , when the loss fraction condition ( [ radis4b ] ) is satisfied . we will show that any lagrangian system ( [ radis2a ] ) with @xmath169 can be overdamped with selective overdamping occurring whenever @xmath34 does not have full rank , i.e. , @xmath344 , which corresponds to a model of a two - component composite with a high - loss and a low - loss component . let us be clear that we study overdamping in this paper only for the case @xmath169 as the case @xmath217 will differ significantly in the analysis and it is not expected that there exists such a critical value for damping when all oscillations cease entirely for any damped oscillation . a study of a proper generalization of overdamping in the case @xmath217 will be carried out in a future publication . we make here statements and provide arguments for overdamping for lagrangian systems . we will study the phenomenon known as overdamping in this section under the following condition : [ no gyrotropy][cndod]for our study of overdamping we assume henceforth that@xmath345 the following proposition is a key result that will be referenced often in our study of overdamping . it describes the spectrum of the @xmath24 matrices @xmath99 , @xmath98 in terms of the spectrum of the @xmath10 matrices @xmath346 , @xmath347 . [ spectra relations][pspre]for the @xmath24 matrices @xmath99 , @xmath98 and the @xmath348 matrices @xmath89 , @xmath90 , @xmath34 we have @xmath349 ( @xmath350 ) and @xmath351 for every @xmath202 . in particular , if @xmath352 and @xmath229 denote the smallest nonzero eigenvalue and the largest eigenvalue of @xmath99 and @xmath98 , respectively , and @xmath353 denotes the smallest positive eigenvalue of @xmath98 then @xmath354 and@xmath355 to study whether overdamping is possible we need to determine conditions under which@xmath356 has a solution for the system operator @xmath357 . but we know from corollary [ cpfsp ] that @xmath358 if and only if @xmath359 hence the solutions to ( [ lod1 ] ) are the solution of@xmath360 thus to study overdamping it suffices to study conditions under which the quadratic eigenvalue problem@xmath361 has a solution . to be explicit we have@xmath362 as it turns out the key to studying overdamping will be to study zero sets of the family of quadratic forms@xmath363 on certain subspaces of the hilbert space @xmath194 . in particular , we find that@xmath364 ^{2}-\frac{\left ( q,\eta q\right ) } { \left ( q,\alpha q\right ) } } .\nonumber\end{aligned}\ ] ] thus we find that a necessary and sufficient condition for an eigenvalue @xmath79 of the system operator @xmath102 to satisfy @xmath343 is@xmath365 now there are some fundamental inequalities that play a key role in our overdamping analysis which are described in the following proposition . we will prove part of this proposition , the rest is proved in section [ sproofs ] , since the proof is enlightening showing how the energy conservation law ( [ radis4 ] ) and the virial theorem [ tvir ] can be used to derive these inequalities . [ fundamental inequalities][pfuineq]the following statements are true . 1 . for any @xmath366 with @xmath209,@xmath367 moreover , if @xmath368 exists then@xmath369 2 . if the matrix @xmath34 has full rank , i.e. , @xmath370 , then for any @xmath366 with @xmath209 , @xmath371 3 . for any @xmath372 @xmath373with @xmath75,@xmath374 4 . if @xmath375 [ or equivalently , @xmath376 then @xmath343 whenever any one of the following two inequalities is satisfied : ( i ) @xmath377 or ( ii ) @xmath368 exists and @xmath378 . 5 . if @xmath70 is an eigenmode of the lagrangian system ( [ radis2a ] ) then energy equipartition , i.e. , @xmath379 , can only hold if @xmath380 or , in the case @xmath368 exists , if @xmath381 . the first two statements are proved in section [ sproofs ] . the third statement will now be proved using the energy conservation law ( [ radis4 ] ) and the virial theorem [ tvir ] , or more specifically , corollary [ tvir_eqp ] on the equipartition of energy . @xmath373with @xmath382 . then @xmath70 is an eigenmode of the lagrangian system ( [ radis2a ] ) . it follows from ( [ radis4 ] ) and corollary [ tvir_eqp ] that for the system energy @xmath383 we have@xmath384 where @xmath385 and @xmath386 are the kinetic and potential energy , respectively , of this state @xmath77 and @xmath387 is the dissipated power . notice that for @xmath55 we have @xmath388 . it now follows from this and ( [ lod7_0c ] ) , ( [ lod7_0d ] ) that@xmath389 the proof of the third statement now follows from these two equalities and the inequality ( [ lod7_0 ] ) . the fourth statement follows immediately from the inequality ( [ lod7_0b ] ) and the inequality ( [ lod7_00 ] ) . the fifth statement follows from the fact that if @xmath390 then the equality ( [ lod7_0f ] ) holds , regardless of whether @xmath75 or not as long as @xmath200 , and so the result follows immediately from the inequalities ( [ lod7_0 ] ) and ( [ lod7_00 ] ) if @xmath200 . and since the existence of @xmath368 prohibits @xmath135 , the proof now follows . this completes the proof . from the inequalities ( [ lod7_0a ] ) , ( [ lod7_0b ] ) one can see clearly that if the matrix @xmath34 has full rank then all the eigenvalues of @xmath270 are purely imaginary once @xmath391 in other words , ( [ lod7_3 ] ) is a sufficient condition for all of the eigenmodes of the canonical system ( [ radis8 ] ) [ or , equivalently , the lagrangian system ( [ radis2a ] ) ] to be overdamped , provided @xmath34 has full rank . but it is also clear that this argument must be refined if @xmath34 does not have full rank , i.e. , @xmath344 , since then @xmath392 and hence @xmath393 . the refinement we use is perturbation theory . in particular , we use our results in sections [ sspsy ] , [ smdhlr ] , [ svir ] on the modal dichotomy and the virial theorem to derive our main results on overdamping . our argument is essentially based on determining which of the eigenmodes can not maintain the energy equipartition in corollary [ tvir_eqp ] , i.e. , the equality between the kinetic and potential energy , when @xmath33 is sufficiently large by either using the inequalities in corollary [ cmdic ] or using the asymptotic expansions in section [ smdhlr ] for the eigenmodes in ( [ pod20 ] ) . this is where the fifth statement in proposition [ pfuineq ] is relevant . we derive the following results on overdamping using the inequalities in proposition [ pfuineq ] . [ complete overdamping][tcovd]if @xmath394 and @xmath395 has full rank , i.e. , @xmath370 , then all the eigenmodes of the canonical system ( [ radis8 ] ) are overdamped , i.e. , for the system operator @xmath187 we have@xmath396 [ overdamped regime is optimal][eodop]the following example shows for the class of system operators @xmath397 under consideration in this paper , the regime @xmath251 is optimal for guaranteeing overdamping in the sense that we can always find an example of a system operator satisfying @xmath398 with @xmath33 as close to @xmath399 as we like such that the set @xmath400 is empty . in particular , it is proven using the following family of system operators@xmath163{cc}0 & -\mathrm{i}\\ \mathrm{i } & 0 \end{array } \right ] , \text { \ \ } b=\left [ \begin{array } [ c]{cc}1 & 0\\ 0 & 0 \end{array } \right ] , \label{eodop1}\\ a\left ( \beta\right ) = \omega-\mathrm{i}\beta b,\text { \ \ } \beta \geq0,\nonumber\end{gathered}\ ] ] derived from the system with one degree of freedom having lagrangian and rayleigh dissipation function @xmath401{l}\dot{q}\\ q \end{array } \right ] ^{\mathrm{t}}m_{\mathrm{l}}\left [ \begin{array } [ c]{l}\dot{q}\\ q \end{array } \right ] , \quad m_{\mathrm{l}}=\left [ \begin{array } [ c]{ll}1 & 0\\ 0 & -1 \end{array } \right ] , \label{eodop2}\\ r=\frac{1}{2}\left ( \dot{q},\beta r\dot{q}\right ) , \text { \ \ } r=1\text{,}\nonumber\end{gathered}\ ] ] which is just is the lagrangian for a one degree - of - freedom spring - mass - damper system with mass @xmath402 , spring constant @xmath402 , and viscous damping coefficient @xmath33 . then we have@xmath403 and therefore@xmath404 in this case , at the boundary of the overdamping regime , where @xmath405 , we have critical damping of the system . now we want to answer the question of which modes of the canonical system ( [ radis8 ] ) are overdamped in the case when @xmath34 does not have full rank , i.e. , @xmath344 . the next theorem gives us a partial answer to the question . [ partial overdamping][tpovd]if @xmath251 and @xmath344 then the modal dichotomy occurs as in theorem [ tmdic ] and corollary [ cmdic ] . in addition , any eigenmode of the canonical system ( [ radis8 ] ) in the invariant subspace @xmath406 ( the high - loss susceptible subspace ) is overdamped , i.e. , @xmath407 now we ask the important and natural question : can any of the modes of the system operator @xmath102 in the low - loss susceptible subspace @xmath273 be overdamped when @xmath34 from the rayleigh dissipative function @xmath31 in ( [ radis1 ] ) does not have full rank , i.e. , in the case when the lagrangian system is a model of a two - component composite with a lossy and a lossless component ? this question will be investigated in the next section . we are interested in solving the problem of describing all the modes of the canonical system ( [ radis8 ] ) [ or , equivalently , the lagrangian system ( [ radis2a ] ) ] , with system operator @xmath245 @xmath117 , that are overdamped when losses are sufficiently large , i.e. , @xmath68 . in the previous section we solved the problem for lagrangian systems with complete rayleigh dissipative , i.e. , @xmath34 has full rank , for then theorem [ tcovd ] tells us that all the modes will become overdamped . on the other hand , for lagrangian systems with incomplete rayleigh dissipative , i.e. , @xmath34 does not have full rank , then theorem [ tpovd ] is a partial solution to the problem since it tells us that , due to sufficiently large losses , modal dichotomy occurs splitting the modes into the two @xmath102-invariant subspaces @xmath273 and @xmath406 ( the high - loss and low - loss susceptible subspaces , respectively ) with the modes belonging to @xmath254 overdamped . thus to solve the problem completely we need to describe which of the modes in the low - loss susceptible subspace @xmath253 can be overdamped when losses are sufficiently large . in this section we solve the problem asymptotically as losses become extremely large , i.e. , as @xmath272 , using the virial theorem [ tvir ] and , in particular , energy equipartition [ tvir_eqp ] for the eigenmodes and the perturbation theory developed in @xcite which we described in section [ smdhlr ] . our description of the selective overdamping phenomenon is based on the modal dichotomy in the high - loss regime @xmath68 as described in section [ smdhlr ] . in particular , we use theorem [ cpodr ] to characterize overdamping in terms of the high - loss and low - loss eigenmodes @xmath408 , @xmath319 of the canonical system ( [ radis8 ] ) which split into the two distinct classes based on their dissipative properties ( [ pod20 ] ) . we say that such an eigenmode @xmath409 is _ overdamped _ for @xmath68 if @xmath410 for all @xmath33 sufficiently large . we say that such an eigenmode _ remains oscillatory _ for @xmath68 if @xmath411 for all @xmath33 sufficiently large . our first result tells us that all of the high - loss eigenmodes @xmath408 , @xmath305 from ( [ pod20 ] ) are overdamped regardless of whether @xmath34 has full rank or not . it is an immediate corollary of theorem [ tpovd ] and theorem [ cpodr ] . [ overdamped high - loss modes][thmod]all of the high - loss eigenmodes from ( [ pod20 ] ) are overdamped for @xmath68 . now we will discuss and give a solution to the problem of determining which of the low - loss eigenmodes @xmath412 , @xmath308 in ( [ pod20 ] ) will be overdamped . in order to do this we work under the following nondegeneracy condition : [ nondegeneracy][cndnd]for our study of the overdamping of the low - loss modes in this section , we assume henceforth that @xmath413 we have two main goals in this section . our first is to find necessary and sufficient conditions to have @xmath414 for @xmath68 . let @xmath415 be the number of low - loss eigenmodes that are overdamped for @xmath68 . our second goal is to calculate this number @xmath415 . we begin by recalling some of the important properties of the low - loss eigenpairs @xmath416 , @xmath417 , @xmath308 in ( [ pod11 ] ) for the system operator @xmath418 that we will need in this section . by ( [ pod14 ] ) and ( [ pod16 ] ) the limits@xmath419 exist and from ( [ pod8 ] ) , ( [ pod15 ] ) the limiting vectors @xmath420 are an orthonormal basis for @xmath421 which also diagonalize the self - adjoint operator @xmath294 from ( [ pod7 ] ) satisfying@xmath422 * overview of main results . * before we proceed , let us now summarize the results in this section . first , its obvious from this discussion that a necessary condition for the low - loss eigenmode @xmath409 to be overdamped for @xmath68 is @xmath423 ( even though it does so with a slow damping of order @xmath329 , as seen in ( [ llod1 ] ) and ( [ pod14 ] ) ) , but what is not obvious is that this is in fact a sufficient condition . although in the case @xmath368 exists ( i.e. , @xmath424 ) , it just follows from the fourth statement in proposition [ pfuineq ] since if @xmath423 then by ( [ llod1 ] ) we have @xmath425 which implies @xmath426 for @xmath427 and hence @xmath428 for @xmath68 . but this argument fails when @xmath429 . thus we must find an alternative method to show that @xmath423 is equivalent to @xmath410 for @xmath68 . the key idea that leads to an alternative method is to realize that by the virial theorem [ tvir ] and the fifth statement in proposition [ pfuineq ] , when @xmath368 exists the modes with @xmath423 are exactly the modes that break the energy equipartition . hence our general method is to show , even in the case @xmath424 , that for a low - loss modes with @xmath423 , there is a breakdown of the virial theorem [ tvir ] since as losses increase , i.e. , as @xmath272 , the equality between the kinetic and potential energy ( see corollary [ tvir_eqp ] ) of the corresponding eigenmode of the lagrangian system ( [ radis2a ] ) can no longer be maintained and must eventually break which forces the modes to be non - oscillatory for @xmath68 , i.e. , overdamped . this proves that @xmath423 is equivalent to @xmath428 for @xmath68 . thus it follows that the low - loss eigenmodes in ( [ pod20 ] ) of the canonical system ( [ radis8 ] ) which are overdamped for @xmath68 , their limiting vectors ( satisfy @xmath430 @xmath431 for @xmath339 fixed ) form a orthonormal basis for @xmath431 . this implies of course that @xmath432 is the number of low - loss eigenmodes that are overdamped for @xmath68 . a major result of this section is that in fact @xmath433 , where @xmath434 . but @xmath435 is also the number of high - loss eigenpairs in ( [ pod11 ] ) . thus we conclude using these results and theorem [ thmod ] that the number of eigenmodes in ( [ pod20 ] ) which are overdamped in the high - loss regime @xmath68 is exactly @xmath436 with @xmath260 of these being low - loss eigenmodes and the rest being all the high - loss eigenmodes . moreover , in the case @xmath344 , the remaining @xmath437 modes are low - loss oscillatory modes with an extremely high quality factor that actually increases as the losses increase , i.e. , as @xmath272 . * main results . * we now give the statements of our main results . the proof of the first theorem we give in this section as it is enlightening . the proofs of the rest of the statements below are found in section [ sproofs ] . [ overdamped low - loss modes][tlmod]a necessary and sufficient condition for a low - loss eigenmode @xmath438 from ( [ pod20 ] ) to be overdamped for @xmath68 is @xmath423 ( or equivalently , @xmath439 ) . let @xmath440 be a low - loss eigenmode from ( [ pod20 ] ) . if it is overdamped for @xmath68 then @xmath410 for @xmath68 ( i.e. , for all @xmath33 sufficiently large ) and so by ( [ llod1 ] ) , ( [ llod2 ] ) we have @xmath423 and @xmath441 ( the nullspace of @xmath294 ) . lets now prove the converse . suppose @xmath423 . then we must show that @xmath428 for @xmath68 . suppose that this was not true . then since @xmath416 was analytic at @xmath302 we must have @xmath442 for @xmath68 . thus by corollary [ cpfsp ] there is a corresponding eigenmode @xmath443 of the lagrangian system ( [ radis2a ] ) such that with respect to the block representation ( [ pfsp8b ] ) the vector @xmath417 is @xmath444{c}-\mathrm{i}\zeta_{j}\left ( \beta\right ) \sqrt{\alpha}q_{j}\left ( \beta\right ) \\ \sqrt{\eta}q_{j}\left ( \beta\right ) \end{array } \right ] . \label{llod3}\ ] ] in particular , @xmath445 and @xmath446 are related by ( [ radis7 ] ) , namely,@xmath447 by section [ seneq ] on the energetic equivalence between the canonical system ( [ radis8 ] ) and the lagrangian system ( [ radis2a ] ) it follows that@xmath448 = \mathcal{t}\left ( \dot{q},q\right ) + \mathcal{v}\left ( \dot{q},q\right ) . \label{llod5}\ ] ] the term on the left is the system energy of the eigenmode @xmath445 of the canonical system defined in ( [ radis11a ] ) whereas the sum on the right is the system energy of the eigenmode @xmath446 of the lagrangian system ( [ radis2a ] ) , i.e. , the sum of its kinetic and potential energy . as proved in section [ pspad ] , for any fixed time @xmath339 we have @xmath449 = \lim_{\beta\rightarrow\infty}\mathcal{v}\left ( \dot{q}_{j}\left ( t\,,\beta\right ) , q_{j}\left ( t\,,\beta\right ) \right ) , \label{llod6}\\ \lim_{\beta\rightarrow\infty}\mathcal{t}\left ( \dot{q}_{j}\left ( t\,,\beta\right ) , \dot{q}_{j}\left ( t\,,\beta\right ) \right ) = 0.\nonumber\end{gathered}\ ] ] this proves that @xmath450 and , consequently , by the energy equipartition ( see corollary [ tvir_eqp ] ) that@xmath451 this completes the proof . [ clmod]the collection of limiting values as @xmath272 ( with @xmath339 fixed ) of the low - loss eigenmodes in ( [ pod20 ] ) of the canonical system ( [ radis8 ] ) which are overdamped for @xmath68 is @xmath452 @xmath453 and is an orthonormal basis for @xmath431 . [ plmod]let @xmath415 denote the number low - loss eigenmodes in ( [ pod20 ] ) of the canonical system ( [ radis8 ] ) which are overdamped for @xmath68 . then @xmath454 where @xmath455 . [ number of overdamped modes][tnlm]the total number of eigenmodes of the canonical system ( [ radis8 ] ) from ( [ pod20 ] ) which are overdamped for @xmath68 is @xmath436 , where @xmath455 . furthermore , @xmath260 of these are low - loss eigenmodes and the rest are all of the high - loss eigenmodes . moreover , if @xmath344 then the @xmath456 modes which are not overdamped as @xmath68 are low - loss eigenmodes which remain oscillatory for @xmath68 . [ selective overdamping][tseod]if @xmath344 then there is exactly @xmath456 low - loss eigenmodes of the canonical system ( [ radis8 ] ) from ( [ pod20 ] ) which remain oscillatory for @xmath68 ( i.e. , @xmath411 for @xmath68 ) and their quality factors as given by ( [ radis12 ] ) , ( [ radis12_1 ] ) have the following property @xmath457 = + \infty\text { for } \beta\gg1\text { or } q\left [ w_{j}\left ( \beta\right ) \right ] \nearrow+\infty\text { as } \beta\rightarrow\infty . \label{tseod1}\ ] ] this section contains the proofs of the results from sections [ svir ] , [ sspan ] and [ sodan ] . proofs are organized in the order in which statements appeared and grouped according to the sections and subsections in which they are located in the paper . all assumptions , notation , and conventions used here will adhere to that previously introduced in this paper . the proofs of statements in section [ svir ] are contained here . [ proof of theorem [ tvir]]for @xmath70 an eigenmode of the lagrangian system ( [ radis2a ] ) , we consider the virial@xmath458 in our proof we will use the real quantities @xmath459 and @xmath460 . if @xmath135 we are done . thus , suppose that @xmath200 . the virial ( [ vir1 ] ) satisfies@xmath461 taking the time - derivative of ( [ vir9 ] ) and using ( [ radis2a ] ) yields the identity@xmath462 where @xmath463 . the energy balance equation ( [ radis4 ] ) implies@xmath464 combining the identities ( [ vir10 ] ) and ( [ vir11 ] ) yields immediately the identity@xmath465 from ( [ vir12 ] ) and since @xmath466 we derive the identity@xmath467 if @xmath75 or @xmath468 if @xmath343 . suppose now that @xmath75 . then , by taking the imaginary and real parts of the identity ( [ vir14 ] ) , we find that@xmath469 the rest of the proof of theorem [ tvir ] now immediately follows from ( [ vir16 ] ) and ( [ radis4_0 ] ) . [ proof of corollary [ tvir_eqp]]corollary [ tvir_eqp ] follows immediately from the virial theorem [ tvir ] . the proofs of statements in section [ sspan ] are contained here . [ proof of proposition [ ppfsp]]the factorization ( [ pfsp8 ] ) follows from formula ( [ mab2 ] ) in appendix [ apxsc ] and ( [ radis6a1 ] ) , ( [ radis6b ] ) , ( [ radis8])([radis8b ] ) , ( [ pfsp2 ] ) since@xmath470{cc}\zeta\mathbf{1}-\omega_{\mathrm{p}}+\mathrm{i}\mathsf{\tilde{r } } & \mathrm{i}\phi^{\mathrm{t}}\\ -\mathrm{i}\phi & \zeta\mathbf{1}\end{array } \right ] \\ = \left [ \begin{array } [ c]{cc}\mathbf{1 } & \zeta^{-1}\mathrm{i}\phi^{\mathrm{t}}\\ 0 & \mathbf{1}\end{array } \right ] \left [ \begin{array } [ c]{cc}\zeta\mathbf{1}-\omega_{\mathrm{p}}+\mathrm{i}\mathsf{\tilde{r}}-\zeta ^{-1}\mathrm{i}\phi^{\mathrm{t}}\left ( -\mathrm{i}\phi\right ) & 0\\ 0 & \zeta\mathbf{1}\end{array } \right ] \left [ \begin{array } [ c]{cc}\mathbf{1 } & 0\\ -\zeta^{-1}\mathrm{i}\phi & \mathbf{1}\end{array } \right ] \\ = \left [ \begin{array } [ c]{cc}k_{\mathrm{p } } & \zeta^{-1}\mathrm{i}\phi^{\mathrm{t}}\\ 0 & \mathbf{1}\end{array } \right ] \left [ \begin{array } [ c]{cc}\zeta^{-1}\mathbf{1 } & 0\\ 0 & \zeta\mathbf{1}\end{array } \right ] \left [ \begin{array } [ c]{cc}c(\zeta,\beta ) & 0\\ 0 & \mathbf{1}\end{array } \right ] \left [ \begin{array } [ c]{cc}k_{\mathrm{p}}^{\mathrm{t } } & 0\\ -\zeta^{-1}\mathrm{i}\phi & \mathbf{1}\end{array } \right ] .\end{gathered}\ ] ] this completes the proof . [ proof of corollary [ cpfsp]]the proof of this corollary follows immediately from the factorization ( [ pfsp8 ] ) using the facts @xmath471 , @xmath472 and@xmath473{cc}k_{\mathrm{p}}^{\mathrm{t } } & 0\\ -\zeta^{-1}\mathrm{i}\phi & \mathbf{1}\end{array } \right ] ^{-1}=\left [ \begin{array } [ c]{cc}\sqrt{\alpha } & 0\\ \zeta^{-1}\mathrm{i}\phi\sqrt{\alpha } & \mathbf{1}\end{array } \right ] .\ ] ] [ proof of proposition [ pssym]]the proof of this proposition follows immediately from elementary properties of the determinant and using the fundamental property of the system operator @xmath275 in ( [ sasp1 ] ) , namely , @xmath474 which implies @xmath475 . [ proof of proposition [ pevbd]]the second statement follows immediately from ( [ sasp1 ] ) since if @xmath206 and @xmath207 then@xmath476 from this and using ( [ sasp1])([sasp3 ] ) , the first and third statements follow immediately from proposition [ apxpebd ] in appendix [ apxebd ] . this completes the proof . [ proof of corollary [ cssym]]the proof of this corollary follows immediately from propositions [ pssym ] and [ pevbd ] . [ proof of theorem [ tmdic]]suppose that @xmath251 . it follows from theorem [ apxtmd ] in appendix [ apxebd ] that the @xmath24 matrix @xmath102 satisfies the hypotheses of the theorem and together with ( [ sasp1 ] ) , ( [ ebmd1 ] ) implies @xmath477 and the sets @xmath478 , @xmath479 in ( [ mdss ] ) satisfy@xmath480 moreover , theorem [ apxtmd ] gives the existence of unique projection matrices @xmath481 , @xmath482 such that in the space @xmath111 the subspaces @xmath483 and @xmath484 have exactly the desired properties in theorem [ tmdic ] with their dimensions satisfying@xmath485 this completes the proof . [ proof of corollary [ cmdic]]first , if @xmath237 is an eigenvector of @xmath270 with corresponding eigenvalue @xmath79 then proposition [ pevbd ] and ( [ sasp3 ] ) imply@xmath486 assume now that @xmath251 . then by ( [ sasp1])([sasp3 ] ) , ( [ pcmdic1 ] ) , and theorem [ tmdic ] the proof of the first part of corollary [ cmdic ] follows immediately from corollary [ apxcmd ] in appendix [ apxebd ] . the rest of the proof now follows immediately from ( [ pcmdic1 ] ) , formula ( [ radis12_1 ] ) , and the first part of corollary [ cmdic ] . this completes the proof . [ proof of theorem [ cpodr]]consider the high - loss and low - loss eigenpairs @xmath416 , @xmath417 , @xmath487 for @xmath68 introduced in section [ smdhlr ] . the have the properties that @xmath488 are a basis of eigenvectors for the system operator @xmath102 satisfying @xmath489 the proof of theorem [ cpodr ] follows from these facts , theorem [ tmdic ] , and corollary [ cmdic ] . the proofs of statements in section [ sodan ] are contained here . recall , that we are assuming from now on that condition [ cndod ] is true i.e. , @xmath490 [ proof of proposition [ pspre]]using the block representation of the matrices @xmath98 , @xmath99 in ( [ radis8a ] ) and the result ( [ mab5 ] ) in appendix [ apxsc ] on the determinant of block matrices with commuting block entries we find that for any @xmath202 we have@xmath491{cc}\zeta\mathbf{1 } & \mathrm{i}\phi^{\mathrm{t}}\\ -\mathrm{i}\phi & \zeta\mathbf{1}\end{array } \right ] \right ) = \det\left ( \zeta^{2}\mathbf{1-}\phi^{\mathrm{t}}\phi\right ) = \det\left ( \zeta^{2}\mathbf{1}-\alpha^{-1}\eta\right ) , \\ \det\left ( \zeta\mathbf{1}-b\right ) & = \det\left ( \left [ \begin{array } [ c]{cc}\zeta\mathbf{1-}\mathsf{\tilde{r } } & 0\\ 0 & \zeta\mathbf{1}\end{array } \right ] \right ) = \zeta^{n}\det\left ( \zeta\mathbf{1-}\mathsf{\tilde{r}}\right ) = \zeta^{n}\det\left ( \zeta\mathbf{1-}\alpha^{-1}r_{\mathrm{h}}\right ) , \end{aligned}\ ] ] since by ( [ radis6a1 ] ) , ( [ radis6b ] ) , ( [ radis8b ] ) we have @xmath492 and @xmath493 . the proof now follows from these facts . [ proof of proposition [ pfuineq]]the last three statements have already been proved . the first two statements are proved using proposition [ pspre ] and the following series of inequalities which just use elementary properties of the operator norm @xmath231 for hermitian matrices @xmath494 where @xmath495 or @xmath34 and the inequality ( [ lod7_1a ] ) requires the hypothesis that @xmath90 or @xmath34 , respectively , be invertible so that @xmath496 . the completes the proof . [ proof of theorem [ tcovd]]suppose @xmath394and @xmath497 , where @xmath216 . if @xmath79 was an eigenvalue of @xmath102 with @xmath75 then by proposition [ pfuineq ] we have @xmath498 a contradiction . therefore , all the eigenvalues of the system operator @xmath102 are real . this completes the proof . [ proof of theorem [ tpovd]]suppose @xmath394and @xmath499 . then by theorem [ tmdic ] and corollary [ cmdic ] , if @xmath500 then @xmath501 and so by proposition [ pfuineq ] we must have @xmath343 . therefore , all the eigenvalues of @xmath270 in @xmath502 are real . [ proof of theorem [ thmod]]if @xmath438 is a high - loss eigenmode from ( [ pod20 ] ) then by the modal dichotomy ( [ cpodr ] ) we must have latexmath:[$\zeta_{j}\left ( \beta\right ) \in\sigma\left ( a\left ( \beta\right ) @xmath68 ( i.e. , for @xmath33 sufficiently large ) . by theorem [ tpovd ] it follows that @xmath428 for @xmath68 . therefore , @xmath409 is overdamped for @xmath68 . this completes the proof . we will now assume in the rest of the proofs that condition [ cndnd ] is true i.e. , @xmath504 . [ proof of theorem [ tlmod]]the proof of this theorem was started in section [ spad2 ] . it remains to show that ( [ llod6 ] ) holds . first , we have that @xmath505 with @xmath506 and @xmath507 by the properties of the low - loss eigenmodes in ( [ pod20 ] ) and since ( [ llod1 ] ) , ( [ llod2 ] ) is true for these modes together with the hypothesis @xmath423 . now we use the representation ( [ llod3 ] ) and ( [ plmodp6 ] ) for @xmath417 and @xmath431 , respectively , to imply that as @xmath272 @xmath444{c}-\mathrm{i}\zeta_{j}\left ( \beta\right ) \sqrt{\alpha}q_{j}\left ( \beta\right ) \\ \sqrt{\eta}q_{j}\left ( \beta\right ) \end{array } \right ] \rightarrow\mathring{w}_{j}=\begin{bmatrix } 0\\ \mathring{\psi}\end{bmatrix } , \text { } -\mathrm{i}\zeta_{j}\left ( \beta\right ) \sqrt{\alpha}q_{j}\left ( \beta\right ) \rightarrow0,\text { } \sqrt{\eta}q_{j}\left ( \beta\right ) \rightarrow\mathring{\psi } \label{tlmod1}\ ] ] for some @xmath508 . it follows from ( [ tlmod1 ] ) and the fact @xmath509 that for any fixed time @xmath339 we have as @xmath272@xmath510 & = \frac{1}{2}\left ( w_{j}\left ( \beta\right ) , w_{j}\left ( \beta\right ) \right ) e^{2\operatorname{im}\zeta_{j}\left ( \beta\right ) t}\rightarrow\frac{1}{2}\left ( \mathring{w}_{j},\mathring{w}_{j}\right ) = \frac{1}{2}.\end{aligned}\ ] ] this proves ( [ llod6 ] ) since by ( [ tlmod1 ] ) we have @xmath511 . this completes the proof . [ proof of corollary [ clmod]]this corollary follows immediately from theorem [ tlmod ] and the fact that the @xmath512 are an orthonormal basis for @xmath421 which diagonalize the self - adjoint operator @xmath294 such that ( [ llod2 ] ) is satisfied . [ proof of theorem [ tnlm]]this theorem follows immediately from theorems [ thmod ] and [ tlmod ] , corollary [ clmod ] , and proposition [ pqaollm ] ( which is proved below ) . [ proof of theorem [ tseod]]this theorem follows immediately from theorems [ tnlm ] , [ tlmod ] and proposition [ pqaollm ] . [ proof of proposition [ plmod]]let @xmath415 denote the number of low - loss eigenmodes of the canonical system ( [ radis8 ] ) in ( [ pod20 ] ) which are overdamped for @xmath68 . then by corollary [ clmod ] we know that@xmath513 we will now prove that @xmath514 where by ( [ radis10a ] ) , @xmath455 . we will use the elementary facts from linear algebra that since @xmath90 , @xmath515 are @xmath10 matrices which are positive semidefinite then @xmath516 another elementary fact from linear algebra which we need is that if @xmath517 and @xmath518 are subspaces in the same hilbert space then@xmath519 where @xmath520 . thus with @xmath521 , @xmath522 we have@xmath523 where the sum is direct by condition [ cndnd ] , i.e. , the hypothesis @xmath334 . in particular , this implies @xmath524 the proof of ( [ plmodp0 ] ) will be carried out in a series of steps which we outline now before we begin . first , we will prove that @xmath525{c}0\\ \psi \end{array } \right ] : \sqrt{\eta}\psi\in\operatorname{ran}\eta\cap\operatorname{ran}r\right\ } , \label{plmodp6}\ ] ] where the block form is with respect to the decomposition @xmath193 , @xmath111 , @xmath526 described in section [ stvspen ] , i.e. , the form ( [ pfsp8b ] ) . next , it follows immediately from ( [ plmodp1 ] ) that @xmath527 finally , we prove that @xmath528 the proof of ( [ plmodp0 ] ) follows immediately from ( [ plmodp6])([plmodp8 ] ) . we begin by computing the representation ( [ plmodp6 ] ) for @xmath431 . first , by its definition from ( [ pod7 ] ) , @xmath294 is the restriction of the operator @xmath98 to the no - loss subspace @xmath280 , that is,@xmath529 where @xmath282 is the orthogonal projection onto @xmath288 in the hilbert space @xmath111 with standard inner product @xmath112 . denote by @xmath530 the orthogonal projection onto @xmath531 in the hilbert space @xmath194 with the standard inner product . then it follows from the block representation of @xmath98 and @xmath99 in ( [ pod7 ] ) with respect to the decomposition @xmath193 that @xmath282 and @xmath532 with respect to this decomposition are the block operators @xmath533 from ( [ plmodp9 ] ) , ( [ plmodp10 ] ) it follows that@xmath525{c}\varphi\\ \psi \end{array } \right ] \in\operatorname{ker}\mathsf{\tilde{r}}\oplus h_{\mathrm{q}}:-\mathrm{i}p_{\mathsf{\tilde{r}}}^{\bot}\phi^{\mathrm{t}}\psi=0\text { and } \mathrm{i}\phi p_{\mathsf{\tilde{r}}}^{\bot}\varphi=0\right\ } .\ ] ] but if @xmath534 with @xmath535 then @xmath536 by definition of @xmath530 and hence @xmath537 and @xmath538 which implies @xmath539 and @xmath540 since by ( [ radis8b ] ) and ( [ radis6a1 ] ) we have @xmath541 , @xmath542 , @xmath543 , and @xmath544 . by the hypothesis of condition [ cndnd ] this implies that @xmath545 . also , since @xmath546 is a hermitian matrix it follows that @xmath547 is the orthogonal projection onto the @xmath548 so that @xmath549 is equivalent to @xmath550 and this is equivalent to @xmath551 @xmath67 by properties ( [ radis6b ] ) . but @xmath551 @xmath67 is equivalent to @xmath551 @xmath552 which by ( [ plmodp1 ] ) is equivalent to @xmath551 @xmath553 . this proves ( [ plmodp6 ] ) . finally , we will now prove ( [ plmodp8 ] ) . it follows from the fact that @xmath90 is a hermitian matrix and ( [ plmodp1 ] ) that the restriction of the operator@xmath88 to @xmath554 , i.e. , @xmath555 , is an invertible operator and that @xmath556 from this it follows that@xmath557 the proof of ( [ plmodp8 ] ) immediately follows from this fact and ( [ plmodp5 ] ) . this completes the proof of the theorem . let @xmath558{cc}p & q\\ r & s \end{array } \right ] \label{mab1}\ ] ] be a square matrix represented in block form where @xmath559 and @xmath560 are square matrices with the latter invertible , that is , @xmath561 . then the following _ aitken block - diagonalization formula _ holds @xcite , @xcite,@xmath558{cc}p & q\\ r & s \end{array } \right ] = \left [ \begin{array } [ c]{cc}\mathbf{1 } & qs^{-1}\\ 0 & \mathbf{1}\end{array } \right ] \left [ \begin{array } [ c]{cc}p - qs^{-1}r & 0\\ 0 & s \end{array } \right ] \left [ \begin{array } [ c]{cc}\mathbf{1 } & 0\\ s^{-1}r & \mathbf{1}\end{array } \right ] , \label{mab2}\ ] ] where the matrix@xmath562 is known as the _ schur complement _ of @xmath560 in @xmath106 . the aitken formula ( [ mab2 ] ) readily implies@xmath563 from this we can conclude that@xmath564 but the latter statement is true even if @xmath565 which is proved using a limiting argument . indeed , by substituting for @xmath560 in ( [ mab1 ] ) the small perturbation @xmath566 which is invertible for @xmath567 and commutes with any matrix that @xmath560 commutes with then statement [ mab5 ] is true for @xmath568 and the limit as @xmath569 yields the desired result . in this appendix we discuss results pertaining to bounds on the eigenvalues of non - hermitian matrices in terms of their real and imaginary parts . we also discuss a phenomenon , which we call _ modal dichotomy _ , that occurs for any non - hermitian matrix with a non - invertible imaginary part which is large in comparison to the real part of the matrix . the bounds and dichotomy described here are extremely useful tools in studying the spectral properties of dissipative systems especially for composite systems with high - loss and lossless components . recall , any square matrix @xmath106 can be written as the sum of a hermitian matrix and a skew - hermitian matrix,@xmath570 the real and imaginary parts of @xmath106 , respectively , and @xmath571 is the conjugate transpose of the matrix @xmath106 . for the set of @xmath572 matrices , denote by @xmath231 the operator norm@xmath573 where @xmath574 and @xmath112 denotes the standard euclidean inner product on @xmath575 column vectors with entries in @xmath576 . denote the spectrum of a square matrix @xmath106 by @xmath577 . recall in the case @xmath106 is a hermitian matrix , i.e. , @xmath578 , or a normal matrix , i.e. , @xmath579 , that @xmath580 where @xmath581 and by convention @xmath582 if @xmath583 @xmath577 . the first result which we will prove here is the following proposition which gives bounds on the eigenvalues of a matrix @xmath106 in terms of @xmath584 and @xmath585 . [ eigenvalue bounds][apxpebd]let @xmath106 be a square matrix . then@xmath586 in other words , the eigenvalues of @xmath106 lie in the union of the closed discs whose centers are the eigenvalues of @xmath587 with radius @xmath588 . moreover , the imaginary part of any eigenvalue @xmath79 of @xmath106 satisfies the inequality@xmath589 for some @xmath590 depending on @xmath79 . before we proceed with the proof we will need the following perturbation result from ( * ? ? ? v.4 , p. 291 4.8 ) for non - hermitian matrices which we prove here for completeness . if @xmath591 is a square matrix with @xmath592 normal then@xmath593 let @xmath591 be a square matrix with @xmath592 normal . then@xmath594 let @xmath595 , i.e. , the resolvent set of @xmath592 . then for any @xmath596 we have@xmath597 thus if @xmath598 then the matrix @xmath599 is invertible since the neumann series @xmath600 converges to @xmath601 for any square matrix @xmath602 satisfying @xmath603 . this implies@xmath604 but the inequality@xmath605 and the fact @xmath592 is a normal matrix implies@xmath606 therefore,@xmath607 this completes the proof . [ proof of proposition [ apxpebd]]let @xmath106 be any square matrix . then @xmath608 and @xmath609 is a normal matrix . by the previous lemma this implies@xmath610 now let @xmath79 be any eigenvalue of @xmath106 . then for any corresponding eigenvector @xmath611 of unit norm@xmath612 moreover , we know there exists an eigenvalue @xmath613 of @xmath585 ( depending on @xmath79 ) such that @xmath614 hence @xmath615 the proof now follows from ( [ papxpebd1 ] ) and ( [ papxpebd2 ] ) . the next result we prove is a proposition on dichotomy of the spectrum of a non - hermitian matrix @xmath106 which we mentioned in the introduction of this appendix . in particular , the following proposition tells us that the spectrum of a matrix @xmath106 will split into two disjoint parts when the imaginary part @xmath585 is non - invertible and large `` in comparison to the real part @xmath584 . the term large '' means that the bottom of the nonzero spectrum of @xmath616 must be greater than twice the top of the spectrum of @xmath617 ( where for a square matrix @xmath232 , @xmath618 as defined in ( * ? ? ? * , p. 196 ) ) , that is , @xmath619 this of course implies @xmath620 is large " in comparison to @xmath588 since it follows from this inequality that@xmath621 [ modal dichotomy][apxtmd]let @xmath106 be any @xmath572 matrix which is non - hermitian such that its imaginary part @xmath585 is non - invertible with rank @xmath622 . denote by @xmath623 , @xmath624 , @xmath625 the nonzero eigenvalues of @xmath585 . if @xmath626 then @xmath627 where@xmath628 furthermore , there exists unique matrices @xmath629 , @xmath630 with the properties@xmath631 where @xmath632 if @xmath633 and @xmath634 if @xmath635 ( @xmath636 ) . moreover , these projection matrices have rank satisfying@xmath637 all the eigenvalues of @xmath585 by hypothesis are @xmath623 , @xmath624 , @xmath638 and @xmath639 . it follows from proposition [ apxpebd ] that@xmath640 and the imaginary part of any eigenvalue @xmath79 of @xmath106 satisfies the inequality@xmath641 for some @xmath642 depending on @xmath79 . in particular , it follows that@xmath643 where @xmath644 now suppose that @xmath645 . we will now show that @xmath646 . suppose this were not true then we could find @xmath647 which would imply for some @xmath648 with @xmath649 we have @xmath650 a contradiction . thus @xmath651 . now it follows from the spectral theory of matrices that the finite - dimensional vector space @xmath652 of @xmath575 column vectors with entries in @xmath576 can be written as the direct sum of two invariant subspaces for the matrix @xmath106,@xmath653 that is , @xmath654 is the union of the generalized eigenspaces of @xmath106 corresponding to eigenvalues in @xmath655 and @xmath656 is the union of generalized eigenspaces of @xmath106 corresponding to the eigenvalues in @xmath657 . denote the projection matrix onto @xmath654 along @xmath656 by @xmath629 , that is , the matrix satisfying @xmath658 , @xmath659 , and @xmath660 . then it follows from the spectral theory of matrices that @xmath629 and @xmath661 are the unique matrices with the properties@xmath662 now it follows that @xmath663 so that to complete the proof of this proposition we need only show that@xmath664 in order to prove this we begin by giving an explicit representation of the matrix @xmath629 in terms of the resolvent of @xmath106 which is defined as @xmath665 to do this we define @xmath666 we will now prove that@xmath667 first , we notice that @xmath668 which proves @xmath669 . next , we have that if @xmath670 then for any @xmath623 with @xmath671 , @xmath672 implying that @xmath673 @xmath674 . this proves @xmath675 . it now follows from the cauchy - riesz functional calculus @xcite that@xmath676 where the contour integral is over the simply closed positively oriented path @xmath677 , @xmath678 . now we consider the family of matrices@xmath679 notice that for @xmath680 we have@xmath681 it follows from this that the results proven so far in this theorem for the matrix @xmath106 apply to the matrix @xmath682 for each @xmath683 $ ] . in particular , for each @xmath684 $ ] we have @xmath685 where@xmath686 furthermore , there exists unique matrices @xmath687 , @xmath688 with the properties@xmath689 moreover , @xmath690 it follows from these facts that @xmath687 is a continuous matrix projection - valued function of @xmath339 in @xmath691 $ ] and@xmath692 where @xmath693 denotes the trace of a square matrix . but @xmath694 is a continuous function of @xmath339 since @xmath687 is a continuous matrix - valued function implying @xmath695 is a continuous function of @xmath339 taking values in the nonnegative integers only . from this we conclude @xmath695 is constant for @xmath696 @xmath691 $ ] . in particular , @xmath697 moreover,@xmath698 but this implies that@xmath699 this completes the proof . we conclude this appendix by giving an alternative characterization of the subsets @xmath655 and @xmath674 from theorem [ apxtmd ] in terms of the magnitude of the imaginary parts of the eigenvalues of @xmath106 . in particular , the next corollary tells us that if the @xmath585 is associated with losses and if the imaginary part of the eigenvalues of @xmath106 are associated with damping then only the eigenmodes with eigenvalues in @xmath674 are susceptible to large damping , relative to the norm of @xmath584 , when losses are large . [ apxcmd]if @xmath645 then@xmath700 by theorem [ apxtmd ] we know that@xmath701 where@xmath702 and @xmath623 , @xmath624 , @xmath638 are the nonzero eigenvalues of @xmath585 . in particular , these facts imply immediately that@xmath703 now if @xmath704 then there exists a nonzero eigenvalue @xmath613 of @xmath585 such that@xmath705 which by the hypothesis @xmath645 implies@xmath706 the corollary follows immediately now from these facts . the term of _ energetics _ refers to a fundamental set up for a system evolution when two energies are defined for any system configuration , namely the _ kinetic energy _ and the _ potential energy_. here is a concise description of the energetics given by h. poincar , @xcite : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the difficulties inherent in the classic mechanics have led certain minds to prefer a new system they call _ energetics_. energetics took its rise as an outcome of the discovery of the principle of the conservation of energy . helmholtz gave it its final form . it begins by defining two quantities which play the fundamental role in this theory . they are _ kinetic energy , or vis viva _ , and _ potential energy_. all the changes which bodies in nature can undergo are regulated by two experimental laws : 1@xmath707 . the sum of kinetic energy and potential energy is constant . this is the principle of the conservation of energy . if a system of bodies is at @xmath232 at the time @xmath708 and at @xmath99 at the time @xmath709 , it always goes from the first situation to the second in such a way that the mean value of the difference between the two sorts of energy , in the interval of time which separates the two epochs @xmath708 and @xmath709 , may be as small as possible . this is hamilton s principle , which is one of the forms of the principle of least action . " _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in other words according to energetics the kinetic and potential energies @xmath48 and @xmath49 that are related to the lagrangian @xmath3 and the total energy @xmath50 by the following relations@xmath710 implying@xmath711 we assume the total energy to be equal to the hamiltonian , that is to be defined by the lagrangian through the legendre transformation@xmath712 then we arrive at the following uniquely defined expressions for the kinetic and potential energies@xmath713 we are particularly interested in the case of the quadratic lagrangian@xmath714 where @xmath89 and @xmath90 are real symmetric matrices which are positive definite and positive semidefinite , respectively , and @xmath9 is a real skew - symmetric matrix , that is@xmath715 then according to ( [ lagham3 ] ) the total energy @xmath50 the takes the form@xmath716 consequently , as it follows from ( [ lagham2 ] ) the kinetic and the potential energies are defined by@xmath717 the euler - lagrange equations corresponding to the quadratic lagrangian ( [ lagham5 ] ) are @xmath718 multiplying the above equation by @xmath719 from the left yields@xmath720 and this equation can be recast recast as@xmath721 combining the above equation with ( [ lagham2 ] ) and ( [ lagham5 ] ) we arrive at the following important identity that holds for any solution @xmath77 to the euler - lagrange equation ( [ lagham9])@xmath722 where the term @xmath723 is often referred to as the _ virial_. let us introduce now the time average @xmath724 for a time dependent quantity @xmath725 defined by@xmath726 and observe that if @xmath123 is bounded for @xmath727 then @xmath728 notice now that in the case of interest when @xmath729 and consequently @xmath730 , all eigenfrequencies of the system are real and hence all solutions to the euler - lagrange equation ( [ lagham10 ] ) are bounded . consequently , applying the relation ( [ ftaver2 ] ) , using the identity ( [ lagham12 ] ) we obtain the virial theorem:@xmath731 * acknowledgment and disclaimer : * the research of a. figotin was supported through dr . a. nachman of the u.s . air force office of scientific research ( afosr ) , under grant number fa9550 - 11 - 1 - 0163 . both authors are indebted to the referee for his valuable comments on our original paper . v. i. arnold , _ mathematical methods of classical mechanics _ , translated from the russian by k. vogtmann and a. weinstein , second edition , graduate texts in mathematics , 60 , springer - verlag , new york , 1989 .
star clusters are vital laboratories for testing theories of stellar dynamics and stellar evolution ( e.g. * ? ? ? the globular clusters orbiting the milky way constitute a sample of about @xmath0 evolved stellar systems with ages @xmath1 years . these provide useful constraints on the long term evolution of clusters but information about the paths which led to these end states can only be inferred via comparisons with the end points of numerical cluster simulations . in contrast , the system of rich star clusters in the large magellanic cloud ( lmc ) includes clusters of comparable mass to galactic globular clusters but which span the entire age range from less than @xmath2 years ( e.g. r 136 ) to about @xmath3 years ( e.g. ngc 2019 ) . this allows us to observe massive star clusters at various epochs in their evolution and makes it possible to identify evolutionary trends , under the assumption that the external processes affecting cluster evolution have not changed significantly over the lifetime of the lmc . with this aim in mind , hst project go7307 ( e.g. * ? ? ? * ) obtained deep images of four pairs of lmc clusters . within each selected pair the clusters are approximately coeval , have similar masses and metallicities and lie at comparable radii from the lmc centre but differ significantly in other properties such as core radius . the pairs range in age from about @xmath4 years up to @xmath3 years . the youngest clusters ( ngc 1805 and ngc 1818 ) are old enough to have expelled all their residual gas and so are pure systems . fig . [ fig : lmccores ] shows a plot of the observed core radii of star clusters in the lmc as a function of their ages , with the clusters in the go7307 hst sample identified by name . there is a clear trend that the spread in core radius is an increasing function of cluster age . this trend was first identified by @xcite from ground - based images and was recently confirmed by @xcite using hst archive data . @xcite calculated core radii for @xmath5 lmc clusters in a uniform manner with the aim of obtaining an unbiased data set . they discuss in some detail the uncertainties in the data and conclude that selection effects are not responsible for the observed trend . they suggest , as @xcite had earlier argued , that the increased spread in core radius is an evolutionary phenomenon : all clusters formed with relatively small core radii ( @xmath6 pc ) and subsequently some clusters experienced core expansion while others did not . @xcite discuss the possibility that inter - cluster variations in the slope of the initial mass function ( imf ) could have produced the observed spread in core radius . they point out that mass loss from the increased numbers of massive stars found in clusters with imfs flatter than the standard salpeter imf could lead to core expansion of the observed magnitude on a timescale of about @xmath7 years . as @xcite notes , however , the imf slope required to explain the largest core radii would lead to disruption of the clusters on a timescale of a few @xmath8 years . more directly , based on a detailed observational analysis of the clusters in the go7307 hst sample , @xcite have shown that the slopes of their luminosity functions are similar to each other within the observational uncertainties , strongly suggesting that their imfs were very similar down to stellar masses of about @xmath9m@xmath10 . the absence of inter - cluster population variations implies that stellar evolutionary mass loss can not be invoked to explain the observations . as a star cluster evolves , encounters lead to a redistribution of energy among the stars . heavier stars and binaries acquire lower velocities , on average , and sink towards the cluster core . this process of mass segregation could enhance the heating effect of a population of primordial binaries because the binaries would tend to fall to the centre where encounters occur more frequently . observations of young clusters ( e.g. * ? ? ? * ) and numerical simulations of star formation ( e.g. * ? ? ? * ) show that so - called primordial mass segregation can also result from the cluster formation process . @xcite have shown that there is strong evidence for mass segregation in all the young and intermediate - age clusters in the go7307 sample . further , the amount of mass segregation appears to be similar in all clusters and can not therefore account for differences in the cluster core radii . the aim of this paper is to investigate , by means of simulations , whether dynamical processes could give rise to the observed core radius age relation . in this context , the pair of clusters ngc 1868 and ngc 1831 are of particular interest . these clusters have very similar ages and metallicities and their masses differ by at most a factor of two ( see table 1 of * ? ? ? however , as fig . [ fig : lmccores ] shows , the core radius of ngc 1831 is almost three times larger than that of ngc 1868 . given that the cluster ages are only around @xmath11myr any process which has affected the core radii must be capable of producing significant evolution even on this relatively short time scale . in this paper , we consider the dynamical effects of ( 1 ) variations in the external tidal field experienced by the clusters and ( 2 ) differences in the initial numbers of dynamically active binary stars in the clusters . given that the crossing times at the half - mass radii of these clusters are about @xmath12 yrs and the times to orbit the lmc are about the same order of magnitude , it is reasonable to suppose that either of these processes could have begun to produce observable effects on cluster structure on the relevant time scale . during the past decade , numerical simulations of star clusters have become increasingly realistic owing to dramatic increases in computational speeds and significant improvements in programming algorithms ( e.g. * ? ? ? the code used in this paper incorporates the stellar evolution models of @xcite . this makes it possible to analyse the output from the simulations in an identical manner to that used for observational data , thus allowing meaningful confrontation of models with observations . this is particularly important when discussing cluster properties such as core radius because the relation between the theoretical and observational definitions of this quantity depend on the conversion between the mass function and luminosity function of the cluster stars . for a real cluster this conversion is non - trivial because it depends on , among other quantities , the cluster metallicity and is further complicated by the possible presence of mass segregation ( see e.g. * ? ? ? * ) . in this paper , all comparisons between models and observations are made in terms of directly observable properties . in the interests of clarity , we include here a note on the definition of the core radius . as @xcite discuss , theoreticians , numerical simulators and observers assign different meanings to the term `` core radius '' , reflecting the quantity of greatest relevance to their purpose . in theoretical work , the core radius @xmath13 refers to the natural length scale of the system under study , for example the scale radius of the widely used @xcite profile . the operational definition of the core radius commonly used in discussions of @xmath14 simulations is that of the density radius @xmath15 . this is the density - weighted average of the distance of each star from the density centre @xcite , where the local density at each star is estimated from the mass within a sphere containing the star s six nearest neighbours and the density centre is defined as the density - weighted mean position of the stars . observationally , since the available information is either in the form of cluster surface brightness profiles or luminosities of individual stars instead of mass densities and stellar masses , the core radius @xmath16 corresponds to the radius at which the projected surface brightness falls to half its central value . in the case of a cluster of equal - mass stars in equilibrium , there is generally a well - defined relation between @xmath13 and @xmath16 . for example , for a @xcite model it is straightforward to show that @xmath17 . it is important to note , however , that during the evolution of the cluster this relation is not static and that there is no general relation between @xmath15 and either @xmath13 or @xmath16 . in this paper , we use the term `` core radius '' exclusively to refer to the observational quantity @xmath16 . in section [ sec : tides_discussion ] we will compare the evolution of @xmath15 and @xmath16 , in order to emphasize the care which must be taken in comparing observations with simulations . of lmc star clusters as a function of age . the clusters in the go7307 hst sample are identified by name , and matched pairs are joined by dotted lines . ( following fig . 1 of @xcite , data taken from @xcite).,width=317 ] in section [ sec : nbody ] we describe briefly the code used in the simulations , as well as the pipeline used to calculate observational properties of the simulated clusters . section [ sec : tides ] considers the effects of time variation of an external tidal field on the evolution of the cluster core radius while section [ sec : binfrac ] deals with clusters moving in a steady tidal field but with different initial binary fractions . in section [ sec : bintide ] we investigate the combined effect of both mechanisms by studying binary - rich clusters in a time - varying tidal field and the results of these simulations are discussed in the context of the lmc clusters in section [ sec : bintidedisc ] . finally in section [ sec : conclusions ] we summarise our conclusions . .summary of runs performed . @xmath19 , @xmath20 and @xmath21 are the initial number of stars , total stellar mass and primordial fraction of hard binaries , respectively . the clusters were all simulated in the tidal field of a point mass representing the lmc , and each simulation ran until time @xmath22 . [ cols="<,^,^,^,^,^,^",options="header " , ] simulations are a very valuable tool in the study of star clusters because they allow many different processes to be studied with a minimum of simplifying assumptions . the simulations discussed below were performed using the nbody4 code ( e.g. * ? ? ? * ) running on the harp-3 and grape-6 special purpose computers @xcite at the institute of astronomy , cambridge and on a grape-6 board at the american museum of natural history . the nbody4 code uses the 4th order hermite integration scheme to follow the orbits of the cluster stars and utilises state - of - the - art regularisation schemes @xcite for the accurate and efficient treatment of the internal motion of binary and higher - order subsystems . it also incorporates a full set of stellar evolution models , in the form of parameterized functions , which are used to follow the evolution of single stars and binaries @xcite . of particular importance to the present study is the treatment of mass loss . any gas which is lost from an evolving star and is not accreted on to a binary companion is assumed to leave the cluster instantaneously , thereby reducing the depth of the cluster potential well . a realistic treatment of mass loss is therefore essential to allow the study of the evolution of such cluster properties as the core radius . within the code the equations of motion are integrated in scaled units such that @xmath23 and the total mass and virial radius of the cluster are also set to unity @xcite . for a star cluster in virial equilibrium the initial energy in these units is @xmath24 and the crossing time is @xmath25 . given the total mass of the cluster in solar masses and an appropriately chosen length scale ( which determines the conversion from length units to physical units ) it is straightforward to obtain the conversion factors for time and velocity from units to myr and @xmath26 , respectively . in order to facilitate direct comparison of the results of our simulations with the observed data , we determine the core radius for each model cluster using the same processing pipeline as that used by @xcite . in summary , the stellar data from the model clusters ( in the form of position and luminosity information for each member star ) are binned radially to obtain a projected surface brightness profile . the three - parameter model @xcite @xmath27 is then fitted to the cluster profile . here , @xmath28 is the central surface brightness , @xmath29 determines the fall - off of the surface brightness profile at large radii and @xmath16 is related to the length scale @xmath30 via @xmath31 . full details of the fitting procedure used and discussion of the associated uncertainties are presented in @xcite . in order to reduce the numerical noise due to the small number of stars in our simulated clusters we observe each cluster along three perpendicular lines of sight and obtain the median core radius from these three estimates . this helps to reduce the impact of individual bright stars which might otherwise skew the core radius estimation . further , each run is performed four times using different random realisations of the initial cluster model and the results are averaged in order to ensure the robustness of any observed trends . initially , all the model clusters are generated according to a @xcite density law @xmath32 following the prescription of @xcite . here @xmath33 is the plummer length scale and @xmath34 is the central mass density . observations of the young clusters in the lmc show that they do not appear to be tidally limited @xcite . in this case , the plummer profile is a reasonable model to use for the initial density profile given that the fall - off in its outer regions lies between that of a tidally truncated king model and the observed profiles of the young lmc clusters . the initial distribution of stellar masses is drawn from the imf of @xcite which is a three - part power law fit to stellar data in the solar neighbourhood . we take the lower and upper mass limits to be @xmath35m@xmath10 and @xmath36m@xmath10 , respectively . we assume solar metallicity for the stars in our clusters . this is an appropriate choice for the younger clusters for example , @xcite have shown that ngc 1805 and ngc 1818 , whose ages are about @xmath37 and @xmath38myr respectively , are well matched by isochrones of approximately solar metallicity . since the pairs of clusters in the go7307 hst sample were chosen to have similar metallicity , we know a priori that the observed range of cluster core radii can not be due to metallicity variations . we have therefore chosen to perform all our simulations at solar metallicity , and postpone discussion of the effects of metallicity variations on star cluster evolution to future work . table [ tab : runs ] summarises the simulations which we have performed . the clusters in these runs initially contain @xmath39 or @xmath40 stars and have a mean mass of about @xmath41 m@xmath10 . clearly , the lmc clusters are more massive and contain greater numbers of stars than our model clusters and therefore direct comparison of the results of our simulations with the observed data must be approached with caution . however , the effects of increasing @xmath18 in our simulations can be estimated by considering the various time scales in the problems we study . detailed discussion of these scaling issues is presented in sections [ sec : tides_discussion ] and [ sec : bintidedisc ] where we argue that none of our conclusions is strongly dependent on the small size of our simulated clusters . we first consider the effects of the external tidal field of the lmc on the evolution of the cluster core radii . theoretical arguments and numerical simulations strongly suggest that tides have played a significant role in the evolution of star clusters and cluster systems in galaxies such as the milky way ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? in addition , observations of tidal tails around at least twenty milky way globular clusters ( e.g. * ? ? ? * ; * ? ? ? * ) provide compelling evidence of the ongoing disruption of star clusters by the milky way . the gravitational field of the lmc ( mass @xmath42m@xmath10 ) is much weaker than that of a galaxy like the milky way ( mass @xmath43m@xmath10 ) . however , as @xcite show ( their fig . 15 ) , there is a deficit of low - mass clusters older than @xmath3 years . if we assume that the cluster population originally contained a spectrum of cluster masses similar to that of the present - day young clusters , their absence at late times strongly suggests that tides have been at work on the cluster system in exactly the manner expected from the models of @xcite . given that the go7307 hst sample of 8 clusters was selected on the basis that they lie at similar distances from the lmc centre but display a wide range of core radii , the steady tidal field experienced by a cluster on a roughly circular orbit can not be the origin of the observed trends . further , given the intrinsic weakness of the lmc field at a radius of @xmath44kpc , where our clusters lie , one might not expect the tidal field to play a significant role in cluster evolution . we therefore consider clusters on non - circular orbits . such clusters experience tidal shocks during their orbits and we investigate whether these impart sufficient energy to the cluster to produce differences in the evolution of @xmath16 . @xcite discusses the possible role of disk shocking in the internal evolution of the intermediate - age lmc clusters , but questions whether any of the clusters is on an appropriate disk - crossing orbit for this to be an important effect . in this section we consider instead the tidal shocking produced by a point mass galaxy . in order to investigate the role of the lmc tidal field in the evolution of its cluster system , it is necessary to incorporate an appropriate external gravitational field into the simulations . we model the lmc potential as that of a point mass of @xmath45m@xmath10 . this is a rather drastic oversimplification of the actual situation . we note , however , that this model exaggerates the effect of the tidal field , as any extended distribution of mass will have a more slowly varying gravitational potential . further , in the radius range @xmath46 kpc , in which our model clusters orbit , the gradient of the potential is within a factor of two of that of the detailed lmc mass model of @xcite . thus our simulations constitute a plausible upper limit to the possible effects of tides . in the simulations , the equations of motion are integrated in an accelerating but non - rotating reference frame centred on the centre of mass of the cluster and whose axes are aligned with fixed spatial directions . this choice of frame simplifies the equations of motion for a cluster on a non - circular orbit despite having less symmetry than the more usual rotating frame used for steady tidal fields whose axes always point towards the attracting centre @xcite . implementation of the time varying tidal field requires some care because mass loss from evolving stars or the escape of stars from the cluster causes the centre of mass to acquire a non - zero velocity relative to the frame of integration . it is therefore necessary to re - centre the frame of reference to coincide with the rest frame of the centre of mass following each such event . all internal and external forces must then be reinitialized to ensure consistency with the re - centred frame during subsequent integrations . we have performed simulations for clusters on elliptical orbits with perigalactic radii @xmath47kpc and apogalactic radii @xmath48kpc , probing the region in which most of our sample clusters lie but largely avoiding the complications of the inner regions of the lmc which may contain a stellar bar ( e.g. * ? ? ? we also modelled clusters on circular orbits at a radius of @xmath49kpc which corresponds to the time - averaged radius of the elliptical orbits . this allows us to isolate the effect of the time variation of the tidal field because both sets of clusters experience the same mean tidal field . we calculate the initial perigalactic tidal radius @xmath50 in the approximation that the cluster is on a circular orbit at a radius @xmath51 . in this case , @xmath50 is determined from the cluster mass @xmath52 , the mass of the point mass galaxy @xmath53 and @xmath51 via ( e.g. * ? ? ? * ) @xmath54 the apocentric tidal radius @xmath55 is determined in a similar way from @xmath56 . the unit of length , which determines the conversion between length units and physical units , is then chosen such that the limiting radius of all clusters is initially equal to @xmath50 . however , at the start of the simulations the clusters are placed at the apocentre of their orbits about the lmc which means that they do not fill their tidal radii . if , instead , the clusters initially extended to @xmath55 and were not truncated at @xmath50 then the removal of stars outside @xmath50 during the first few perigalactic passages could mask any heating effects at radii interior to @xmath50 due to the time variation of the tidal field . in contrast to a king model , the plummer density law ( [ eqn : plummer ] ) formally extends to infinity and so the outer radius of a cluster is not well - defined . the length unit we use yields a pericentric tidal radius close to the ( random ) position of the outermost star in our initial clusters . the internal stellar velocities are scaled to ensure virial equilibrium at the start of the simulation . the tidal radius of a cluster on an elliptical orbit does not have a unique value . it varies continuously both as the cluster moves along its orbit and when the cluster loses mass through stellar evolution or the escape of member stars . however , for clusters whose internal relaxation times are longer than their orbital periods , tidal stripping tends to truncate the star cluster at @xmath57 @xcite . these two time scales are of similar magnitude for our simulated clusters so we expect two - body relaxation processes to lead to some expansion of the cluster beyond @xmath50 . we take account of this by defining escaping stars to be those stars outside twice the apogalactic tidal radius @xmath55 . stars whose orbits carry them outside this radius are generally lost to the cluster ( e.g. * ? ? ? stars are thus free to populate the region between @xmath50 and @xmath58 . the choice of length unit determines the physical density of the cluster and hence the physical rate at which two - body relaxation processes affect the evolution of the cluster . our aim is to investigate the influence of a time - varying external tidal field so it is important to ensure that all internal evolutionary processes proceed in an identical manner for each of our simulated clusters . we therefore use the same length unit for the clusters on circular orbits as we chose for those on elliptical orbits . this means that the clusters on circular orbits also do not initially fill their tidal radii and so necessarily allows expansion of these clusters . however , since the circular orbits are placed at the time - averaged radius of the elliptical orbits this expansion should on average be the same for both sets of clusters . any differences in the rates of expansion are due to the time variation in the tidal field . the effect of steady tidal fields on star cluster evolution has been the subject of papers by a number of authors ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? however , until recently the problem of clusters moving in time - varying tidal fields had received relatively little attention from simulations . @xcite uses the results of some simulations of single - mass star clusters orbiting a point mass to determine the appropriate choice of escape radius for simulations of mass loss . more recently , @xcite presented an extensive survey of the dynamical evolution of star clusters in external tidal fields . our investigation is complementary to theirs because we are looking at the evolution of the observable global structure of the cluster while @xcite focus their attention mostly on the evolution of the stellar mass function within the cluster . the results of all our simulations are presented in terms of quantities which can be compared directly with those calculated from the observed data , thereby enabling us to discuss how different dynamical processes impact on observations . we follow the evolution of the clusters for about @xmath59 ( where @xmath60 is the half - mass relaxation time ) this corresponds to @xmath61gyr in most cases . by this time , the clusters on elliptical orbits have lost more than @xmath36 per cent of their stars making estimates of properties such as the core radius increasingly noisy . at regular intervals during the evolution , the data from the clusters are recorded and all relevant properties of the clusters are calculated . ( upper curves ) and @xmath62myr ( lower curves ) . there is clear evidence for mass segregation developing in the cluster.,width=317 ] the solid curve in fig . [ fig : circecc ] presents the evolution of the core radius for clusters on circular orbits about the lmc at a distance of about @xmath49kpc from the centre . the broken curve shows the evolution of @xmath16 for clusters on elliptical orbits with the same time - averaged galactocentric radius . it is apparent from this figure that the core radius evolution in both cases is indistinguishable , especially in view of the noise associated with the output of small-@xmath18 simulations . during the first @xmath63yrs , the core radius is seen to decrease slightly from its initial value . this is because the calculation of @xmath16 is dominated by the light of the most massive stars which rapidly become more centrally concentrated due to the effects of mass segregation . as fig . [ fig : massseg ] shows , after @xmath64myr the half - mass radius of the more massive stars is already @xmath65 per cent smaller than that of the less massive stars . after its initial contraction phase , the core radius begins to increase . this occurs because the clusters do not fill the tidal radius corresponding to their galactocentric positions at the start of the simulations . a combination of mass loss from the rapid evolution of massive stars and two - body relaxation can therefore drive the overall expansion of the cluster seen in fig . [ fig : massseg ] . the top panel of fig . [ fig : velevol ] shows that the cluster s half - mass radius @xmath66 is increasing with time ( see fourth panel of fig . [ fig : internal_props ] for long - term @xmath66 behaviour ) . during this period its @xmath15 ( not shown ) is roughly constant at about @xmath67pc . ( solid curve ) and core radius @xmath16 ( broken curve ) ; middle : plot of the virial ratio @xmath68 ( whose equilibrium value is @xmath69 ) ; bottom : velocity anisotropy parameter @xmath70 of stars within the inner @xmath37 per cent lagrangian radius ( broken curve ) and between the @xmath71 per cent and @xmath72 per cent radii ( solid curve ) . see text for discussion.,width=317 ] the expansion of @xmath66 is driven by a combination of two processes which are illustrated in the lower panels of fig . [ fig : velevol ] . the middle panel shows the evolution of the virial ratio @xmath68 . initially the cluster is in equilibrium with @xmath73 . after about @xmath74myr , the cluster is temporarily displaced from equilibrium ( @xmath75 ) by mass loss from the most massive stars . this is accompanied by the increase in the half - mass radius of the cluster shown in the top panel of fig . [ fig : velevol ] . it is interesting to note that the @xmath16 of the cluster does not change as a result of this mass loss . stellar evolution can not account for the sustained expansion of the cluster , however . the cluster loses about @xmath76 per cent of its mass as a result of the above mass loss events . using the virial theorem it is straightforward to show that the gravitational radius @xmath77 of a stellar system with initial radius @xmath78 and mass @xmath52 which loses an amount of mass @xmath79 is given by @xcite @xmath80 a @xmath76 per cent mass loss should lead to an increase of less than @xmath37 per cent in the size of the cluster which is insufficient to account for the cluster expansion seen in figs . [ fig : circecc ] and [ fig : massseg ] . the bottom panel of fig . [ fig : velevol ] shows the evolution of the velocity anisotropy parameter @xmath81 @xcite inside the @xmath37 per cent lagrangian radius of the cluster ( the radius containing @xmath37 per cent of the total mass of the cluster ) and in the outer parts ( between the @xmath71 per cent and @xmath72 per cent lagrangian radii ) . initially , the velocity distribution throughout the cluster is isotropic ( @xmath82 ) . while the central parts of the cluster retain this isotropy , in the outer parts the distribution becomes increasingly radially anisotropic ( @xmath83 ) . this indicates that the continued growth of @xmath66 is being fueled by encounters in the inner parts of the cluster which produce a population of stars on extended radial orbits . the velocity distribution in the outer parts of isolated clusters simulated by @xcite exhibited similar behaviour during post core - collapse evolution ; those authors similarly concluded that the growth of the outer lagrangian radii at late times was due to two - body processes in the inner regions . given the similarity between the core radius evolution of clusters on circular and elliptical orbits , we now proceed to investigate the extent to which the clusters on elliptical orbits have been affected by the variation in the tidal field . fig . [ fig : internal_props ] presents the evolution of certain key internal properties of two typical clusters , one on a circular orbit and one on an elliptical orbit . the top panel shows the galactocentric radius @xmath84 along the elliptical orbit to emphasize the clear correlation between perigalactic passages and changes in the internal properties of the cluster on the elliptical orbit . the second panel shows that , as expected , during a perigalactic passage the rms velocity @xmath85 of the stars in the cluster increases sharply . this velocity kick ejects the most energetic stars from the cluster . as a result of this mass loss , @xmath85 settles to a value slightly below that of the cluster on the circular orbit once it returns to virial equilibrium . the third and fourth panels of fig . [ fig : internal_props ] show how @xmath60 and @xmath66 evolve during the cluster orbit . the most noticeable effect is that the relaxation time for the cluster on the elliptical orbit is significantly reduced after the second perigalactic passage , so that the cluster evolves more rapidly following this encounter . this is due both to the decrease in the total number of stars in the cluster and to the increase in the mean stellar mass @xmath86 ( shown in the bottom panel of fig . [ fig : internal_props ] ) . the increase in @xmath86 occurs because , on average , low - mass stars acquire larger velocities during pericentre passages and are hence more likely to escape from the cluster . the value of @xmath66 is seen to increase during the perigalactic passage but rapidly settles back to a value slightly below the circular - orbit value . thus while the tidal shocks experienced by the clusters on elliptical orbits do not affect the evolution of @xmath16 , they are nonetheless strong enough to modify other important properties of the cluster . as fig . [ fig : internal_props ] shows , the impact of the shocks on the cluster properties increases with time this is due to the decreasing number of bound stars in the cluster . with time for a cluster on an elliptical orbit ( broken curve ) . the solid line shows the position of the circular orbit of equal time - averaged radius . ( lower four panels ) evolution of rms velocity @xmath85 , half - mass relaxation time @xmath60 , half - mass radius @xmath66 and mean stellar mass @xmath86 . curves for two cases are shown : ( 1 ) a cluster on a circular orbit ( solid curves ) and ( 2 ) a cluster on an elliptical orbit ( broken curves ) . note that the half - mass radius is calculated using all stars within a radius of @xmath58 of the cluster centre . in the simulations all such stars are classified as non - escapers , although at any given time a fraction of them are formally unbound.,width=309 ] before drawing our final conclusions based on these simulations , we now investigate whether the lack of impact on @xmath87 is due to the weakness of the shocks we have simulated or whether the value of @xmath16 is intrinsically robust to tidal shocking . the strength of a tidal shock is generally presented in terms of the dimensionless parameter @xmath88 given by ( e.g. * ? ? ? * ) @xmath89 where @xmath90 is an estimate of the destruction time of the cluster owing to energy input from the gravitational shocks . the time scale @xmath90 is given by @xmath91 where @xmath92 is the energy of the cluster evaluated at the half - mass radius and @xmath93 is the rate at which stars at the half - mass radius acquire energy due to the shock . in a simulation containing a finite number of stars , evaluation of this expression requires some care . averaging the energy changes of stars near the half - mass radius does not produce robust results because individual stars may gain and lose energy during the perigalactic passage through other means than interaction with the external tidal field . instead we calculate the total energy binding the cluster as a function of time and estimate the energy input to the cluster during perigalactic passages . [ fig : energy ] compares the time evolution of the binding energy @xmath94 of a cluster on a circular orbit with that of a cluster on an elliptical orbit . for a cluster in an external potential @xmath94 is defined as ( e.g. * ? ? ? * ) @xmath95 where @xmath96 and @xmath97 are the total kinetic and potential energy of the cluster , respectively , and @xmath98 is the energy in the tidal field . the binding energy must be negative for the cluster to remain bound . the effects of the external field are visible as dramatic changes in @xmath94 at about @xmath99 and @xmath100myr . note that the sudden , temporary change in @xmath94 for the circular run at about @xmath101myr is due to a collision between two binaries that results in the the break - up of one of the binaries and the subsequent ejection of its components . the ratio of the energy produced by this event to the binding energy of the cluster emphasises the relatively fragile nature of the clusters we are simulating . . curves for two cases are shown : ( 1 ) a cluster on a circular orbit ( solid curve ) and ( 2 ) a cluster on an elliptical orbit ( broken curve ) . energies are given in standard units ( @xmath102).,width=317 ] for clusters on circular ( solid curve ) and elliptical ( broken curve ) orbits in units of the initial mass @xmath20 . enhanced mass loss during the period of revirialisation which follows each perigalactic passage is seen in the elliptical orbit case.,width=317 ] during perigalactic passages , some stars acquire velocities above the escape velocity of the cluster and this leads to an enhanced mass - loss rate ( see fig . [ fig : masstime ] ) . however , the finite time required for such stars to escape means that @xmath94 can temporarily become positive because it is calculated by summation over all stars within twice the tidal radius of the cluster . after the cluster has passed pericentre , stars continue to leave across the tidal boundary and @xmath94 returns to a negative value indicating that the remaining cluster is still gravitationally bound . following each shock , the cluster undergoes a revirialisation process during which the energy acquired from the tidal interaction is redistributed among the cluster stars . this process results in further loss of mass from the cluster because , on average , the revirialisation process gives velocity kicks to stars which are three times larger than those arising from the initial shock ( see e.g. * ? ? ? * ) . , @xmath65 , @xmath36 , @xmath103 and @xmath72 per cent of the cluster mass . the heating effect from the external tidal field is evident , even at the radius containing only @xmath65 per cent of the cluster mass.,width=317 ] we take the energy input by the shock to be the difference between the energy of the cluster after this revirialisation process is complete and the energy of the cluster on the circular orbit at this time ( about @xmath104myr ) . the time between the shocks is @xmath105myr , which leads to @xmath106myr . at this time @xmath107 for the cluster is about @xmath108myr , so the shock strength is about @xmath109 . comparing this value with fig . 13 of @xcite and noting that the initial concentrations @xmath110 of our clusters are about @xmath111 ( where we assume that @xmath112 is the initial outer limit of the clusters ) , we find that our clusters lie approximately on the borderline of the region where shocks are important . we conclude that the shocks our clusters are experiencing would significantly affect their evolution on long time scales . indeed , as fig . [ fig : massradcomp ] illustrates , even the @xmath65 per cent lagrangian radius of the cluster is affected by the tidal shocks , indicating energy input into the cluster over a wide range of radii . nevertheless , on a timescale of up to @xmath113 yrs , the effects of tidal heating are insufficient to modify the observed @xmath16 of the clusters significantly . ( broken curve ) with the observational core radius @xmath16 ( solid curve ) for a single cluster . top : un - smoothed results from the simulations . middle : curves smoothed by averaging over five time bins . bottom : ratio of @xmath16 to @xmath15 calculated using the smoothed curves from the middle panel . see text for discussion.,width=317 ] the top panel of fig . [ fig : velevol ] illustrates an important point which is worth emphasizing in the context of comparisons between simulated data and observations , namely that @xmath16 and @xmath114 are not , in general , related in a simple way . in particular circumstances , for example at the start of our simulations where the model clusters are in strict dynamical equilibrium , it is easy to show that , for a plummer model composed of equal mass stars , the two quantities are related by @xmath115 . however , fig . [ fig : velevol ] shows that this relation quickly breaks down as the cluster evolves away from its initial conditions . thus , in the absence of further information , observational estimation of @xmath87 must not be assumed to imply knowledge of the mass distribution in the cluster . the same is true of the density radius @xmath15 defined in section [ sec : introduction ] . [ fig : corecomp ] shows the evolution of @xmath16 ( solid curve ) and @xmath15 ( broken curve ) for one of the clusters on an elliptical orbit . as the top panel of the figure shows , in a small-@xmath18 simulation , both @xmath16 and @xmath15 are intrinsically noisy quantities : this is emphasised by the middle panel where we have smoothed the curves from the top panel using a window function of width five time bins . the bottom panel of fig . [ fig : corecomp ] , which shows the ratio of @xmath16 to @xmath15 for the smoothed data , demonstrates that the relation between the two quantities is not a simple proportionality and that the evolutionary trends in the two quantities can be quite different . the complicated relation between light and mass in star clusters is the result of a variety of processes such as stellar evolution and mass segregation which modify the mass and light distributions in different ways . given the absence of simple relations between the different core radii , great care must be exercised when we interpret observational data through comparison with numerical simulations or theoretical models . the simulations we have performed contain relatively low numbers of stars and are smaller and of lower total mass than the actual lmc clusters . however , the mean densities of our clusters are very similar to those of the lmc clusters and it is this which determines the extent to which an external tidal field can influence the internal dynamics of a cluster . on this basis , we can be confident that conclusions based on our simulations are relevant to the clusters in the lmc . our simulations have shown that , on a timescale of @xmath116gyr , which corresponds to several @xmath60 for our model clusters , tidal forces are not able to produce a spread in core radius similar to that observed among the lmc clusters . there are two obvious avenues for building on the simulations we have performed . first , simulations of clusters containing a larger numbers of stars @xmath18 could be performed . this would allow us to achieve cluster masses comparable to those of the real lmc clusters . however , increasing @xmath18 in a cluster while holding the physical density of the cluster constant leads to an increase in @xmath60 . the expansion seen in @xmath16 for our simulated clusters is driven primarily by two - body processes ( which proceed on a time scale @xmath60 ) rather than by tidal heating and it will therefore be slowed by increasing @xmath18 . further , the disruptive effect of pericentre passages depends on the ability of a cluster to re - fill the region outside its pericentric tidal radius @xmath50 during the time between close passages . as @xmath60 increases this re - filling is reduced . the cluster will therefore have fewer stars outside @xmath50 and hence will be less affected by shocks than small-@xmath18 clusters ( see fig . 2 of @xcite ) . a cluster on an elliptical orbit and which , as in our simulations , initially filled its pericentric tidal radius would remain truncated at this radius throughout its evolution because there would not be sufficient time between perigalactic passages for the region outside @xmath50 to be re - filled @xcite . thus shocks should have less impact on the @xmath16 of larger-@xmath18 clusters and the observed expansions of clusters on both circular and elliptical orbits should remain similar . the second possible improvement would be to use a more realistic model of the lmc potential , for example one based upon the recent mass models of @xcite . however , the actual lmc mass distribution is more extended than the keplerian potential we have been using in our simulations and therefore the gradients in its gravitational field are smaller . consequently , the associated strength of its tidal effects on clusters moving on elliptical orbits is reduced . thus , we conclude that neither of the above modifications to our simulations would tend to increase the effect of the tidal field . the fact that the time - varying tidal field in our simulations did not result in changes in the evolution of @xmath16 in the extreme case of a low - mass cluster in orbit about a point - mass galaxy leads us to conclude that tides have not yet had time to play a significant role in the internal evolution of the intermediate - age lmc clusters . in this section we present results for models moving in a steady tidal field but with different numbers of primordial binaries , in order to isolate the impact of large binary fractions on the evolution of the cluster properties . the clusters discussed here move on the same orbits as the circular orbit runs of section [ sec : tides ] . based on an analysis of the observed binary sequence in the young cluster ngc 1818 , @xcite have shown that the binary fraction @xmath117 in the centre of this cluster is about @xmath118 . here @xmath119,where @xmath120 and @xmath121 are the numbers of single stars and binary systems , respectively . to date , the binary fractions of the older clusters in our sample have not been determined . thus , at present the binary fractions of these clusters are essentially free parameters . heggie s @xcite law implies that only hard binaries are dynamically important in the evolution of a star cluster . a binary is said to be hard when the magnitude of its binding energy @xmath122 exceeds the kinetic energy of a typical cluster star @xcite . in the case of a binary with component masses @xmath123 and @xmath124 at a separation @xmath30 moving in a cluster with mean mass @xmath86 and velocity dispersion @xmath125 the hardness condition is @xmath126 ( we note that the nbody4 code uses a dispersion of @xmath127 in this equation to determine the hardness limit for binaries , in order to take account of the larger velocity dispersion in the core of a centrally concentrated cluster . ) in our simulations the boundary between hard and soft binaries occurs at a separation of about @xmath128au for binaries containing stars of the mean stellar mass in the cluster . hard binaries act as heat sources because they become more tightly bound as a result of interactions with single stars and thereby impart kinetic energy to the stars they encounter . @xcite found that the density radii of clusters containing primordial binaries experienced shallower contraction during core collapse and subsequently tended to be larger than those of clusters containing only single stars . this result , based on simulations of single - mass clusters with binary fractions of about @xmath129 , motivates our investigation of clusters containing large numbers of primordial binaries . hereinafter , we will use @xmath117 to refer to the fraction of _ hard _ binaries in the cluster . as an extreme case , we consider first clusters with initial @xmath130 and with the hard binaries randomly distributed throughout the cluster . in each binary the mass of the primary star is drawn from the imf of @xcite and the mass of its companion is obtained by randomly choosing a mass ratio for the binary from a uniform distribution constrained by the single star mass limits ( for a more detailed discussion , see * ? ? ? the binary orbital parameters are generated from the distribution of @xcite . the clusters are otherwise generated exactly as in the single star runs of section [ sec : tides ] . we note that the inclusion of a large number of hard primordial binaries dramatically increases the computation time required for each simulation . per cent , @xmath131 per cent and @xmath36 per cent hard primordial binaries ( short - long dashed , long - dashed and solid curves , respectively ) . all clusters are on circular orbits about the lmc . see text for discussion.,width=317 ] ( top panel ) and the energy binding the cluster @xmath94 ( middle panel ) . two cases are shown ( 1 ) a cluster without primordial binaries ( broken curves ) and ( 2 ) a cluster containing @xmath36 per cent primordial binaries ( solid curves ) . the total mass is normalised by its initial value @xmath20 and @xmath94 is given in standard units . the bottom panel shows the evolution of the internal binding energy of the primordial binaries @xmath132 ( in standard units ) in the run containing primordial binaries . the decrease in @xmath132 is the result of interaction - induced binary hardening . note that @xmath133 includes the internal energy of escaped binaries . see text for discussion.,width=317 ] per cent primordial binaries ( solid curves ) and with single stars only ( broken curves ) . from bottom to top , the curves show the radii containing @xmath37 , @xmath65 , @xmath36 , @xmath103 and @xmath72 per cent of the cluster mass . the heating effect of the primordial binaries is visible at all radii.,width=317 ] fig . [ fig : sinbin ] compares the evolution of @xmath16 for clusters containing @xmath134 single stars and no primordial binaries with that for clusters containing a significant initial binary population . an increased binary fraction produces a noticeable effect on the evolution of @xmath16 . after @xmath116gyr , the core radii of clusters containing @xmath36 per cent primordial binaries are about a factor of two larger than those of clusters with no binaries . the results presented in fig . [ fig : sinbincomp ] give some insight into the mechanism by which a large binary fraction can drive an expansion of @xmath16 . the top panel shows that the rate at which mass is lost from a cluster with initial @xmath135 is significantly greater than in the case of a cluster with no initial binary population . after @xmath116gyr this cluster has lost almost @xmath131 per cent more mass than the cluster containing only single stars . the evolution of the energy binding the cluster @xmath94 is shown in the second panel of fig . [ fig : sinbincomp ] . in a cluster with a large initial binary fraction , @xmath94 rapidly attains a value significantly higher than that of a cluster of single stars . the enhanced mass - loss rate due to the presence of the binaries decreases the depth of the cluster s potential - well ; the magnitude of the binding energy is further reduced by the heating of the remaining bound stars through weak encounters . as fig . [ fig : sinbinpercent ] shows , the energy input due to binary heating is distributed throughout the cluster . all the lagrangian radii of the cluster rapidly attain larger values when the cluster contains primordial binaries . we note that , as was discussed in section [ sec : tides_results ] , the growth of the clusters containing only single stars is due to the fact that the clusters do not fill their tidal radii at the start of the simulations . we use identical initial conditions for the clusters containing binaries the excess heating observed in those clusters is due to the presence of the hard binary population . the third panel of fig . [ fig : sinbincomp ] shows the internal binding energy of the primordial hard binaries @xmath132 as a function of time . in calculating @xmath132 we include the binding energy of those hard binaries which have escaped from the cluster because we are interested in the rate at which encounters lead to the hardening of binaries . this plot demonstrates that , as expected , the hard binaries become harder with time and so impart energy to the cluster . it is well known that energy generation by hard binaries halts core collapse ( e.g. * ? ? ? however , fig . [ fig : sinbincomp ] demonstrates that even in the very early stages of a cluster s evolution , the heating effect of a large primordial hard binary population can have a significant impact on the energy budget of the cluster . we note that the occasional large changes in @xmath132 are the result of stellar evolutionary processes and therefore do not contribute to the direct heating of the cluster . for example , at @xmath136myr the components of one binary system merge to form a single star : prior to the merger the binary is hardened by tidal interactions between its components . the massive star thus produced explodes as a supernova soon after and the velocity kick received by the remnant as a result of the explosion is reflected in the spike in @xmath137 . the prominence of this event is due to the relatively small number of stars in the simulation . [ fig : sinbin ] shows that the @xmath16 evolution of clusters with @xmath138 is very similar to that of clusters with @xmath135 initially . this suggests that the heating effect of primordial binaries saturates above a certain value of the binary fraction @xmath139 . for larger values of @xmath117 , the @xmath16 evolution is unchanged , although the mass - loss rate is greater for increased @xmath117 . from fig . [ fig : sinbin ] we see that @xmath139 for the clusters we have simulated appears to lie in the range @xmath140 because an initial binary fraction of @xmath37 per cent has a noticeably smaller impact on @xmath16 than one of @xmath131 per cent . we are currently investigating this effect in more detail , including its dependence on the number of stars in the simulation : the results will be presented elsewhere . here we note only that the saturation may arise from the fact that at early times only a relatively small number of binaries take part in the heating process . in our simulations only about three per cent of the binaries are dynamically hardened during the first @xmath141myr . we find that this fraction appears to be independent of the value of @xmath117 . whether a particular binary experiences an encounter which leads to hardening depends on the masses and separation of the components as well as the local stellar density in the neighbourhood of the binary . for a fixed binary mass , however , the hardest binaries have the smallest separations and hence are least likely to interact with other stars . this is reflected in our simulations : for a given binary mass , all the binaries which undergo early hardening are initially among the more weakly bound pairs . in our final set of simulations we examine clusters containing large numbers of primordial binaries and moving on non - circular orbits about the lmc . we have considered the extreme case of @xmath135 but , as was shown in the previous section , these results also apply to somewhat smaller values of @xmath117 . as fig . [ fig : eccbincorecomp ] shows , the core radius of a cluster on an elliptical orbit and containing primordial binaries exhibits stronger growth than that of a cluster of single stars on a circular orbit : after about @xmath142myr the core radius of the former cluster is almost a factor of two larger . the increase in core radius is solely due to the presence of binaries . as in the case of clusters containing only single stars ( section [ sec : tides ] ) , a cluster with the same primordial binary fraction but on a circular orbit exhibits almost identical core radius evolution . however , the increased mass - loss rate which results from binary heating leaves the cluster on the elliptical orbit more vulnerable to tidal disruption . after only @xmath142myr , the cluster has lost almost @xmath143 per cent of its mass and the core radius determination is consequently very uncertain . by @xmath116gyr it is virtually disrupted and retains only about @xmath37 per cent of its initial mass while the circular orbit cluster still contains about @xmath144 per cent of its initial mass . our simulations demonstrate that the presence of a large population of hard binaries can produce significant expansion of the core radius of a cluster . in agreement with many previous studies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) we have thus shown that binaries constitute a dynamically significant population in star clusters . in particular , we have found that primordial binaries can rapidly affect observable cluster properties such as @xmath16 . in the context of the lmc clusters , however , it is clear that the presence of binaries alone can not account for the the observed trends in @xmath16 and in particular for the factor of three difference in core radius between ngc 1831 and ngc 1868 . even in the extreme case in which one cluster has no primordial binaries at all while the other cluster has a hard binary fraction of @xmath145 per cent , the difference in @xmath16 over the lifetime of the intermediate - age clusters is at most a factor of two . the core radius of a cluster containing @xmath37 per cent binaries expands by only a factor of @xmath146 over the same time scale . further , the similarity of the observed present - day luminosity functions of the clusters in the go7307 sample @xcite argues against large inter - cluster variations in the imf and hence makes significant differences in binary fraction unlikely . the initial radial distribution of hard binaries has an impact on their heating effect on the cluster stars ( e.g. * ? ? ? * ) and may also have a bearing on the saturation of the heating effect which we observed in section [ sec : binfrac ] . a binary population which was more centrally concentrated than the single stars as a result of either primordial or dynamical mass segregation would be expected to produce more heating because of the increased interaction rates in the inner parts of the cluster . in our future simulations we will consider alternatives to the uniform distribution used in the present runs . however , based on our current simulations it is possible to estimate the significance of centrally concentrating the binary population . the clusters with @xmath36 per cent binary fractions contain approximately the same number of binaries within @xmath147 as a cluster with a @xmath37 per cent overall binary fraction in which all the binaries are located within @xmath147 . given that none of our simulations exhibited @xmath147 growth of a magnitude sufficient to match the observations , a core binary fraction in excess of @xmath36 per cent would be required to produce greater @xmath147 growth . the observed similarity of the degree of mass segregation in the intermediate - age clusters @xcite render it unlikely that the radial distribution of binaries could vary so greatly between clusters . as in the case of the simulations in section [ sec : tides ] , it is natural to ask whether our conclusions are also valid for clusters containing larger numbers of stars than those in our current simulations . the simulation of large-@xmath18 clusters containing significant numbers of primordial binaries has only recently become feasible and we are currently investigating the impact of binaries on the evolution of such clusters ; the results of this study will be presented elsewhere . the present simulations show that the disruption time for a small-@xmath18 cluster in a time - varying external tidal field is significantly reduced if it contains a large fraction of primordial binaries . it is not trivial to extrapolate these results to larger-@xmath18 clusters . however , binary heating of the cluster is essentially a two - body process and therefore proceeds on the relaxation time scale @xmath60 . given that a larger-@xmath18 cluster with the same mean physical density has a larger @xmath60 , the physical rate at which the cluster is heated by its binary population will be reduced . thus , while larger-@xmath18 clusters would obviously survive for longer than those in our present simulations , their physical rate of @xmath16 expansion would be reduced . large variations in primordial binary fraction are therefore unlikely to be the explanation of the lmc observations . for the lmc cluster sample . clusters with @xmath16 larger ( smaller ) than @xmath148 pc are shown as open ( filled ) circles . data are taken from @xcite.,width=317 ] the core radii of star clusters in the lmc display a strong trend with age , namely that the older clusters exhibit a much greater spread in core radius than their younger counterparts . in particular , the @xmath16 of the intermediate - age clusters ngc 1868 and ngc 1831 differ by about a factor of three . it has been suggested @xcite that the trend represents a physical evolution of the clusters and that the clusters all formed with similar , small core radii which were modified by subsequent dynamical evolution . the possibility that this might be caused by variations in the cluster imf has been ruled out based on hst photometry of a sample of eight clusters @xcite . in this paper we have used simulations of star clusters initially containing up to @xmath134 stars to investigate some possible alternative explanations . we have considered two possible dynamical effects which could produce significant @xmath16 evolution : time - varying external tidal fields and a varying population of primordial binaries . throughout this paper we have analysed our simulations using the same procedures as were applied by @xcite to the observed data . we emphasised the importance of such an approach by comparing the evolution of the observational core radius to that of the half - mass radius and the density radius . as has been pointed out by a number of authors ( e.g. * ? ? ? * ) , in general there are no simple relations between these quantities . thus observations of simulated data sets are essential to ensure that comparisons between observed and simulated data are made in a consistent manner . we considered clusters on circular and elliptical orbits of equal time - averaged radius about a point mass representing the lmc . the clusters are scaled so that their limiting radii correspond to the pericentric tidal radius of the clusters on elliptical orbits . however , at the start of each simulation the clusters lie well within the tidal radius which corresponds to their initial galactocentric position ( apocentre for the clusters on elliptical orbits and the circular radius for the clusters on circular orbits ) . initially , the clusters expand due to stellar mass loss but the competing effect of mass segregation means that @xmath16 decreases slightly . at later times , two - body relaxation processes drive a further expansion of the clusters because the clusters are not tidally limited . this expansion is mirrored by a growth in @xmath87 . however , the rate of @xmath16 expansion is found to be identical irrespective of the orbit on which the cluster is moving . we conclude that differences in the cluster orbits can not be responsible for the observed differences in @xmath16 among the lmc clusters . given that the binary fractions of all but the youngest clusters are currently unconstrained by observations , we compared the evolution of the @xmath16 of a cluster containing no primordial binaries with that of a cluster with a large initial hard binary fraction . the presence of significant numbers of binaries leads to increased growth of @xmath16 compared with the evolution of clusters containing only single stars . binary fractions of @xmath36 per cent and @xmath131 per cent produce similar degrees of @xmath16 expansion , while a @xmath37 per cent binary fraction results in noticeably less growth . even in the most extreme case of a cluster with a binary fraction of @xmath36 per cent on an elliptical orbit , however , @xmath16 increases by at most a factor of two . not only is this smaller than the increase required to explain the lmc observations but the required spread in primordial binary fraction is already difficult to reconcile with the observed similarity of the cluster stellar luminosity functions @xcite . our simulations have demonstrated that neither large variations in primordial binary fraction nor differences in cluster orbits about the lmc can account for the @xmath16age relation seen in the lmc clusters . the origin of the spread in @xmath16 among the intermediate - age clusters therefore remains an open issue . we now return briefly to the observed data on the lmc cluster system as a whole to see whether there are any further clues to the origin of the trend . [ fig : rhom ] presents evidence for a systematic difference in the central density of the large-@xmath16 clusters relative to the rest of the cluster population : these clusters have systematically lower central densities for a given cluster mass . this could be indicative of variations in star formation efficiency ( sfe ) between the clusters : low sfe leads to rapid cluster expansion and consequently a drop in the central density ( e.g. * ? ? ? simulations of cluster expansion due to residual gas expulsion from clusters with low sfe @xcite do not appear to show sustained core expansion of the magnitude seen in the intermediate - age lmc clusters . however , the core radii quoted in @xcite are those of the mass distribution in the simulated clusters which , as was discussed in section [ sec : tides ] , may not accurately reflect the changes in the light distribution . we are currently investigating the long - term evolution of clusters which experience different degrees of primordial gas loss owing to sfes in the range @xmath149 , required to explain the range of core radii seen in the youngest lmc clusters @xcite . we are also simulating the effects of primordial binaries on larger-@xmath18 clusters . of particular interest is the possibility that the heating effects of a large binary population could enable a cluster to retain a large initial @xmath16 produced by gas expulsion . miw acknowledges financial support from pparc . jrh acknowledges financial support from a kalbfleisch fellowship and thanks the institute of astronomy ( cambridge , uk ) for supporting a visit during this work . adm would like to acknowledge the support of a trinity college ers grant and a british government ors award . cat thanks churchill college for a fellowship . we are very grateful to sverre aarseth for assistance with the nbody4 code and for numerous useful discussions . we thank pete bunclark at the institute of astronomy , cambridge , for handling the installation and setup of the grape-6 hardware . we also thank hongsheng zhao for helpful discussions and richard de grijs and douglas heggie for useful discussions and comments on an earlier version of this paper .
quark gluon plasma ( qgp ) , strongly interacting matter composed of quarks and gluons , has been created in high - energy nuclear collisions at the relativistic heavy ion collider ( rhic ) at brookhaven national laboratory ( bnl ) and the large hadron collider ( lhc ) at cern @xcite . one of the main discoveries at the rhic was that the qgp behaves like a nearly perfect fluid @xcite , and since then the transport properties of the qgp , especially its shear and bulk viscosity coefficients , have been under active investigation . the qgp created in high - energy nuclear collisions expands , cools down , and turns into a hadron gas . since hadrons scatter with each other via the strong interaction , information about the qgp stage is , in general , contaminated by rescatterings in the hadronic stage . to probe the qgp more directly , thermal photons and dileptons were proposed as penetrating probes @xcite . once emitted , photons and dileptons propagate to detectors without rescattering since they interact only electromagnetically and their mean free path is therefore longer than the typical size of the system . however , photons and dileptons are not perfect probes of the qgp either . they are emitted during all the stages of high - energy nuclear collisions such as the primary hard scatterings , pre - equilibrium stage , and late hadronic stage @xcite , and it is not possible to distinguish photons and dileptons emitted from the qgp from those originating from the other processes . in this paper , we follow up the previous studies @xcite about multistrange hadrons , and claim that their distributions allow to probe the qgp immediately after hadronization . multistrange hadrons , in particular @xmath0 mesons and @xmath1 baryons , have small scattering cross - sections with pions since they do not form any resonances unlike other hadrons . this is a unique property of the multistrange hadrons : their distributions reflect the properties of the system mainly at a specific stage deep inside the fireball , unlike photons and dileptons which come from the all stages . by using a hybrid model in which hydrodynamic description of the qgp fluid is followed by a hadronic cascade model , violation of the mass ordering of the differential elliptic flow parameter @xmath2 was predicted @xcite . this phenomenon was observed recently by the star collaboration @xcite , which indicates that @xmath0 mesons are less affected by hadronic rescatterings and , consequently , that they are good probes deep inside the matter . in this paper , we investigate the hadronic rescattering effects on observables more systematically by employing an integrated dynamical approach which we have previously used to analyze various observables at the rhic and the lhc energies @xcite . in the previous study of the violation of mass ordering @xcite , a first order phase transition model for the equation of state ( eos ) was employed and only hadrons up to the mass of @xmath3 were taken into account in the hadron phase . in the present paper , we use a more realistic equation of state @xcite connecting a parametrized lattice qcd eos at large temperatures to a hadron resonance gas eos at low temperatures . the hadron resonance gas in the eos contains all the resonances in the hadron cascade model jam ( jet @xmath4 microscopic transport model ) @xcite used as the last stage of the integrated dynamical approach . hence , in addition to @xmath0 , we are able to investigate hadronic rescattering effects on heavier multistrange hadrons such as @xmath5 and @xmath1 . in the following we introduce the integrated dynamical model in sec . [ sec : model ] . we compare transverse momentum ( @xmath6 ) spectra and @xmath6-differential elliptic flow parameters @xmath7 with the star data in sec . [ sec : results ] . after we confirm we reasonably reproduce the data by using this model , we discuss the hadronic rescattering effect in sec . [ sec : rescatteringeffects ] . in this section , we first revisit the violation of mass ordering of @xmath7 . we next focus on the mean transverse momentum @xmath8 and @xmath9-averaged @xmath10 calculated with or without hadronic rescatterings and investigate how much these final observables reflect the information when the hydrodynamic stage finishes . we summarize our results in sec . [ sec : summary ] . we describe the space - time evolution of high - energy nuclear collisions by an integrated dynamical approach on an event - by - event basis . we divide the whole reaction into three separate stages : initial stage , hydrodynamic stage , and transport stage . the initial entropy density distribution after the collision of energetic heavy nuclei is parametrized by using a monte - carlo version of the glauber model . the subsequent expansion of the matter is described by relativistic ideal hydrodynamics . when the system becomes sufficiently dilute due to strong expansion , we switch the description of the system from hydrodynamics to kinetic theory . in the following we briefly overview each part of the model and the corresponding interfaces . for further details , see ref . @xcite . hydrodynamics is a macroscopic description and an effective theory of the system s long wavelength / time behavior . the hydrodynamical equations of motion are obtained from the conservation of energy and momentum : @xmath11 where @xmath12 is energy - momentum tensor . in the ideal fluid approximation , the energy momentum tensor can be decomposed as @xmath13 where @xmath14 , @xmath15 , @xmath16 , and @xmath17 are energy density , pressure , flow velocity , and the minkowski metric , respectively . to close the system of partial differential equations we need to know the equation of state ( eos ) of the fluid to obtain pressure @xmath15 as a function of energy density @xmath14 . once the eos is known , the space - time evolution of thermodynamic quantities and flow velocity is determined for given initial conditions , i.e. , energy density distribution and flow field at initial time @xmath18 . we ignore the baryon number and its continuity equation since at the collider energies , net baryon density at midrapidity is tiny . when we solve eq . ( [ eq : hydro ] ) , temperature is not needed since it is merely a parameter to connect energy density and pressure . nevertheless , we also follow the space - time evolution of temperature to decide where to switch from fluid to cascade . we employ the _ s95p_-v1.1 parametrization for the eos . the _ s95p _ parametrization @xcite connects the lattice qcd based eos @xcite in the high temperature region to a hadron resonance gas eos in the low temperature region . the hadronic part of _ s95p_-v1.1 contains the same hadronic species as the jam hadronic cascade model @xcite described in the next subsection . we numerically solve eq . ( [ eq : hydro ] ) in the milne coordinates @xmath19 which is appropriate for the description of the evolution at relativistic energies @xcite . here @xmath20 is longitudinal proper time , @xmath21 $ ] is space - time rapidity , and @xmath22 and @xmath23 are transverse coordinates perpendicular to the collision axis . in the numerical calculations , we employ the piecewise parabolic method @xcite which is known to be able to describe shock waves . for details about this numerical algorithm , see also ref . @xcite . we employ the hadronic cascade model jam @xcite to describe the space - time evolution of hadron gas after particlization " , i.e. , after switching from fluid to particles . at the late stage of collisions , the system is too dilute to maintain equilibrium . we assume this happens around temperature , which we call the switching temperature @xmath24 , and switch description from hydrodynamics to kinetic theory on @xmath25 isosurface @xcite . within the current model , @xmath24 is rather adjustable parameter which controls the final particle ratios . we choose @xmath26 mev to reproduce the observed @xmath27 spectra of pions , kaons , and protons + anti - protons in the low @xmath27 region @xcite at the full rhic energy . at @xmath28 , we calculate the single particle phase space distributions @xmath29 for all hadrons included in the eos by employing the cooper - frye prescription @xcite @xmath30 here @xmath31 is a degeneracy of the hadronic species @xmath32 , @xmath33 is four - momentum of a particle , and @xmath34 is a normal vector of the @xmath28 hypersurface . it is well known that at some momenta the cooper - frye formula ( [ eq : cf ] ) gives negative contribution for space - like hypersurface element ( or even time - like hypersurface element with negative time component ) . these negative numbers correspond to in - coming particles , which one can not treat in the hadronic cascade model . so we just neglect these contributions and consider only out - going particles . we sample particles by going through all the hypersurface elements and generate ensembles of hadrons on an event - by - event basis . note that we do not oversample particles to gain statistics for each event . this means each hydrodynamic event corresponds to one ensemble of particles . among all the calculations in the integrated dynamical approach , sampling particles from particlization hypersurface is numerically most expensive , which is crucial in event - by - event simulations . in ref . @xcite , we discussed in detail how to sample particles and integrate the cooper - frye formula ( [ eq : cf ] ) over momentum at less numerical cost . in hadronic cascade models , experimental hadronic cross section data are implemented when available . however , there are many hadronic scattering processes where data do not exist . in such a case , we use the additive quark model @xcite to describe the scattering cross section @xmath35 here @xmath36 is the total nucleon - nucleon cross section , @xmath37 is the number of constituent quarks in a hadron , and @xmath38 is the number of strange quarks in a hadron . thus hadrons containing strange quark(s ) have relatively small total cross section in this model . moreover , @xmath0 mesons and @xmath1 baryons do not form resonances when scattering with pions and therefore have much smaller cross sections than non - strange hadrons . in the simulations , dynamics of the system is described by a sum of incoherent two - body collisions . two particles collide with each other if their minimum distance @xmath39 in the center of mass frame is smaller than the distance given by the total cross section @xmath40 this is a classical , geometrical interpretation of the two - body collision cross section . note that in jam the properties of hadrons are those of free particles . the medium modifications of hadron properties such as a broadening and mass shift @xcite have been extensively studied by various further developments of transport models such as the relativistic vlasov - uehling - uhlenbeck ( rvuu ) model @xcite , the hadron - string - dynamics ( hsd @xcite ) and parton - hadron - string - dynamics ( phsd @xcite ) transport models , the giessen boltzmann - uehling - uhlenbeck ( gibuu ) @xcite , and the isospin quantum molecular dynamics ( iqmd ) @xcite . nevertheless , since the main difference in the cross sections of multistrange particles and the other hadrons is due to the ( non-)existence of resonance channels , we do not expect the in - medium modifications to significantly affect this difference , and the use of free particle properties to provide a conservative baseline is justified . as mentioned , the interactions in jam are described as two particle scatterings , and multi - particle scatterings are not included . it has been argued that multi - pion fusion processes in particular would have a significant effect on the yields of proton - antiproton pairs @xcite , but subsequent calculations have shown that regeneration is a @xmath41% level effect at the lhc @xcite . nevertheless , the multi - meson fusion effects for @xmath1 baryons are much smaller than for protons or @xmath42 @xcite , and thus their inclusion would make the later - discussed differences between rescatterings of non - strange and multistrange particles even larger . as a default setting strong interactions and decays are simulated in jam . however , we switch off the decay channels for @xmath0 mesons in the present study to investigate hadronic rescattering effects on @xmath0 mesons in an efficient way . otherwise , we would need to perform mass reconstruction of @xmath0 mesons from two kaons after numerical simulations . since the lifetime of @xmath0 mesons ( @xmath43 fm/@xmath44 ) @xcite is somewhat larger than the typical life time of the system ( @xmath41 fm/@xmath44 ) , the daughter kaons from @xmath0 meson decays ( or lack of them ) are not expected to affect the bulk evolution . as we will discuss in sec . [ sec : results ] , the yield of @xmath42 measured by star contains the feed down from electromagnetic @xmath45 decays . this process does not happen in the default setting in jam and , therefore , is simulated separately when we analyze @xmath42 spectra . we also have an option in jam to deactivate all the hadronic rescatterings and/or resonance decays to investigate how these affect final observables . once initial conditions for the hydrodynamic evolution are fixed , the subsequent evolution is determined , and we are able to obtain final particle distributions which can be compared with experimental data . the pre - thermalization stage between the first contact of colliding nuclei and the initial time of hydrodynamic evolution may be described by non - equilibrium field theory , although how exactly it could be done is one of the open issues in the physics of relativistic heavy ion collisions . thus , we do not even try to describe it and simply rely on the monte - carlo ( mc ) glauber @xcite and the modified brodsky - gunion - kuhn ( bgk ) models @xcite to give the density distribution at @xmath18 in the transverse plane and in the longitudinal direction , respectively . the mc glauber model enables us to calculate the number density of participants and of binary collisions on an event - by - event basis . our assumption is that the initial entropy density is proportional to a linear combination of these two densities . even at very high collision energies matter profile in the longitudinal direction depends on space - time rapidity , and boost invariance does not hold . furthermore , in non - central collisions the longitudinal density gradient at mid space - time rapidity may be non - zero due to the different thicknesses of the colliding nuclei at each transverse position . this can be deduced from ( pseudo-)rapidity distribution of hadrons in p(d)-a collisions . in the modified bgk model smooth initial longitudinal profile is parametrized by taking account of the difference in the local thickness of colliding nuclei . it is noted that we do not employ the monte - carlo kharzeev - levin - nardi model @xcite which was used to calculate initial conditions in our previous hydrodynamic studies @xcite since , in the present study , we focus on the hadronic rescattering effects on transverse dynamics , which are not supposed to be sensitive to the choice of initial conditions . assuming local thermalization , energy density distribution @xmath46 and pressure @xmath47 are obtained from the entropy density distribution through the _ s95p_-v1.1 eos . initial flow velocity at @xmath48 is supposed to be bjorken s scaling flow @xcite , namely , @xmath49 and @xmath50 . throughout this study , we fix @xmath51 fm/@xmath44 . we choose initial parameters to reproduce the @xmath27 distributions of pions , kaons , and protons + anti - protons in au+au collisions at @xmath52 gev measured by the phenix collaboration @xcite . the centrality dependence of multiplicity is controlled by a fraction of soft ( participants ) and hard ( binary collisions ) component in the mc glauber model , whereas the multiplicity in central collisions is fixed by the overall normalization parameter of the initial entropy distribution . on the other hand , as mentioned particle ratios are controlled by the switching temperature . the parameters controlling the shape of the rapidity distribution of particles are kept the same as in our previous study @xcite . since here we concentrate on observables at midrapidity , we do not discuss them further . in hydrodynamic simulations , one needs to convert the position distribution ( a sum of delta functions in the transverse plane ) of collision points in mc glauber model to a hydrodynamic density distribution . in our model we obtain the density at a point @xmath22 by counting the number of collision points or participants within radius @xmath53 around the point , and divide by the inelastic cross section in @xmath54+@xmath54 collisions at @xmath55 gev , @xmath56 . for further details , see ref . @xcite . in this section , we compare invariant @xmath6 spectra and differential elliptic flow parameter @xmath57 of identified hadrons with experimental data from the star collaboration . in the present study , we analyzed 0.6 million `` minimum bias '' events ( defined as @xmath58 in the mc glauber calculations for hydrodynamic initial conditions ) . centrality is defined using an event distribution of the charged hadron multiplicity in @xmath59 as done by the star collaboration @xcite . note that error bars in the theoretical plots denote the statistical errors . in figs . [ fig : dist_pi&k ] , [ fig : dist_pro&phi ] , and [ fig : dist_hyperon ] , we show @xmath6 distributions of identified hadrons around midrapidity in au+au collisions at @xmath60 . for clarity , the results and the experimental data points are scaled up or down by a common factor . figure [ fig : dist_pi&k ] shows the @xmath27 spectra of ( a ) positive pions ( @xmath61 ) and ( b ) positive kaons ( @xmath62 ) compared with the star data ( @xmath63 for pions @xcite and @xmath64 for kaons @xcite ) . the jacobian between @xmath65 and @xmath23 is taken into account to obtain the invariant @xmath6 spectra in our results . even if we tuned the initial conditions and the switching temperature in the model using pion , kaon , and proton + anti - proton @xmath27 spectra from the phenix collaboration in our previous study , the results are overall in good agreement with the star data too . in particular , the model reproduces the pion data up to @xmath66 gev/@xmath44 in central ( 0 - 12% ) collisions . above this , our result gradually deviates from the data due to appearance of ( semi-)hard components such as recombination and jet fragmentation . the more peripheral the collision , the lower the @xmath6 where our result begins to deviate from the data . due to the limited @xmath6 range of the data , the same behavior is not seen in the kaon spectra . if one looks at the pion spectrum in peripheral collisions ( 60 - 80% centrality ) carefully , one sees that experimental data are systematically larger than the results . this is because in peripheral collisions there is a small discrepancy between phenix @xcite and star @xcite pion yields . nevertheless we did not further tune the parameters to reproduce the star data because we want to keep the same framework as in our previous study @xcite . in fig . [ fig : dist_pro&phi ] , we compare the @xmath27 spectra in @xmath61 for ( a ) protons and ( b ) @xmath0 mesons with the star data @xcite . due to limited statistics , we do not show some data points of the proton spectra in high @xmath27 regions ( above @xmath67 gev/@xmath44 ) in 20 - 40 , 40 - 60 , and 60 - 80% centrality . the proton spectra from star are corrected for the @xmath42 and @xmath68 feed down . weak decays do not occur in the default setting in jam , so we are able to compare our results directly with the star corrected data . since we neglect baryon chemical potential in our model , our results for protons are slightly below the experimental data at some @xmath27 range . nevertheless , overall slopes are in good agreement with the data below @xmath66 gev/@xmath44 . on the other hand , similar to pions , deviation between the experimental data and the results gradually increases with @xmath27 , and the more peripheral the collision , the lower the @xmath9 where this deviation appears . as mentioned in the previous section , we switch off the decays of @xmath0 mesons in the hadronic cascade to be able to analyze @xmath0 meson spectra directly without resorting to mass reconstruction from kaons . tendencies seen in @xmath0 meson spectra are similar to those in proton spectra . figure [ fig : dist_hyperon ] shows the @xmath27 spectra of strange baryons , ( a ) @xmath69 , ( b ) @xmath70 , and ( c ) @xmath71 , compared with the star data @xcite . similar to the proton case , the more peripheral the collisions , the worse the agreement between experimental data and the results , especially above @xmath72 - 2.0 gev/@xmath44 . as expected the measured @xmath42 and @xmath5 yields are larger than the calculated yields . we had chosen the switching temperature @xmath24 to reproduce the observed pion to kaon and pion to proton ratios , which leads to a lower temperature than the statistical model fit to all observed hadrons @xcite , and thus to lower yields of hyperons . on the other hand it has been argued that the statistical model @xcite with the common chemical freezeout temperature does not reproduce the proton and @xmath42 yields simultaneously at the lhc and rhic energies @xcite . it was claimed @xcite that earlier switching from hydrodynamics to urqmd at energy density @xmath73 mev/@xmath74 , which corresponds to higher temperature than @xmath75 mev , would be a key to resolve the issue , but our results do not improve significantly if we use higher switching temperature @xmath76 mev . it remains to be seen whether the origin of this discrepancy lies in a different description of the proton - antiproton annihilations in jam and urqmd cascades , or in different expansion dynamics in collisions at rhic and lhc energies . note that the experimental yield of @xmath42 in star contains the feed down from electromagnetic @xmath77 decays ( @xmath78 ) @xcite . since this is the dominant mode of @xmath45 decay with the branching ratio @xmath79 and @xmath80 m @xcite , the feed down correction of it is experimentally challenging . we simulate this decay in jam and include this contribution in the final @xmath42 spectra in fig . [ fig : dist_hyperon ] ( a ) . since the mass of @xmath45 is close to that of @xmath42 , the yield of the primordial @xmath45 is expected to be of the same order as the yield of the primordial @xmath42 . the ratio of @xmath77 to @xmath42 at switching temperature can be estimated as @xmath81 after correcting for the effects of resonance decay and rescatterings , we find that the final particle ratio of @xmath77 to @xmath42 is around 0.3 . we have included both sources of @xmath42 in our results , but in general this should be kept in mind when comparing theoretical results with the data . although the number of data points from star is limited for @xmath1 baryons , their yields and slopes are consistent with our results . it is worth noticing here that the recent lattice qcd calculations suggest the existence of resonances in general @xcite , or strange baryon resonances in particular @xcite , which have not been discovered yet . such resonances , whether they are of hagedorn kind @xcite , or those predicted by quark models ( see e.g. refs . @xcite ) , would contribute to the yields of strange hadrons , and might somewhat change the dynamics of the late hadronic stage @xcite . some of those resonances could be formed in the scattering of @xmath0 mesons or @xmath1 baryons , but it is unknown whether such resonances would significantly affect the average scattering cross section of @xmath0 mesons and @xmath1 baryons in temperatures below the switching temperature . investigation of this possibility is beyond the scope of the present paper . in fig . [ fig : v2_pikp1 ] the two - particle cumulant @xmath7 of pions , kaons , and protons in the midrapidity region ( @xmath82 ) are compared with star @xmath83 data @xcite at three centralities ( 10 - 20% , 30 - 40% , and 50 - 60% ) . for pions , we reasonably reproduce the experimental data for each centrality class . due to the limited number of events , it is hard to obtain @xmath84 with smaller statistical errors , and especially difficult this would be for protons . thus some points with large error bars , in particular in low @xmath6 region , are not shown in these figures for clarity . we also calculated @xmath10\{rp } ( not shown ) , namely @xmath10 with respect to theoretically - known reaction plane , to reduce the statistical errors and found that @xmath10\{rp } of pions , kaons , and protons from our model are in reasonable agreement with the star data in mid - central collisions . however , @xmath10\{rp } lacks initial fluctuations of the event plane angle and , consequently , is significantly smaller than the data in both central and peripheral collisions . it is noted that , from the mc glauber model analysis , the fluctuation effect on eccentricity is large in central and peripheral collisions , but almost negligible in mid - central collisions . e.g. _ , fig . 3 ( a ) in ref . @xcite . at the centralities shown in fig . [ fig : v2_pikp1 ] a clear mass ordering behavior among pions , kaons , and protons , namely larger @xmath10 for smaller mass at fixed @xmath6 , is seen both in our results and in the star data . in fig . [ fig : v2_hyperon ] , we show the event plane @xmath7 of @xmath0 , @xmath42 , @xmath5 , and @xmath1 at some centrality classes from the integrated dynamical approach . in comparison , star data from the event plane method @xmath10\{ep } @xcite are also shown in the figure . we calculate @xmath10\{ep } by employing the @xmath65-subevent method using hadrons in @xmath85 as done by the star collaboration @xcite , @xmath86 \rangle.\ ] ] the event plane angle @xmath87 ( @xmath88 ) is defined for the particles with positive ( negative ) pseudorapidity . in the actual experimental analysis , an pseudorapidity gap of @xmath89 between two subevents was introduced to suppress the nonflow effects @xcite . we found , however , that introduction of the gap does not change final results within error bars in our analysis . the resolution factor in this case is calculated as @xmath90 \rangle } $ ] . in calculating the @xmath91 vector , @xmath92 needed for the event plane angle @xmath93 a weight factor @xmath94 was also introduced @xcite . however , within statistical errors the results are similar to those obtained using @xmath95 . therefore , we employ @xmath95 in all the results shown in fig . [ fig : v2_hyperon ] . for @xmath42 and @xmath5 our results are systematically larger than the star data . for @xmath0 and @xmath1 our results show similar behavior , but since the data have large errors , no firm conclusion can be made . our results of elliptic flow parameters for identified hadrons are systematically larger than the star data , which suggests that there is room for finite , but perhaps small , viscosity in the fluid - dynamical stage . however , we do not go into details about the qgp viscosity since the purpose of the present study is to investigate the effects of rescattering in the late hadronic stage , which are not affected by the qgp viscosity nor by the initial state of the system . in this section , we investigate hadronic rescattering effects on final observables , in particular , how much final observables reflect the properties of the system at particlization in the integrated dynamical approach . if some observables do not change much during the late kinetic stage , these observables can be utilized as `` penetrating '' probes . namely , their distributions reflect the properties of the system immediately after hadronization , and are not contaminated by later hadronic rescatterings . to quantify the effects of hadronic rescattering we perform simulations of nuclear collisions with the following three options in the hadronic cascade calculations : ( i ) full dynamical evolution ( default ) , ( ii ) deactivating hadronic rescatterings , and ( iii ) deactivating both hadronic rescatterings and resonance decays . to simplify the calculations we calculate elliptic flow parameters with respect to the theoretically - known reaction plane . in figs . [ fig : v2_noscat_scat ] ( a ) and ( b ) , we show @xmath7 of pions , kaons , protons , and @xmath0 mesons in minimum bias au+au collisions at @xmath96 from the integrated dynamical approach . figure [ fig : v2_noscat_scat ] ( a ) represents the results without hadronic rescatterings ( option ( ii ) ) . these results exhibit the mass ordering behavior , namely @xmath97 for @xmath98 , in the low @xmath27 region . in general the mass ordering results from collective flow in which all the components of the fluid flow at a common fluid velocity @xcite . on the other hand , fig . [ fig : v2_noscat_scat ] ( b ) shows a violation of mass ordering between protons and @xmath0 mesons below @xmath99 gev/@xmath44 : @xmath100 even though @xmath101 . if one compares these two plots , @xmath102 decreases below @xmath1031.5 gev/@xmath44 during the hadronic rescattering stage . on the other hand , @xmath104 stays almost unchanged during the hadronic rescattering stage . as a consequence , the order is inversed between protons and @xmath0 mesons . to see this behavior more clearly , we also show the ratio @xmath105 . below 1 gev/@xmath44 , this ratio becomes clearly positive indicating the violation of mass ordering . the violation of mass ordering was predicted in ref . @xcite and observed recently by the star collaboration @xcite . it should be noted that a larger switching temperature ( @xmath106 mev ) and a first order phase transition in the eos were employed in the calculations of ref . we confirm in the present study that the violation of mass ordering is also observed when a more realistic eos from lattice qcd and a lower switching temperature ( @xmath75 mev ) are used . we can interpret the violation of mass ordering as a result of evolution of mean @xmath27 and @xmath27-averaged @xmath10 during the late transport stage . the slope of the @xmath27-differential @xmath10 can be approximated by the ratio of @xmath9-averaged @xmath10 to mean @xmath27 @xcite , @xmath107 if @xmath57 is linearly proportional to @xmath27 , it is easy to show that eq . ( [ eq : v2slope ] ) holds exactly . thus a change of slope in @xmath57 due to hadronic rescatterings could be attributed to changes of @xmath10 and/or @xmath8 , which provides us with more intuitive understanding . we next calculate @xmath8 and @xmath6-averaged @xmath108 for each identified hadron in @xmath109 in minimum bias au+au collisions at @xmath96 with the three different options mentioned above . we show @xmath8 and @xmath108 as functions of hadron mass in figs . [ fig : pt&v2_versus_mass ] ( a ) and ( b ) , respectively . default results using the option ( i ) are shown with closed squares . to see the effect of hadronic rescatterings , we use the option ( ii ) in the hadronic cascade calculation . we also use the option ( iii ) to obtain the results excluding the contributions of both hadronic rescatterings and resonance decays . in the default option ( i ) in which both hadronic rescatterings and resonance decays are included , @xmath8 is not a linear function of the hadron mass . by switching off hadronic rescatterings by employing the option ( ii ) , we are able to extract the hadronic rescattering effects on evolution of mean @xmath6 in the late hadronic stage . the difference of @xmath8 between with and without hadronic rescatterings are relatively smaller for @xmath110 , @xmath0 , @xmath5 , and @xmath1 than for the others . in order to confirm a pure flow effect we also calculate @xmath8 with option ( iii ) in which resonance decay contributions are also not included . as is well - known in analysis using the blast wave model @xcite , @xmath8 is a linear function of the hadron mass in this case @xcite . we also confirm that it is the case even in our fully dynamical calculations . @xmath6-averaged @xmath10 for three options ( i ) , ( ii ) , and ( iii ) are compared with each other in fig . [ fig : pt&v2_versus_mass ] ( b ) . by comparing the results ( i ) and ( ii ) one finds that due to hadronic rescatterings @xmath10 for hadrons other than pions increases little by at most @xmath1036% but @xmath10 for pions increases by @xmath10325 even though elliptic flow parameters are commonly believed to be sensitive to the early stage of the reaction . on the other hand , the effect of hadronic rescattering on pion mean @xmath27 is almost absent @xcite . it is interesting to note that the effect of resonance decays on pion @xmath10 is also significant , see fig . [ fig : pt&v2_versus_mass ] ( b ) . how resonance decays reduce pion @xmath10 was discussed in ref . @xcite . as we expected , neither @xmath111 nor @xmath10 for multistrange hadrons is affected by hadronic rescatterings , but at least one of these observables is affected for all other particles . to see how much hadronic rescatterings affect the final observables , we show ratios of @xmath8 and @xmath10 at particlization but after decays ( option ( ii ) ) to the ones in the final state ( option ( i ) ) @xcite for each identified hadron in fig . [ fig : ratio ] . for comparison , we also show the ratio obtained from a @xmath112 scaling ansatz ( open square ) @xcite in the case of mean @xmath6 . in this ansatz , one can parametrize @xmath6 distribution as @xmath113 where @xmath114 , @xmath115 , and @xmath116 are freeze - out temperature , particle mass , and flow velocity , respectively . parameters were chosen to reproduce the slope of the proton @xmath6 spectrum in option ( ii ) ( @xmath117 and @xmath118 ) and in option ( i ) ( @xmath119 and @xmath120 ) . if all the particles interact strongly with each other and move at a common collective velocity during the hadronic stage , mass ordering would appear . in fact , the result from the @xmath112 scaling depends monotonically on particle mass , and pions , kaons , and protons follow this monotonic tendency obtained from @xmath112 scaling ansatz . however , multistrange hadrons ( including `` hidden '' strangeness ) ( @xmath0 , @xmath5 , and @xmath1 ) in the integrated dynamical approach deviate from the pattern . this means that multistrange hadrons freeze out earlier than other hadrons and do not participate in the radial flow during the transport stage . multistrange hadrons couple weakly with the system consisting mostly of pions since they have small scattering cross sections and hardly rescatter with pions in the late hadronic stage . mean transverse momentum @xmath8 for dominant constituents of the medium , namely pions , decreases during the hadronic rescattering stage since @xmath121 work done in the longitudinal direction reduces the transverse energy per unit rapidity @xcite but the number of pions is fixed . in this way , @xmath122 in fact , the two parameter sets in the @xmath112 scaling ansatz mentioned above are chosen to obey this inequality ( [ eq : meanptineq ] ) . as for @xmath10 in fig . [ fig : ratio ] ( b ) , about 20% of final @xmath10 of pions is generated during the late transport stage . in other words , only @xmath10380% of final @xmath10 reflects elliptic flow generated during the fluid - dynamical stage . obviously , this could depend on centrality ( and perhaps collision energy ) and the number can be regarded as an average value at the rhic energy . whereas hadronic rescatterings affect little final @xmath10 for the other hadrons . by combining these two results in fig . [ fig : ratio ] , we conclude that multistrange hadrons are less affected by hadronic rescatterings and , therefore , can be used to probe the hadronization stage in high - energy nuclear collisions . coming back to violation of mass ordering in differential @xmath10 shown in fig . [ fig : v2_noscat_scat ] and also in fig . 5 in ref . @xcite , one can interpret this intriguing phenomenon as follows . for pions @xmath6-averaged @xmath10 increases but @xmath8 changes little in the hadronic stage , so slope of @xmath123 gets steeper as seen from eq . ( [ eq : v2slope ] ) . on the other hand , for protons , @xmath6-averaged @xmath10 does not change so much , but @xmath8 increases in the hadronic stage , so @xmath102 decreases . however , for @xmath0 mesons , both @xmath6-averaged @xmath10 and @xmath8 after the hydrodynamic stage are almost the same in the final state and , consequently , @xmath124 is unchanged . finally we show the normalized freeze - out time ( defined as the longitudinal proper time @xmath125 of the last interaction in hadronic cascade simulations ) distributions around mid - rapidity ( @xmath126 ) in minimum bias au+au collisions at @xmath127 . in fig . [ fig : dndtau ] the results are shown in two panels , ( a ) for mesons and ( b ) for baryons , for clarity . the shaded areas represent the particlization time distribution for charged hadrons . since the particlization time does not depend on scattering cross sections , it is almost equal for all hadrons within error bars . a prominent peak with small width is seen below @xmath128 fm/@xmath44 , corresponding to the time of particlization at different parts of the system . the freeze - out time distributions of @xmath0 and @xmath1 do not differ from the particlization time distribution significantly , which reflects their small scattering cross sections . on the other hand , other hadrons have broader freeze - out time distributions . for @xmath5 , a peak also exists at an early time like for @xmath0 and @xmath1 but its height is lower and the peak is accompanied by a broad tail . this is because @xmath5 has a contribution from long - living resonance @xmath5(1530 ) ( decay width @xmath129 mev @xmath103 1/(20 fm/@xmath44 ) @xcite ) whose decay forms a long tail in the freeze - out time distribution . so the primordial @xmath5 freezes out as early as @xmath0 and @xmath1 do . these results support our findings above that multistrange hadrons freeze out just after the particlization due to less rescatterings in the late kinetic stage . we have studied the effects of the hadronic rescatterings especially on multistrange hadrons and claimed they can be used to probe the hadronization stage in high - energy nuclear collisions . we have employed an integrated dynamical model in which the ideal hydrodynamic model is combined with the hadronic cascade model , jam , and confirmed that our model fairly reproduces the experimental data of @xmath6 spectra and differential @xmath10 not only for pions , kaons , and protons but also for multistrange hadrons . within this approach , we have simulated the hadronic stage with or without the rescatterings and compared the calculated mean transverse momentum and @xmath6-averaged @xmath10 in these two cases to investigate the hadronic rescattering effects . we have found that the multistrange hadrons are less affected by the rescatterings than non - strange hadrons . because of the small scattering cross sections , the multistrange hadrons do not fully participate in the radial flow during the hadronic stage and freeze out earlier than non - strange hadrons . with these results we have shown how to interpret behaviors of mass ordering and its violation in @xmath7 . changes of slope of @xmath7 during the hadronic rescattering stage result from interplay between changes of mean @xmath6 and @xmath6-averaged @xmath108 . we have also showed the freeze - out time distributions for identified hadrons and confirmed that the multistrange hadrons freeze out soon after they are `` particlized '' . in future , we also plan to investigate the hadronic rescattering effects on observables for multistrange hadrons at the lhc energy . although elliptic flow parameters are expected to be less affected by the hadronic rescatterings at the lhc , it would be important to quantify how much they do change , and thus how much the observed @xmath108 at lhc reflects the qgp evolution . m. gyulassy , arxiv : nucl - th/0403032 . lee , nucl . phys . a * 750 * , 1 ( 2005 ) . m. gyulassy and l. mclerran , nucl . a * 750 * , 30 ( 2005 ) . shuryak , nucl . a * 750 * , 64 ( 2005 ) . t. hirano and m. gyulassy , nucl . a * 769 * , 71 ( 2006 ) . feinberg , nuovo cim . a * 34 * , 391 ( 1976 ) . shuryak , phys . b * 78 * , 150 ( 1978 ) . bratkovskaya , nucl . phys . a * 931 * , 194 ( 2014 ) . a. shor , phys . lett . * 54 * , 1122 ( 1985 ) . m. he , r.j . fries , and r. rapp , phys . c * 85 * , 044911 ( 2012 ) . x. zhu , f. meng , h. song , and y.x . liu , phys . c * 91 * , 034904 ( 2015 ) . m. nasim [ star collaboration ] , nucl . a * 904 - 905 * , 413c ( 2013 ) . t. hirano , p. huovinen , k. murase , and y. nara , prog . phys . * 70 * , 108 ( 2013 ) . p. huovinen and p. petreczky , nucl . a * 837 * , 26 ( 2010 ) . bass , a. dumitru , m. bleicher , l. bravina , e. zabrodin , h. stoecker , and w. greiner , phys . c * 60 * , 021902 ( 1999 ) . d. teaney , j. lauret , and e.v . shuryak , phys . rev . lett . * 86 * , 4783 ( 2001 ) . d. teaney , j. lauret , and e.v . shuryak , arxiv : nucl - th/0110037 . t. hirano , u.w . heinz , d. kharzeev , r. lacey , and y. nara , phys . b * 636 * , 299 ( 2006 ) . c. nonaka and s.a . bass , phys . c * 75 * , 014902 ( 2007 ) . h. petersen , j. steinheimer , g. burau , m. bleicher and h. stocker , phys . c * 78 * , 044901 ( 2008 ) . f. cooper and g. frye , phys . d * 10 * , 186 ( 1974 ) . c.m . ko , v. koch , and g.q . li , ann . nucl . part . 47 * , 505 ( 1997 ) . fang , c.m . ko , g.q . li , and y.m . zheng , nucl . a * 575 * , 766 ( 1994 ) . w. cassing and e.l . bratkovskaya , phys . * 308 * , 65 ( 1999 ) . w. cassing , eur . j. st * 168 * , 3 ( 2009 ) . o. buss , t. gaitanos , k. gallmeister , h. van hees , m. kaskulov , o. lalakulich , a.b . larionov , t. leitner , j. weil , and u. mosel , phys . rept . * 512 * , 1 ( 2012 ) . c. hartnack , h. oeschler , y. leifels , e.l . bratkovskaya , and j. aichelin , phys . rept . * 510 * , 119 ( 2012 ) . r. rapp and e.v . shuryak , phys . * 86 * , 2980 ( 2001 ) . j. steinheimer , j. aichelin , and m. bleicher , phys . * 110 * , 042501 ( 2013 ) . miller , k. reygers , s.j . sanders , and p. steinberg , ann . nucl . part . sci . * 57 * , 205 ( 2007 ) . drescher and y. nara , phys . c * 75 * , 034905 ( 2007 ) . drescher and y. nara , phys . c * 76 * , 041903 ( 2007 ) . j. adams _ et al . _ [ star collaboration ] , phys . rev . lett . * 98 * , 062301 ( 2007 ) . j. cleymans , b. kampfer , m. kaneta , s. wheaton , and n. xu , phys . rev . c * 71 * , 054901 ( 2005 ) . p. huovinen , eur . j. a * 37 * , 121 ( 2008 ) . c. adler _ et al . _ [ star collaboration ] , phys . lett . * 89 * , 092301 ( 2002 ) . a. majumder and b. muller , phys . * 105 * , 252002 ( 2010 ) . a. wiranata , v. koch , m. prakash and x.n . wang , phys . c * 88 * , no . 4 , 044917 ( 2013 ) . r. hagedorn , nuovo cim . suppl . * 3 * , 147 ( 1965 ) . u. loring , b. c. metsch and h.r . petry , eur . j. a * 10 * , 395 ( 2001 ) . u. loring , b. c. metsch and h.r . petry , eur . phys . j. a * 10 * , 447 ( 2001 ) . c. greiner , p. koch - steinheimer , f.m . liu , i.a . shovkovy and h. stoecker , j. phys . g * 31 * , s725 ( 2005 ) . j. noronha - hostler , j. noronha , g.s . denicol , r.p.g . andrade , f. grassi and c. greiner , phys . c * 89 * , no . 5 , 054904 ( 2014 ) . j. noronha - hostler and c. greiner , arxiv:1405.7298 [ nucl - th ] . p. huovinen , p.f . kolb , u.w . heinz , p.v . ruuskanen , and s.a . voloshin , phys . b * 503 * , 58 ( 2001 ) . e. schnedermann , j. sollfrank , and u.w . heinz , phys . c * 48 * , 2462 ( 1993 ) . h. heiselberg and a.m. levy , phys . c * 59 * , 2716 ( 1999 ) . t. hirano , phys . rev . lett . * 86 * , 2754 ( 2001 ) . t. hirano , p. huovinen , and y. nara , phys . c * 83 * , 021902 ( 2011 ) . m. gyulassy and t. matsui , phys . d * 29 * , 419 ( 1984 ) . ruuskanen , phys . b * 147 * , 465 ( 1984 ) .
magnetic - doped quantum dots@xcite have attracted considerable theoretical and experimental interests over the last two decades . diluted magnetic ii - vi and iii - v semiconductor ( dms ) quantum dots were fabricated with single - electron control@xcite . a rich variety of different magnetic and optical properties were discovered@xcite . in such structures one can explore the physical properties coming from inter - carrier interactions and the interaction of the carriers with the magnetic ion . this system promises to be relevant for future quantum computing devices , where for instance the spin of the magnetic ion is used as a quantum bit . more recently , electrically active devices were fabricated in which a single manganese ion is inserted into a single quantum dot @xcite with control of the amount of charge in the quantum dot and consequently the possibility of the control of magnetism of single mn - doped quantum dots . investigation of the exact electronic structure of a two - dimensional quantum dot confined by a parabolic potential containing several electrons and a single magnetic impurity ( in this paper @xmath1 ) in the presence of an applied magnetic field is a new topic . in a recent investigation@xcite , a three - dimensional ( 3d ) cd(mn)se quantum dot containing several electrons , where only the low - energy levels of the single - electron problem were taken into account , was investigated in the presence of a magnetic field . here we will extend this work and include all relevant energy levels in order to obtain a convergent solution for the ground state ( and also excited spectrum ) of the system . it is known that in the absence of a magnetic ion , an external magnetic field is able to change the spin polarized state of weakly interacting electrons @xmath2 in a quantum dot in such a way that in the ground state it maximizes the total spin of the system : i.e. @xmath3 . if the inter - particle interaction is strong , even without an applied magnetic field the electrons may already be polarized . however , with increasing magnetic field and in the case the inter - particle interaction is strong , the total spin of the system can be unusually reduced by the magnetic field@xcite . in the present study , we investigate theoretically the few - electron two - dimensional confined quantum dot system that contains a single magnetic ion in the presence of an external magnetic field taking into account a sufficient large number of single - particle orbitals such that numerical exact " results are obtained . we explore how sensitive the whole system is to the position of the magnetic ion in the quantum dot and to the presence of a magnetic field and investigate the competition between the following three energies : i ) the interaction of the magnetic ion with the electrons , ii ) the interaction of the magnetic ion with the magnetic field , and iii ) the interaction of the external field with the electrons . these terms affect the spin polarization of the @xmath2 electrons in the quantum dot . explicit studies of a @xmath2-correlated - electron system interacting with a single magnetic ion in nonzero magnetic field are very rare in the literature . recent theoretical@xcite and experimental work@xcite has focused either on a small number of electrons using the exact diagonalization approach at zero field ( refs.@xcite ) for a @xmath4d quantum dot or at nonzero field@xcite including only the lowest single - particle states for a 3d system or the exciton states relevant for optical spectroscopy of self - assembled magnetic - doped quantum dots@xcite . here , we will examine thoroughly the exact properties of the system containing several correlated electrons and a single magnetic impurity in the presence of a magnetic field . in our numerical exact " diagonalization approach we include an arbitrary number of single - particle states to guarantee the accuracy of our results . we investigate the influence of the strength of the inter - particle interaction and the position of the magnetic ion on the ground state of the system . we predict the interesting phenomenon that the magnetic ion ferromagnetically couples with the electrons in a region below a critical magnetic field and antiferromagnetically with the electrons above this critical field . thermodynamic properties as magnetization , susceptibility , and the heat capacity are investigated as function of magnetic field and temperature . ) , and the coulomb interaction strength @xmath5 . the inset is the first derivative of @xmath6 with respect to @xmath7 that highlights the non - linearity of @xmath6 in certain regions of the magnetic field.,width=347 ] this paper is organized as follows . section ii introduces the model and the numerical method . in section iii , we present our numerical results for the many - particle ground state and investigate correlations through the appearance of vortices in the many - electron wave function . iv addresses the many - particle spectrum and in sect . v we present results for different thermodynamic quantities . our discussion and conclusions are presented in sect . a quantum dot containing @xmath2 electrons with spins @xmath8 confined by a parabolic potential and interacting with a single magnetic ion ( @xmath1 ) with spin @xmath9 and a magnetic field is described by the following hamiltonian : @xmath10 + \sum_{i < j=1}^{n_e}\frac{e^2}{4\pi\epsilon_0\epsilon\|\overrightarrow{r_i}-\overrightarrow{r_j}\| } \nonumber\\+ \frac{1}{2}\hbar\omega_c\left(g_e m^ * { s}_z+g_{mn } m^ * m_z\right ) -j_c\sum_{i=1}^{n_e}\overrightarrow{s_i}\cdot \overrightarrow{m } \delta(\overrightarrow{r_i}-\overrightarrow{r}).\end{aligned}\ ] ] the vector potential @xmath11 is taken in the symmetric gauge : @xmath12 where the magnetic field @xmath13 points perpendicular to the plane of the interface . the confinement frequency @xmath14 is related to the confinement length by : @xmath15 . @xmath16 and @xmath17 are the land g - factor of the host semiconductor and the magnetic ion , respectively . the dimensionless coulomb strength is defined as : @xmath18 with @xmath19 the effective bohr radius . the cyclotron frequency is : @xmath20 . @xmath21 , @xmath22 , and @xmath23 are the projections of the total angular momentum of the electrons , their spins , and the magnetic ion in the direction of the magnetic field . the electrons and the magnetic impurity in the quantum dot interact via the contact exchange interaction with strength @xmath24 . we use the set of parameters@xcite that is applicable to cd(mn)te which is a ii(mn)vi quantum dot with typical lateral size of about tens of nanometers . the dielectric constant @xmath25 , effective mass @xmath26 , @xmath27 @xmath28 , @xmath29 , @xmath30 , @xmath31 , and @xmath32 about tens of nanometers ( @xmath33 corresponding to tens of @xmath34 ) . for example , with @xmath35 @xmath34 gives @xmath36 . we rewrite the hamiltonian in second - quantized form : @xmath37,\end{aligned}\ ] ] where the first term is the single - particle energies @xmath38 for an electron in state @xmath39 with spin @xmath40 and the second term is the coulomb interaction . the third term is the electron and magnetic ion zeeman energy . the last sum is the electron-@xmath41 interaction in which the first term describes the difference between the number of spin up and spin down electrons , and the last two terms describe the energy gained by flipping the electron spin along side with flipping the spin of the magnetic ion . @xmath42 , and @xmath43 are the z - projection , raising and lowering operators , respectively , of the magnetic ion spin ( we consider @xmath44 ions which have a spin of size @xmath45=5/2 ) . the single - particle states in a parabolic confinement potential define a complete basis of fock - darwin orbitals @xmath46 and spin functions @xmath47 : @xmath48 with the fock - darwin orbitals : @xmath49 in hamiltonian ( [ e : secondquantized ] ) , @xmath39 denotes a set of quantum numbers @xmath50 with @xmath51 the radial and azimuthal quantum numbers , respectively . the effective length @xmath52 in the presence of a magnetic field is defined through the hybrid frequency @xmath53 where @xmath54{1+(\omega_c/2\omega_0)^2}$ ] . the single - particle orbital energy is given by : @xmath55 the interaction parameters between the electrons and the magnetic ion in the quantum dot is expressed : @xmath56 as a product of two fock - darwin orbitals calculated at the position of the magnetic ion . we construct the many - particle wave function following the configuration interaction ( ci ) method . @xmath57 where @xmath58 is the @xmath59-th state of the non - interacting many electron wave function determined by @xmath2 electrons with @xmath2 different sets of quantum numbers @xmath60 and the single scatterer with one of the six states of the @xmath44 . @xmath61=@xmath62 stands for the coordinates and spin of a single electron . in second quantization representation , the state @xmath58 , which is a slater determinant composed of single - electron states , can be translated into a ket vector @xmath63 grouping a total of @xmath2 electrons into @xmath64 electrons with z - component of spin up and @xmath65 electrons with z - component of spin down : @xmath66 here @xmath67 and @xmath68 are the indices of the single - electron states for which each index is a set of two quantum numbers ( radial and azimuthal quantum numbers ) , as mentioned above . ) and @xmath69 . the upper inset is the first derivative of the main plot with respect to @xmath7 to visualize more clearly the cusps . the lower inset is a zoom of the rectangular regions of the main plot . ( b ) the z - projection of the total spin of the magnetic impurity ( solid black curve ) , of the electrons ( dashed red curve ) , and the total angular momentum ( dash - dotted green curve).,width=340 ] the ci method , which is in principle exact if a sufficient number of states are included , is limited to a small number of electrons due to computational limitations . for a larger number of electrons and/or magnetic ions , other approaches that e.g. are based on spin - density functional theory ( sdft ) using e.g. the local spin density approximation ( lsda ) as was used in ref.@xcite is able to handle a large number of electrons and/or magnetic ions . the lsda is exact only in the case of the homogeneous electron gas , and in practice , works well also in most inhomogeneous systems . however , in really highly correlated few - particle systems as discussed in this paper , the lsda might fail or be at least less accurate . and the coulomb strength @xmath69 . the upper inset is a zoom of the small magnetic field region and the lower one presents the average of @xmath23 , @xmath22 , and @xmath21 for the ground state at very small magnetic fields.,width=347 ] we first explore the magnetic field dependence of the total zeeman energy : @xmath70 which is the difference in energy in the presence and without a magnetic ion . it consists of three terms : the difference of the zeeman energy of the electrons @xmath71 , the zeeman energy describing the interaction of the magnetic ion having spin @xmath72 with the magnetic field , @xmath73 , and the exchange interaction of the ion with the electrons , @xmath74 . @xmath75 is just the so called local zeeman splitting term " as discussed in ref.@xcite . this sum is basically the difference in the zeeman energy of the electrons between the cases with and without a magnetic ion plus the energy contribution of the magnetic ion . , @xmath76 , and @xmath77 for the coulomb strength @xmath69 . the horizontal line at @xmath78 separates the @xmath75 plane into ferromagnetic ( fm ) and antiferromagnetic ( afm ) regions . , width=340 ] for @xmath79 , we find a total zeeman energy that appears linear in magnetic field . a similar linear behavior is also found in the cases with @xmath80 but with different slopes ( see fig . [ localzeeman1234e ] ) . let us suppose , in hamiltonian ( [ e : hamiltonian ] ) , that the contribution from the last term ( the local zeeman energy or the exchange interaction term ) is zero . for instance this is the case when a magnetic ion is located at the center of the quantum dot having three - electrons in the partially filled @xmath81 shell , in which the first two - electrons fully fill the @xmath82 shell and the remaining one is in either of the orbitals of the @xmath81 shell . then a perfect linear behavior of the total zeeman energy is found . a closer look to @xmath6 gives us a slightly different picture , as is provided by taking the derivative ( see the inset of fig . [ localzeeman1234e ] ) . notice that the total zeeman term @xmath6 has pronounced cusps and the number and positions of these cusps is different for different number of electrons @xmath2 . there exists one at @xmath83 for a two - electron quantum dot , one at @xmath84 for a three - electron quantum dot , and two at @xmath85 and @xmath86 for the four - electron quantum dot , with the magnetic ion located at @xmath87 . the three - electron system exhibits a much richer behavior when we increase the coulomb interaction strength to @xmath69 as seen in fig . [ zeeman3e0.5l0andlzmzsz](a ) where we placed the magnetic ion at @xmath76 . cusps , which are highlighted in the two insets in fig . [ zeeman3e0.5l0andlzmzsz](a ) , appear when the total angular momentum and/or the total z - projection of the spin of the electrons change abruptly with magnetic field . notice that the total zeeman energy of a two - electron quantum dot in the presence of the magnetic ion does not produce a similar behavior due to the fact that the z - projection of the total spin is zero making the main contribution ( from the zeeman spin term of the @xmath44-impurity ) negligible . . the inset in the upper plot are the averages of the three quantities : @xmath23 , @xmath22 , and @xmath21 with the ion located at @xmath76 at small magnetic fields . the horizontal line separates the plane into ferromagnetic ( fm ) and antiferromagnetic ( afm ) regions in both two plots . the schematic diagram in ( b ) explains why the exchange energy is almost zero in the case @xmath88 . , width=340 ] the coulomb strength and the position of the magnetic impurity affects the total zeeman energy and influences the number and the position of the cusps . the first pronounced cusp appears at lower magnetic field for larger coulomb interaction strength . this is a consequence of the competition between the coulomb energy and the energy gap of the single - particle problem . larger coulomb strength ( smaller energy gap ) leads to stronger electron - electron correlation and consequently the electrons are more clearly separated from each other . it results into a high probability for finding the electrons to occupy higher energy states . that also means that the system transfers to a configuration with larger @xmath22 and @xmath21 at smaller applied field . . the horizontal line separates the plane into ferromagnetic ( fm ) and antiferromagnetic ( afm ) regions . the inset shows the total @xmath89-projection of the @xmath23 , @xmath22 , and @xmath21 of the electron system with the magnetic ion located at the center of the dot.,width=340 ] . the horizontal line separates the plane into ferromagnetic ( fm ) and antiferromagnetic ( afm ) regions . the inset shows @xmath22 alone vs. magnetic field.,width=340 ] we also found that the ground - state energy is sensitive to the presence of the magnetic field . in zero magnetic field , the ground state receives contributions from many different configurations having z - projection of the magnetic ion , @xmath23 , from @xmath90 to @xmath91 . when a magnetic field is applied , the ground state favors states with projection of the spin of the magnetic ion down and the states with @xmath92 give the main contribution to the ground state . , @xmath93 , and @xmath4 at @xmath94 . the horizontal dotted line at @xmath78 separates the plane into ferromagnetic ( fm ) and antiferromagnetic ( afm ) regions.,width=347 ] fig . [ zeeman3e0.5l0andlzmzsz](b ) shows the average of the three quantities @xmath23 , @xmath22 , and @xmath21 of the three - electron quantum dot as a function of the magnetic field for a coulomb interaction strength @xmath69 . we realize that with increasing @xmath95 the @xmath96 ( @xmath97 ) exhibits jumps at smaller critical @xmath7 ( compare green dash - dot - dot curve in fig . [ localzeeman1234e ] ) . now we direct our attention to the very small magnetic field behavior . there exists a very small region of the magnetic field where the total spin of electrons and the total spin of the magnetic ion are oriented parallel , we found this earlier in ref.@xcite for the zero magnetic field case . these results are now extended to nonzero magnetic field . this is made more clear in the upper inset of fig . [ twoterms_3e_0.5l0 ] where the crossing point of the two terms : the zeeman energy of the magnetic ion ( @xmath98 ) and the exchange interaction ( @xmath99 ) , occurs at @xmath100 ( converted to @xmath101 for the considered system ) . this ferromagnetic coupling extends further , up to @xmath102 ( see the lower inset of fig . [ twoterms_3e_0.5l0 ] ) . ( black squares referred to ( 1 ) ) and ( @xmath103 ) ( red circles referred to ( 2)).,width=347 ] it is worth noticing that this ferromagnetic coupling is extended to a much larger magnetic field range ( up to @xmath104 ) if we move the magnetic ion to the center of the quantum dot ( see fig . [ exchange_3e_pos ] ) . this can be understood as follows . when the magnetic ion is located at the center of the dot and the magnetic field is very small the absolute value of @xmath98 always dominates over @xmath99 . this is opposite to the case when the magnetic ion is located at @xmath76 . recall that in ref.@xcite we found for zero field that the exchange zeeman energy is minimum when the magnetic ion is at the center of the quantum dot and approximately zero at positions very far from the center of the quantum dot . ( black squares referred to ( 1 ) ) and @xmath105 ( red circles referred to ( 2 ) ) . , width=347 ] fig . [ exchange_3e_pos ] tells us that the magnetic field where the antiferromagnetic coupling between electron and the magnetic ion starts depends on the position of the magnetic ion in the quantum dot . the system with the magnetic impurity located at @xmath77 exhibits an antiferromagnetic coupling for @xmath106 , that is larger than @xmath107 in case of @xmath76 . we have discussed the appearance of antiferromagnetism in a three - electron quantum dot . now we go back to the two simpler cases with the number of electrons @xmath108 ( see fig . [ exc1e2e ] ) . let us first discuss the results for @xmath79 as given in fig . [ exc1e2e](a ) . the antiferromagnetic coupling between the electron and the magnetic ion starts at smaller magnetic field as the magnetic ion is moved . this is different from the previous results for @xmath109 . the reason is as follows : for the quantum dot with a single electron , the electron tends to accommodate permanently the @xmath82-shell with @xmath110 ( see the inset of fig . [ exc1e2e](a ) ) in the ground state while the exchange parameter in the @xmath82 shell ( @xmath111 ) is found to be maximum right at the center of the quantum dot . moving the magnetic ion away from the center of the dot , this @xmath111 is found to be smaller and as a consequence the exchange electron-@xmath44 interaction becomes smaller than the electron zeeman energy , leading to an antiferromagnetic coupling at smaller magnetic field . the story for @xmath88 electrons ( see fig . [ exc1e2e](b ) ) is now interesting since the two electrons accommodate the @xmath82 shell with spins antiparallel making the total spin of the electron zero in the ground state with almost unit probability . this leads to zero contribution to the first term written in the last line in [ [ e : secondquantized ] ] for diagonal elements . therefore , the main contribution ( even very small ) to the exchange energy is now expected to come from coupling with configurations where one of the electrons ( spin down ) stays in the @xmath82 level and the other occupies higher level ( see the schematic diagram in fig . [ exc1e2e](b ) ) . in this diagram , the magnetic ion is assumed to be located at the center of the dot with spin down ( @xmath90 ) . the coupling of the electron ( spin up ) in the @xmath82 orbital with an electron from either of the @xmath81 shell is zero . the only non - zero coupling is with an electron with the quantum numbers ( 1,0 ) of the @xmath112 shell ( as shown in the diagram and this quantum state would change if the ion is located away from the center of the dot ) with the amount of about @xmath113 . this picture remains valid until the magnetic field is high enough to excite one electron from the @xmath82 shell to a higher quantum state forming the ground state with two up spins antiferromagnetically coupling with the magnetic ion . for smaller coulomb interaction strength , the antiferromagnetic behavior occurs at larger magnetic field since the two electrons repel each other less and consequently they stay longer antiparallel in the @xmath82 shell . calculated for a three - electron quantum dot with the magnetic ion ( white cross ) located at @xmath76 in ( a ) , ( b ) and @xmath114 in ( c ) when the coulomb strength @xmath69 in a magnetic field @xmath115 ( see ( a ) ) and @xmath116 ( see ( b ) and ( c ) ) . ( d ) is the magnetic ion position dependence of @xmath74 with the magenta circles indicating the respective calculated positions of the magnetic ion in ( a ) , ( b ) , and ( c ) . red color region is for highest value of radial density.,width=340 ] from the @xmath88 result we may ask a question : whether a four - electron quantum dot has similar properties since the number of electrons in both cases are even and one may have a situation where the total spin of the electrons is zero . indeed , if the magnetic ion is located at the center of the quantum dot even though for @xmath117 the outer shell is half - filled , this is possible as illustrated in fig . [ exc4e ] where the antiferromagnetic coupling occurs at @xmath118 at which the total spin of the electrons reaches the maximum value @xmath119 . in this case , the first two electrons will occupy the @xmath82 shell and the remaining two will occupy two of the five orbitals of the @xmath81 and @xmath112 shells . this picture holds at small magnetic field . however , there is a big difference in the exchange energy as compared to the previous case of @xmath88 when the ion is shifted away from the center of the dot , e.g. in this plot at @xmath76 . the exchange energy is much larger than the result obtained for @xmath120 because when the ion is out of the center of the quantum dot , the two remaining electrons at higher orbitals have a non - zero contribution in the diagonal exchange elements dominating the exchange energy of the ground state . this is the reason why the antiferromagnetic transition occurs at smaller magnetic field ( @xmath85 ) as compared to the case when the ion is at the center of the dot ( @xmath118 ) although the pictures of the @xmath23 and @xmath21 transition in these cases are similar . to complete the picture for few - electron quantum dot system , we will discuss the afm behavior for the system with the highest number of electrons , @xmath121 , where we were able to obtain accurate numerical results . we focus on the small magnetic field region ( see fig . [ exc5e ] ) . for the magnetic ion in the center of the dot , the fm coupling is dominant in the shown magnetic field region because the diagonal exchange matrix elements dominate over the zeeman energies of the electrons and of the magnetic ion . this is different for the cases with the magnetic ion displaced a bit from the center of the quantum dot . the fm - afm occurs at @xmath122 and @xmath123 for @xmath124 and @xmath125 , respectively . it is similar to the cases for the system with @xmath126 due to the zero coupling between the orbitals from the @xmath127 shell with the @xmath82 orbital . to observe the afm behavior for the system with the magnetic ion located at the center of the dot where the diagonal exchange elements are almost zero , it is crucial to include enough quantum orbitals ( that rapidly increases the size of the hamiltonian matrix resulting in very time consuming calculations ) so that one allows the electrons to jump to higher energy levels and having parallel spins as previously shown for the case @xmath117 ( see fig . [ exc4e ] ) . in that case , the four - electron system exhibits an anti - ferromagnetic coupling with the magnetic ion at the magnetic field where the total @xmath89-projection of the spin is maximum @xmath128 . the system is strongly polarized . for the case @xmath121 , up to @xmath129 , the total @xmath130 and the total @xmath131 . the inset in fig . [ exc5e ] supports the afm behavior for the out of center @xmath44 as obtained in the main plot . now we change the coulomb interaction strength and explore the magnetic behavior as function of the position of the magnetic ion . from fig . [ exchangepos_3e_severallamb ] , it is clear that when reducing the coulomb interaction the system undergoes a ferromagnetic to antiferromagnetic transition at gradually larger magnetic fields for @xmath1 positions that are closer to the center of the dot . we see that @xmath132 has a peak structure with a maximum at some specific position of the magnetic ion , e.g. see the peak for the case @xmath5 ( the blue full circles ) . however , it is certain that at high magnetic field , the system is always antiferromagnetic . the fm - afm phase diagram for a three - electron quantum dot in ( @xmath133 ) space is shown in fig . [ pdlamb - omegac ] for two different positions of the magnetic ion . when the magnetic ion is in the center of the quantum dot ( black curve with squares in fig . [ pdlamb - omegac ] ) the critical magnetic field increases as the coulomb interaction strength decreases . the reason is that increasing the coulomb interaction helps the electrons to approach closer the magnetic ion and therefore the critical magnetic field for the system to transit to the antiferromagnetic phase decreases . calculated for the quantum dot of fig . [ denlambda1 ] where one electron ( orange arrow ) is pinned at one of the three most - probable positions of the electrons as obtained in fig . [ denlambda1 ] . it is at @xmath134 in ( a ) for fig . [ denlambda1](a ) , @xmath135 in ( b ) for fig . [ denlambda1](b ) , and @xmath136 in ( c ) for fig . [ denlambda1](c ) , respectively . ( b ) the same correlation function for the case the position of the fixed electron is closer to the ion ( at @xmath137 ) as compared to ( b ) . the position of the magnetic ion is indicated by the white cross.,width=294 ] now we move the magnetic ion away from the center of the quantum dot and we obtain the phase diagram as shown by the red curve ( with solid circles ) in fig . [ pdlamb - omegac ] . for @xmath138 , the stability of the fm phase with respect to an applied magnetic field is strongly reduced and a small magnetic field turns the three - electron system into the afm phase . notice that for sufficient strong electron - electron interaction ( i.e. @xmath139 ) we obtain practically the same fm - afm phase diagram as for the case the @xmath44-ion is located in the center of the quantum dot . a remarkable re - entrant behavior is found in the region @xmath140 and @xmath141 where with increasing @xmath95 we go from an antiferromagnetic to ferromagnetic and back to antiferromagnetic phase . this unusual behavior is understood as follows . as the impurity is moved away from the center of the quantum dot the exchange matrix will have many nonzero off - diagonal terms that leads to a smaller fm - afm critical transition magnetic field . now let us turn our attention to the region @xmath138 . for very small coulomb interaction strength the electrons will repel each other only weakly and are therefore pulled towards the magnetic ion ( the nonzero exchange matrix elements increase strongly ) resulting in a very small fm - afm magnetic field . for @xmath139 , the electrons become more strongly correlated and the critical field stays about @xmath142 from the result for @xmath143 . if one moves the ion further and further away from the center , the @xmath144 transition line moves to larger @xmath95 values . for example , for @xmath69 and the magnetic ion located at @xmath76 the fm - afm critical transition occurs at @xmath145 , which is much smaller than @xmath146 found for @xmath147 . and the coulomb interaction strength is @xmath5.,width=359 ] the dependence of the ferromagnetic - antiferromagnetic transition of a three - electron quantum dot system on the position of the magnetic ion is summarized in the phase diagram shown in fig . [ pdlamb - rmn ] for two different magnetic fields @xmath148 . we can predict that with slightly larger ( smaller ) magnetic field the slope of the curve will be larger ( smaller ) . from fig . [ pdlamb - omegac ] , we already learned that the fm - afm transition magnetic field is largest for the ion in the center of the quantum dot as is also seen in fig . [ pdlamb - rmn ] . the re - entrant behavior of the afm phase as function of @xmath95 is found for small @xmath149 values , i.e. when the @xmath44-ion is not too far from the center of the quantum dot , in case the magnetic field is not too small . the critical point ( @xmath150)=(@xmath151 ) for @xmath105 moves down ( up ) with increasing ( decreasing ) magnetic field . in high magnetic field , the magnetic ion tends to attract electrons because they are oppositely polarized . because the exchange interaction is small as compared to the coulomb interaction , the electrons and magnetic ion are arranged in such a way that the electrons repel each other and also try to be as close to the magnetic ion as possible . this picture holds above the fm - afm critical magnetic field . to show this behavior explicitly , we studied the radial density and the radial pair correlation functions . their respective operators are defined as : @xmath152 and @xmath153 we plot in fig . [ denlambda1 ] the radial density of a three electron quantum dot that is polarized in high magnetic field for the case that the coulomb strength is @xmath69 and the magnetic ion is located at two different positions for two magnetic fields . the electrons and the magnetic ion are antiferromagnetically coupled . the strength of that coupling can be seen from fig . [ denlambda1](d ) in which we plot the magnetic ion s position dependence of the exchange energy at two magnetic fields @xmath154 and @xmath155 . those magnetic fields are typical in the sense that the exchange term is found to be very large ( @xmath115 ) or the correlation between the electrons very high ( @xmath116 ) . density plots are shown for @xmath156 at @xmath76 and ( @xmath157 ) . we observe three distinct peaks of maximum probability . they are found at : @xmath134 , @xmath158 , and @xmath159 in fig . [ denlambda1](a ) ; @xmath135 , @xmath160 , and @xmath161 in fig . [ denlambda1](b ) ; @xmath136 , @xmath162 , and @xmath163 in fig . [ denlambda1](c ) . these figures show clearly the interplay effect where the three electrons on the one hand try to be close to the magnetic ion and on the other hand repel each other via the coulomb potential energy . it results in the merging of the radial density such that the higher the exchange energy the larger the merging of the local maxima in the electron density and the smaller the correlations . [ denlambda1](d ) gives an idea about the variation of @xmath74 with the position of the magnetic ion and reaches a maximum at @xmath76 for @xmath116 . in fig . [ denlambda1](c ) , the three electrons are less attracted to the magnetic ion via the antiferromagnetic coupling as compared to that in fig . [ denlambda1](b ) . this is due to the fact that the @xmath74 for the case shown in fig . [ denlambda1](b ) is larger than that in fig . [ denlambda1](c ) . the electrons are therefore found more correlated in the latter case presented by the extended red region in fig . [ denlambda1](c ) . thereby , correlation between electrons in fig . [ denlambda1](c ) is expected to be the highest and in fig . [ denlambda1](a ) the smallest . the position of the magnetic ion affects the ground - state property as is made clear in fig . [ paircorrpos ] . we fix the spin state and the position of one electron ( indicated by the orange arrow ) and the position of the magnetic ion ( the white cross ) . the magnetic field is such that @xmath115 in fig . [ paircorrpos](a ) and @xmath116 in the others . it also reflects the fact that the system in fig . [ denlambda1](a ) exhibits the smallest correlation as compared to the other two . this illustrates the point raised above about the density . at the magnetic field @xmath116 , the electrons are strongly polarized resulting in the red regions of the up - up spin pair correlation function that tends to surround the magnetic ion . we see that the three electrons are most likely to localize around some specific positions defining a triangle with the three electrons at the three vertices , while they are attracted to the magnetic ion . when we locate one electron at a position closer to the magnetic ion , see fig . [ paircorrpos](b ) , the two peaks decrease in amplitude as compared to those in fig . [ paircorrpos](b ) . the addition energy ( often called the chemical potential ) is defined as the increase of the energy of the quantum dot system when an electron is added : @xmath164 . this quantity can be measured experimentally and is plotted in fig . [ chimicalpot ] as function of the magnetic field . with the magnetic ion ( the white cross ) located at @xmath76 in ( a),(b ) and at @xmath114 in ( c ) for coulomb strength @xmath69 . two fixed electrons ( indicated by the two magenta circles ) are located at the two peaks appearing in the radial density : @xmath135 and @xmath160 in ( a ) , ( c ) ; and @xmath135 and @xmath161 in ( b ) . , width=347 ] there are several cusps appearing in the addition energy curves as a consequence of changes in the ground state . these changes are due to variations in the z - projection of the total spin of the electrons and/or the z - projection of the total angular momentum of the system when the magnetic field increases beyond some specific values . the presence of the magnetic ion leads to more cusps and the position of these cusps is also influenced by the number of electrons and the position of the magnetic ion . the cusps are from either of the two systems in the study . for instance , the green triangles in fig . [ chimicalpot ] are for @xmath165 has two cusps at @xmath83 and @xmath84 . the cusp at the point @xmath83 comes from the change in the configuration of the average of the total z - projection spin and the total z - projection of angular momentum @xmath166 of the two - electrons in the quantum dot from ( 0,0 ) to ( 1,1 ) . while the other cusp @xmath84 comes from the change of the phase of the three - electron quantum dot from @xmath167 to @xmath168 . it is similar to the case for @xmath169 ( the blue left pointing triangles ) where the cusp appears at @xmath84 . at this point , we observed a change from configuration ( @xmath170 ) to ( @xmath171 ) . the remaining one , @xmath172 , is from the four - electron case when its configuration changes from ( @xmath171 ) to ( @xmath173 ) . another way to obtain information on the correlations that are present in the many - particle wave function is to investigate the vortex structure . at a vortex the many - body wave function is zero and is characterized by a change of phase of @xmath174 when we go around this point . levels of the three ( the top ) , four ( the middle ) , and five ( the bottom ) electron quantum dot with the magnetic impurity located at @xmath76 . the coulomb strength is @xmath69 with magnetic fields @xmath175 ( the blue squares ) and @xmath94 ( the magenta circles ) . the inset in the top figure shows the average of @xmath23 ( the dark - blue triangles ) and @xmath22 ( the red triangles ) as a function of @xmath7 of the sixth energy level.,width=321 ] the zeros of the wave function are similar to flux quanta when e.g. the wave function corresponds to the order parameter in a superconductor . the fixed electrons and the zero of the wave function follows closely the displaced electron and one may say that the electron plus its zero form a composite - fermion object . the composite - fermion@xcite ( and references therein ) is a collective quasi - particle that consists of one electron bound to an even number of vortices ( flux quanta ) . the composite - fermion concept introduces a new type of quasi - particle that is used to understand the fractional quantum hall effect in terms of the integer quantum hall effect of these composite fermions . to obtain the zeros of the wave function of the system with @xmath2 electrons , we fix @xmath176 electrons at some positions inside the quantum dot and leave the remaining one free . the resulting reduced wave function gives the probability to find the remaining electrons at different positions in the quantum dot and the zeros of this function are those points where the phase of the wave function changes by @xmath174 . as an example , we investigate the situation of a three - electron quantum dot . levels are shown ) of the one - electron quantum dot with the magnetic impurity located at @xmath76 and @xmath69 ( the inset is a zoom of the first @xmath177 levels at low magnetic fields and the inset in the inset shows the first @xmath178 levels , it is a zoom of the rectangular region.,width=347 ] figs . [ vortice_lamb1 ] shows the vortex pictures of a three - electron quantum dot containing a magnetic ion located at positions ( see the white cross ) that are identical to its positions in fig . [ denlambda1](a ) . two , among three , electrons are fixed at the respective peaks in the electron density . red and black regions are referring to the highest ( 2@xmath179 ) and lowest phase @xmath180 , respectively . those plots show that there are always two vortices near the pinned electrons positions . for example , the number of vortices pinned to each electron in the case @xmath109 at @xmath181 is @xmath4 describing the system at filling factor @xmath182 . notice also that one of the vortices appears to be pinned at a position very close to the @xmath44 ion . we realize that moving the magnetic impurity to a different position changes the relative positions of the vortices that are pinned to the electrons with respect to one another , as shown in figs . [ vortice_lamb1](a ) and ( c ) . as the electrons are antiferromagnetically coupling to the magnetic ion this kind of movement consequently depends on the position of the magnetic ion . but now for the two - electron quantum dot . the inset in the inset is a zoom of levels @xmath155th to @xmath177th for magnetic field close to zero.,width=347 ] in the case @xmath69 , we found that the average of the maximum z - projection of the total angular momentum is @xmath183 and the two vortices appear at the external field @xmath184 . apparently , the larger @xmath95 the smaller @xmath7 for which the first two vortices appear at the pinned electrons . in the presence of an external magnetic field , the many - fold degeneracy of the energy spectrum of the system is lifted . fig . [ spectrum3eimp0.5l0](a ) illustrates that point for the case of three electrons . in the absence of the interaction between the electrons and the magnetic ion and in the absence of a magnetic field ( blue squares ) , the energy spectrum is @xmath155-fold degenerate for the first @xmath155 lowest energy levels , the next level is then @xmath185-fold degenerate , and the next @xmath155-fold degenerate , and so on . the origin of this was explained in ref.@xcite and is due to the coupling of the electrons and the magnetic ion . when the magnetic field is different from zero , see red circles in fig . [ spectrum3eimp0.5l0 ] , the degeneracy is lifted . in the inset of fig . [ spectrum3eimp0.5l0 ] we plot @xmath186 ( the magenta triangles ) and @xmath97 ( the dark - blue ones ) as a function of magnetic field for the sixth level . the average of @xmath186 and @xmath97 change abruptly as compared to those found for the ground - state energy , e.g. see fig . [ zeeman3e0.5l0andlzmzsz ] . @xmath23 and @xmath97 of the sixth state jump between two different values , e.g. @xmath187 and @xmath188 for @xmath186 and @xmath93 and @xmath189 for @xmath97 as function of the field . this is a consequence of anti - crossings of energy levels as will be apparent later . the result for four- and five - electron quantum dots are also shown in fig . [ spectrum3eimp0.5l0 ] . we see the degeneracy of @xmath190 , @xmath191 , @xmath192 , and @xmath178 for the first @xmath193 levels in the case @xmath117 and of @xmath155 , @xmath185 , @xmath155 , @xmath185 , @xmath185 , and level @xmath193 has the same degeneracy with the next energy level beyond the first @xmath193 for the case @xmath121 in @xmath194 tesla . but now for the three - electron quantum dot.,width=347 ] to have a clearer picture of the energy spectrum of the quantum dot system we plot in figs . [ spectrum - b1eimp0.5l0 ] , [ spectrum - b2eimp0.5l0 ] , and [ spectrum - b3eimp0.5l0 ] the magnetic field dependence of the first @xmath195 energy levels for @xmath196 , respectively . the spectra at small magnetic fields is enlarged ( see insets ) to see the zeeman splitting and the nearly - linear behavior of the energy levels . remember that this is due to the coupling of the electron spins with the magnetic ion spin . for @xmath79 the first two levels for @xmath194 are @xmath155- and @xmath185-fold degenerate ( @xmath155-fold degenerate is due to the ferromagnetic coupling of the @xmath82-shell electron spin @xmath197 with the magnetic ion spin @xmath91 and @xmath185-fold degenerate of that electron now with spin @xmath198 to the magnetic ion with spin @xmath91 ) , respectively . a closer inspection ( see fig . [ spectrum - b1eimp0.5l0 ] ) tells us that these @xmath178-levels are exchange split into two bundles of @xmath155- and @xmath185-levels ( the inset in the inset in fig . [ spectrum - b1eimp0.5l0 ] ) . notice that there is a first large energy gap at very small fields between the first @xmath178 levels and the next @xmath177 ones as seen in fig . [ spectrum - b1eimp0.5l0 ] while that kind of gap appears between the first @xmath191 and the next @xmath199 for @xmath88 ( fig . [ spectrum - b2eimp0.5l0 ] ) . for @xmath109 ( fig . [ spectrum - b3eimp0.5l0 ] ) , this kind of gap appears after the first @xmath177 energy levels . the origin is the coupling of the third electron , which can reside at either two states of the @xmath81 shell while the @xmath82 shell is already fully filled , with the magnetic ion with @xmath191 @xmath89-components of the spin at very small fields , i.e. the intra - shell ( @xmath81 ) exchange interaction . for @xmath88 the electron ground state corresponds with a filled @xmath82-shell , i.e. @xmath200 , and therefor for @xmath194 only a @xmath191-fold degeneracy , as shown in fig . [ spectrum - b2eimp0.5l0 ] , is found due to the @xmath89-component of the @xmath44-spin . the next level is @xmath190-fold degenerate at @xmath194 ( @xmath190 come from the ferromagnetic coupling of the two - electron system with total spin @xmath201 to the magnetic ion with spin @xmath91 ) ( see the inset in the inset in fig . [ spectrum - b2eimp0.5l0 ] ) , etc . for three values of the coulomb strength @xmath5 , @xmath93 , and @xmath202.,width=328 ] with increasing magnetic field , we see that for @xmath79 there is periodically an opening of energy gaps in the spectrum . similar energy gaps have been found earlier ( as an example see e.g ref.@xcite ) for a quantum dot without a magnetic impurity and are a consequence of the electron with two - fold spin degeneracy filling the equally - gaped - energy single - particle quantum states with different sets of the radial and angular quantum numbers . notice that for @xmath203 , these gaps have disappeared . and @xmath5.,width=328 ] the spectra exhibit a lot of crossings and anti - crossings , the number of them has increased as compared to the quantum dot case without a magnetic ion because of the zeeman splitting of the @xmath44-spin . when the applied field increases the gaps in the spectrum of @xmath79 are still open and appear more often than in the cases of @xmath203 . once again , we see a lot of cusps in the energy levels and that reminds us to abrupt changes in the configuration of the system with magnetic field as discussed before for the ground state . we first calculate the magnetization and susceptibility of the system : @xmath204 and @xmath205 at zero temperature . the magnetization of a quantum dot with the magnetic ion located at @xmath76 with @xmath203 electrons is plotted in fig . [ mag2and3e ] . we see several jumps that are a consequence of changes in the ground state , e.g. changes in @xmath96 ( see previous section ) . for example , the magnetization of the three - electron quantum dot as plotted in fig . [ mag2and3e](b ) for the case @xmath5 and the magnetic ion at @xmath76 has a step at @xmath206 . consequently , the susceptibility also has a peak at @xmath206 . the same thing happens at @xmath207 and @xmath208 for @xmath209 in the magnetization and the susceptibility . for non - zero temperature , the temperature dependence of the magnetization and susceptibility is defined by : @xmath210 , @xmath211 , respectively . the statistical average @xmath212 is calculated as : @xmath213 where the sum is over the energy levels as displayed in e.g. fig . [ spectrum3eimp0.5l0 ] . these quantities are explored in figs . [ magtempo3e ] for @xmath109 and a few different temperatures ( including the zero temperature case ) . with increasing temperature the jumps become smoother . a very low magnetic field peak shows up because for @xmath214 we have @xmath215 at @xmath216 . an important quantity that is related to the storage of energy is the heat capacity : @xmath217 the heat capacity is investigated as a function of the coulomb strength @xmath218 , temperature @xmath219 , magnetic field , and the position of the impurity @xmath156 . and the magnetic field @xmath220 ( the three lower curves ) and @xmath221 ( the three upper curves).,width=328 ] we plot in fig . [ cvseveralpos_lamb1 ] the specific heat for two values of the magnetic field , i.e. @xmath220 and @xmath222 , and three typical positions of the magnetic ion . for weak fields , the three electrons start to polarize and we see that the position of the main peak moves towards higher temperature as the magnetic ion is moved away but not too far from the center of the dot . for the high magnetic field case the three electrons are strongly polarized and we see a different behavior in the shift of the main peak . this results from the change of the statistical average of the energy levels at different fields . , @xmath5 , at several temperatures : @xmath223 , @xmath224 , @xmath225 and @xmath226 k.,width=302 ] now we examine the behavior of the heat capacity at a specific temperature as function of magnetic field . figs . [ cv - b_lamb0.4_severalt ] , [ cv - btemp=5.8and11.6 ] , and [ cv - bseveralt_4e ] are the plots of the magnetic field dependence of the heat capacity of three ( at two coulomb interaction regimes ) and four - electron quantum dots at some specific temperatures and two different @xmath227 . the peak at small magnetic fields broadens and moves to higher fields with increasing temperature . the heat capacity exhibits a number of peaks and a clear minimum around e.g. @xmath228 as shown in figs . [ cv - b_lamb0.4_severalt ] . remember that this field corresponds to a cusp in the energy versus magnetic field behavior as discussed previously in subsection iii.a . at very low temperatures , this cusp still affects the heat capacity through the sharpness of the minimum as shown in the figure and this gradually becomes small at high temperatures . in fig . [ cv - btemp=5.8and11.6 ] , we see a very interesting behavior of the heat capacity at @xmath229 : the single peak becomes a double peak as temperature increases from @xmath230 to @xmath231 . this is due to the cusps now occurring around this field in the low - energy levels of the spectrum of the three - electron quantum dot system as observed in fig . [ spectrum - b3eimp0.5l0 ] . besides , the structure of the heat capacity is more complex ( more peaks ) with increasing @xmath95 . this is made clear if one looks back to the previous discussion related to figs . [ localzeeman1234e ] and [ zeeman3e0.5l0andlzmzsz](a ) . , @xmath69 , and the temperatures are @xmath232 and @xmath233.,width=321 ] for the case @xmath117 , the heat capacity exhibits more peaks as compared to the case @xmath109 and the behavior of the peaks with increasing temperature is also very different . temperature affects the heat capacity of the system in the sense that it increases the peak values and separates them in magnetic field . the coulomb interaction strength changes the structure of the magnetic field dependence of the heat capacity and is illusive through figs . [ cv - b_lamb0.4_severalt ] and [ cv - btemp=5.8and11.6 ] . the peak of the heat capacity for the case with smaller coulomb interaction strength appears at higher magnetic field as compared to the case with larger one . due to the presence of the magnetic ion ( and electron - electron interaction ) , electrons in the ground state do not always completely polarize in the presence of an external magnetic field . the configurations are mixed consisting of electrons having spins up and down . but for very large magnetic field , the magnetic ion tends to pull the electrons closer to the ion forming a ring - like electron density profile . these are the consequences of the interplay of several effects such as the zeeman effect ( on the electrons and the magnetic ion s spins ) , the coulomb repulsion , the spin exchange interaction , etc . this competition results in a crossover from ferromagnetic- to antiferromagnetic coupling between the electrons and the magnetic ion at some specific magnetic field . interestingly , this effect is observed to appear at higher magnetic field when we move the magnetic ion further from the origin of the quantum dot . a re - entrant behavior of the fm - afm transition is found as function of the coulomb interaction strength when the magnetic ion is moved out ( but not too far ) from the center of the quantum dot . , @xmath5 plotted for several temperatures.,width=302 ] the energy levels exhibit cusps as function of the magnetic field which correspond to changes of the configuration of the system as expressed by the values of ( @xmath234 ) . these cusps move to lower magnetic field with increasing coulomb interaction strength . the number of cusps increases with increasing number of electrons . these cusps show up in the addition energy . the transformation of the electron system to those of composite fermions is studied . in high magnetic fields , the electrons attach an even number of quantized vortices which we made clear by examining the many - body ground - state wave function in the presence of a magnetic ion . unlike the case without a magnetic ion where all the vortices are tightly bound to the electrons , when we fix the electrons at different positions the system of vortices stays pinned to the electrons and moves with the electrons but the relative positions of the vortices are modified . the contribution of the local zeeman splitting energy to the total energy of the system in large external fields is very small as compared to the contributions from the other parts . however , a slight movement of the position of the magnetic ion inside the quantum dot affects the result , slightly . with increasing applied magnetic field , each time the system jumps to a different @xmath235 configuration leads to the appearance of a peak in the thermodynamic quantities as e.g. the susceptibility and the heat capacity . in the presence of the magnetic ion , the structure of peaks in the heat capacity changes with the position of the magnetic ion . as temperature increases , these peaks split into two peaks and become smoother . this work was supported by fwo - vl ( flemish science foundation ) , the eu network of excellence : sandie , and the belgian science policy ( iap ) . 99 j. k. furdyna , j. appl . phys . * 64 * , r29 ( 1988 ) . d. l. klein , r. roth , a. k. l. lim , a. p. alivisatos , and p. l. mceuen , nature ( london ) * 389 * , 699 ( 1997 ) . y. f. chen , j. h. huang , w. n. lee , t. s. chin , r. t. huang , f. r. chen , j. j. kai , and h. c. ku , appl . . lett . * 90 * , 022505 ( 2007 ) . k. chang , j. b. xia , and f. m. peeters , appl . . lett . * 82 * , 2661 ( 2003 ) . a. o. govorov , phys . b * 72 * , 075358 ( 2005 ) . paul i. archer , steven a. santangelo , and daniel r. gamelin , nano lett . * 7 * , 1037 ( 2007 ) . j. fernndez - rossier and l. brey , phys . * 93 * , 117201 ( 2004 ) ; l. besombes , y. lger , l. maingault , d. ferrand , h. mariette , and j. cibert , phys . 93 * , 207403 ( 2004 ) ; y. lger , l. besombes , l. maingault , d. ferrand , and h. mariette , phys . * 95 * , 047403 ( 2005 ) ; @xmath236 phys . b * 72 * , 241309(r ) ( 2005 ) ; l. maingault , l. besombes , y. lger , c. bougerol , and h. mariette , appl . 89 * , 193109 ( 2006 ) . t. schmidt , m. scheibner , l. worschech , a. forchel , t. slobodskyy , and l. w. molenkamp , j. appl . phys . * 100 * , 123109 ( 2006 ) . p. wojnar , j. suffczyski , k. kowalik , a. golnik , g. karczewski , and j. kossut , phys . b * 75 * , 155301 ( 2007 ) . j. fernndez - rossier and ramn aguado , phys . * 98 * , 106805 ( 2007 ) . y. lger , l. besombes , j. fernndez - rossier , l. maingault , and h. mariette , phys . lett . * 97 * , 107401 ( 2006 ) ; l. maingault , l. besombes , y. lger , h. mariette , and c. bougerol , phys . ( c ) * 3 * , 3992 ( 2006 ) ; m. m. glazov , e. l. ivchenko , l. besombes , y. lger , l. maingault , and h. mariette , phys . b * 75 * 205313 ( 2007 ) . m. b. tavernier , e. anisimovas , f. m. peeters , b. szafran , j. adamowski , and s. bednarek , phys . b * 68 * , 205305 ( 2003 ) ; m. b. tavernier , e. anisimovas , and f. m. peeters , phys . rev . b * 70 * , 155321 ( 2004 ) ; @xmath236 phys . rev . b * 74 * , 125305 ( 2006 ) ; t. stopa , b. szafran , m. b. tavernier , and f. m. peeters , phys . rev . b * 73 * , 075315 ( 2006 ) . shun - jen cheng , phys . b * 72 * , 235332 ( 2005 ) . fanyao qu and pawel hawrylak , phys . lett . * 95 * , 217206 ( 2005 ) ; _ ibid_. * 96 * , 157201 ( 2006 ) . nga t. t. nguyen and f. m. peeters , phys . b * 76 * , 045315 ( 2007 ) . ramin m. abolfath , pawel hawrylak , and igor zuti , phys . lett . * 98 * , 207203 ( 2007 ) . j. k. jain , phys . * 63 * , 199 ( 1989 ) ; @xmath236 phys . b * 41 * , 7653 ( 1990 ) ; @xmath236 * 42 * , 9193(e)(1990 ) . h. saarikoski , a. harju , m. j. puska , and r. m. nieminen , phys . lett . * 93 * , 116802 ( 2004 ) . s. tarucha , d. g. austing , t. honda , r. j. van der hage , and l. p. kouwenhoven , phys . * 77 * , 3613 ( 1996 ) .
we are extremely pleased to be able to add our contribution to this collection of works , by many eminent authors in a wide spectrum of fields , in honor of rolf landauer s lifetime of contributions to our understanding of the physical world . we will say a lot more about what those contributions have meant for us below , but we might note one important motivation which we have been given by rolf s many battles with the producers of conference books and special volumes such as this one ! rolf believes strongly that the written word should be taken seriously , and that contributions , seriously written , should not languish on an editor s shelf waiting for slaggart authors or testy co - editors to do their neglected duty . so we have , first and foremost , been assiduous in delivering this contribution to prof . datta by the announced deadline ! as will be evident in our technical discussion below , we are largely mathematical physicists . we would say that rolf generally takes a jaundiced view of such creatures ; for him , there is an absolute need for _ explanation _ and _ understanding _ of scientific truths in the full human sense , not in the purely formal and mathematical sense with which those of our species often content themselves . rolf has challenged us throughout our careers , from our green function studies ( as mere children ) of conductance fluctuations and aharanov - bohm effects in mesoscopic conductors , to our very recent brazen assertions about the efficacy of error correction techniques in quantum information processing he has challenged us to explain , with as much clarity and insight as we could muster , the basic underlying reasons why the assertions that we were making were really true . and , as we feel we have come to understand better as we struggle towards our own scientific maturity , it is in the answers to these questions , posed relentlessly to everyone ( we suspect ) whose article appears in this volume , and not the successful summations of a diagram series , from which true , significant scientific progress is produced . so , we thank rolf greatly for these promptings which he has given to our own work . as one specific illustration of the remarkable flowering of scientific progress which has resulted from the landauer way of thinking , we would like to tell the story , in which one of us ( ddv ) has been a player , of the development of the theory of quantum error correction . this theoretical area , which today has many ardent practitioners ( some more about them below ) , is remarkable in that it was believed to be a strictly non - existent subject as little as two years ago ( at this writing , september 1997 ) . rolf contributed to the belief that quantum error correction could not be done@xcite , as indeed it could not in the original conceptions of quantum computing . his message seemed to be compelling quantum information is a form of analog information . in many respects this statement is entirely correct : what we mean by a qubit is a two - state quantum system which , unlike the conventional bit , can be a continuum of possible states as specified by its wavefunction : @xmath0 here @xmath1 and @xmath2 are arbitrary complex numbers , apart from the normalization condition . ( actually , the permissible state of a qubit is more general ; it can be in a mixed state described by a density operator . ) the availability of this continuum of states is intrinsic to quantum computation ; it appears that any attempt to restrict the qubit to one of a discrete set of states will nullify the great potential capabilities of quantum algorithms . it would make no physical sense to do so in any case , since the unitary evolution of the quantum state under the action of a hamiltonian is naturally a continuous - time process . noting this feature in the earliest speculations about quantum computation , landauer challenged the workers in this infant community with the statement that quantum computation could not be error corrected , and thus lacked a crucial element that defines computation itself . he based these criticisms on the well known defect of classical analog computing : since all states of such a device , and of the quantum computer , are legal computational states , it is argued that there would be no way to distinguish a state to which some noise had been added . thus , there is no correction mechanism , and the computation immediately starts to drift off track . the imagery which has been used by landauer of the situation in digital computation is that of a particle moving along a track of the `` standard '' digital state , with very high potential - energy walls continually `` restandardizing '' the state as the computation proceeded ( by the movement of the ball in this potential - energy maze ) . rolf quite correctly saw no hint of restandardization in any of the quantum computer implementations which were initially discussed , and he offered up detailed criticisms of several of the schemes ; for instance@xcite , in a computation scheme proposed by benioff in which the computation advanced by the propagation of a wavepacket in a one - dimensional periodic potential , landauer pointed out that the phenomenon of one - dimensional localization made it exponentially likely that before completing , the wave packet / computation would be turned around by localization and relection . thus landauer s criticism hung as a bit of a pall over the earlier days of quantum computation , sprinkling a little rain on the otherwise cheerful and naive quantum computation parade . but landauer s criticisms had an extremely important effect ; it got a couple of very original minds like peter shor and andrew steane to _ think _ about whether restandardization _ could _ , despite appearances , be performed in quantum computation . ( berthiaume _ et al._@xcite and bennett _ et al._@xcite were also pursuing lines which was not too far distant from what turned out to be the correct one . ) part of the discovery which shor@xcite , steane@xcite , and then many others developed was a relatively obvious one , namely that quantum states could be _ encoded_. in classical information theory , coding just refers to the use of a string of bits to stand in for the value of one bit ( or perhaps a smaller block of bits ) . the idea is that the redundancy in this encoding allows errors ( at least some errors ) to be caught and repaired ; such is the standard practice in digital communications . it should , of course , be not at all obvious how redundancy can be of any use in quantum computation . redundancy is apparently not very useful in conventional analog computation ; in addition , since quantum states can not be `` cloned '' or copied@xcite , it would seem that even the simplest kind of redundancy is not even possible in principle . what shor and steane discovered was an ingenious way to use an entirely different fundamental quantum property , _ entanglement _ , in the service of redundancy and error correction . entanglement , introduced into quantum physics by schrdinger@xcite in 1935 , is at one level a fairly prosaic mathematical feature of the wavefunction of two ( or more ) particles . it refers to the fact that the composite wavefunction may not be expressible as the product of the states of the two individual particles . for example , for two qubits one may have a state like @xmath3 this inescapable feature of the fundamental principles of quantum physics has a variety of consequences which , depending on your point of view , are deep , profound , bizarre , ridiculous , or some combination of the four . ( indeed , schrodinger recognized the peculiarity of all this ; we highly recommend his brilliantly written articles@xcite from 1935 - 6 on the subject . ) one simple consequence was touched on above : the state of an individual particle can not be described by a pure state @xmath4 in general ; the density operator is given by tracing the full state @xmath5 over the other part of the quantum system . it may be said that all randomness in quantum physics , as described by the probability amplitudes in the density operator , arises from entanglement . so , what shor and steane started with was the idea that the logical @xmath6 and @xmath7 of a qubit could be coded as two orthogonal entangled quantum states ; the simple example which is central to both of their analyses is the coding into the three - qubit states @xmath8 and @xmath9 ; that is , @xmath10 is coded as @xmath11 . the second , more important point has to do with the role of the entanglement of these states it is easy to see that the density operator of any one of the three qubits in the codeword is an equal mixture of @xmath12 and @xmath9 , whether the coded qubit is in the @xmath6 or @xmath7 state . as steane says it , the information about which state the coded qubit lies in is not contained in any single one of the coding qubits ; it is spread out into a `` multi - particle interference , '' which is set up as a result of the entanglement of the encoded state . one would like to think that this lack of information in any one particle means that the coded qubit could be recovered after any interaction by the environment with one of the three coded qubits , since the way that a qubit gets disturbed is by a successful external measurement of its state . this turns out not to work for the three - qubit example above , but it is very definitely on the right track . in fact , this reasoning leads us to what s wrong with the three - qubit code ; it is easy to see that , while a demon making measurements on one qubit can not learn whether the coded qubit is a @xmath6 or a @xmath7 , it can easily learn whether the coded qubit is in the state @xmath13 or @xmath14 . this is so because of the simple algebraic fact that @xmath15 thus , the coding of these states involves product states rather than entangled states , and is therefore quite ineffective at hiding the state of the coded qubit from the environment which is `` measuring '' in this diagonal basis . this result suggests that a quantum code which is completely effective against single - qubit error must be able to make the information about the coded state suitably recondite in _ any _ basis ; one might presume that this will work by making the state appropriately entangled in any basis . one might also have a brief worry that there will some fundamental feature of quantum mechanics which makes the right kind of entanglement unavailable . fortunately this is not the case , and indeed there was no point in the history of the subject when this worry held sway , as shor and steane both immediately found solutions inspired by the three - qubit entangled states . their discoveries , however , involved the invention of new forms of multi - particle entangled states , in shor s case a pair of nine - bit entangled states , and in steane s case a pair of seven - bit states . just in case the reader would like to contemplate these new sorts of entangled states explicitly , we would like to write down a pair of states which were discovered some time after by laflamme _ et al._@xcite and independently by us@xcite ; they are five - qubit states , and they are the smallest states for which one - bit error correction is fully effective : @xmath16 and @xmath17 these new states turn out not only to have the important features we have mentioned for error correction , but they also all appear to exhibit a type of quantum non - locality which is of great interest in the foundations of quantum theory , and may indeed be viewed as the natural extension of the work starting with the seminal paper of einstein , podolsky and rosen , and continuing to the discovery of the greenberger - horne - zeilinger states@xcite . a remarkable feature of the error - correcting properties of these states is that , despite landauer s and other s expectations , the process of error correction has an essentially digital rather than analog character@xcite . we mean this in a rather straightforward operational sense : we find ( divincenzo and shor@xcite give the details for the five - qubit code ) that error detection involves the performance of a series of binary - valued quantum measurements . then these bit values provide an instruction for an error detection step , which involves a discrete rotation of a specific one of the five qubit states . it seems that we can say that the reason for this essentially digital character , despite the analog structure of the state space , is that the code states arrange that any error which the environment can cause by operating ( in any arbitrary way ) on a single qubit acts in a subspace _ orthogonal _ to the state space of the coded qubit itself . thus , the complex coefficients are , to very high accuracy , completely untouched by the error process , and all the error detection and correction steps can work in a way which is oblivious to their values . thus , landauer has turned out to be wrong on this point ; but by his challengings , he has opened up an entirely unanticipated line of investigation in the fundamental properties of quantum mechanics . we will not pursue the details here of how this story has further developed , although we have told just the beginning of what has been a tremendous development in the last two years : the formal conditions have been constructed for entangled states to be effective as correctable code states@xcite , which in turn led to the discovery of a powerful group - theoretic framework which permitted the classification of essentially _ all _ interesting quantum codes of arbitrary block size@xcite ; it has been found how error correction can be implemented fault tolerantly@xcite , that is , in such a way that it is insensitive to errors that occur during the error detection operations themselves ; finally , this has in turn led to a discovery of protocols that permit fully general quantum computation to be performed in the presence of errors and decoherence@xcite . all of this has amounted to a revolution in the way we theorists think about the future prospects for quantum computation . as of this writing , we would say that rolf , while being quite impressed by these developments , and being convinced that they really make quantum computing not such a hopeless enterprise as he once had thought , takes a slightly jaundiced view of these developments . he will make sure that we do nt loose sight of the fact that this theoretical development emphatically does _ not _ solve all ( or perhaps even any ) of the problems which stand in the way of making progress in the laboratory today on the construction of a quantum computer . we expect that rolf s wry commentaries on our efforts will continue to nudge our mathematical efforts in the direction of real progress . we would like to turn now to a different aspect of quantum information and its physical embodiment , which also has its roots in the promptings of rolf landauer . early in the development of the theory of quantum computation , benioff@xcite , feynman@xcite , and deutsch@xcite suggested , by way of existence proofs , some abstract models of quantum systems whose dynamics would result in the execution of some computation . landauer criticized this work very pointedly ; he emphasized that computers are made out of physical apparatus and not out of hamiltonians . and he made the point , as we discussed above , that it does not constitute a serious model of computation if the imperfections in these apparatuses are not dealt with in the analysis he showed how several of the abstract systems which had been proposed , if taken seriously , would exhibit very severe flaws which would preclude their being a serious physical basis for computation . so the situation sat in around 1994 , when shor revolutionized the field@xcite and raised to an altogether higher level of significance the question of whether quantum information really could be embodied physically if a quantum computer to do factoring really could be built . we believe that the existing commentaries which had been put in the literature by rolf made the exploration of possible physical embodiments a much soberer and realistic undertaking than it would have been . models which were current at that time included the spin - polymer concept of lloyd@xcite , and the atomic force microscope `` clockwork '' computer of divincenzo@xcite . both were motivated by the landauer criterion of proposing actual physical embodiments of instruments which could manipulate quantum information , and which were thus subjectable to realistic criticisms based on the criteria of experimental physics . however , both the lloyd and divincenzo efforts look , at the remove of three years , very naive and incomplete compared with the much more impressive recent proposals to realize quantum computation in the laboratory . one enormously impressive stream of ideas and proposals have flowed forth from the group of zoller and cirac@xcite . these workers , armed with a very deep understanding of the state of the art in atomic physics and quantum optics , and informed by the very perceptive general formulations of quantum information processing as conditional dynamics produced by artur ekert at a very important conference in 1994@xcite , have , we think , largely passed the landauer test of giving a thoroughly complete , analyzable , and testable scenario ( actually several different ones ) for quantum computation . here would not be the place to give a long technical description of the cirac - zoller proposals . however , we would like to describe an exercise@xcite which we went through subsequent to their work , and partly motivated by it : we have tried to codify , in a comprehensive but general way , a set of sufficient criteria that any physical system should satisfy if it is to be suitable as a realization of a quantum computer . we would like to review our five criteria , discuss briefly how cirac and zoller have succeeded in meeting these criteria , and finally discuss a new proposal which we have made@xcite for implementing quantum computation in a coupled quantum dot system , discussing how we envision the criteria could be met in this arena . here , then , are the five essential criteria which we perceive for the physical implementation of quantum computation : \1 ) hilbert space control . the available quantum states must be precisely enumerated , and it must be known how to confine the state vector of the quantum system to this part of hilbert space . in addition , the hilbert space should be extendable , preferably with a simple tensor - product structure , by adding particles to the system . for example , @xmath18 spin-1/2 particles have a simple spin hilbert space of @xmath19 dimensions . \2 ) state preparation . within this hilbert space , it must be possible to set the state vector initially to a simple fiducial starting state . a simple example of this , in the spin system , would be to set all the spins in the spin - down state . frequently this only requires being able to bring the system to sufficiently low temperature that it is in its ground state . this is more difficult in some examples than in others . \3 ) low decoherence . this is the criterion most closely tied to the topic of the first part of this paper . the coupling to the environment ( i.e. , to all the rest of the hilbert space of the world ) should be sufficiently weak that quantum interference in the computational hilbert space is not spoiled . given our current understanding of error correction and fault tolerant quantum computation , and given fairly benign assumptions about the nature of the decoherence ( e.g. , that it acts independently on each quantum bit ) reliable computation is possible if the decoherence time exceeds the switching time by @xmath20@xcite . more clever fault - tolerant techniques@xcite may well succeed in making this rather demanding threshold number more relaxed in the future . \4 ) controlled unitary transformations . this is the fairly obvious central requirement of quantum computing : it must be possible to subject the computational quantum system to a controlled sequence of precisely defined unitary transformations . the precision requirements are closely related to the decoherence threshold ; imprecision of unitary operations is a form of decoherence . for convenience of programming , it is very desirable that the elementary unitary transformations be implementable as discrete one- and two - qubit operations . \5 ) state - specific quantum measurements . the readout of a quantum computation , which would consist of some ordinary bit string , is to be the result of a sequence of quantum measurements performed on the computational quantum system . it is very desirable ( although _ not _ necessary ) that these measurements be the textbook projection measurements of individual quanta . it is essential that these measurements can be made on specific , identified subsystems of the computational state ; in the simplest case , this means that it should be possible to do a projection measurement on each qubit individually . recent work in nuclear - magnetic - resonance computation has shown that certain aspects of this criterion can be relaxed@xcite : if many identical copies of the quantum computer are available , then weaker , ensemble measurements , rather than projection measurements , are adequate . it is still necessary that these ensemble measurements be subsystem - specific , though . the cirac - zoller proposal of the ion - trap quantum computer@xcite has done a beautiful job of satisfying these criteria : 1 ) the hilbert space which they employ , the low - lying electronic and spins states of the ions , and the quantized states of vibration of the ions in the trap , have been thoroughly mapped out by atomic physicists in a series of careful experiments spanning many years . extendibility is achieved by adding more atoms in a line to the trap . 2 ) laser cooling techniques have enabled experimentalists to place this system in the ground state . ( there are questions about how well this can be done when many ions are added to the trap . ) 3 ) long coherence times are well known for the internal states of the ions ( although coherence times are a bit more problematic for the vibrational states of the ions , and the recent modifications of the proposals made by cirac , zoller , and collaborators have been partly designed to circumvent this decoherence problem ) . 4 ) precision spectroscopic manipulation of the ion s internal - plus - vibrational states is thoroughly demonstrated , and a complete set of quantum logic operations is known to be achievable . 5 ) the availability of quantum - jump spectroscopy implies that virtually ideal , strong , quantum - specific measurements are available@xcite . it would appear from all this that the landauer plea , that quantum computing be considered at the level of real - world devices , has been completely satisfied . not quite , though : despite its plausibility , the cirac - zoller device is a great extrapolation in scale , and in its simultaneous achievement of a variety of experimental capabilities from any existing experiment in ion - trap physics , and some have questioned whether these extrapolations , especially to very large scale quantum computation ( c. 1000 qubits ) will really be possible . there are other technical objections having to do with the fact that the machine as envisioned does not permit parallel operations , quantum gate operations performed simultaneously on different parts of the device . this is important because , from the theoretical point of view , all the powerful results of fault - tolerant quantum computation need this parallel capability@xcite . so , despite the brilliance of the ion - trap proposal , we have remained motivated to propose other platforms , from very different areas of physics , that have the potential for satisfying the five criteria for the implementation of a quantum computer . it is often speculated that a solid - state physics approach will be the only plausible arena for the massive scale - up of quantum computation which will ultimately be desirable . this is indeed a debatable proposition : our current ability to hold and process untold millions of ordinary bits on a silicon chip in no way translates into a corresponding capability to have and hold a large number of qubits . still , it is a fact that there is a great deal of basic research , projecting forth from the fabulous successes of microelectronics , to understand the quantum properties of small solid state devices , and it is on this fact that we hang our hopes . we have thus been encouraged to work on a proposal for a solid - state quantum computer , one based on quantum dots@xcite . we are sure that the quantum behavior of such structures will be the subject of many other articles in this volume , as the whole topic of quantum interference in mesoscopic structures is another one which sprang largely from the brain of rolf landauer . but we think that the proposal which we have made prescribes a much deeper use of the quantum properties of these structures than has had been contemplated before . our proposal has been outlined in detail elsewhere@xcite , and we will just give a summary here , but with a couple of added features which have come to light recently . we will give this outline by discussing how we envision our five criteria for the realization of quantum computation to be satisfied . for pedagogical reasons , we will visit these five items out of their order above . \1 ) hilbert space : we propose using the real _ spin _ states of electrons confined in an array of quantum dots . gaining control over this hilbert space requires , first , that the number of excess electrons confined to each quantum dot be precisely controlled , and in particular that the electron number be _ odd _ so that the dot has an excess spin-1/2 . at this stage in development of experiments in this area , many groups have succeeded in maintaining some electron number in a dot exactly ; fewer groups have the capability of fixing the _ absolute _ number of excess electrons , but we hope that this will be readily doable in many of the quantum dot experiments which are envisioned . we also require that the electrons populate , with reasonable probability , the lowest - lying electron orbital ; in other words , we want only spin degrees of freedom to be available , but not charge degrees of freedom . this should be achievable by a combination of strong confining potentials ( i.e. , small dots ) and low temperatures . it seems that for dots substantially below 100 nm in size , conventional cryogenics ( necessary for many other aspects of the proposal ) should be sufficient . \2 ) state preparation : not much need be said on this point : any conventional method of preparing the set of spins in a highly spin - polarized state ( as simple , for example , as cooling the spins in a strong magnetic field ) would be satisfactory . \5 ) strong quantum measurement : it is necessary ( unless we adopt the ensemble approach introduced in nuclear - magnetic - resonance quantum computing ) to be able to measure whether the spin of any individual dot is up or down with respect to some quantization axis . single - spin measurements in the solid state are still in the future , but such measurements have been the holy grail of quantum magnetism experiments for many years , and we feel confident that eventually such a measurement will be achieved by some means . we might highlight one suggestion which we have made@xcite for how to do this which integrates well with the technology of single - electron quantum - dot experiments . suppose a tunneling barrier could be introduced into the system whose barrier potential is spin dependent ; such barriers are well known in some areas of magnetic physics , although it has not yet been contemplated how to incorporate them into the processing used to create quantum dots . the gating of such a barrier between two quantum dots , one containing the spin state to be measured , and the other containing no excess electron , could , at some desired instant in time , make it possible for the electron to tunnel through the barrier only if it is in one of the two spin states . then , the presence or absence of the excess electron in the second dot , which can be done by well understood and perfected single - electron electrometry techniques , would serve as the desired measurement of whether the electron had been in the spin up or spin down state . if this technique turns out to be infeasible , we are confident that experimentalists will use their ingenuity to solve this problem in a much more practical way than we can ever envision . \4 ) gate operations : this is at the heart of our quantum dot proposal ; we discuss a few recent further advances in our thinking on this in sec . we envision a variety of mechanisms for effecting one - qubit and two - qubit gate operations on the spin qubits of the quantum dots . our proposals begin with the recent development of the experimental capability to controllably couple or decouple the states of neighboring dots by externally controlling an electric potential barrier between them@xcite . in the present experiments , this capability is used only to demonstrate that the dots can go from a regime where a single added electron enters one of the two dots ( the decoupled situation ) to a regime in which an added electron goes into a delocalized state of the two dots ( the coupled situation ) . we propose using this capability in a more subtle way : it is well known that virtual tunneling of electrons between two spin - degenerate sites leads to an effective exchange coupling between the spins of the two electrons . by turning on tunneling ( by lowering the potential barrier ) for a controlled length of time , a specific two - qubit gate operation could be achieved . the exchange interaction leads to a quantum gate of the `` swap '' type ; for a particular duration of the interaction ( or any odd multiple of the fundamental duration ) , the exchange is complete and gate is just a complete swapping or interchange of the two spin states . this does not constitute a very useful two - qubit quantum gate ; but if the interaction is left on for half of this fundamental swapping time , the resulting `` square - root of swap '' operation , in conjunction with other gates which we will discuss next , would provide an efficient basis for programming any desired quantum computation . square - root - of - swap is still not as powerful a quantum gate as is needed theoretically , because it respects rotational symmetry in spin space . thus , it leaves the total angular momentum of the spin system , and its projection on any quantization axis , unchanged . but in quantum computation it is desirable to be able to rotate the state vector from any state to any other state . for this reason , it is very desirable to supplement the two - bit swap - type gate with other gates which break the spin - space rotational symmetry . this is very easy to do from a theoretical point of view , it just requires adding a simple family of one - qubit gates . unfortunately , the experimental implementation of such gates is surprisingly problematic , as it involves the application of magnetic fields locally to an individual spin ( and not to the surrounding ones ) . this is a daunting technical requirement , which , we readily admit , would require quite heroic experimental efforts to achieve ; we hope that there would be other ways to achieve this which we are still investigating , perhaps if it were possible to perform local electron - spin - resonance operations on the system . but for the time being , we offer a few tentative ideas for how this application of a localized magnetic field might be conceivable , with apologies for not having been able to see how to make it any simpler . since , in the measurement scheme which we have proposed above , we have suggested the incorporation of magnetic materials into the system ( to make the spin - dependent tunneling barrier ) , we could envision using such materials to accomplish the one - bit gates . for example , if a piece of such a magnetized material were placed near the quantum dot , such that by lowering a gate potential , the electron could be made to , at some desired time , partially penetrate the magnetized barrier , the electron spin would precess around this internal magnetization and the one - bit operation could be achieved . a magnetized dot which the spin state could be swapped into and out of could have the same effect . if magnetic materials were undesirable , one could envision various local coil arrangements or magnetized probe tips which could also give the desired one - bit operations . hopefully , more ingenuity will lead to more elegant solutions to this problem . \3 ) coherence times : consideration of this criterion for quantum computation also leads us into speculative territory , but one which we are reasonably hopeful about . in the usual mesoscopic experimental regimes , it was rare to find decoherence times even as long as one nanosecond ; in mesoscopics , however , it was always the decoherence of a _ degree of freedom which was being studied . there is every reason to believe , from a theoretical point of view , that the coherence of electron _ spin _ states should , under favorable circumstances , be much longer . generally speaking , the coupling of the environment to spin is weaker than to charge . there is as yet very little experimental indication of how long these spin coherence times could be . et al._@xcite have observed free induction decay for a population of photoexcited electron spins in a quantum well . the @xmath21 measured in this decay , which is a lower limit on the decoherence time for the spins , was seen to be several nanoseconds . we may in addition consider the decoherence time for the spins of itinerant electrons in a 2d quantum well to be a lower limit on the time for electrons in a similar material but confined to a zero - dimensional structure . for these reasons , we believe that the kikkawa observations should be just considered a very early starting point in the search for long spin coherences , and that increases in these times of many orders of magnitude would not be out of the question . it would indeed be extremely desirable to find a system with a decoherence time of , say @xmath22 sec . , since the speed of the desired gate operations would be scaled to this time ; in any foreseeable experiment it would be very interesting to make the gate times a few orders of magnitude faster than the decoherence time . this would mean gate operations going at a mhz rate , which would we think be a fairly comfortable regime for ac manipulations of low temperature electronic systems . we hope very much to engage in a dialog , in the landauer style , with experimentalists and other interested parties to improve this quantum dot proposal through critical discussion . we are certain that the solutions which we have proposed for satisfying the criteria of quantum computation are not optimal , and perhaps on further examination they will prove to be laughable ; but we can not see any `` show stoppers '' at this point , and we remain optimistic that solid state quantum computation will indeed be possible and will indeed be a very exciting line for fundamental experiments in quantum physics . to obtain a more quantitative understanding of the origin of the exchange coupling occurring in the effective two - spin heisenberg model and to determine its magnetic and electric field dependence , we have begun recently@xcite to investigate coupled quantum dots from a more microscopic viewpoint . in the following we wish to report on these preliminary findings . our investigations have been motivated by recent advances in the physics of semiconductor quantum dots that were fabricated in a 2deg gaas system by tarucha et al . @xcite these experimentalists have demonstrated that such dots are well - described by a parabolic confinement potential ( of energy @xmath23mev ) and that one can fill in one electron after the other ( starting with an empty dot ) in a controlled way . armed with this information it is now reasonable to expect that it should be within experimental reach ( as envisioned in point ( 4 ) of sec . iii ) to couple two such dots ( containing only a few electrons ) via a tunable or non - tunable barrier ( as has already been achieved in bigger dots @xcite ) . the physics of such a system can then be described by adopting the lines of reasoning used in molecular theory . to put it in other words , in the same way as one can consider an isolated quantum dot as an artificial atom that obeys e.g. analogs of hund s rule when electrons are added to the shells , one can now consider the problem of coupled quantum dots as the problem of artificial molecules , or more generally as the problem of quantum dot chemistry . " like in ordinary chemistry , we can use techniques such as the heitler - london method or more refined approaches such as the hund - mullikan ansatz , hybridization etc . to obtain the low - lying energies . one of the main differences between ordinary atoms and quantum dots is that the attractive forces between nuclei and electrons are now replaced by the parabolic confining potential that can be controlled externally by changing the gate voltage . the associated bohr radius @xmath24 ( @xmath25 is the effective electron mass ) is typically in the range of tens of nanometers and thus much larger than in real atoms . one important consequence of this is that ( coupled ) quantum dots are much more sensitive to external magnetic and electric fields . as we will see below it is this field sensitivity which allows one to tune the exchange constant " to zero as a function of uniformly applied external fields , the strengths of which are easily accessible in standard set - ups . to be specific let us consider the simplest case , namely two circular quantum dots of radius @xmath26 lying in the same plane and whose centers are separated by 2@xmath26 . each dot contains one electron of spin 1/2 which interact via the ( possibly screened ) coulomb interaction . it is then straightforward to write down an explicit hamiltonian that captures the physics just described and that will allow us to perform some more concrete evaluations : @xmath27 where @xmath28 + g\mu_b { \bf s}_i\cdot{\bf b}/\hbar \,\ , , \nonumber \\ v_{ij}&=&{e^2\over{\|{\bf r}_i-{\bf r}_j\|}}\,e^{-\mu @xmath29 describes the effect of the constant magnetic field @xmath30 , and @xmath31 is some electric field applied along the @xmath32-axis connecting the dots . these dots are located at * a@xmath33 . the coupling of the dots is described in terms of an @xmath34-potential with @xmath35 given by the parabolic confining energy of a single isolated dot . the change of barrier height between the dots can then be described by changing the interdot distance @xmath36 . the last term in @xmath37 is the zeeman term . the coulomb interaction is described by @xmath38 with @xmath39 being some effective screening parameter . the motion of the electrons is assumed to be planar , i.e. @xmath40 . this hamiltonian can not be solved exactly but we can make progress with the help of variational ( or numerical ) techniques to find for instance the exchange constant given by the difference between singlet and triplet energies . in particular , in the heitler - london approximation and making use of the darwin - fock solution for the isolated dots we find ( omitting all details of the calculation @xcite ) , @xmath41\,\ , , \label{exchange}\ ] ] where @xmath42 is the ratio of coulomb energy to confinement energy , and @xmath43 the zeroth - order bessel function . for @xmath44 mev @xcite we have @xmath45 nm and thus @xmath46 . further , the parameter @xmath47 , with @xmath48 being the larmor frequency , describes the effect of the orbital diamagnetism , it becomes appreciable in the tesla range since @xmath49 ( b / t ) . the dimensionless distance between the dots is given by @xmath50 . for the moment we have set @xmath51 and assumed a bare coulomb interaction ( which is a reasonable assumption for the two electron system ) . note that the energy scale of the exchange coupling is given by the confinement energy @xmath52 . a plot of @xmath53 versus magnetic field is given in fig . 1 ( for @xmath54 ) . the most interesting feature of the heitler - london result is the fact that due to the influence of the orbital diamagnetism the exchange @xmath55 passes through zero ( at a field value of about 1 t ) and thus changes from antiferromagnetic ( @xmath56 ) to ferromagnetic ( @xmath57 ) coupling . this suggests again a novel mechanism with which one can tune the exchange coupling j to zero . of course , @xmath55 can also be tuned to zero asymptotically . we stress that the magnetic field is not local but extends uniformly over the two dots , and such a uniform field can be easily accounted for in the xor operation . finally we note that for vanishing b - field and for c values with @xmath58 ( which is in the range of experimental interest ) j is positive for all distances @xmath59 ( i.e. the singlet state is lower in energy than the triplet state ) , as it must of course be the case for a two - electron system on general grounds . the heitler - london approximation for @xmath55 breaks down ( i.e. @xmath55 becomes negative even for @xmath60 ) for certain @xmath59 s when @xmath61 exceeds 2.8 . next , adding an electric field @xmath62 will lead to a simple shift of @xmath55 , @xmath63 this expression is valid for not too large electric fields with @xmath64 not exceeding one . thus , within the heitler - london approximation we find that such a field ( or biasing voltage ) can then be used to also tune the exchange constant . both of these tuning mechanisms could be used alternatively to or in conjunction with a gate between the dots by which one can tune the barrier height by varying the gate voltage ( in our model , this tuning mechanism can be accounted for by varying the dot distance @xmath59 ) . it is interesting to examine the magnetization @xmath65 ( along the @xmath66-axis ) of the coupled dots , as this quantity can give independent information about intrinsic parameters such as the exchange coupling but also about the interplay between orbital diamagnetism and spin paramagnetism . this quantity can also be calculated in the heitler - london approximation@xcite . however , we shall not write down the lengthy expression here and content ourselves with a plot of the magnetization versus b - field , see fig . 2 , where we stay in the low temperature regime of ref.@xcite with @xmath67 k@xmath68 t , which corresponds to an electron temperature of t@xmath69mk . the striking feature to be noted here is the initial diamagnetic response ( with the spins being antiparallel ) followed by a sudden jump at about 1 tesla . indeed , this jump can be traced back to the change of sign in the exchange constant . after the jump , the response becomes diamagnetic again ( with the spins being now parallel ) and finally approaches saturation asymptotically . thus the heitler - london approximation suggests that the sudden switch around 1 tesla allows one to get direct information about the exchange constant from the magnetization . it does not need to be stressed of course that it will be rather difficult to measure the magnetization of only two electrons , as the magnetization is only of the order of a few bohr magnetons . still , in a first set of experiments one can again envision ( as in @xcite ) a scaling up to many independent systems of two coupled quantum dots . also , the present status of cantilever technology is capable of measuring magnetic moments on the order of a single bohr magneton ! it would be interesting to explore the possibility whether one could use magnetic force microscopes etc . to measure such magnetization effects . the above analysis can ( and should ! ) be refined by making use of the hund - mullikan ( or lcao ) method and by including sp - type of hybridization effects ( which however are balanced by orbital field effects ) . these calculations become rather involved @xcite , and we will report on them elsewhere . in principle , it is possible to solve this problem to arbitrary accuracy by making use of powerful numerical techniques developed in molecular physics . [ it is amusing to mention parenthetically that in these numerics one approximates the atomic wave functions by gaussians mainly for technical reasons ; here in our case of quantum dots with parabolic confinement this would in fact be exact and a much better convergence can be expected . ] it is worth mentioning that spin - orbit effects can essentially be neglected in the case of only very few electrons per dot . indeed , the spin - orbit interaction in a quantum dot with parabolic confinement takes the form @xmath70 , where @xmath71 is the angular momentum of the confined electron which is of the order of @xmath72 . thus we can estimate that @xmath73 for above values and with an effective electron mass found in gaas . it is clear by now that the above analysis can be extended to situations with more than one electron per dot , although the complexity of the problem increases rapidly . we hope to report soon on our progress in this direction . finally , it is a most rewarding aspect of this area of research , which we have the privilege to bring to rolf landauer s attention , that even apart from our ultimate goal of building a working quantum computer there is plenty of fascinating and novel physics to be discovered on our way that will keep us ( and hopefully our experimental colleagues ! ) quite busy for a while . r. landauer , `` is quantum - mechanically coherent computation useful ? '' , in _ proc . drexel-4 symposium on quantum nonintegrability quantum - classical correspondence _ , philadelphia , pa , 8 september 1994 ( eds . d. h. feng and b .- hu , international press , boston , to be published ) . r. landauer , `` information is physical , '' proc . workshop on physics and computation physcomp 92 ( ieee comp . press , los alamitos , 1993 ) pp . 1 - 4 . a. berthiaume , d. deutsch , and r. jozsa , `` the stabilisation of quantum computation , '' proc . workshop on physics and computation physcomp 92 ( ieee comp . press , los alamitos , 1993 ) pp . 60 - 62 . c. h. bennett , g. brassard , s. popescu , b. schumacher , j. a. smolin , and w. k. wootters , `` purification of noisy entanglement and faithful teleportation via noisy channels , '' phys . lett . 76 * , 722 ( 1996 ) . c. h. bennett , d. p. divincenzo , j. a. smolin , and w. k. wootters , `` mixed - state entanglement and quantum error correction , '' phys . a * 54 * , 3824 ( 1996 ) ; see http://xxx.lanl.gov/abs/quant-ph/9604024 . p. w. shor , `` scheme for reducing decoherence in quantum computer memory , '' phys . a * 52 * , 2493 ( 1995 ) . a. m. steane , `` error correcting codes in quantum theory , '' phys . lett * 77 * , 793 ( 1996 ) . w. k. wootters and w. h. zurek , nature ( london ) * 299 * , 802 ( 1982 ) . c. h. bennett , g. brassard , and n. d. mermin , phys . * 68 * , 557 ( 1992 ) . e. schrdinger , `` die gegenwrtige situation in der quantenmechanik , '' naturwissenschaften * 23 * , 807 - 812 ; 823 - 828 ; 844 - 849 ( 1935 ) . translation : j. d. trimmer , `` the present situation in quantum mechanics : a translation of schrdinger s cat paradox paper , '' proc . . soc . * 124 * , 323 ( 1980 ) . e. schrdinger , `` discussion of probability relations between separated systems , '' cambridge phil . soc . * 31 * , 555 ( 1935 ) ; `` probability relations between separated systems , '' _ ibid . _ , * 32 * , 446 ( 1936 ) . r. laflamme , c. miquel , j. p. paz , and w. h. zurek , `` perfect quantum error correction code , '' phys . lett . * 77 * , 198 ( 1996 ) . see d. p. divincenzo and a. peres , `` quantum code words contradict local realism , '' phys . a * 55 * , 4089 ( 1997 ) . see d. p. divincenzo , j. a. smolin , and p. w. shor , `` quantum channel capacity of very noisy channels , '' phys . a , in press ; see http://xxx.lanl.gov/abs/quant-ph/9706061 . d. p. divincenzo and p. w. shor , `` fault - tolerant error correction with efficient quantum codes , '' phys . lett . * 77 * , 3260 ( 1996 ) ; see http://xxx.lanl.gov/abs/quant-ph/9605031 . e. knill and r. laflamme , `` theory of quantum error correcting codes , '' phys . a * 55 * , 900 ( 1997 ) ; see http://xxx.lanl.gov/abs/quant-ph/9604034 . d. gottesman , `` class of quantum error correcting codes saturating the quantum hamming bound , '' phys . a * 54 * , 1862 ( 1996 ) ; see http://xxx.lanl.gov/abs/quant-ph/9604038 . a. r. calderbank , e. m. rains , p. w. shor , and n. j. a. sloane , `` quantum error correction and orthogonal geometry , '' phys . rev . lett . * 78 * , 405 ( 1997 ) ; see http://xxx.lanl.gov/abs/quant-ph/9605005 . p. w. shor , `` fault - tolerant quantum computation , '' proc . 37th symposium on the foundations of computer science ( ieee computer society press , los alamitos , 1996 ) , pp . 56 - 65 ; see http://xxx.lanl.gov/abs/quant-ph/9605011 . j. preskill , `` reliable quantum computers , '' proc . london , in press ; see http://xxx.lanl.gov/abs/quant-ph/9705031 . p. benioff , `` quantum mechanical hamiltonian models of turing machines , '' j. stat * 29 * , 515 ( 1982 ) . r. feynman , `` quantum - mechanical computers , '' found * 16 * , 507 ( 1986 ) . d. deutsch , `` quantum computational networks , '' proc . a * 425 * 73 ( 1989 ) . p. w. shor , `` algorithms for quantum computation : discrete log and factoring , '' proc . 35th annual symposium on the foundations of computer science ( ieee computer society , los alamitos , 1993 ) , p. 124 . s. lloyd , `` a potentially realizable quantum computer , '' science * 261 * , 1589 ( 1993 ) ; `` envisioning a quantum supercomputer , '' _ ibid . _ * 263 * , 695 ( 1994 ) . d. p. divincenzo , `` two - bit gates are universal for quantum computation , '' phys . a * 51 * , 1015 ( 1995 ) ; see http://xxx.lanl.gov/abs/cond-mat/9407022 . this work started with j. i. cirac and p. zoller , `` quantum computations with cold trapped ions , '' phys . lett . * 74 * , 4091 ( 1995 ) . a. ekert , `` quantum computation , '' in _ atomic physics 14 14th international conference on atomic physics _ , boulder , co , 1994 , aip conference proceedings 323 ( aip press , new york , 1995 ) , p. 450 . d. p. divincenzo , `` topics in quantum computers , '' in _ mesoscopic electron transport _ , ed . g. schoen ( kluwer , 1997 , to be published ) ; proceedings of the nato advanced study institute on mesoscopic electron transport , curacao , june 24-july 6 , 1996 . see http://xxx.lanl.gov/abs/cond-mat/9612126 . d. loss and d. p. divincenzo , `` quantum computation with quantum dots , '' phys . a * 57 * , 120 ( 1998 ) . see http://xxx.lanl.gov/abs/cond-mat/9701055 . kitaev , `` fault tolerant quantum computation by anyons , '' http://xxx.lanl.gov/abs/quant-ph/9707021 . n. a. gershenfeld and i. l. chuang , `` bulk spin resonance quantum computation , '' science * 275 * , 350 ( 1997 ) . a good source for discussion of general progress in the ion - trap computer may be found in r. j. hughes _ et al . _ , `` the los alamos trapped ion quantum computer experiment , '' to be published in fortschritte der physik ; see http://xxx.lanl.gov/abs/quant-ph/9708050 . one experiment which features this capability is c. livermore _ et al . _ , `` the coulomb blockade in coupled quantum dots , '' science * 274 * , 1332 ( 1996 ) . j. m. kikkawa , i. p. smorchkova , n. samarth , and d. d. awschalom , `` room - temperature spin memory in two - dimensional electron gases , '' science * 277 * , 1284 ( 1997 ) . s. tarucha , d.g . austin , t. honda , r.j . van der hage , and l.p . kouwenhoven , phys . lett . * 77 * , 3613 ( 1996 ) . g. burkard , d.p . divincenzo , and d. loss , in preparation .
it is an important task to visualize , understand and then extract useful information in complex continuous dynamical systems from many science and engineering fields including flight simulation @xcite , wave propagation @xcite , bio - inspired fluid flow motions @xcite , ocean current model @xcite , and etc . it is , therefore , natural to see many existing tools to study dynamical systems . for example , a lot of work in dynamical systems focuses on understanding different types of behaviors in a model , such as elliptical zones , hyperbolic trajectories , chaotic attractors , mixing regions , to name just a few . one interesting tool is to partition the space - time domain into subregions based on certain quantity measured along with the passive tracer advected according to the associated dynamical system . because of such a lagrangian property in the definition , the corresponding partition is named the lagrangian coherent structure ( lcs ) . one commonly used quantity is the so - called finite time lyapunov exponent ( ftle ) @xcite which measures the rate of separation of a passive tracer with an infinitesimal perturbation in the initial condition , over a finite period of time . another popular diagnostic of trajectory separation in dynamical systems is the finite - size lyapunov exponent ( fsle ) @xcite which is especially popular in the applications from the oceanography . instead of introducing an infinitesimal perturbation in the initial condition of the fluid particles , fsle considers finite perturbation . it does not measure the rate of separation but computes the time required to separate two adjacent tracers up to certain distance . since all these quantities are long treated as a lagrangian property of a continuous dynamical system , most , if not all , numerical methods are developed based on the traditional lagrangian ray tracing method by solving the ode system using any well - developed numerical integrator . these approaches , however , require the velocity field defined at arbitrary locations in the whole space depending on the location of each individual particle . this implies that one has to in general implement some interpolation routines in the numerical code . unfortunately , it could be numerically challenging to develop an interpolation approach which is computationally cheap , monotone , and high order accurate . an eulerian approach to the flow map computation for the ftle has been first proposed in @xcite which has incorporated the level set method @xcite . based on the phase flow method @xcite , we have developed a backward phase flow method for the eulerian ftle computations in @xcite . based on these partial differential equation ( pde ) based algorithms , more recently , we have developed in @xcite an efficient eulerian numerical approach for extracting invariant sets in a continuous dynamical system in the extended phase space ( the @xmath1 space ) . in @xcite , we have proposed a simple algorithm called vials which determines the average growth in the surface area of a family of level surfaces . this eulerian tool relates closely to the ftle and provides an alternative for understanding complicated dynamical systems . in this paper , we are going to further develop these eulerian tools for continuous dynamical systems . there are several main contributions of this paper . we will first improve the numerical algorithm for _ forward _ flow map construction . the papers @xcite have proposed to construct the _ forward _ flow map defined on a fixed cartesian mesh by solving the level set equations or the liouville equations in the _ backward _ direction . in particular , to compute the forward flow map from the initial time @xmath2 to the final time @xmath3 , one needs to solve the liouville equations _ backward _ in time from @xmath3 to @xmath2 with the initial condition given at @xmath3 . this implementation is numerically inconvenient , especially when incorporating with some computational fluid dynamic ( cfd ) solvers , since the velocity field is loaded from the current time @xmath3 backward in time to the initial time . this implies that the whole field at all time steps has to be stored in the desk which might not be practical at all . in this paper , we first propose an eulerian approach to compute the _ forward _ flow map _ on the fly _ so that the pde is solved _ forward _ in time . developed using this improved algorithm , we propose a simple eulerian approach to compute the isle . for each of the separation factor @xmath0 in the definition of the isle , typical lagrangian method requires to shoot an individual set of rays and to monitor the variations in these trajectories . if the scale @xmath0 is changed , these methods will require to restart the whole computations all over again . in this work , we are going to develop an efficient algorithm which provides the isle field for any arbitrary separation factor @xmath0 . for an individual value of @xmath0 , the method requires to extract only an isosurface of a volumetric data . this can be done easily by any well - developed interpolation algorithm or simply the function isosurface in matlab . another main contribution of this work is to provide a theoretical link between the ftle and the isle . even though these two quantities are measuring different properties of the flow , it has been reported widely , such as @xcite , that they give visually very similar solutions in many examples . we are going to show in particular that their ridges can be identified through the ridge of the largest eigenvalue of the associated cauchy - green deformation tensor of the flow . to the best of our knowledge , this theoretical result is the first one quantitatively revealing the relationship between the ftle ridges and the isle ridges . indeed , comparisons between ftle and fsle have been made in various studies . the paper @xcite has pointed out that the ftle might be unable to recognize the boundaries between the chaotic and the large - scale mixing regime . the article @xcite has argued that the fsle has several limitations in lagrangian coherence detection including aspects from local ill - posedness , spurious ridges and intrinsic jump - discontinuities . nevertheless , the work @xcite has stated that these two concepts might yield similar results , if properly calibrated , and could be even interchangeable in many flow visualizatons such as oceanography and chemical kinetics . in this paper , however , we would like to emphasize on the computational aspect of using these lyapunov exponents , rather than on comparing the advantages and the disadvantages of these different methods . the preference for one or the other could be based on the tradition in various fields , or the availability of the computational resources . this paper is organized as follows . in section [ sec : background ] we will give a summary of several important concepts including the ftle , fsle and isle , and also our original eulerian formulations for computing the flow maps and the ftle . then , our proposed eulerian algorithms will be given in section [ sec : proposal ] . we will then point out some theoretical properties and relationships between the ftle and the isle in section [ sec : relation ] . finally , some numerical examples will be given in section [ sec : examples ] . in this section , we will summarize several useful concepts and methods , which will be useful for the developments that we are proposing . we first introduce the definition of various closely related lyapunov exponents including the finite time lyapunov exponent ( ftle ) @xcite , the finite size lyapunov exponent ( fsle ) @xcite and also the infinitesimal size lyapunov exponent ( isle ) @xcite . then we discuss typical lagrangian approaches and summarize the original eulerian approaches as in @xcite . the discussions here , however , will definitely not be a complete survey . we refer any interested readers to the above references and thereafter . in this paper , we consider a continuous dynamical system governed by the following ordinary differential equation ( ode ) @xmath4 with the initial condition @xmath5 . the velocity field @xmath6 is a time dependent lipschitz function where @xmath7 is a bounded domain in the @xmath8-dimensional space . to simplify the notation in the later sections , we collect the solutions to this ode for all initial conditions in @xmath9 at all time @xmath10 and introduce the flow map @xmath11 such that @xmath12 represents the arrival location @xmath13 of the particle trajectory satisfying the ode ( [ eqn : ode ] ) with the initial condition @xmath14 at the initial time @xmath15 . this implies that the mapping will take a point from @xmath16 at @xmath15 to another point @xmath13 at @xmath17 . the finite time lyapunov exponent ( ftle ) @xcite measures the rate of separation between adjacent particles over a finite time interval with an infinitesimal perturbation in the initial location . mathematically , consider the initial time to be 0 and the final time to be @xmath18 , we have the change in the initial infinitesimal perturbation given by @xmath19 the leading order term of the magnitude of this perturbation is given by @xmath20^ * \nabla \phi_0^t(\mathbf{x } ) \delta \mathbf{x}(0 ) \right > } \ , .\ ] ] with the cauchy - green strain tensor @xmath21^ * \nabla \phi_0^t(\mathbf{x})$ ] , we obtain the maximum deformation @xmath22 } } \\| \mathbf{e}(0)\\| = e^ { \sigma_0^t({\mathbf{x } } ) \|t\| } \\|\mathbf{e}(0 ) \\| \ , , \ ] ] where @xmath23 is the eigenvector associated with the largest eigenvalue of the deformation tensor . using this quantity , the finite - time lyapunov exponent ( ftle ) @xmath24 is defined as @xmath25 } = \frac{1}{\|t\| } \ln \sqrt{\lambda_0^t({\mathbf{x } } ) } \ , . \label{eqn : ftle}\ ] ] where @xmath26 denotes the largest eigenvalue of the cauchy - green tensor . the absolute value of @xmath18 in the expression reflects the fact that we can trace the particles either _ forward _ or _ backward _ in time . in the case when @xmath27 , we are measuring the maximum stretch _ backward _ in time and this corresponds to the maximum compression _ forward _ in time . to distinguish different measures , we call @xmath24 the _ forward _ ftle if @xmath28 and the _ backward _ ftle if @xmath27 . a related concept is a widely used concept in oceanography called the finite size lyapunov exponent ( fsle ) which tries to measure the time it takes to separate two adjacent particles up to certain distance @xcite . there are two length scales in the original definition of the fsle . one is the initial distance between two particles which is denoted by @xmath29 . the other one is the so - called separation factor , denoted by @xmath30 . for any trajectory with initial location @xmath31 with @xmath32 , we first determine the shortest time @xmath33 so that @xmath34 the fsle at the point @xmath35 is then defined as @xmath36 numerically , on the other hand , the maximization over the constraint @xmath32 might require special considerations and the corresponding optimization problem might not be easily solved by typical optimization algorithms . one simple approximation is to consider only @xmath37-neighbors of @xmath35 with each point @xmath38 obtained by perturbing one coordinate of @xmath35 with distance @xmath29 . to avoid introducing two separate length scales ( @xmath29 and @xmath0 ) and the approximation by @xmath37 neighbors in the final optimization step , one theoretically takes the limit as the length scale @xmath29 tends to 0 . the fsle can then be reduced to the infinitesimal size lyapunov exponent ( isle ) as defined in @xcite given by @xmath39 where @xmath40 is the shortest time for which @xmath41 . even with tremendous theoretical development , there are not many discussions on the numerical implementations . because the quantity is defined using lagrangian particle trajectories , most numerical algorithms for computing the ftle or the fsle are developed based on the ray tracing method which tries to solve the flow system ( [ eqn : ode ] ) using a numerical integrator . in the two dimensional case , for example , the flow map is first determined by solving the ode system with the initial condition imposed on a uniform mesh @xmath42 . then one can apply simple finite difference to determine the deformation tensor and , therefore , its eigenvalues . the computation of the ftle is a little easier . for a given final time , one simply solves the ode system up to that particular time level . then the ftle at each grid location can be computed directly from these eigenvalues . the isle , on the other hand , is a little more involved since for a particular separation factor , @xmath0 , one has to continuously monitor the change in these eigenvalues and record the precise moment when it reaches the thresholding value . computational complexity could be one major concern of these numerical tools . let @xmath43 and @xmath44 be the number of grid points in each physical direction and in the temporal direction , respectively . the total number of operations required to compute the flow map @xmath45 is , therefore , @xmath46 . note however that such complexity is needed for determining the ftle or the isle on one single time level . for all time levels , the computational complexity for these usual lagrangian implementation is @xmath47 . there are several nice ideas to improve the computational time . one natural idea is to concentrate the computations near the locations where the solutions have large derivatives by implementing an adaptive mesh strategy @xcite . another interesting approach is to avoid the computation of the overall map @xmath48 but to numerically decompose it into sub - maps @xmath49 @xcite . then , the extra effort from @xmath50 to @xmath51 , given by @xmath52 are simply some interpolations . all these proposed improvements are only on the flow map computations , and therefore the ftle . for the fsle or the isle computations on one single time level and one particular @xmath0 , say @xmath53 , one has to go through all @xmath54 mesh points in the temporal direction . for a different time level or even a different separation factor @xmath0 , we have to go though the same procedure all over again . we do not aware of many efficient algorithms for the fsle or the isle implementation . ( a ) @xcite . ( a ) lagrangian ray tracing from a given grid location @xmath55 at @xmath2 . note that @xmath56 might be a non - grid point . ( b ) eulerian values of @xmath57 at a given grid location @xmath56 at @xmath3 gives the corresponding take - off location at @xmath2 . note the take - off location might not be a mesh point.,title="fig:",width=217 ] ( b ) @xcite . ( a ) lagrangian ray tracing from a given grid location @xmath55 at @xmath2 . note that @xmath56 might be a non - grid point . ( b ) eulerian values of @xmath57 at a given grid location @xmath56 at @xmath3 gives the corresponding take - off location at @xmath2 . note the take - off location might not be a mesh point.,title="fig:",width=217 ] we briefly summarize the eulerian approach based on the level set method and the louville equation . we refer interested readers to @xcite and thereafter . we define a vector - valued function @xmath58 . at @xmath2 , we initialize these functions by @xmath59 these functions provide a labeling for any particle in the phase space at @xmath2 . in particular , any particle initially located at @xmath60 in the extended phase space can be * implicitly * represented by the intersection of @xmath8 codimension-1 surfaces represented by @xmath61 in @xmath62 . following the particle trajectory with @xmath63 as the initial condition in a given velocity field , any particle identity should be preserved in the lagrangian framework and this implies that the material derivative of these level set functions is zero , i.e. @xmath64 this implies the following level set equations , or the liouville equations , @xmath65 with the initial condition ( [ eqn : levelsetequationic ] ) . the above * implicit * representation embeds all path lines in the extended phase space . for instance , the trajectory of a particle initially located at @xmath66 can be found by determining the intersection of @xmath8 codimension-1 surfaces represented by @xmath67 in the extended phase space . furthermore , the forward flow map at a grid location @xmath63 from @xmath2 to @xmath3 is given by @xmath68 where @xmath56 satisfies @xmath69 . note that , in general , @xmath56 is a non - mesh location . the typical two dimensional scenario is illustrated in figure [ fig : forwardbackward ] ( a ) . the solution to ( [ eqn : levelsetequation ] ) contains much more information than what was referred to above . consider a given mesh location @xmath56 in the phase space at the time @xmath3 , as shown in figure [ fig : forwardbackward ] ( b ) , i.e. @xmath70 in the extended phase space . as discussed in our previous work , these level set functions @xmath71 defined on a uniform cartesian mesh in fact give the backward flow map from @xmath3 to @xmath2 , i.e. @xmath72 . moreover , the solution to the level set equations ( [ eqn : levelsetequation ] ) for @xmath73 provides also backward flow maps for all intermediate times , i.e. @xmath74 . to compute the forward flow map , on the other hand , @xcite has proposed to simply reverse the above process by initializing the level set functions at @xmath3 by @xmath75 and solving the corresponding level set equations ( [ eqn : levelsetequation ] ) backward in time . a typical algorithm of this type is given in algorithm 1 . note that in the original lagrangian formulation , the attracting lcs or the unstable manifold is obtained by backward time tracing , while the repelling lcs or the stable manifold is computed by forward time integration . in this current eulerian formulation , on the other hand , forward time integration of the liouville equations gives the attracting lcs and backward time marching provides the repelling lcs . ( 4,0 ) ( -0.5,0)(1,0)12 + algorithm 1 : computing the forward flow map @xmath76 : 1 . discretize the computational domain to get @xmath77 . 2 . initialize the level set functions on the last time level @xmath78 by @xmath79 3 . solve the liouville equations for each individual level set function @xmath80 @xmath81 from @xmath82 down to @xmath2 using any well - developed high order numerical methods like weno5-tvdrk2 @xcite with the boundary conditions @xmath83 where @xmath84 is the outward normal of the boundary . 4 . assign @xmath85 . ( 4,0 ) ( -0.5,0)(1,0)12 in this section , we introduce two new eulerian algorithms for flow visualizations . the first algorithm is to improve the eulerian numerical method for computing the forward flow map . instead of solving the pde _ backward _ in time as in section [ subsec : oldeulerian ] and @xcite , we develop an eulerian pde algorithm to construct the forward flow map from @xmath2 to the final time @xmath3 so that we do not need to load the time - dependent velocity field from the terminal time level backward to the initial time . in other words , we construct the forward flow map _ on the fly_. this is computationally more natural and convenient . because of this simple algorithm , we propose to develop an efficient and systematic pde based method for computing the isle field . typical challenge in obtaining such field is that the choice of the separation factor @xmath0 is in general flow dependent . one has to vary @xmath0 to extract the necessary information . usual lagrangian methods unfortunately require shooting different sets of initial rays for each individual separation factor @xmath0 . this is computationally very inefficient . the proposed eulerian formulation requires either some simple interpolations or only one single level surface extraction which can be easily implemented using any well - developed contour extraction function such as in . one disadvantage about the previous eulerian approach for forward flow map computation as discussed in section [ subsec : oldeulerian ] and @xcite is that the level set equation @xmath86 has to be first solved _ backward _ in time from @xmath3 to @xmath2 . once we have the final solution at the time level @xmath2 , one can then identity the flow map @xmath76 by @xmath87 . this could be inconvenient especially when we need to access the intermediate forward flow map . in this section , we propose a new algorithm to construct the forward flow map _ on the fly_. consider the forward flow map from @xmath2 to @xmath3 . suppose the time domain @xmath88 $ ] is discretized by @xmath89 discrete points @xmath90 , where @xmath91 and @xmath92 . then on each subinterval @xmath93 $ ] for @xmath94 , we use the same method as described in section [ subsec : oldeulerian ] to construct the forward flow map @xmath49 . in particular , we solve the level set equation ( [ eqn : levelset ] ) _ backward _ in time from @xmath95 to @xmath96 with the terminal condition @xmath97 imposed on the time level @xmath95 . then the forward flow map is given by @xmath98 . once we have obtained this one step _ forward _ flow map @xmath49 , the _ forward _ flow map from @xmath2 to @xmath3 can be obtained using the composition @xmath99 . and this can be easily done by typical numerical interpolation methods . to prevent extrapolation , we could enforce @xmath100 \times [ y_{\min},y_{\max}]$ ] . here we emphasize that even though we solve equation ( [ eqn : levelset ] ) _ backward _ in time for each subinterval , we access the velocity field data @xmath101 prior to @xmath102 and therefore the forward flow map is indeed obtained _ on the fly_. as a simple example , we consider the first order euler method in the temporal direction . the forward flow map @xmath49 can be obtained by first solving @xmath103 with the terminal condition @xmath104 , and then assigning @xmath105 . higher order generalization is straight - forward . for example , if we use the tvd - rk2 method in the temporal direction @xcite , we can compute the forward flow map @xmath49 by @xmath106 with the terminal condition @xmath104 , and the functions @xmath107 and @xmath108 are two intermediate predicted solutions at the time levels @xmath109 and @xmath90 , respectively . then the flow map from the time level @xmath96 to @xmath95 is given by @xmath110 . it is rather natural to decompose the flow map @xmath111 into a composition of maps @xmath112 . for example , such idea has been used recently in @xcite to improve the computational efficiency of the lagrangian ftle construction between various time levels . define the interpolation operator by @xmath113 , which returns the interpolated mesh values from the non - mesh values . then the flow map @xmath48 can be decomposed into @xmath114 . if the maps @xmath115 for @xmath116 are all stored once they are computed , one can form @xmath48 by determining only @xmath117 . the idea in this work shares some similarities with what we are implementing . our approach can also re - use all intermediate flow maps , the work @xcite however has concentrated only on improving the computational efficiency of the lagrangian approach . we are using the idea of flow map decomposition to propose a forward computational strategy for the eulerian formulation . to obtain a stable evolution in the flow map constructions , theoretically we require the interpolation scheme between two local flow maps to be monotone , i.e. the interpolation scheme should preserve the monotonicity of the given data points @xcite . this is especially important in the context of the phase flow maps for autonomous flows @xcite since the overshooting due to the interpolation could be significantly amplified in the time doubling strategy @xmath118 ^ 2 \ , .\ ] ] in this section , we propose an efficient eulerian approach to compute the isle function for an arbitrary separation factor @xmath0 based on the techniques developed in the previous section . to compute the isle at @xmath119 , one has to determine the minimum time @xmath120 at each location @xmath119 for which @xmath121 . it is , therefore , necessary to keep track of all intermediate values @xmath122 for all @xmath123 s during the flow map construction . at each grid point @xmath124 and each intermediate time step @xmath90 , we construct a quantity @xmath125 by @xmath126\ ] ] with @xmath127 . once we determine the flow map @xmath128 , we have also constructed an increasing sequence at each grid point @xmath124 in time @xmath129 , i.e. @xmath130 for each @xmath124 . now , we interpret this quantity as a spatial - temporal volumetric data in the @xmath1 space . due to such monotonicity property in the temporal direction , the isosurface @xmath131 forms a graph of @xmath119 , i.e. for each @xmath119 we have a unique @xmath132 so that @xmath133 . and more importantly , the value @xmath134 gives the shortest time so that @xmath135 . this implies that @xmath136 and so the level surface in the extended phase space , i.e. @xmath137 can be used to define @xmath138 . finally , the isle function can then be computed using @xmath139 the computational algorithm is summarized in algorithm 2 . ( 4,0 ) ( -0.5,0)(1,0)12 + algorithm 2 ( our proposed eulerian method for computing the isle ) : 1 . discretize the computational domain @xmath140 2 . set @xmath141 . 3 . for @xmath94 , 1 . solve the level set equation ( [ eqn : levelset ] ) for @xmath142 with the condition @xmath143 . 2 . define @xmath144 . 3 . interpolate the solutions to obtain @xmath145 . 4 . compute @xmath146 and determine @xmath147 \ , .\ ] ] 4 . given an arbitrary separation factor @xmath0 . for each @xmath124 , search for a time step @xmath90 such that @xmath148 . then determine @xmath138 by applying the piecewise linear interpolation using the values @xmath125 and @xmath149 . + the isle function is computed as @xmath150 ( 4,0 ) ( -0.5,0)(1,0)12 + there are several major advantages of the proposed eulerian approach . it provides an efficient yet simple numerical implementation of @xmath151 for multiple separation factors @xmath0 . the construction of the quantity @xmath152 can be first done in the background . for each particular given value of @xmath0 , one has to perform only one single isosurface extraction in the extended phase space , i.e. the @xmath1 space . unlike the typical lagrangian implementation where one has to repeatedly keep track of _ all _ particle trajectories for the constraint @xmath153 , the proposed eulerian algorithm provides a systematic way based on a simple thresholding strategy . note that in our implementation , if it happens that @xmath154 for all @xmath90 , we simply set @xmath155 . we conclude this section by discussing the computational complexity of our proposed algorithm for a two dimensional flow , i.e. @xmath156 . let @xmath157 and @xmath54 be the discretization size of one spatial dimension and time dimension respectively . since the liouville equation is a hyperbolic equation , we have @xmath158 by the cfl condition . at each time step @xmath90 , a short time flow map @xmath49 is obtained by solving the liouville equations from @xmath159 to @xmath160 , the computational effort is @xmath161 . then in computing the long time flow map @xmath162 , an interpolation @xmath163 takes @xmath161 operations . therefore the construction of flow map at each time step requires @xmath164 operations . computing @xmath165 and @xmath149 also needs @xmath161 operations , the complexity order is kept at @xmath161 . summing up this procedure in all time steps , the overall computational complexity is @xmath166 . this new eulerian method has significantly improved the overall computational complexity in the application to the isle computations . since the calculations involve the forward flow map from @xmath167 to @xmath168 for all @xmath169 , the original eulerian approach as discussed in section [ subsec : oldeulerian ] requires to solve the liouville equation from each individual time level @xmath168 backward in time to @xmath167 . the overall computational complexity in obtaining all flow maps is therefore given by @xmath170 which is one order magnitude larger than the newly developed eulerian method . with algorithms proposed in section [ sec : proposal ] , we can finally develop an efficient numerical approach to compute the _ forward _ ftle field by solving corresponding pdes _ forward _ in time and the isle field for arbitrary separation factor @xmath0 . the ftle and the isle ( or the fsle ) are indeed measuring different properties of a given flow , even though both quantities depend on the same function @xmath171 which relates the growth of an infinitesimal perturbation . in particular , ftle measures how much @xmath171 grows over a finite time span @xmath172 $ ] , while the isle measures the time required for the quantity @xmath171 to reach a threshold value prescribed by the separation factor . they are two different tools to study how rapid @xmath171 grows . a careful comparison of these two quantities can be found in , for example , @xcite . in many numerical experiences , on the other hand , the ftle fields can show striking visual resemblance with the isle fields of certain separation factors . however , there has not been much studies on the theoretical relationship between these two quantities . in this section , we are going to demonstrate how these two quantities relate to each other by considering their corresponding _ ridges_. of course , the way to define a _ ridge _ is not unique and one can pick a convenient definition for a particular application @xcite . in this work , we adapt the following simple definition . for a given scalar function @xmath173 , we define a @xmath174-ridge to be a codimension - one compact and @xmath175 surface , denoted by @xmath176 , such that for every point @xmath177 , @xmath173 is a local maximum along the normal direction to @xmath176 at @xmath119 . at these locations , @xmath173 is also called a generalized maximum in some literatures . next , we define a particular neighbourhood of a ridge for later discussion . suppose @xmath178 is a codimension - one , compact and smooth surface . we define @xmath179 , the tubular neighbourhood of @xmath178 with radius @xmath180 , as the collection of any point @xmath181 that can be uniquely expressed as @xmath182 , for some point @xmath183 , a real number @xmath184 , and @xmath185 is an unit normal vector of @xmath178 at @xmath119 . to ensure that any point @xmath181 on @xmath179 can be uniquely expressed as such , one can equivalently require that the boundary of @xmath179 does not intersect itself . for a smooth surface , this can always be guaranteed by choosing a small enough positive number @xmath186 . the definitions of both the ftle and the isle involve @xmath187 . it is therefore natural to link them through this particular quantity . the following simple theorem which identifies the @xmath188-ridges with the @xmath189-ridges , will be the very first step in building up our analysis . [ prop : ftle&lambda ] let @xmath188 and @xmath189 be the ftle function and the largest eigenvalue of the cauchy - green deformation tensor from time @xmath190 to time @xmath18 , respectively . then , a @xmath188-ridge is also a @xmath189-ridge . since @xmath191 is a just a scaling factor and the natural logarithm function is strictly increasing , all generalized maxima of @xmath188 as defined in ( [ eqn : ftle ] ) and @xmath189 coincide . for convenience , in the time window @xmath172 $ ] , we impose @xmath192 if @xmath193 for all @xmath194 $ ] . if an isle function @xmath195 on an open set @xmath9 of the spatial domain , there is obviously no visible isle feature on @xmath9 , corresponding to this separation factor @xmath0 . the following theorem establishes a fundamental result that for certain choices of @xmath0 , the isle function @xmath196 can be positive on a tubular neighbourhood of a @xmath189-ridge . this explains the non - trivial isle features around the location of a ftle ridge , when one chooses a suitable separation factor @xmath0 . [ prop : funprop1 ] let @xmath189 be the largest eigenvalue of the cauchy - green deformation tensor from time @xmath190 to time @xmath18 . suppose @xmath189 is continuous everywhere in the spatial domain , and @xmath176 is a @xmath189-ridge with @xmath197 . then for any positive number @xmath198 , there is a tubular neighbourhood @xmath199 such that the isle function is positive on @xmath199 . for each @xmath177 , using the definition of @xmath200 , we have @xmath201 . then by the continuity of @xmath189 at @xmath119 , we can find an open ball @xmath202 with center @xmath119 and radius @xmath203 such that @xmath204 for all @xmath205 . collect all these open balls at every point on @xmath176 into @xmath206 , which is clearly an open cover of @xmath176 . by the compactness of @xmath176 , we can find a finite subcover @xmath207 from @xmath206 . take @xmath208 . without loss of generality , assume @xmath186 is small enough such that the tubular neighbourhood @xmath199 is contained in the union of sets in @xmath209 . now for any @xmath210 , it is clear that @xmath211 . because the function @xmath212 is a continuous function of @xmath213 , we apply the intermediate value theorem and conclude that it must have reached @xmath0 at some @xmath214 $ ] . therefore @xmath215 is positive . the next theorem is to identify a @xmath189-ridge with a @xmath196-ridge of a certain isle function @xmath196 . to justify the condition , the points on the location of a @xmath189-ridge , are expected to exhibit highly chaotic behavior . it is reasonable to assume that we can find a tubular neighbourhood of a @xmath189-ridge , where the time monotonicity of @xmath216 holds for @xmath217 $ ] , and that no other points reach the minimum value of @xmath218 , except those on @xmath176 . [ prop : funprop2 ] let @xmath189 be the largest eigenvalue of the cauchy - green deformation tensor from time @xmath190 to time @xmath18 . suppose @xmath176 is a @xmath189-ridge with @xmath219 . if there exists a tubular neighbourhood @xmath199 of @xmath176 such that 1 . for any fixed @xmath210 , the quantity @xmath220 is increasing in @xmath217 $ ] , and 2 . @xmath221 , then @xmath176 is also a @xmath222-ridge for the isle function @xmath222 . let @xmath177 , and @xmath223 be a point in the normal direction of @xmath176 at @xmath119 . by the second condition , we have @xmath224 . and since @xmath220 is increasing with @xmath213 , it follows that @xmath225 for @xmath217 $ ] . therefore @xmath226 by our convention . we can conclude that no other points on the normal direction of @xmath176 at @xmath119 has positive value of @xmath222 , except @xmath119 itself . this makes @xmath119 a generalized maximum of @xmath222 . theorem [ prop : funprop2 ] is a restricted result , in the sense that the whole @xmath189-ridge is preserved for a particular @xmath196 function . but it is straightforward to extend it to the following corollary , which looks for a certain portion of a @xmath189-ridge that can be conserved in a larger class of isle functions . [ prop : funcor ] let @xmath189 be the largest eigenvalue of the cauchy - green deformation tensor from time @xmath190 to time @xmath18 . suppose @xmath176 is a @xmath189-ridge with @xmath219 , and @xmath227 is a connected subset of @xmath176 with @xmath228 . for any positive number @xmath229 $ ] , if there exists a tubular neighbourhood @xmath230 of @xmath227 such that 1 . for any fixed @xmath231 , the quantity @xmath220 is increasing in @xmath217 $ ] , and 2 . @xmath232 , then @xmath227 is a @xmath196-ridge for the isle function @xmath196 . the proof is parallel to that of theorem [ prop : funprop2 ] , with @xmath176 and @xmath200 being substituted by @xmath227 and @xmath0 , respectively . through theorem [ prop : funprop1 ] to corollary [ prop : funcor ] , we have obtained a close relationship between the @xmath189-ridges and the isle ridges . by the equivalence of the ftle ridges and the @xmath189-ridges as proved in theorem [ prop : ftle&lambda ] , we can then link the ftle ridges and the isle ridges . in practice , the estimated location of the isle ridges can be inferred by the location of the ftle ridges , while @xmath189 plays an important role for estimating a suitable separation factor defined in the isle ridge . to demonstrate this , suppose that there is a ftle ridge @xmath176 satisfying all conditions in theorem [ prop : funprop2 ] . then @xmath176 is an isle ridge with the separation factor @xmath233 . although the value of @xmath200 is actually an unknown , we can determine a number @xmath234 that is slightly less than @xmath235 manually . from the definition of the ftle , we have @xmath236 > e^{lt } \ , .\ ] ] now , with any separation factor @xmath237 , theorem [ prop : funprop1 ] implies that there is a tubular neighborhood around @xmath176 with positive isle values . we can then increase the separation factor from @xmath238 approaching to @xmath200 , so that the tubular neighborhood containing @xmath176 becomes narrower , and thus leads to a better approximation of @xmath176 . in practice , the value of @xmath234 has to be chosen in a trial - by - error fashion and has to be determined case - by - case . based on the eulerian algorithm we developed in section [ subsec : neweulerian ] and section [ subsec : islecomp ] , we are now able to efficiently construct the isle field of each individual separation factor using one single isocontour extraction . in this section , we present numerical results to show the feasibility of the proposed eulerian approach . we will compare the solution from the new approach with the previous eulerian approach discussed in section [ subsec : oldeulerian ] . it is noteworthy that for a dynamical system in a fixed time window , we can now construct the isle function with an arbitrary choice of the separation factor in only one single numerical flow simulation . therefore , we can easily verify the theoretical connection between ftle and isle , developed in section [ sec : relation ] . this first example is a simple flow taken from @xcite to describe a periodically varying double - gyre . the flow is modeled by the following stream - function @xmath239 \sin(\pi y ) \ , , \ ] ] where @xmath240 in this example , we follow @xcite and use @xmath241 , @xmath242 . we discretize the domain @xmath243\times[0,1]$ ] using 513 grid points in the @xmath244-direction and 257 grid points in the @xmath245-direction . this gives @xmath246 . in figure [ fig : ftle ] , we compare the ftle field @xmath247 of the double - gyre flow with @xmath248 computed using the lagrangian approach and our proposed eulerian approach , respectively . these two solutions match extremely well . we have also checked the @xmath249-error of the forward flow map @xmath250 using different @xmath251 s ranging from @xmath252 to @xmath253 . since we do not have the exact solution for this flow , the _ exact _ solutions are computed using the lagrangian ray tracing with a very small time step . figure [ fig : convergence ] shows the errors in the flow map @xmath254 with respect to different @xmath251 s . we find that the flow map computed using the proposed eulerian approach is approximately second order accurate . figures [ fig : isle1 ] shows the isle fields @xmath255 with several separation factors @xmath0 . in the implementations , the flow map is computed from @xmath2 up to @xmath256 . recall that we impose zero isle value at points whose local separation rate is less than the given separation factor . with the increasing separation factor , the regions with non - zero isle values , get smaller . nevertheless , the regions with non - zero or high isle values , always concentrate near the location of the ftle ridges shown in figure [ fig : ftle ] . we denote by @xmath176 the most prominent ridge originating from the central bottom of the domain . note that a rough estimate of the minimum ftle value on @xmath176 is approximately 0.3 . and so we have the estimate @xmath257 . using theorem [ prop : funprop1 ] , we can always find a tubular neighborhood near @xmath176 of non - zero isle using a separation factor @xmath0 smaller than approximately 20 . in figure [ fig : isle1 ] , we have computed the isle using various separation factors from @xmath258 to @xmath259 . as @xmath0 increases , we can see that there are indeed non - zero isle regions around the major ridge @xmath176 . and evidently , the isle ridges lie within these regions . to better compare the ftle field and the isle field , we plot these two solutions together in figure [ fig : isosurface ] . the @xmath244- and the @xmath245- axis represent the computational domain @xmath9 while the @xmath260 axis denotes the temporal direction . in ( a ) , we plot the ftle field @xmath261 at @xmath262 and @xmath263 at @xmath264 . in ( b - d ) , we plot the isosurfaces of the isle field @xmath265 with @xmath266 , respectively . the @xmath260-values denote the time required for the local separation ratio to achieve @xmath0 for the first time , i.e. the @xmath260-values are the @xmath267 we defined before . near the major ftle ridge , we can see the values of @xmath120 are significantly lower than the remaining part of the domain . therefore the corresponding isle values are generalized maxima . this shows us the isle ridges match with the ftle ridges very well . we consider a simple analytical field @xcite given by @xmath268 and @xmath269 in this example . the computational domain is @xmath270 ^ 2 $ ] and we discretize the domain with @xmath271 . we impose the fixed inflow boundary condition and non - reflective outflow boundary condition as in @xcite . figure [ fig : ftle2 ] shows the ftle field @xmath272 computed using our original eulerian approach and the proposed eulerian approach . it is clear that the new approach can successfully handle the inflow and outflow of particles on the boundaries , just like the previous approach does . next , we are going to apply the results in section [ sec : relation ] to analyze the isle ridges . figure [ fig : ftle2 ] suggests that there is only one ftle ridge . and the minimum value of @xmath273 on the ridge is close to @xmath274 . by the definition of ftle , we have @xmath275 \ , .\ ] ] therefore we can conclude that @xmath276 is a reasonable estimate of the minimum value of @xmath277 on the ridge . we have tested with several separation factors that are near @xmath278 , and the results can be plotted in figure [ fig : isle2a ] . to better visualize the isle field , we plot in figure [ fig : isle2 ] the time @xmath267 required for different separation factors @xmath0 . the shorter the time , the bigger the corresponding isle . [ ! htb ] we consider the dynamical system governed by a duffing and a van der pol oscillator , as in @xcite . the system is given by @xmath279 the computational domain is @xmath280\times [ -1.5,1.5]$ ] , with mesh size @xmath281 . we simulate the flow from @xmath2 to @xmath256 . the discretization size is rather coarse for such a complicated dynamics , but our new approach can still demonstrate the fine details of the transport barrier as shown in figure [ fig : ftle3 ] . there are several major ftle ridges that are easily spotted . the minimum value of @xmath282 on those ftle ridges are very close to @xmath283 . so @xmath284 . theorem [ prop : funprop1 ] implies that the separation factor @xmath285 can approximately single out the isle ridges . we pick several separation factors that are close to @xmath286 , and we find that the isle ridges corresponding to @xmath287 , as shown in figure [ fig : isle3](c ) , have the greatest resemblance to the ftle ridges in figure [ fig : ftle3 ] . [ ! htb ] to further demonstrate the effectiveness and robustness of our proposed approach , we test our algorithm on a real life dataset obtained from the ocean surface current analyses real - time ( oscar ) . the oscar data were obtained from jpl physical oceanography daac developed by esr . it measures ocean surface flow in the region from @xmath288 to @xmath289 in latitude and from @xmath290 to @xmath291 longitude . the time resolution is about 5 days and the spatial resolution is @xmath292 in each direction . in this numerical example , we consider the area near the line islands in the region from @xmath293 to @xmath294 in latitude and from @xmath295 to @xmath296 longitude . to have a better visualization of both the ftle and isle field , we interpolate the velocity data to obtain a finer resolution of @xmath297 days in the temporal direction and @xmath298 in each spatial direction . then we look at the ocean surface current within the first 50 days in the year 2014 to obtain the forward ftle field @xmath299 and the isle field @xmath300 . as demonstrated in figure [ fig : oscar_ftle ] , our new approach works well in extracting fine details of the transport barrier even for real data . there are a few major ftle ridges spotted in the figure . according to theorem [ prop : funprop1 ] , we found that the separation factor @xmath0 roughly equals to @xmath301 to @xmath302 can single out the isle ridges . we have tested several separation factors that are close to @xmath303 , as shown in figure [ fig : oscar_isle ] . we find that the isle ridges corresponding to @xmath259 , in figure [ fig : oscar_isle](b ) , have the greatest resemblance to the ftle ridges in figure [ fig : oscar_ftle ] . as a comparison , we have also plotted the isle field @xmath300 for @xmath304 and @xmath305 in figure [ fig : oscar_isle ] ( a ) and ( c ) , respectively . these results verify the theory developed in section [ sec : relation ] . to summarize , we emphasize the novelty and importance of both the two eulerian algorithms we developed in this paper . based on the first algorithm , we are now able to determine the _ forward _ flow map _ on the fly _ so that the pde is solved _ forward _ in time . unlike the original eulerian algorithm as developed in @xcite , there is no need to store the whole velocity field at all time steps at the beginning of the computations . this makes the forward flow map computations practical . more importantly , because of the first algorithm , we are now able to access all intermediate forward flow maps which is necessary in computing the isle field computations . this eulerian approach is extremely efficient in computing the isle field . to the best of our knowledge , this is the first efficient numerical approach for the isle computations . theoretically , this paper provides a rigorous mathematical analysis to explain the similarity between the ftle ridges and the isle ridges based on @xcite , which has been verified by the proposed eulerian approach . the work of you was supported by the talents introduction project of nanjing audit university , and the natural science foundation of jiangsu higher education institutions of china ( no.16kjb110012 ) . the work of leung was supported in part by the hong kong rgc grants 16303114 and 16309316 . c. w. shu . essentially non - oscillatory and weighted essentially non - oscillatory schemes for hyperbolic conservation laws . in b. cockburn , c. johnson , c.w . shu , and e. tadmor , editors , _ advanced numerical approximation of nonlinear hyperbolic equations _ , volume 1697 , pages 325432 . springer , 1998 . lecture notes in mathematics . w. tang , p.w . chan , and g. haller . accurate extraction of lagrangian coherent structures over finite domains with application to flight data analysis over hong kong international airport . , 20:017502 , 2010 .
lattice simulations are used in many areas of quantum physics , ranging from nuclear and particle physics to atomic and condensed matter physics @xcite . in such calculations the system is solved numerically using a discrete spacetime lattice over a finite volume . in practice this finite volume is usually taken to be a cubic box with periodic boundaries . when simulating composite objects such as bound states , the boundaries of the periodic box will modify quantum wave functions . this leads to finite volume shifts in the binding energies , and detailed knowledge of these finite volume effects is necessary to improve high precision lattice calculations . in ref . @xcite , lscher derived a formula for the finite volume mass shift of s - wave bound states of two particles with reduced mass @xmath1 interacting via a potential with finite range @xmath2 . when a bound state with energy @xmath3 is put in a periodic cubic box of length @xmath4 , its energy in the rest frame is shifted by an amount @xmath5 where @xmath6 is the binding momentum and @xmath7 is the asymptotic wave function normalization defined by @xmath8 for @xmath9 . for potentials with exponential fall - off , @xmath10 for large @xmath11 , the formula is modified by exponentially small corrections when the binding momentum @xmath12 is smaller than @xmath13 . bound states moving in a finite periodic volume also have a topological phase correction to the energy @xcite . this topological phase correction contains information about the number and mass of the constituents of the bound states . it must be included when determining scattering phase shifts for composite objects in a finite volume . in ref . @xcite , we briefly discussed the generalization of lscher s formula ( [ eq : delta_m - luscher ] ) for the finite volume mass shift to higher partial waves . in this paper , we present the full derivation of these results . moreover , we give the general mass shift for states with angular momenta up to @xmath14 . in general , the mass shift for a given state depends on its transformation properties with respect to the symmetry group of the cubic box . in addition to reducing finite volume effects from precision lattice calculations , our finite volume results can also be used as a diagnostic tool to probe the angular momentum and radial structure of the bound state wave function . a summary of our results is given by eq . together with table [ tab : results ] at the end of sec . [ sec : results ] . we also derive a general mass shift formula for two - dimensional systems in a finite area with periodic boundaries . in our discussion here we only consider stable bound states . however , resonances above continuum thresholds can also be described from the volume dependence of avoided level crossings @xcite . our results are universal and can be applied to a wide range of systems . we will discuss a few examples below . in particle physics , there is some interest in hadronic molecules with angular momentum @xcite . in the case of s - waves , the deuteron and some exotic weakly bound states such as the h - dibaryon were recently studied in lattice qcd @xcite . similar investigations for exotic bound states with angular momentum appear feasible in the future . in atomic physics , several experiments have investigated strongly - interacting p - wave feshbach resonances in @xmath15li and @xmath16k @xcite , which can be tuned to produce bound p - wave dimers . if such systems are simulated in a finite volume , our results can be used to describe the finite volume dependence of the dimers . halo nuclei constitute a special class of weakly - bound nuclei with molecular character . they consist of a compact core and one or more nucleons with low separation energy @xcite . among halo nuclei there are some systems with nonzero orbital angular momentum . a well - known example of a p - wave halo state is the @xmath17 excited state in @xmath18be . the electromagnetic properties of the low - lying states in this nucleus can be well described in a two - body halo picture of a @xmath19be core and a neutron @xcite . a similar strategy was applied to radiative neutron capture on @xmath20li @xcite . a related class of systems is given by nuclei with an @xmath21-cluster structure such as @xmath22be and excited states of @xmath23c @xcite . the hoyle state in @xmath23c was recently calculated for the first time in an _ ab initio _ approach using chiral effective field theory and nuclear lattice calculations @xcite . understanding the finite volume corrections is crucial in such a calculation . due to the @xmath21-cluster structure , the volume dependence of the hoyle state is governed by the volume dependence of a three - body system . finally , we note that the asymptotic normalization coefficient of the bound state wave function appears in the mass shift formula . hence , our results can be used to extract this quantity from lattice calculations at finite volume . the asymptotic normalization is directly connected to zero - energy capture reactions @xcite , which play an important role in nuclear astrophysics . the paper is organized as follows . we start with some prerequisites and a general discussion of the finite volume mass shift in section [ sec : boundstates ] . lscher s result for s - waves is reviewed in section [ sec : s - wave ] , while our extension to higher partial waves is given in section [ sec : p - wave - and - higher ] . in particular , we discuss the mass shift for the irreducible representations of the cubic group , relate the sign of the shift to the leading parity , and derive a trace formula for the multiplet - averaged mass shift for states with arbitrary angular momentum @xmath24 . in section [ sec : numtest ] , we verify our results numerically for two model systems . the case of two spatial dimensions is treated in section [ sec:2d ] . finally , a brief summary and outlook is given in section [ sec : conclusion ] . in the appendix , we discuss the relation of the asymptotic normalization constant to scattering parameters and derive explicit shallow binding relations between the effective range , the binding momentum and the asymptotic normalization coefficient . as a starting point , we first review several results from @xcite . we closely follow lscher s derivation , but consider a more general system with arbitrary angular momentum and non - local interactions . we start with some definitions and basic identities about bound states in rotationally symmetric potentials @xmath25 . this will also help to define our notation used throughout the discussion . for a bound state with energy @xmath26 and angular - momentum quantum numbers @xmath27 , we separate the wave function @xmath28 as a product of the radial wave function @xmath29 and spherical harmonics @xmath30 , @xmath31 this leads to the radial schrdinger equation @xmath32u_\ell(r ) = 0 \ , , \label{eq : sg - rad}\ ] ] with @xmath33 and @xmath34 . the mass is denoted by @xmath1 since in later applications it will be the reduced mass for a two - particle system . the normalized radial wave function satisfies @xmath35 for vanishing potential , we have two linearly independent solutions of the free wave equation . these are the riccati bessel functions @xmath36 and riccati neumann functions @xmath37 , which can be expressed in terms of ordinary bessel functions . it is often more convenient to work with the riccati hankel functions @xmath38 these have the asymptotic form @xmath39 as @xmath40 . for future reference , we give the explicit expressions for @xmath41 for @xmath42 : @xmath43 in order to discuss the asymptotic form of bound state wave functions we define solutions @xmath44 of that fulfill the condition @xmath45 the bound - state solution normalized according to can then be written as @xmath46 where @xmath47 and @xmath48 is the asymptotic normalization coefficient . the factor @xmath49 adjusts the phase such that @xmath29 is a real function . if the potential has a finite range @xmath2 , _ i.e. _ , @xmath50 for @xmath9 , we have the exact identity @xmath51 the asymptotic normalization @xmath7 is itself an interesting quantity . in low - energy astrophysical reactions it determines the capture rate of a single proton or neutron at zero relative energy @xcite . in the shallow binding limit it is also directly related to low - energy scattering parameters . for @xmath52 , [ eq : gamma - r ] @xmath53 whereas for angular momenta @xmath54 one has @xmath55 we note that @xmath7 will in general depend upon the binding momentum @xmath12 . we will derive the relations and in appendix [ sec : shallowbinding ] . the s - wave formula is well known and given in @xcite , although without the @xmath56 correction estimate . for higher angular momentum the situation is qualitatively different due to the divergence of @xmath57 at @xmath58 . we now consider a system of two spinless particles with reduced mass @xmath1 with zero total momentum . we will work with relative coordinates @xmath59 . the interaction is given by a potential @xmath60 , which we in general allow to be non - local in configuration space . it is assumed to be hermitian , rotationally symmetric , and to have a finite range @xmath2 , _ i.e. _ , @xmath61 we consider the case when the system has a bound state @xmath62 with angular quantum numbers @xmath27 . the schrdinger equation , @xmath63 can be written as @xmath64 in configuration space . we note that for a local potential , @xmath65 eq . reduces to the familiar form @xmath66\,\psi_b({\mathbf{r } } ) = -{e_b}\,\psi_b({\mathbf{r } } ) \ , . \label{eq : sg - local}\ ] ] regardless of the locality of the interaction , the wave function @xmath67 has the asymptotic form @xmath68 where @xmath6 is the binding momentum . we now consider what happens when the two - body system is put into a cubic periodic box with length @xmath69 . for this problem it is convenient to define a periodic extension of the potential @xmath70 we take @xmath71 to be an exact periodic solution of the finite volume schrdinger equation , @xmath72 with the finite - volume hamiltonian @xmath73 . it is clear that @xmath74 approaches the infinite volume eigenvalue @xmath75 and @xmath76 as @xmath77 . we now derive a formula for the finite volume mass ( energy ) shift , @xmath78 let us define a state @xmath79 by adding together periodic copies of the infinite volume wave function in , @xmath80 this clearly satisfies the periodicity condition . acting upon this state with the finite volume hamiltonian , we get @xmath81 the final result can be written as @xmath82 where we have defined @xmath83 as @xmath84 with the substitution @xmath85 for each term in the sum , this can be rewritten as @xmath86 due to the finite range of the potential we only get contributions from the domain @xmath87 . we note that @xmath88 when @xmath89 and @xmath90 . therefore we can use the asymptotic form of the wave function and find that @xmath91 . this means that @xmath79 is an approximate solution of the finite volume schrdinger equation for large @xmath4 . motivated by this , we write the exact finite volume solution @xmath71 explicitly as @xmath92 we take @xmath71 to be unit - normalized per volume @xmath93 . the same is true of @xmath79 up to corrections of order @xmath94 . we will choose @xmath21 such that @xmath95 consider now the matrix element @xmath96 . acting with @xmath97 on @xmath79 , we get @xmath98 according to and . on the other hand , acting with @xmath97 on @xmath99 yields @xmath100 combining these two results we find @xmath101 we first consider the numerator in this expression . obviously , @xmath102 we note that the factor of @xmath21 here will cancel the @xmath21 in the denominator of . we can now simplify further starting with @xmath103 for each @xmath104 we can make the substitutions @xmath105 and @xmath85 . these leave the integrals invariant , and we get @xmath106 setting @xmath107 and @xmath108 yields @xmath109 where @xmath110 counts the number of repeated periodic copies . the fact that @xmath110 diverges simply reflects the fact that we are working with periodic wave functions with normalization measured per volume @xmath93 , and @xmath110 will cancel in the final result . for the integral to be non - zero , both @xmath111 and @xmath112 have to be close to @xmath113 due to the finite range of the potential . from the assumption @xmath69 it then follows that all terms with @xmath114 are suppressed by at least a factor of @xmath115 , and we have @xmath116 the possible nonvanishing values of @xmath117 are @xmath118 . we therefore arrive at @xmath119 for the denominator in , an analogous procedure yields @xmath120 \label{eq : psi0-psi0-final}\ ] ] with the same constant @xmath110 as above . combining and , the constant @xmath110 cancels and we get @xmath121 where we have renamed @xmath122 back to @xmath104 . is a general result valid for any angular momentum . the dependence of the mass shift upon quantum numbers @xmath27 will emerge from the wave function @xmath123 and the resulting overlap integrals in . in the following , we explore this dependence in detail and denote the mass shift as @xmath124 . for @xmath52 the asymptotic wave function is given as @xmath125 with @xmath126 due to the finite range @xmath90 of the potential we only have contributions with @xmath127 in . hence we can insert the asymptotic form for @xmath128 and get @xmath129 we can furthermore use the schrdinger equation to eliminate the potential . doing this and then renaming @xmath130 , we get @xmath131 \psi_b^\big(\|{\mathbf{r}}\|\big)\right\ } \frac{{\mathrm{e}}^{-\kappa\|{\mathbf{r}}+{\mathbf{n}}l\|}}{\|{\mathbf{r}}+{\mathbf{n}}l\| } + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big ) \\ & = \frac\gamma{\sqrt{4\pi}}\sum\limits_{\|{\mathbf{n}}\|=1 } \int{\mathrm{d}^3r}\,\psi_b^\big(\|{\mathbf{r}}-{\mathbf{n}}l\|\big ) \frac{1}{2\mu}\left[{\delta}_r-\kappa^2\right]\frac{{\mathrm{e}}^{-\kappa r}}{r } + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big ) \ , . \end{split}\ ] ] in the second line we have shifted the integration variable and used partial integration to let the laplacian act on @xmath132 . finally , we use the fact that @xmath133 is a green s function for the operator @xmath134 , @xmath135\frac{{\mathrm{e}}^{-\kappa r}}{4\pi r } = -{\delta^{(3)}}({\mathbf{r } } ) \ , . \label{eq : helmholtz - gf}\ ] ] this allows us to perform the integral and arrive at @xmath136 in the last step we have inserted the asymptotic form of the wave function for @xmath137 , and the sum yields a factor of six . this is just lscher s result as given in the introduction . we now discuss the generalization of the mass shift formula to arbitrary angular momentum . the general form for the asymptotic wave function is @xmath138 inserting this into and performing steps analogous to those presented above for the s - wave case , we find @xmath139\!\psi_b^({\mathbf{r}}-{\mathbf{n}}l)\right\}\ , { y}_\ell^m(\theta,\phi)\frac{{\mathrm{i}}^\ell\gamma\hat{h}_\ell^+({\mathrm{i}}\kappa r)}r + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big)\ , . \label{eq : delta_m - int - asympt}\ ] ] the crucial ingredient is the relation @xmath140 \ , , \label{eq : sh - hankel}\ ] ] where @xmath141 are the solid harmonics defined via @xmath142 . the derivation of follows from lemma b.1 in ref . @xcite , which proves that @xmath143 for any smooth function @xmath144 . we obtain by using the relation , @xmath145 and noting that @xmath146 and @xmath147 , where @xmath148 is a spherical hankel function of the first kind . we can illustrate using an example . for the case @xmath149 and @xmath150 we have @xmath151 and @xmath152 . a straightforward calculation shows that indeed @xmath153 \,,\ ] ] with @xmath154 . using to rewrite , we get @xmath155\!\psi_b^({\mathbf{r}}-{\mathbf{n}}l ) \vphantom{\left[\frac{{\mathrm{e}}^{-\kappa r}}{r}\right]}\right\ } \left\{r_\ell^m\left(-\frac1\kappa{\boldsymbol{\nabla}}_r\right ) \left[\frac{{\mathrm{e}}^{-\kappa r}}{r}\right]\right\ } + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big ) \,.\ ] ] we now integrate by parts and pass the laplacian through the differential operator @xmath156 . since the operators both consist of partial derivatives , this is not a problem when the wave function is smooth . we assume that this is the case , with the possible exception of a measure zero region that can be omitted from the integral . the partial integrations give a factor @xmath157 . we can now proceed in exactly the same way as for s - waves . we perform one more integration by parts so that the laplacian acts on @xmath132 . this yields a delta function times a factor of @xmath158 , and the final result is then @xmath159 for @xmath160 we can insert the asymptotic form since it is evaluated in the asymptotic region . for @xmath149 , we find the same result for all three p - wave states , @xmath161 when compared to the s - wave case , the sign of the p - wave mass shift is reversed while the magnitude is the same . qualitatively , this means that s - wave bound states are more deeply bound when put in a finite volume while p - wave bound states are less bound . this will be discussed in more detail later . we next discuss the results for @xmath162 . from , we find @xmath163 where @xmath164 we note that the size and even the sign of the mass shift depends on the quantum number @xmath165 . to understand this effect , we need to take into account that our cubic finite volume breaks the rotational symmetry group down to a cubic subgroup . the cubic symmetry group @xmath166 is a finite subgroup of @xmath167 with 24 elements . there are five irreducible representations of @xmath166 . they are conventionally called @xmath168 , @xmath169 , @xmath170 , @xmath171 and @xmath172 , and their dimensionalities are @xmath173 , @xmath173 , @xmath174 , @xmath175 and @xmath175 , respectively . irreducible representations @xmath176 of the rotation group @xmath167 are reducible with respect to @xmath166 for @xmath177 . for further details about the decomposition see , for example , ref . @xcite . in our discussion we assume that the infinite volume system has no partial wave mixing , such that orbital angular momentum @xmath24 is a good quantum number . we also assume that there are no accidental degeneracies in the bound state spectrum , so we can use @xmath24 as a label for the family of cubic representations split apart at finite volume . parity invariance remains unbroken by the cubic volume , and we have @xmath178 just as in the infinite volume case . for clarity , however , we will indicate parity explicitly with @xmath179 superscripts in the following . with our assumptions , an s - wave state in infinite volume will map onto an @xmath180 state at finite volume . also a p - wave triplet will map onto the three elements of the @xmath181 representation at finite volume . for d - waves , however , the five d - wave states are split into a @xmath182 triplet and an @xmath183 doublet , @xmath184 in the following we use the notation @xmath185 , @xmath186 , for the basis vectors of the irreducible cubic representations . we can rewrite the finite volume mass shift in eq . as @xmath187 where @xmath188 is the translation operator by displacement @xmath189 . we can also calculate the matrix elements of @xmath190 in the @xmath27 basis . in this case there will be off - diagonal matrix elements connecting @xmath27 and @xmath191 when @xmath165 and @xmath192 are equivalent modulo 4 . according to ref . @xcite , the unitary transformation between the two basis sets for the five d - wave states is @xmath193 and @xmath194 so , for example , we have @xmath195 where we have defined @xmath196 as a straightforward generalization of . as expected from cubic symmetry , the mass shift is the same for all three @xmath182 states , and the same for both @xmath183 states . to summarize our results , we write the mass shift for a state belonging to irreducible representation @xmath197 with angular momentum @xmath24 as @xmath198 we list the coefficients @xmath199 for @xmath200 in table [ tab : results ] . .coefficient @xmath201 in the expression for the finite volume mass shifts for @xmath200 . @xmath197 indicates the corresponding representation of the cubic group . [ cols="^,^,^",options="header " , ] the sign of the finite volume mass shift can be understood in terms of the parity of the wave function . in infinite volume the tail of each bound state wave function must vanish at infinity . in the finite volume , however , the bound state wave functions with even parity along a given axis can remain nonzero everywhere . only the derivative needs to vanish , and the kinetic energy is lowered by broadening the wave function profile . on the other hand , a wave function with odd parity along a given axis must change sign across the boundary . in this case the wave function profile is compressed and the kinetic energy is increased . we have illustrated both cases for a one - dimensional square well potential in fig . [ fig : wf ] . in three dimensions , the situation is slightly more complicated , which can be seen from the fact that for @xmath162 the sign of the mass shift depend on the representation of the cubic group even though the parity is just @xmath202 for all states . in order to understand this , we consider the basis polynomials for the cubic representations . these basis polynomials are obtained by decomposing the cubic basis vectors in terms of solid harmonics which are homogeneous polynomials in @xmath203 , @xmath204 and @xmath205 . for @xmath206 the basis polynomials are also given explicitly in @xcite . for a given polynomial @xmath207 , we define its _ leading parity _ as @xmath208 where @xmath209 is the maximum degree of @xmath210 with respect to any one of the three variables . it is this leading parity that determines the asymptotic behavior of the mass shift as @xmath211 . more precisely , we have @xmath212 for the @xmath199 in eq . [ eq : delta_m - gamma ] . it can easily be checked that this relation holds for all results presented in table [ tab : results ] . for @xmath162 , for example , we have the basis polynomials @xmath213 and hence @xmath214 for the @xmath182 representation and @xmath215 for the @xmath183 representation . the expressions for the finite volume mass shift become simpler when we sum over all @xmath165 for a given @xmath24 . we can rewrite as @xmath216 inserting the asymptotic form of the wave function , @xmath217^ * \bigg\|_{{\mathbf{r}}={\mathbf{n}}l } \,,\ ] ] and using a second time yields @xmath218 \bigg\|_{{\mathbf{r}}={\mathbf{n}}l } \\ + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big ) \,.\ ] ] now , from the well - known relation @xmath219 and @xmath220 we get an analogous expression for the solid harmonics , which then carries over to @xmath221 for any sufficiently smooth function @xmath144 . finally , we have @xmath222 which follows from eq . . putting everything together , we arrive at @xmath223 \bigg\|_{{\mathbf{r}}={\mathbf{n}}l } + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big ) \\ & = ( -1)^{\ell+1}(2\ell+1)\cdot3\|\gamma\|^2\frac{{\mathrm{e}}^{-\kappa l}}{\mu l } + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big)\ , , \end{split } \label{eq : avshift-1}\ ] ] where the sum just yields a factor of six . dividing by @xmath224 , we obtain the average mass shift for states with angular momentum @xmath24 , @xmath225 apart from the alternating sign , this average shift is independent of @xmath24 . eq . can be verified explicitly for the the results presented in sec . [ sec : results ] ( _ cf . _ table [ tab : results ] ) . for @xmath162 , for example , one has to average over the three - dimensional representation @xmath182 and the two - dimensional representation @xmath183 . in order to verify our predictions numerically , we put the schrdinger equation on a discrete spatial lattice such that the hamiltonian becomes an ordinary matrix . we then calculate the corresponding energy eigenvalues and eigenvectors . we use a hat symbol to denote dimensionless lattice units . for example , we have @xmath226 where @xmath227 denotes the lattice spacing . the free lattice hamiltonian is given by @xmath228 \label{eq : h - latt}\ ] ] where @xmath229 and @xmath230 are creation and annihilation operators for a lattice site @xmath231 and @xmath232 is a unit vector in the @xmath233-direction . the corresponding lattice dispersion relation is @xmath234 with the lattice function @xmath235\ ] ] and the lattice momenta @xmath236 the binding momentum for a bound state with energy @xmath237 is determined by @xmath238 the lattice green s function for the hamiltonian is @xmath239 we impose periodic boundary conditions by defining the distance @xmath240 to the origin as @xmath241 we calculate the mass shift using three different methods : 1 . as a direct difference in energies , eq . , where we use a very large volume ( @xmath242 ) to approximate the infinite volume result 2 . from the overlap formula . 3 . using discretized versions of and , which we obtain by replacing @xmath132 with the lattice green s function . more precisely , we write the asymptotic bound state wave function as @xmath243 and replace the continuum green s function @xmath244 with the lattice version @xmath245 effectively , this amounts to the replacement @xmath246 in the mass shift formula . the lattice green s function is also used to calculate the asymptotic normalization @xmath7 from the lattice data . this procedure has the advantage of avoiding large lattice discretization errors . in the following we report physical quantities in units where the reduced mass @xmath1 is set to 1 . we first use a gaussian potential , @xmath247 with @xmath248 and @xmath249 . this potential does not have a finite range in a strict mathematical sense , but the range corrections can be entirely neglected in comparison with other errors in our numerical calculation . the smoothness of the gaussian potential helps to minimize lattice discretization artifacts . in fig . [ fig : shifts - sp - gauss ] we show the s- and p - wave mass shifts obtained with the three methods described in sec . [ sec : methods ] above . the results from the three different methods described above agree well for both s- and p - waves . in order to compare the dependence on the box size @xmath4 with the predicted behavior , we have plotted @xmath250 against @xmath4 ( we use the absolute value of @xmath190 since the s - wave mass shift is negative ) . for both s- and p - waves , the expected linear dependence is clearly visible . as functions of the box size @xmath4 ( in lattice units ) for a gaussian potential . we show the results obtained from the direct difference eq . ( crosses ) , evaluation of the overlap integral eq . ( squares ) , and discretized versions of eqs . , ( circles ) . the dashed lines show linear fits to the overlap integral results.,scaledwidth=50.0% ] when we perform a linear fit to the overlap integral data ( dashed lines in fig . [ fig : shifts - sp - gauss ] ) we obtain @xmath251 , @xmath252 for the s - wave results and @xmath253 , @xmath254 for the p - wave results . the values for the asymptotic normalization are in good agreement with the results @xmath255 ( s - wave ) and @xmath256 ( p - wave ) that are obtained directly from the @xmath257 data . inserting the corresponding energy eigenvalues into the lattice dispersion relation , we find @xmath258 ( s - wave ) and @xmath259 ( p - wave ) , again in quite good agreement with the fit results . the remaining small discrepancies can be attributed to the mixing with higher partial waves induced by the lattice discretization and the fact that we have not performed a continuum extrapolation to vanishing lattice spacing . for a simple step potential , @xmath260 which we use with @xmath261 and @xmath262 , the numerical calculation becomes more difficult since the discontinuous shape introduces considerable lattice artifacts . yet we discuss it here because it has a strict finite range and we find that for a small lattice spacing of @xmath263 the results are satisfactory . in fig . [ fig : shifts - sp - step ] , we show a plot analogous to the one presented for the gaussian potential . again , the results from the different methods agree well and the expected linear behavior is clearly visible . furthermore , the results from the three methods agree well with each other already for smaller @xmath4 ( compared to the results for the gaussian potential ) , as expected from the fact that the step potential does not have a tail . as functions of the box size @xmath4 ( in lattice units ) for a simple step potential . the symbols are as in fig . [ fig : shifts - sp - gauss].,scaledwidth=50.0% ] from fitting to the overlap integral data ( dashed lines in fig . [ fig : shifts - sp - step ] ) we obtain @xmath264 , @xmath265 for the s - wave results and @xmath266 , @xmath267 for the p - wave results . from the @xmath257 data we find @xmath268 , @xmath269 ( s - wave ) and @xmath270 , @xmath271 ( p - wave ) . given that we do not have error estimates for the @xmath257 results , the overall agreement is quite good . finally , we also check our result for the d - wave mass splittings using the step potential with @xmath263 . in fig . [ fig : shift - d - t2-e - step ] , we show the mass shift for the d - wave states in both the @xmath182 and the @xmath183 representation . due to the polynomial coefficients @xmath199 ( see eq . and table [ tab : results ] ) one does not expect a linear dependence on @xmath4 for @xmath250 . hence , we simply plot @xmath190 as a function of @xmath4 directly and do not perform a fit . nevertheless , we see that ( except for very small @xmath4 , where obviously the condition @xmath272 is not satisfied ) the agreement between the three methods to calculate @xmath190 is very good and hence conclude that our mass shift formula indeed gives the right result also for @xmath162 . for @xmath182 rep . ( left panel ) and @xmath183 rep . ( right panel ) as a function of the box size @xmath4 ( in lattice units ) for a simple step potential . the symbols are as in fig . [ fig : shifts - sp - gauss].,title="fig:",scaledwidth=49.0% ] for @xmath182 rep . ( left panel ) and @xmath183 rep . ( right panel ) as a function of the box size @xmath4 ( in lattice units ) for a simple step potential . the symbols are as in fig . [ fig : shifts - sp - gauss].,title="fig:",scaledwidth=49.0% ] in this section we derive a formula for the finite - volume ( or rather finite - area ) mass shift of bound states in two - dimensional systems . the results can be used , for example , in lattice simulations of cold atomic systems , which can be prepared experimentally to be effectively two - dimensional @xcite . in two dimensions , the schrdinger equation is @xmath273 with @xmath274\psi_b({\mathbf{r}})\ ] ] in polar coordinates . states are described by a single angular momentum quantum number @xmath275 , and for the wave function we have the separation @xmath276 with @xmath277 the two linearly independent solutions of the free radial equation @xmath278 are just the bessel and neumann functions @xmath279 and @xmath280 . for a bound state , we have @xmath281 , and the wave function has the asymptotic form @xmath282 where @xmath283 is the modified bessel function of the second kind . it is related to the hankel function of the first kind , @xmath284 via @xmath285 as in the three - dimensional case , @xmath7 is the asymptotic normalization . inserting into yields a form which is more similar to the three - dimensional expression , eq . . to make the analogy to the calculations in sec . [ sec : p - wave - and - higher ] as explicit as possible , we will use the hankel function in the following intermediate steps and only express the final results in terms of the modified bessel function . nearly all of the three - dimensional calculation carries over and we just replace all exponential terms by hankel functions . the overlap integral for the mass shift is now @xmath286 from the asymptotic form of the hankel function , @xmath287 it is clear that in principle we still have an exponential behavior . in deriving eq . we have used this to write @xmath288 in the following , we will simply write the correction terms as @xmath289 , as in the three - dimensional case . the two - dimensional analogue of the relation is @xmath290 where @xmath291 . this follows from @xmath292 and @xmath293 the proof for eq . can be carried out in the same manner as the three - dimensional proof in lemma b.1 in @xcite , where one uses the expansion of @xmath294 ( 2d vectors ) in terms of bessel functions . as the final ingredient , we have @xmath295 \frac{{\mathrm{i}}}4 h_0^{(1)}({\mathrm{i}}\kappa r ) = -\delta^{(2)}({\mathbf{r } } ) \ , . \label{eq : helmholtz - gf-2d}\ ] ] using all this in steps completely analogous to those in three dimensions , we get @xmath296 for @xmath150 ( two - dimensional s - waves ) , this directly yields @xmath297 in fact , eq . can be simplified further . inserting the asymptotic form for the wave function for @xmath298 and using a second time gives @xmath299 \bigg\|_{{\mathbf{r}}={\mathbf{n}}l } \\ + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big ) \,.\end{gathered}\ ] ] from and @xmath300 it is clear that @xmath301which then yields @xmath302 for any sufficiently smooth @xmath144 . this is essentially the same relation that we used to derive the trace formula in the three - dimensional case , only that here we do not have to sum over different @xmath165 . together with the two - dimensional analogue of , @xmath303 we then get @xmath304\bigg\|_{{\mathbf{r}}={\mathbf{n}}l } + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big ) \\ & = ( -1)^{m+1}\cdot\frac{2\|\gamma\|^2}{\mu } k_0(\kappa l ) + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big ) \ , . \end{split } \label{eq : delta_m - general-2d-2}\ ] ] as we shall see in the following , this is the final result for @xmath150 and any odd @xmath165 , whereas for even @xmath305 things become slightly more complicated . in general , we have to take into account that the finite volume breaks the original planar rotational symmetry of the system down to the symmetry group of a square . we find that states with the same absolute value of @xmath165 may mix to form good eigenstates in the finite volume . more precisely , we have the symmetric and antisymmetric combinations @xmath306 for @xmath305 . when we calculate the mass shift for these states ( in the same way as described in sec . [ sec : results ] ) , we get mixing terms of the form @xmath307 since the condition for the mixing of states is @xmath308 they do not play a role for odd @xmath165 ( in fact , they vanish in this case ) . for even @xmath165 , however , we have to take them into account and find @xmath309 as our final result . as an illustration , we give the explicit results for @xmath310 : @xmath311 + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big ) \,,\ ] ] @xmath312 + { \mathcal{o}}\big({\mathrm{e}}^{-{\sqrt2\kappa l}}\big ) \,.\ ] ] in this work we have derived explicit formulae for the mass shift of p- and higher - wave bound states in a finite volume and discussed their decomposition into states transforming according to the representations of the cubic group . we have compared our numerical results for @xmath313 with numerical calculations of the finite - volume dependence for lattice gaussian and step potentials and found good agreement with the predictions . for @xmath314 , the mass shift of a given state @xmath27 depends on the angular momentum projection @xmath165 due to the breaking of rotational symmetry . when we average over all @xmath165 in a multiplet , however , the absolute value of the mass shift is even independent of @xmath24 . the mass shift for states in representations of the cubic group , however , is the same for all states . the sign of the mass shift can be understood from the leading parity of the representations . finally , we have derived corresponding expressions for the finite volume mass shift in two - dimensional systems . lattice calculations also provide a method to extract asymptotic normalization coefficients , which are of interest in low - energy astrophysical capture reactions . we have pointed out how the asymptotic normalization and binding momentum of a shallow bound state can be used to extract the effective range from a simulation . our work provides a general framework for future lattice studies of molecular states with angular momentum in systems with short - range interactions . applications to nuclear halo systems and molecular states in atomic and hadronic physics appear promising . an important next step would be to include coulomb effects into the framework in order to investigate the volume dependence of bound states of charged particles , which are much easier to treat experimentally . in particular , this extension of the formalism is important if one wants to describe proton - halo nuclei . another interesting direction would be to analyze the volume dependence of resonances along the lines of refs . @xcite . this research was supported in part by the dfg through sfb / tr 16 `` subnuclear structure of matter '' , the bmbf under contract no . 06bn9006 , and by the us department of energy under contract no . de - fg02 - 03er41260 . s.k . was supported by the `` studienstiftung des deutschen volkes '' and by the bonn - cologne graduate school of physics and astronomy . in this section , we derive the relation between the asymptotic normalization of the bound state wave function and the effective range of the corresponding two - particle scattering process that we have quoted in sec . [ sec : prereq ] . it was already pointed out by lscher in @xcite that the asymptotic normalization @xmath7 of the bound state wave function is related to scattering parameters . more precisely , the analytically - continued elastic scattering amplitude in forward direction has a pole at the bound state energy , and the residue of this pole is proportional to @xmath315 . in the limit of shallow bound states , _ i.e. _ , @xmath316 , it is possible to make a more direct connection to the effective range in the corresponding scattering channel . the crucial ingredients for this are given in a general investigation of wigner causality bounds @xcite . we express the elastic scattering amplitude @xmath317 in terms of the scattering phase shift @xmath318 , @xmath319 } \,,\ ] ] for which we have the well - known effective range expansion @xmath320 where @xmath321 and @xmath322 are the scattering and effective range parameters , respectively . a scattering state is described by a wave function @xmath323 that is a solution of the radial schrdinger equation for positive real center - of - mass momentum @xmath324 . as it is done in @xcite , we choose the normalization such that outside the range of the potential ( @xmath9 ) we have @xmath325 \,.\ ] ] for a bound state , we have @xmath47 and @xmath326 . hence , the wave function is @xmath327 \\ = { \mathrm{i}}^\ell \kappa^\ell \hat{h}_\ell^+({\mathrm{i}}\kappa r)\ ] ] in this case . we immediately see that @xmath328 where @xmath29 is the radial bound state wave function from the previous sections ( normalized to @xmath173 ) . note that the phase convention is chosen such that the wave functions are real . from @xcite we have the relation @xmath329 ^ 2 \label{eq : er-1}\ ] ] with @xmath330 for @xmath52 , we get @xmath331 note that @xmath332 as @xmath316 , and since we only consider finite - range potentials , we can in fact write @xmath333 . using the normalization of the wave function we can rewrite as @xmath334 ^ 2\right\ } \ , . \label{eq : er-2}\ ] ] the remaining integral can be expressed as @xmath335 ^ 2 = \kappa^{2\ell}\int\limits_r^\infty\mathrm{d}r\ , \left[{\mathrm{i}}^\ell\hat{h}_\ell^+({\mathrm{i}}\kappa r ) \right]^2 \ , . \label{eq : int - exterior - intermediate}\ ] ] for @xmath52 , we get @xmath335 ^ 2 = \frac{{\mathrm{e}}^{-2\kappa r}}{2\kappa } = \frac1{2\kappa } - r + { \mathcal{o}}(\kappa ) { { \ \ } \text{as}{\ \ } } \kappa\to0 \,.\ ] ] together with this yields @xmath336 up to the given order this is equivalent to the relation @xmath337 from @xcite . for @xmath54 , the integral is @xmath335 ^ 2 = \frac{\gamma\left(\ell-\frac12\right)\gamma\left(\ell+\frac12\right)}\pi \left(\frac r2\right)^{-2\ell+1 } + { \mathcal{o}}(\kappa ) { { \ \ } \text{as}{\ \ } } \kappa\to0 \ , . \label{eq : int - exterior}\ ] ] we see that the leading term exactly cancels the one in such that we arrive at @xmath338 in all expressions above , the asymptotic normalization of course depends on the binding momentum , _ i.e. _ , @xmath339 , only in general this dependence can not be calculated analytically . using the results that we have derived in this paper , it is possible to determine both @xmath12 and @xmath7 from the volume dependence of a given bound state . if this state is sufficiently shallow , the formulae derived in this chapter can then be used to get an estimate for the effective range of the interaction . d. lee , prog . part . nucl . * 63 * ( 2009 ) 117 [ arxiv:0804.3501 [ nucl - th ] ] . a. bazavov _ et al . _ , rev . * 82 * ( 2010 ) 1349 [ arxiv:0903.3598 [ hep - lat ] ] . s. r. beane , w. detmold , k. orginos and m. j. savage , prog . * 66 * ( 2011 ) 1 [ arxiv:1004.2935 [ hep - lat ] ] . m. lscher , commun . * 104 * ( 1986 ) 177 . s. bour , s. knig , d. lee , h .- w . hammer , and u .- g . meiner , [ arxiv:1107.1272 [ nucl - th ] ] . c. a. regal , c. ticknor , j. l. bohn and d. s. jin , phys . * 90 * ( 2003 ) 053201 [ arxiv : cond - mat/0209071 ] . c. h. schunck _ et al . _ , phys . a * 71 * ( 2005 ) 045601 [ arxiv : cond - mat/0407373 [ cond-mat.soft ] ] . s. typel and g. baur , phys . * 93 * ( 2004 ) 142502 [ arxiv : nucl - th/0406068 ] . w . hammer and d. r. phillips , nucl . phys . a * 865 * ( 2011 ) 17 [ arxiv:1103.1087 [ nucl - th ] ] . g. rupak and r. higa , phys . * 106 * ( 2011 ) 222501 [ arxiv:1101.0207 [ nucl - th ] ] . a. tohsaki , h. horiuchi , p. schuck and g. rpke , phys . rev . lett . * 87 * ( 2001 ) 192501 [ arxiv : nucl - th/0110014 ] . j. r. taylor , _ scattering theory : the quantum theory of nonrelativistic collisions _ , dover publications , inc . m. l. goldberger and k. m. watson , _ collision theory _ , john wiley & sons ,
over the last fifteen years there has been a surge of studies in anomalous diffusion , characterized by the deviation of the mean squared displacement ( msd ) @xmath5 of a stochastic process with the probability density function @xmath1 to find the particle at position @xmath6 at time @xmath7 , from the linear time dependence @xmath8 of ordinary brownian motion @xcite . anomalous diffusion is usually characterized in terms of the power - law form @xmath9 with the anomalous diffusion coefficient @xmath10 of physical dimension @xmath11 and the anomalous diffusion exponent @xmath12 . depending on the value of @xmath12 we distinguish subdiffusion ( @xmath13 ) and superdiffusion ( @xmath14 ) @xcite . since the milestone discoveries of superdiffusion in turbulence already in 1926 @xcite and of subdiffusion in amorphous semiconductors @xcite the recent vast increase of interest in anomalous diffusion is due to its discovery in numerous microscopic systems , in particular in biological contexts . the cytoplasm of biological cells is heavily crowded with various obstacles , including proteins , nucleic acids , ribosomes , the cytoskeleton , as well as internal membranes compartmentalizing the cell @xcite . diffusion of natural and artificial tracers in this complex environment is often subdiffusive . similar situations are encountered in cell membranes @xcite . the experimental evidence for subdiffusion in the crowded cytoplasm of living cells ranges from the motion of small labeled proteins @xcite over mrna molecules and chromosomal loci @xcite , lipid and insulin granules @xcite , virus particles @xcite , as well as chromosomal telomeres @xcite and cajal bodies @xcite inside the nucleus . subdiffusion was observed for membrane resident proteins experimentally @xcite and for membrane lipid molecules in computer simulations @xcite . in controlled _ in vitro _ experiments with artificial crowders , anomalous diffusion was consistently observed @xcite . on larger scales anomalous diffusion was observed , for instance , for the motion of bacteria in a biofilm @xcite . the observed subdiffusion was ascribed to various physical mechanisms @xcite . apart from the apparent transient anomalous diffusion caused by a crossover from free normal diffusion to the plateau value of the msd @xcite , typically , three main families of anomalous diffusion processes are considered : ( i ) diffusion in a fractal environment where dead ends and bottlenecks slow down the motion on all scales @xcite ; ( ii ) motion in a viscoelastic environment , in which the effective anomalous motion of the tracer particle in the correlated many - body environment shows long - ranged antipersistent motion @xcite . the latter process is associated with fractional brownian motion ( fbm ) and generalized langevin equation motion with a power - law memory form of the friction kernel @xcite . ( iii ) and continuous time random walk ( ctrw ) models , in which the moving particle is successively trapped by binding events to the environment or caging effects for waiting times @xmath15 distributed like a power - law @xmath16 with @xmath13 @xcite . all three mechanisms lead to the power - law msd ( [ msd ] ) and they were indeed identified as processes generating the motion of different tracers in different cellular environments @xcite or in colloidal systems _ in vitro _ @xcite . these three anomalous diffusion mechanisms characterized by anomalous diffusion with a constant in time generalized diffusivity were identified in experiments involving fairly large endogenous as well as artificial tracers . recent experiments on eukaryotic cells @xcite using considerably smaller tracer proteins sampling over much larger subvolumes of the cell indicated a systematic variation of the cytoplasmic diffusivity with the separation from the cell nucleus . these spatial diffusivity gradients are due partly to the non - uniform distribution of crowders in the cytoplasm . this distribution can non - trivially affect the diffusion of tracers of different sizes in the cell cytoplasm . _ in vitro _ , fast gradients of the diffusivity can be realized , for instance , via a local variation of the temperature in thermophoresis experiments @xcite , as sketched in fig . [ fig - model ] . the diffusion of brownian particles in explicit solvents with temperature gradients was recently studied by multi - particle collision dynamics @xcite . fluctuation - dissipation relations and spurious drift effects in systems with spatially varying friction coefficient were recently treated theoretically in ref . @xcite . on larger scales , the diffusion of water molecules monitored by diffusive magnetic resonance imaging in the brain white matter was demonstrated to be heterogeneous and strongly anisotropic @xcite . the anisotropy is due to the presence of some spatially - oriented structures in the tissue and obstacles which give rise to a tensorial character of the apparent diffusion coefficient . the existence of a population splitting into two pools of water molecules with slow and fast diffusivities was shown for brain white matter @xcite . finally , spatial heterogeneities are also abundant in the completely different context of anomalous diffusion in subsurface hydrology @xcite . . the diffusivity of the random walker is coupled to the spatial variation of the temperature of the environment , as symbolized by the speed of the runner.,width=302 ] in the present study we examine the effects of the strength of the diffusivity gradient as defined by the scaling exponent @xmath4 and of the initial particle position @xmath17 in heterogeneous diffusion processes ( hdps ) with space - dependent diffusion coefficient @xmath18 . we pay particular attention to a phenomenon , that recently received considerable attention for its immediate relevance to the surging field of single particle tracking experiments in microscopic systems , namely , the so - called _ weak ergodicity breaking_. this is the distinct disparity between physical observables depending on whether they are evaluated in the conventional ensemble sense or , from measured time series @xmath19 of the particle position , in terms of time averages @xcite . we find that the hdp process gives rise to weakly non - ergodic behavior with a pronounced amplitude scatter of the time averaged msd of individual trajectories . we obtain details of the distribution of this scatter as well as the frequently used ergodicity breaking and non - gaussianity parameters . our analysis uncovers remarkable similarities of the hdp process with those of ctrw motion . in particular , we study the behavior of the hdp at the critical value of the scaling exponent of the diffusivity . the paper is organized as follows . we introduce the hdp model with @xmath6-dependent diffusivity in sec . [ sec - model ] . the main results for the evolution of the msd , the time averaged msd , the probability density function , the ergodicity breaking and non - gaussianity parameters in the whole range of the model parameters are then presented in sec . [ sec - main - results ] . we discuss our results and point out the directions for future research in sec . [ sec - discussion ] . hdps are defined in terms of the multiplicative yet markovian langevin equation @xcite @xmath20 where @xmath18 is the position - dependent diffusion coefficient and @xmath21 represents white gaussian noise . in what follows we concentrate on the power - law form @xmath22 for the diffusivity , where the amplitude @xmath23 has dimension @xmath24 . logarithmic and exponential forms for @xmath18 were considered in ref . the offset @xmath25 in eq . ( [ eq - dx ] ) avoids either divergencies of @xmath18 ( @xmath26 ) or stalling of the particle ( @xmath27 ) around @xmath28 in the simulations . in the following calculations we use the scaling form @xmath29 . we interpret the langevin equation ( [ lang ] ) in the stratonovich sense @xcite . the msd following from the stochastic equation ( [ lang ] ) with diffusivity ( [ eq - dx ] ) takes on the power - law form @xcite @xmath30 where we introduced the scaling exponent @xmath31 which denotes superdiffusion for @xmath32 and subdiffusion for @xmath26 . for @xmath27 the diffusivity grows away from the origin , leading to a progressive acceleration of the particle as it ventures into more distant regions from the origin , and vice versa for @xmath26 . in the special case @xmath3 , the theory developed in ref . @xcite breaks down as the scaling exponent ( [ pexp ] ) diverges . in refs . @xcite an exponential growth of the msd was found , see the discussion below . for even larger values of @xmath4 the anomalous diffusion exponent @xmath33 becomes negative , i.e. , we observe a strong localization , see the discussion below . the probability density function ( pdf ) of the hdp is given by the stretched or compressed gaussian @xcite @xmath34 for positive @xmath4 it exhibits a cusp , while for negative @xmath4 the pdf features a dip to zero at the origin . in single particle tracking experiments one measures the time series @xmath19 of the particle position for a time span @xmath35 . it is usually evaluated in terms of the time averaged msd @xcite @xmath36 ^ 2dt , \label{eq - tamsd}\ ] ] where @xmath37 is the lag time . brownian motion is ergodic and for sufficiently long measurement times @xmath35 the equality @xmath38 holds @xcite . anomalous diffusion ( [ msd ] ) described by fbm or fractional langevin equation motion is asymptotically ergodic @xcite but may feature transiently non - ergodic behavior @xcite in confinement as well as transient aging @xcite , the explicit dependence on the time span elapsing between initial preparation of the system and start of the recording of the position , @xmath19 . ctrw processes with diverging time scales of the waiting time distribution @xmath39 show weakly non - ergodic behavior for all @xmath37 . namely , despite the scaling ( [ msd ] ) of the msd , @xmath40 scales _ linearly _ with @xmath37 . more precisely , if we average over sufficiently many trajectories , the quantity @xmath41 for ctrws scales like @xmath42 @xcite . this linear dependence on @xmath37 is preserved for aging ctrws @xcite . apart from this linear @xmath37-dependence we also observe the dependence on the process time @xmath35 , a signature of aging @xcite . interestingly , the hdp with diffusivity ( [ eq - dx ] ) displays weak ergodicity breaking of the form @xcite @xmath43 this behavior is analogous to that of scale - free ctrw motion , despite of the fact that the increment correlation function of hdps are ( anti)persistent in analogy to the ergodic fbm @xcite . we note that also other processes such as aging and correlated ctrws @xcite as well as scaled brownian motion with time - dependent diffusivity @xcite exhibit the duality between ( [ msd ] ) and a linear @xmath37-dependence of @xmath40 . ( [ eq - tamsd - scaling ] ) shows the @xmath44-scaling as function of the process time @xmath35 . for subdiffusion with @xmath45 , that is , the effective diffusivity of the process decays over time , as the particle ventures into low - diffusivity areas . conversely , for @xmath46 the diffusivity increases over time , the particle discovers areas with increasingly higher @xmath18 . in the current paper we perform a detailed analysis of the msd and the time averaged msd in the entire range of @xmath4 , including the critical value @xmath3 . we are particularly interested in the extent of the weakly non - ergodic behavior , especially the fluctuations of the time averaged msd around the mean value @xmath47 characterized by the ergodicity breaking parameter . moreover we analyze the non - gaussianity of the process . in our analysis we study effects of the scaling exponent @xmath4 of @xmath18 as well as the initial position of the particle . the latter is known to affect the time - scales at which anomalous diffusion becomes significant @xcite . we combine analytical and numerical approaches . the simulations scheme for hdps was introduced in ref . in this section we start with the analysis of the msd and the time averaged msd with its amplitude fluctuations , the latter being quantified by the corresponding scatter distribution . we then analyze the ergodicity breaking and non - gaussianity parameters . finally , we study the pdf of the function that quantifies the degree of particle dispersion . fig . [ fig - msd - tamsd - alfa ] shows the results from computer simulations for the msd @xmath48 and the time averaged msd @xmath40 from individual realizations along with the average @xmath49 taken over all single time traces . the values for the scaling exponent @xmath4 of the diffusivity ( [ eq - dx ] ) studied in fig . [ fig - msd - tamsd - alfa ] cover both the sub- and superdiffusive domains and include , in particular , the critical value @xmath3 where the scaling exponent @xmath33 of the msd ( [ msd ] ) diverges . for each @xmath4 we show @xmath50 single trajectories . in all cases , apart from the critical point , the scaling of the msd @xmath51 and both individual ( @xmath40 ) and mean ( @xmath47 ) time averaged msds agree well with the expected analytical behavior shown by the dashed lines : the scaling exponent of the msd varies consistently with @xmath4 , while that of the time averaged msd remains unity throughout . the initial discrepancy between theory and msds is due to the choice for the initial position @xmath52 , whose influence relaxes on a time scale depending on both @xmath17 and @xmath4 . the deviations of individual time traces @xmath40 at long lag times from the predicted behavior is due to unavoidable , bad statistics when the lag time gets close to the overall length @xmath35 of the time series . irreproducibility of time averages of physical observables such as the msd is an intrinsic property of weakly non - ergodic processes @xcite . for ctrw processes with scale - free waiting time distribution individual traces @xmath19 contain different , few dominating waiting time events that cause the amplitude scatter between different trajectories or , in other words , fluctuations of the apparent effective diffusion constant . this phenomenon occurs no matter how long the measurement time @xmath35 is taken . for hdps the scatter is due to the difference in the extent of excursions to regions of significantly different diffusivity . the amplitude scatter between individual realizations @xmath40 in our hdp shown in fig . [ fig - msd - tamsd - alfa ] varies significantly with the value of the scaling exponent @xmath4 : away from the brownian value @xmath53 in both subdiffusive ( @xmath26 ) and superdiffusive ( @xmath27 ) cases the fluctuations of @xmath40 become more pronounced when the modulus of @xmath4 increases . at the critical point the fluctuations of @xmath40 appear to diverge , while beyond this critical point , the fluctuations decrease again . moreover , a population splitting in a faster ( steeper slope of @xmath54 ) and slower ( shallower slope ) fraction of trajectories appears @xcite , see especially the panel for @xmath55 . in terms of the dimensionless variable @xmath56 the amplitude scatter distribution @xmath57 reflects the randomness of individual time averages of the msd . for a sub- and a superdiffusive @xmath4 it was analyzed in ref . as shown here for a whole spectrum of @xmath4 values , there is a clear trend towards extreme fluctuations at the critical point @xmath3 , but even for considerably smaller values such as @xmath58 the fluctuations around the mean @xmath59 are significant . the width of the fluctuations of @xmath40 in each panel varies only marginally with the lag time @xmath37 , apart from the behavior at @xmath60 . as studied in ref . @xcite , the relative amplitude scatter distribution @xmath57 can be fitted with the generalized gamma - distribution . in particular , it tends to zero at @xmath61 , in contrast to subdiffusive ctrw processes , for which @xmath62 is always positive , indicating completely stalled trajectories @xcite . we note that when the exponent @xmath4 approaches the critical value @xmath3 , the number of steps @xmath63 necessary to approach the theoretically predicted asymptote ( [ eq - msd - scaling ] ) increases significantly . thus , as seen from fig . [ fig - msd - tamsd - alfa ] , for @xmath52 only few simulation steps , @xmath64 are needed for negative @xmath4 with larger modulus . it increases to @xmath65 steps for @xmath66 , @xmath67 for @xmath68 , and already @xmath69 for @xmath58 . once the anomalous diffusion exponent @xmath33 becomes negative , that is , for values of @xmath4 larger than the critical value @xmath3 , the msd ( [ eq - msd - scaling ] ) becomes a decreasing function of time . this agrees with our simulations if @xmath17 is chosen sufficiently large to enable the relaxation to the theoretical asymptote . for instance , for the extreme value @xmath70 and values of the initial position of @xmath71 and above the msd indeed follows the theoretical prediction @xmath72 ( not shown ) . in this region @xmath73 the diffusivity grows fast away from the origin the decreasing msd corresponds to the localization of particles in regions of slow diffusivity , see the detailed discussion in sec . [ sec - spread ] below . in the limit @xmath74 , due to the huge spread one or few extremely large amplitudes @xmath40 of the time averaged msd may substantially affect the mean @xmath47 . the msd in this limit follows an exponential growth @xmath75 , as indicated by the dashed curve in the panel for @xmath3 in fig . [ fig - msd - tamsd - alfa ] . such a fast increase of the msd is consistent with the divergence of the scaling exponent @xmath33 as function of @xmath4 in eq . ( [ eq - msd - scaling ] ) as well as with the exponential msd growth predicted for a parabolically space - varying diffusivity in refs . this property can be straightforwardly inferred from the diffusion equation , @xmath76.\ ] ] multiplying both sides with @xmath77 and integrating over @xmath6 one arrives at @xcite @xmath78.\ ] ] this also rationalizes the observation that the scatter of @xmath40 is maximal for @xmath3 as the particles perform extremely far - reaching excursions relative to other @xmath4-values . as introduced in refs . @xcite the fluctuations of the time averaged msd @xmath40 can be quantified by their variance , the ergodicity breaking parameter @xmath79 when @xmath80 , it means that the process is perfectly reproducible and all time averages over sufficiently long trajectories give the same value . this case corresponds to the sharp scatter distribution @xmath81 @xcite . for the canonical brownian motion the ergodicity breaking parameter reaches the zero value in the form @xmath82 at finite ratio @xmath83 . weakly non - ergodic processes have a positive value of @xmath84 . an alternative , weaker condition for ergodicity is when the msd ( [ msd ] ) and the time averaged msd ( [ eatamsd ] ) coincide . to measure the relative deviations from ergodicity , the @xmath85 parameter was introduced @xcite @xmath86 for ergodic dynamics its value is unity . on the lag time @xmath37 . the brownian asymptote @xmath87 is shown as the dashed curve in panel a ) . the asymptotes ( [ eq - eb2 ] ) are shown in panel b ) as dashed lines of the corresponding color . the initial position of the particle was @xmath52 and @xmath88 traces were used for each value of @xmath4 , all other parameters are the same as those used in fig . [ fig - msd - tamsd - alfa ] . the symbols in both panels correspond to the same parameters.,width=302 ] , eq . ( [ pexp ] ) , of @xmath89 in panels a ) and the magnification in panel b ) as well as @xmath90 in panel c ) . the theoretical predictions from ref . @xcite correspond to the dashed black curve in panels a ) and b ) . the ergodicity breaking parameter of the subdiffusive ctrw given by eq . ( [ eq - eb - ctrw ] ) is shown by the dot - dashed red curve in panel b ) . to compute each point in the graphs for @xmath50 traces of @xmath91 steps takes some two hours on a standard 3 ghz core workstation . different sets of the shown points correspond to varying particle initial conditions @xmath17 indicated by different colors . note the alternative scale for the power exponent @xmath92 on the top axes . parameters are the same as in fig . [ fig - msd - tamsd - alfa].,width=264 ] to calculate the @xmath84 parameter analytically is not always an easy task . to see this , note that the msd follows from the change of the stochastic variable to the standard wiener process in the form @xmath93^{-1/2}$ ] for the time averaged msd @xmath40 the calculation is already more complicated , as it involves the two - point position correlation function . the latter is expressed via fox @xmath94-functions for the hdp @xcite . the analytical derivation of the ergodicity breaking parameter @xmath84 , however , requires fourth - order moments , whose calculation is a formidable task . so far only approximate methods are known for hdps @xcite . for subdiffusive ctrw processes it is possible to obtain the @xmath84 parameter more easily from the conjectured and numerically proven equivalence @xmath95 of @xmath96 with the ratio of the number of steps @xmath97 in an individual realization and the average @xmath98 @xcite . as for the limiting distribution for @xmath99 is known @xcite , this allows straightforward calculation of @xmath57 and its moments @xcite . for the multiplicative process studied here such a scheme does not work . we also note that another parameter involving the fourth moment of the particle displacement is the non - gaussianity parameter @xmath100 @xcite discussed below . due to the lack of an analytical theory , a major reason for the current simulations study is to explore the behavior of these two parameters in the whole range of the model parameters . figs . [ fig - eb - lag ] and [ fig - eb - p ] display the dependence of the ergodicity breaking parameters on the lag time @xmath37 , the initial position @xmath17 , as well as the scaling exponent @xmath33 . we find that for standard brownian motion with @xmath101 the ergodicity breaking parameter eb follows the known asymptote @xmath102 from above . concurrently @xmath103 as expected for ergodic motion , except for very small @xmath37 values because of the initial relaxation of the influence of the initial position @xmath17 . as we depart from the value @xmath53 of brownian motion , the magnitude of @xmath84 grows and its functional dependence on the lag time becomes less pronounced , see fig . [ fig - eb - lag]a . approaching the critical point @xmath104 , due to the huge fluctuations of @xmath40 and the ensuing possibility of extreme events the ergodicity breaking parameter also explicitly depends on the number of traces @xmath105 used for the averaging and reaches values of @xmath106 and higher . on both sides of the critical point , corresponding to large positive or negative values of @xmath33 , the values of the parameter @xmath107 approach one another , as seen in fig . [ fig - eb - p]a , while @xmath108 exhibits a jump , fig . [ fig - eb - p]b . the ratio of the different msds given by @xmath85 also depends on the power exponent @xmath4 and the initial value @xmath17 . as anticipated already from fig . [ fig - msd - tamsd - alfa ] , for the chosen initial condition @xmath52 the magnitude of @xmath85 decreases as @xmath4 gets progressively negative , while @xmath85 grows as @xmath4 increases from 0 to 2 , reaching very large values @xmath109 at @xmath3 . this is illustrated in fig . [ fig - eb - lag]b which is also consistent with the scaling @xmath110 with the lag time at different @xmath111 values . because of the initial condition @xmath17 , similar to our statements for @xmath47 this asymptote is approached later when @xmath74 . in fig . [ fig - eb - p ] we also analyze the effects of the initial position @xmath17 and show the dependence of the ergodicity breaking parameters for long traces or short lag times , i.e. , when @xmath112 which is practically equivalent to @xmath113 , versus the anomalous diffusion exponent @xmath33 and the scaling exponent @xmath4 . we observe that in the region @xmath46 the values of @xmath107 obtained from simulations is in good agreement with the analytical estimate @xmath113 from ref . @xcite , represented by the dashed curve in fig . [ fig - eb - p]a and b. these approximations are based on the asymptotic scaling of @xmath18 and do not consider the effect of the initial positions @xmath17 , assumed to be relaxed in the relevant @xmath114 limit . to see their actual impact on the dynamics we study the ergodicity breaking parameters numerically . we observe that the value of @xmath107 grows from the small pre - asymptotic brownian value at @xmath115 to progressively larger values at larger modulus of @xmath33 , as @xmath4 approaches the critical value @xmath3 from below and above . the functional dependence of @xmath116 obtained from our simulations follows quite well the predictions from the approximate calculation of @xmath84 @xcite , particularly in the range @xmath117 , see fig . [ fig - eb - p]b . the deviations for even more pronounced variation of @xmath18 as @xmath4 tends to 2 are likely due to insufficient statistics . as we show in fig . [ fig - eb - p]a by the circles and triangles , the value @xmath118 reveals measurable deviations for @xmath119 as compared to @xmath50 traces used for the averaging , while for less extreme @xmath4 values the two sets yield nearly identical results . this is a further indication towards a divergence of the fluctuations at the critical point . namely , the chance to find an even larger amplitude of @xmath40 in a given realization grows with the number of simulated trajectories . as already anticipated in ref . @xcite the hdp process considered here is in some aspects reminiscent of subdiffusive ctrw processes with scale - free , power - law waiting time distributions with regard to the scaling of the msd and the time averaged msd , compare also sec . [ sec - discussion ] . apparently , the ergodicity breaking properties of these two processes also reveal similar features : the fluctuations of @xmath40 increase when the anomalous diffusion exponent @xmath33 becomes successively smaller than the brownian value 1 . indeed , as shown in fig . [ fig - eb - p]b for the relevant region @xmath45 there exists even a close agreement of the quantitative behavior of ctrw and hdp . we show the ergodicity breaking parameter obtained for the ctrw process @xcite @xmath120 where @xmath13 is the exponent in the pdf @xmath16 of the waiting times @xmath15 @xcite . for the ctrw process the exponent @xmath12 also occurs in the msd ( [ msd ] ) . we note that in contrast to hpds in superdiffusive ctrw processes the nonergodic behavior is merely ultraweak , effecting a different prefactor in the time averaged msd compared to the msd @xcite . for @xmath121 we find that @xmath122 for initial positions @xmath17 close to the origin , in agreement with the results presented in ref . @xcite . for @xmath73 or @xmath123 the analytical model of ref . @xcite no longer applies but the simulations yield another region of growth for the value of @xmath107 . note that @xmath107 also depends on the trace length @xmath35 ( not shown ) , as discussed for two - dimensional hdps @xcite . as we show in fig . [ fig - eb - p]c , the magnitude of @xmath108 strongly varies with @xmath17 . in the region @xmath124 the ratio of the msds grows as @xmath125 indicated by the dashed asymptote in fig . [ fig - eb - p]c , while for large negative @xmath33 we observe @xmath126 . we finally note that @xmath108 reveals a much weaker dependence on the number @xmath105 of simulated traces , see the circles and triangles in fig . [ fig - eb - p]c . the reason is that @xmath85 involves only the second moment of the time averaged msd , while the @xmath84 is defined in terms of the fourth order moment , which is more sensitive to large variations . as function of the lag time , plotted for the parameters of fig . [ fig - eb - lag ] with @xmath50 trajectories for each @xmath4 value . the approximately zero - valued trace @xmath127 for brownian motion ( @xmath53 ) is only shown partially . color coding is the same as in fig . [ fig - eb - lag ] . * b ) * the value of @xmath128 evaluated for varying power exponent @xmath4 and initial position @xmath17 . the color scheme is the same as in fig . [ fig - eb - p].,width=317 ] the non - gaussianity parameter @xmath100 is a sensitive experimental indicator that often enables one to distinguish the type of particular diffusion processes observed in single - particle tracking experiments @xcite . it is related to the stationarity of increments of the diffusion process and involves the fourth moment along the time - averaged trajectory . for the diffusion process in an embedding space of dimension @xmath129 it is defined via the experimentally relevant time averaged quantities as @xcite @xmath130 where , in analogy to eq . ( [ eq - tamsd ] ) , the fourth moment is defined via @xmath131 ^ 4~dt.\ ] ] for brownian and fractional brownian motion , both gaussian processes , one consistently finds @xmath132 . for diffusion processes revealing a transient anomalous diffusion behavior , this parameter deviates substantially from zero , see , e.g. , the discussion in ref . @xcite . we systematically examine the behavior of @xmath127 in fig . [ fig - non - gauss]a . corroborating the results for the ergodicity breaking parameter , for brownian motion we obtain approximately zero values . as the power @xmath4 in @xmath133 deviates from zero , the non - gaussianity parameter reveals a rich behavior as a function of the lag time @xmath37 and the initial position @xmath17 . more specifically , in the region @xmath26 the non - gaussianity parameter progressively grows and reaches considerably large values for large @xmath2 , see the curve for @xmath134 in fig . [ fig - non - gauss]a . the value @xmath127 systematically decreases with the lag time @xmath37 along the trace for negative @xmath4 . in the region @xmath135 the value of @xmath100 also grows with @xmath4 but , in contrast to the case @xmath26 , the function @xmath127 stays rather constant with @xmath37 . as we approach the critical @xmath3 the non - gaussianity parameter reaches very high values . in the region @xmath73 the value of @xmath127 decreases again . these features of the functional behavior of @xmath136 correspond with the properties of @xmath137 , compare the curves in fig . [ fig - non - gauss]a and [ fig - eb - lag]a . note that similar to @xmath84 the statistics required to obtain a smooth curve for @xmath127 strongly depends on @xmath4 : the result from @xmath50 traces in fig . [ fig - non - gauss]a acquires pronouncedly higher fluctuations when we approach the critical value @xmath3 . we find that variations in the initial positions @xmath17 only have a marginal effect on @xmath138 . in fig . [ fig - non - gauss]b we illustrate the range of @xmath100 values for exponents in the range @xmath139 . even for the wide range from @xmath71 to @xmath140 the value of @xmath128 does not reveal any appreciable variation , in stark contrast to the strong @xmath17-dependence of @xmath141 . this indicates that for the hdp the non - gaussianity parameter is more robust than @xmath84 with respect to the choice of the initial condition . in fig . [ fig - non - gauss]b the dashed line indicates the proportionality to @xmath142 , which nicely matches the measured shape of @xmath138 . we finally analyze the time evolution of the pdf of the particle position in fig . [ fig - spread ] for three different values of the scaling exponent @xmath4 , distinguishing superdiffusion ( top row for @xmath68 , i.e. , ballistic motion with @xmath143 ) , normal diffusion ( middle row for @xmath53 and @xmath115 ) , and subdiffusion ( bottom row for @xmath144 , i.e. , @xmath145 ) . in each row the leftmost panel represents a very early evolution of the pdf close to the initial condition , while in the rightmost panel the pdf in some of the cases has almost reached the diffusion limit , in which the analytical asymptote ( [ pdf - hdp ] ) is valid . we observe that for hdps with positive @xmath4 , i.e. , when the diffusivity @xmath18 grows with the modulus of the position , the pdf for particles with an initial position far away from the origin , an asymmetric shape of the pdf is effected . namely , they progressively move towards regions of small diffusivity and accumulate there . this focusing due to the quenched nature of the diffusivity erases any memory of the initial condition and the common asymptote ( [ pdf - hdp ] ) of stretched gaussian shape ( @xmath135 ) is approached . in the opposite case with negative @xmath4 the focusing of particles in lower diffusivity regions applies again , albeit at longer times than shown here . this time , however , instead of the cusp at the origin for positive @xmath4 , the pdf drops down to zero at the origin and acquires the bimodal shape predicted by ( [ pdf - hdp ] ) , a compressed gaussian . for the brownian case with vanishing @xmath4 , no focusing takes place . in the final panel the distribution is already so broad that the individual , shifted pdfs appear on top of each other . similarly to the evolution of the msd shown in fig . [ fig - msd - tamsd - alfa ] , the relaxation time required for the system of diffusing particles to approach this long - time limit , @xmath1 depends on the power @xmath4 of the diffusivity in eq . ( [ eq - dx ] ) . for more negative @xmath4 values the traces are not yet relaxed to the theoretical long - time shape even after @xmath91 steps . the particles starting at larger @xmath17 remain trapped for the whole length of the simulation . conversely , when the diffusivity increases fast away from the origin ( moderate positive @xmath4 values ) , the equilibration to the long - time pdf is relatively fast , see the panel for @xmath68 in fig . [ fig - spread ] . for hpd processes whose diffusivity @xmath18 varies in power - law form with the distance @xmath6 from the origin and which give rise to anomalous diffusion we analyzed in detail the weakly non - ergodic behavior for varying power exponents @xmath4 and initial positions @xmath17 of the particles . in particular , we examined the functional dependencies and magnitudes of the variation of the ensemble and time averaged characteristics of such hdps with these parameters . the fluctuations of the amplitude of the time averaged msd of individual time traces were shown to systematically grow with increasing departure of the anomalous diffusion exponent @xmath33 from the brownian value @xmath115 , corresponding to the scaling exponent @xmath53 in the power - law form for @xmath18 . the fluctuations increase dramatically when the scaling exponent approaches its critical value @xmath3 . at that critical point we corroborated the turnover from the power - law scaling in time of the msd to an exponential growth . beyond the critical point the fluctuations of the time averaged msd become smaller again . concurrently we observe a distinct population splitting into a faster and slower fraction of time averaged msds . we paid particular emphasis on several parameters used to classify the departure from ergodic behavior , namely the two ergodicity breaking parameters @xmath84 and @xmath85 as well as the non - gaussianity parameter @xmath100 . while @xmath146 is simply defined as the ratio of the mean time averaged msd versus the msd , both @xmath84 and @xmath100 are based on the fourth order moments of the particle position and are known analytically only from approximate theories . a detailed numerical analysis of their properties was therefore used to obtain more concrete information on their behavior for different values of the scaling exponent @xmath4 and the initial particle position @xmath17 in the heterogeneous environment @xcite . within the analyzed parameter range we find that the behavior of @xmath147 is indeed in agreement with the heuristic theoretical analysis from ref . the parameter portraits for the ergodicity breaking and non - gaussianity parameters obtained from our numerical analysis will be useful for actual data evaluation . we also explored the detailed behavior of the hdp dynamics at the critical value @xmath3 and its vicinity , in particular with respect to the dramatic values reached by @xmath84 reflecting the dramatic fluctuations of the time averaged msd . a systematic numerical analysis of the particle pdf for varying initial conditions @xmath17 sheds additional light on hdps with different exponent @xmath4 , in particular , the visualization of the particle focusing in low diffusivity regions . in hdps the non - ergodic behavior arises due to the heterogeneity of the environment . physically , this represents a space - dependent mobility or temperature . it is not a property of the particle , and each time the particle revisits a given point @xmath6 in space it has the same diffusivity @xmath18 . in a random walk sense , this scenario could also be translated into a local dependence of the waiting time for a jump event . in that interpretation the hdp corresponds to a motion in a quenched energy landscape @xcite , albeit a deterministic ( in contrast to random ) one . as such , the process is intrinsically different from renewal ctrw processes , which correspond to the motion in an annealed environment @xcite . nevertheless we observed that subdiffusive hdps share a number of features with subdiffusive ctrws with scale - free , power - law waiting time distribution . these features include the scaling laws for the ensemble and time averaged msds , and the form of the ergodicity breaking parameter @xmath85 ( in the relevant range @xmath45 ) . from the sole analysis of the msds @xmath51 and @xmath40 as well as the ergodicity breaking parameters , a significant distinction between ctrw and hdp is therefore difficult . yet there exist some crucial differences between the hdp and ctrw processes , that can be measured . thus , in hdps with @xmath45 the distribution @xmath57 of the relative amplitude @xmath40 of individual realizations decays to zero at @xmath61 , in contrast to the ctrw s finite fraction of immobile particles reflected in the finite value @xmath148 . indeed the scatter distribution @xmath57 was previously advocated as a good diagnosis tool for different stochastic processes @xcite , complementing other stochastic analysis methods @xcite . hdps are physically meaningful alternatives to the standard stochastic models for anomalous diffusion processes , namely , the ctrw process with long - tailed waiting time distribution , fbm and fractional langevin equation motion based on gaussian yet long - ranged in time correlations , diffusion in fractal environments , or their combinations @xcite . hpds are weakly non - ergodic , sharing some yet not all properties of ctrw processes . it is therefore important to explore the properties of hdps in more detail in the future as well as to develop more sophisticated tools to interpret measured time series and with confidence identify the underlying stochastic mechanism . in particular , we will study the properties of confined and aged hdps as well as spatially varying diffusivities with a random component , and the coupling of hdps with active motion . i. bronstein , y. israel , e. kepten , s. mai , y. shav - tal , e. barkai , and y. garini , phys . lett . * 103 * , 018102 ( 2009 ) ; k. burnecki , e. kepten , j. janczura , i. bronshtein , y. garini , and a. weron , biophys . j. * 103 * , 1839 ( 2012 ) . g. r. kneller , k. baczynski , and m. pasenkiewicz - gierula , j. chem . phys . * 135 * , 141105 ( 2011 ) ; j .- jeon , h. martinez - seara monne , m. javanainen , and r. metzler , phys . * 109 * , 188103 ( 2012 ) ; m. javanainen , h. hammaren , l. monticelly , j .- h . jeon , m. s. miettinen , h. martinez - seara monne , r. metzler , and i. vattulainen , faraday disc . * 161 * , 397 ( 2013 ) . s. c. kou , ann . * 2 * , 501 ( 2008 ) ; e. lutz , phys . e * 64 * , 051106 ( 2001 ) ; r. kupferman , j. stat . phys . * 114 * , 291 ( 2004 ) ; w. min , g. luo , b. j. cherayil , s. c. kou , and x. s. xie , phys . lett . * 94 * , 198302 ( 2005 ) . q. xu , l. feng , r. sha , n. c. seeman , and p. m. chaikin , phys . 106 * , 228102 ( 2011 ) ; i. y. wong , m. l. gardel , d. r. reichman , e. r. weeks , m. t. valentine , a. r. bausch , and d. a. weitz , _ ibid . _ * 92 * , 178101 ( 2004 ) . bouchaud , j. phys . i ( paris ) * 2 * , 1705 ( 1992 ) ; g. bel and e. barkai , phys . lett . * 94 * , 240602 ( 2005 ) ; a. rebenshtok and e. barkai , _ ibid . _ * 99 * , 210601 ( 2007 ) ; m. a. lomholt , i. m. zaid , and r. metzler , phys . lett . * 98 * , 200603 ( 2007 ) . v. tejedor and r. metzler , j. phys . a * 43 * , 082002 ( 2010 ) ; m. magdziarz , r. metzler , w. szczotka , and p. zebrowski , phys . rev . e * 85 * , 051103 ( 2012 ) . see also the alternative approach in a. v. chechkin , m. hofmann , and i. m. sokolov , phys . e * 80 * , 031112 ( 2009 ) ; j. h. p. schulz , a. v. chechkin , and r. metzler , j. phys . a. * 46 * , 475001 ( 2013 ) . g. zumofen and j. klafter , physica d * 69 * , 436 ( 1993 ) ; a. godec and r. metzler , phys . e * 88 * , 012116 ( 2013 ) ; d. froemberg and e. barkai , _ ibid . _ * 87 * , 030104(r ) ( 2013 ) ; _ ibid . _ * 88 * , 024101 ( 2013 ) ; t. akimoto , phys . * 108 * , 164101 ( 2012 ) . m. magdziarz , a. weron , k. burnecki , and j. klafter , phys . . lett . * 103 * , 180602 ( 2009 ) ; m. magdziarz and j. klafter , phys . e * 82 * , 011129 ( 2010 ) ; y. meroz , i. eliazar , and j. klafter , j. phys . a * 42 * , 434012 ( 2009 ) ; y. meroz , i. m. sokolov , and j. klafter , phys . rev . lett . * 110 * , 090601 ( 2013 ) ; s. condamin , v. tejedor , r. voituriez , o. bnichou , and j. klafter , proc . usa * 105 * , 5675 ( 2008 ) ; v. tejedor , o. bnichou , r. voituriez , r. jungmann , f. simmel , c. selhuber - unkel , l. oddershede , and r. metzler , biophys . j. * 98 * , 1364 ( 2010 ) ; m. bauer , r. valiullin , g. radons , and j. kaerger , j. chem . phys . * 135 * , 144118 ( 2011 ) ; j .- h . jeon , e. barkai , and r. metzler , j. chem . phys . * 139 * , 121916 ( 2013 ) .
in order to study complex systems such as the world wide web ( www ) we propose a model for generating a simple directed random graph with prescribed degree distributions . the ability to match degree distributions to real graphs is perhaps the first characteristic one would desire from a model , and although several models that accomplish this for undirected graphs have been proposed in the recent literature @xcite , not much has been done for the directed case . in the www example that motivates this work , vertices represent webpages and the edges represent the links between them . empirical studies ( e.g. , @xcite ) suggest that both the _ in - degree _ and _ out - degree _ , number of links pointing to a page and the number of outbound links of a page , respectively , follow a power - law distribution , a characteristic often referred to as the _ scale - free _ property . the model we propose in this paper is closely related to the work in @xcite for undirected graphs , where given a probability distribution @xmath0 , the goal is to provide an algorithm to generate a simple random graph whose degree distribution is approximately @xmath0 . two of the models presented in @xcite , as well as the model in @xcite , are in turn related to the well - known configuration model @xcite , where vertices are given stubs or half - edges according to a degree sequence @xmath2 and these stubs are then randomly paired to form edges . to obtain a prescribed degree distribution , the degree sequence @xmath2 is chosen as i.i.d . random variables having distribution @xmath0 . this method allows great flexibility in terms of the generality of @xmath0 , which is very important in the applications we have in mind . the most general of the results presented here require only that the degree distributions have finite @xmath3th moment , and are therefore applicable to a great variety of examples , including the www . for a directed random graph there are two distributions that need to be chosen , the in - degree and out - degree distributions , denoted respectively @xmath4 and @xmath5 . the in - degree of a node corresponds to the number of edges pointing to it , while the out - degree is the number of edges pointing out . to follow the ideas from @xcite , we propose to draw the in - degree and out - degree sequences as i.i.d . observations from distributions @xmath0 and @xmath1 . unlike the undirected case where the only main problem with this approach is that the sum of the degrees might not be even , which is necessary to draw an undirected graph , in the directed case the corresponding condition is that the sum of the in - degrees and the sum of the out - degrees be the same . since the probability that two i.i.d . sequences will have the same sum , even if their means are equal , converges to zero as the number of nodes grows to infinity , the first part of the paper focuses on how to construct valid degree sequences without significantly destroying their i.i.d . properties . once we have valid degree sequences the problem is how to obtain a simple graph , since the random pairing may produce self - loops and multiple edges in the same direction . this problem is addressed in two ways , the first of which consists in showing sufficient conditions under which the probability of generating a simple graph through random pairing is strictly positive , which in turn suggests repeating the pairing process until a simple graph is obtained . the second approach is to simply erase the self - loops and multiple edges of the resulting graph . in both cases , one must show that the degree distributions in the final simple graph remain essentially unchanged . in particular , if we let @xmath6 be the probability that a randomly chosen node from a graph of size @xmath7 has in - degree @xmath8 , and let @xmath9 be the corresponding probability for the out - degree , then we will show that , @xmath10 as @xmath11 . we also prove a similar result for the empirical distributions . the question of whether a given pair of in- and out - degree sequences @xmath12 is graphical , i.e. , from which it is possible to draw a simple directed graph , has been recently studied in @xcite , where algorithms to realize such graphs have also been analyzed . random directed graphs with arbitrary degree distributions have been studied in @xcite via generating functions , which can be used to formalize concepts such as in - components " and out - components " as well as to estimate their average size . models of growing networks that can be calibrated to mimic the power - law behavior of the www have been analyzed using statistical physics techniques in @xcite . the approach followed in this paper focuses on one hand on the generation of in- and out - degree sequences that are close to being i.i.d . and that are graphical with high probability , and on the other hand on providing conditions under which a simple graph can be obtained through random pairing . the directed configuration model with ( close to ) i.i.d . degree sequences , although not a growing network model , has the advantage of being analytically tractable and easy to simulate . the rest of the paper is organized as follows . in section [ s.degreesequences ] we introduce a model to construct in- and out - degree sequences that are very close to being two independent sequences of i.i.d . random variables having distributions @xmath0 and @xmath1 , respectively , but whose sums are the same ; in the same spirit as the results in @xcite we also show that the suggested method produces with high probability a graphical pair of degree sequences . in subsection [ ss.repeat ] we prove sufficient conditions under which the probability that the directed configuration model will produce a simple graph will be bounded away from zero , and show that conditional on the resulting graph being simple , the degree sequences have asymptotically the correct distributions . in subsection [ ss.erase ] we show that under very mild conditions , the process of simply erasing self - loops and multiple edges results in a graph whose degree distributions are still asymptotically @xmath0 and @xmath1 . as mentioned in the introduction , the goal of this paper is to provide an algorithm for generating a random directed graph with @xmath7 nodes with the property that its in - degrees and out - degrees have some prespecified distributions @xmath0 and @xmath1 , respectively . moreover , we would like the resulting graph to be _ simple _ , that is , it should not contain self - loops or multiple edges in the same direction . the two models that we propose are based on the so - called configuration or pairing model , which produces a random undirected graph from a degree sequence @xmath13 . in @xcite the prescribed degree distribution is obtained by drawing the degree sequence @xmath2 as i.i.d . random variables from that distribution . more details about the configuration model can be found in section [ s.configurationmodel ] . following the same idea of using a sequence of i.i.d . random variables to generate the degree sequence of an undirected graph , the natural extension to the directed case would be to draw two i.i.d . sequences from given distributions @xmath0 and @xmath1 . we note that in the undirected setting the two main problems with this approach are : 1 ) that the sum of the degrees may be odd , in which case it is impossible to draw a graph , and 2 ) that there may not exist a simple graph having the prescribed degree sequence . the first problem is easily fixed by either sampling the i.i.d . sequence until its sum is even ( which will happen with probability 1/2 asymptotically ) , or simply adding one to the last random number in the sequence . the second problem , although related to the verification of graphicality criteria ( e.g. , the erds - gallai criterion @xcite ) , turns out to be negligible as the number of nodes goes to infinity , as the work in @xcite shows . for directed graphs a graphicality criterion also exists , and the second problem turns out to be negligible for large graphs just as in the undirected case . nonetheless , the equivalent of the first problem is now that the potential in - degree and out - degree sequences must have the same sum , which is considerably harder to fix . before proceeding with the formulation of our proposed algorithm we give some basic definitions which will be used throughout the paper . we denote by @xmath14 a directed graph on @xmath7 nodes or vertices , @xmath15 , connected via the set of directed edges @xmath16 . we say that @xmath14 is _ simple _ if any pair of nodes are connected by at most one edge in each direction , and if there are no edges in between a node and itself . the in - degree @xmath17 , respectively , out - degree @xmath18 , of node @xmath19 is the total number of edges from other nodes to @xmath20 , respectively , from @xmath20 to other nodes . the pair of sequences @xmath21 of nonnegative integers is called a _ bi - degree - sequence _ if @xmath17 and @xmath18 correspond to the in - degree and out - degree , respectively , of node @xmath20 . [ d.bi-degree-sequence ] a bi - degree - sequence @xmath22 is said to be _ graphical _ if there exists a simple directed graph @xmath14 on the set of nodes @xmath23 such that the in - degree and out - degree sequences together form @xmath22 . in this case we say that @xmath24 realizes the bi - degree - sequence . in view of these definitions our goal is to generate the sequences @xmath25 and @xmath26 from i.i.d . samples of given distributions @xmath4 and @xmath27 , respectively . both @xmath0 and @xmath1 are assumed to be probability distributions with support on the nonnegative integers with a finite common mean @xmath28 . note that although the strong law of large numbers ( slln ) guarantees that if we simply sample i.i.d . random variables @xmath29 from @xmath0 and , independently , i.i.d . random variables @xmath30 from @xmath1 , then @xmath31 it is also true that in general @xmath32 one potential idea to fix the problem is to sample one of the two sequences , say the in - degrees , as i.i.d . observations @xmath29 from @xmath0 and then sample the second sequence from the conditional distribution @xmath1 given that its sum is @xmath33 . this approach has the major drawback that this conditional distribution may be ill - behaved , in the sense that the probability of the conditioning event , the sum being equal to @xmath34 , converges to zero in most cases . it follows that we need a different mechanism to sample the degree sequences . the precise algorithm we propose is described below ; we focus on first sampling two independent i.i.d . sequences and then add in- or out - degrees as needed to match their sums . the following definition will be needed throughout the rest of the paper . we say that a function @xmath35 is slowly varying at infinity if @xmath36 for all fixed @xmath37 . a distribution function @xmath0 is said to be regularly varying with index @xmath38 , @xmath39 , if @xmath40 with @xmath35 slowly varying . we will also use the notation @xmath41 to denote convergence in distribution , @xmath42 to denote convergence in probability , and @xmath43 to refer to the positive integers . we assume that the target degree distributions @xmath0 and @xmath1 have support on the nonnegative integers and have common mean @xmath44 . moreover , suppose that there exist slowly varying functions @xmath45 and @xmath46 such that @xmath47 for all @xmath48 , where @xmath49 . we refer the reader to @xcite for all the properties of slowly varying functions that will be used in the proofs . however , we do point out here that the tail conditions in ensure that @xmath0 has finite moments of order @xmath50 for all @xmath51 , and @xmath1 has finite moments of order @xmath50 for all @xmath52 . the constant @xmath53 will play an important role throughout the paper . the algorithm is given below . 1 . fix @xmath54 . 2 . sample an i.i.d . sequence @xmath29 from distribution @xmath0 ; let @xmath55 . 3 . sample an i.i.d . sequence @xmath56 from distribution @xmath1 ; let @xmath57 . 4 . define @xmath58 . if @xmath59 proceed to step 5 ; otherwise repeat from step 2 . choose randomly @xmath60 nodes @xmath61 without remplacement and let @xmath62 where @xmath63 \(i ) this algorithm constructs a bi - degree - sequence @xmath64 having the property that @xmath65 . ( ii ) note that we have used the capital letters @xmath66 and @xmath67 to denote the in - degree and out - degree , respectively , of node @xmath68 , as opposed to using the notation @xmath17 and @xmath18 from definition [ d.bi-degree-sequence ] ; we do this to emphasize the randomness of the bi - degree - sequence itself . ( iii ) clearly , neither @xmath69 nor @xmath70 are i.i.d . sequences , nor are they independent of each other , but we will show in the next section that asymptotically as @xmath7 grows to infinity they have the same joint distribution as @xmath71 . ( iv ) we will also show that the condition in step 4 has probability converging to one . ( v ) note that we always choose to add degrees , rather than fixing one sequence and always adjust the other one , to avoid having problems with nodes with in- or out - degree zero . we now provide some results about the asymptotic behavior of the bi - degree - sequence obtained from the algorithm we propose . the first thing we need to prove is that the algorithm will always end in finite time , and the only step where we need to be careful is in step 4 , since it may not be obvious that we can always draw two independent i.i.d . sequences satisfying @xmath72 in a reasonable amount of time . the first lemma we give establishes that this is indeed the case by showing that the probability of satisfying condition @xmath72 converges to one as the size of the graph grows . all the proofs in this section can be found in subsection [ ss.degreesequencesproofs ] . [ l.differencelimit ] define @xmath73 , then @xmath74 since the sums of the in - degrees and out - degrees are the same , we can always draw a graph , but this is not enough to guarantee that we can draw a simple graph . in other words , we need to determine with what probability will the the bi - degree - sequence @xmath64 be graphical , and to do this we first need a appropriate criterion , e.g. , a directed version of the erds - gallai criterion for undirected graphs . the following result ( corollary 1 on p. 110 in @xcite ) gives necessary and sufficient conditions for a bi - degree - sequence to be graphical ; the original statement is for more general @xmath75-graphs , where up to @xmath75 parallel edges in the same direction are allowed . the notation @xmath76 denotes the cardinality of set @xmath77 . [ thm : graphical - condition ] given a set of @xmath7 vertices @xmath78 , having bi - degree - sequence @xmath79 , a necessary and sufficient condition for @xmath22 to be graphical is 1 . @xmath80 , and 2 . @xmath81 for any @xmath82 . we now state a result that shows that for large @xmath7 , the bi - degree - sequence @xmath64 constructed in subsection [ ss.algorithm ] is with high probability graphical . related results for undirected graphs can be found in @xcite , which includes the case when the degree distribution has infinite mean . [ thm : always - graphical ] for the bi - degree - sequence @xmath64 constructed in section [ ss.algorithm ] we have @xmath83 the second property of @xmath64 that we want to show is that despite the fact that the sequences @xmath84 and @xmath85 are no longer independent nor individually i.i.d . , they are still asymptotically so as the number of vertices @xmath7 goes to infinity . the intuition behind this result is that the number of degrees that need to be added to one of the i.i.d . sequences @xmath86 or @xmath87 to match their sum is small compared to @xmath7 , and therefore the sequences @xmath84 and @xmath85 are _ almost _ i.i.d . and independent of each other . this feature makes the bi - degree - sequence @xmath64 we propose an approximate equivalent of the i.i.d . degree sequence considered in @xcite for undirected graphs . [ t.weakconvergence ] the bi - degree - sequence @xmath64 constructed in subsection [ ss.algorithm ] satisfies that for any fixed @xmath88 , @xmath89 as @xmath11 , where @xmath90 and @xmath87 are independent sequences of i.i.d . random variables having distributions @xmath0 and @xmath1 , respectively . to end this section , we give a result that establishes regularity conditions of the bi - degree - sequence @xmath64 which will be important in the sequel . [ p.mainconditions ] the bi - degree - sequence @xmath64 constructed in subsection [ ss.algorithm ] satisfies @xmath91 @xmath92 , \qquad \frac{1}{n } \sum_{i=1}^n d_i \stackrel{p}{\longrightarrow } e[\xi_1 ] , \qquad \text{and } \qquad \frac{1}{n } \sum_{i=1}^n n_id_i \stackrel{p}{\longrightarrow } e[\gamma_1 \xi_1],\ ] ] as @xmath11 , and provided @xmath93 < \infty$ ] , @xmath94 , \qquad \text{and } \qquad \frac{1}{n } \sum_{i=1}^n d_i^2 \stackrel{p}{\longrightarrow } e[\xi_1 ^ 2],\ ] ] as @xmath11 . in the previous section we introduced a model for the generation of a bi - degree - sequence @xmath64 that is close to being a pair of independent sequences of i.i.d . random variables , but yet has the property of being graphical with probability close to one as the size of the graph goes to infinity . we now turn our attention to the problem of obtaining a realization of such sequence , in particular , of drawing a simple graph having @xmath64 as its bi - degree - sequence . the approach that we follow is a directed version of the _ configuration model_. the configuration , or _ pairing model _ , was introduced in @xcite and @xcite , although earlier related ideas based on symmetric matrices with @xmath95 entries go back to the early 70 s ; see @xcite for a survey of the history as well as additional references . the configuration model is based on the following idea : given a degree sequence @xmath96 , to each node @xmath20 , @xmath97 , assign @xmath18 stubs or half - edges , and then pair half - edges to form an edge in the graph by randomly selecting with equal probability from the remaining set of unpaired half - edges . this procedure results in a multigraph on @xmath7 nodes having @xmath98 as its degree sequence , where the term multigraph refers to the possibility of self - loops and multiple edges . although this algorithm does not produce a multigraph uniformly chosen at random from the set of all multigraphs having degree sequence @xmath98 , a simple graph uniformly chosen at random can be obtained by choosing a pairing uniformly at random and discarding the outcome if it has self - loops or multiple edges @xcite . the question that becomes important then is to estimate the probability with which the pairing model will produce a simple graph . for the undirected graph setting we have described , such results were given in @xcite for regular @xmath99-graphs ( graphs where each node has exactly degree @xmath99 ) , and in @xcite for general graphical degree sequences . from the previous discussion , it should be clear that it is important to determine conditions under which the probability of obtaining a simple graph in the pairing model is bounded away from zero as @xmath11 . such conditions are essentially bounds on the rate of growth of the maximum ( minimum ) degree and/or the existence of certain limits ( see , e.g. , @xcite ) . the set of conditions given below is taken from @xcite , and we include it here as a reference for the directed version discussed in this paper . [ cond.regularityundirected ] given a degree sequence @xmath96 , let @xmath100}$ ] be the degree of a randomly chosen node , i.e. , @xmath101 } = k ) = \frac{1}{n } \sum_{i=1}^n 1(d_i = k).\ ] ] 1 . _ weak convergence . _ there exists a finite random variable @xmath102 taking values on the positive integers such that @xmath103 } \rightarrow d , \qquad n \to \infty.\ ] ] 2 . _ convergence of the first moment . _ @xmath104 } ] = e[d].\ ] ] 3 . _ convergence of the second moment . _ @xmath105})^2 ] = e[d^2].\ ] ] it is straightforward to verify that if the degree sequence is chosen as an i.i.d . sample @xmath70 from some distribution @xmath0 on the positive integers having finite first moment , then parts ( a ) and ( b ) of condition [ cond.regularityundirected ] are satisfied , and if @xmath0 has finite second moment then also part ( c ) is satisfied ; the adjustment made to ensure that the sum of the degrees is even , if needed , can be shown to be negligible . condition [ cond.regularityundirected ] guarantees that the probability of obtaining a simple graph in the pairing model is bounded away from zero ( see , e.g. , @xcite ) , in which case we can obtain a uniformly simple realization of the ( graphical ) degree sequence @xmath2 by repeating the random pairing until a simple graph is obtained . when part ( c ) of condition [ cond.regularityundirected ] fails , then an alternative is to simply erase the self - loops and multiple edges . these two approaches give rise to the _ repeated _ an _ erased _ configuration models , respectively . having given a brief description of the configuration model for undirected graphs , we will now discuss how to adapt it to draw directed graphs . the idea is basically the same , given a bi - degree - sequence @xmath22 , to each node @xmath20 assign @xmath17 inbound half - edges and @xmath18 outbound half - edges ; then , proceed to match inbound half - edges to outbound half - edges to form directed edges . to be more precise , for each unpaired inbound half - edge of node @xmath20 choose randomly from all the available unpaired outbound half - edges , and if the selected outbound half - edge belongs to node , say , @xmath106 , then add a directed edge from @xmath106 to @xmath20 to the graph ; proceed in this way until all unpaired inbound half - edges are matched . the following result shows that conditional on the graph being simple , it is uniformly chosen among all simple directed graphs having bi - degree - sequence @xmath22 . all the proofs of section [ s.configurationmodel ] can be found in subsection [ ss.configurationmodelproofs ] . [ p.uniformity ] given a graphical bi - degree - sequence @xmath22 , generate a directed graph according to the directed configuration model . then , conditional on the obtained graph being simple , it is uniformly distributed among all simple directed graphs having bi - degree - sequence @xmath22 . the question is now under what conditions will the probability of obtaining a simple graph be bounded away from zero as the number of nodes , @xmath7 , goes to infinity . when this probability is bounded away from zero we can repeat the random pairing until we draw a simple graph : the repeated model ; otherwise , we can always erase the self - loops and multiple edges in the same direction to obtain a simple graph : the erased model . these two models are discussed in more detail in the following two subsections , where we also provide sufficient conditions under which the the probability of obtaining a simple graph will be bounded away from zero . we end this section by mentioning that another important line of problems related to the drawing of simple graphs ( directed or undirected ) is the development of efficient simulation algorithms , see for example the recent work in @xcite using importance sampling techniques for drawing a simple graph with prescribed degree sequence @xmath2 ; similar ideas should also be applicable to the directed model . in this section we analyze the directed configuration model using the bi - degree - sequence @xmath64 constructed in subsection [ ss.algorithm ] . in order to do so we will first need to establish sufficient conditions under which the probability that the directed configuration model produces a simple graph is bounded away from zero as the number of nodes goes to infinity . since this property does not directly depend on the specific bi - degree - sequence @xmath64 , we will prove the result for general bi - degree - sequences @xmath22 satisfying an analogue of condition [ cond.regularityundirected ] . as one may expect , we will require the existence of certain limits related to the ( joint ) distribution of the in - degree and out - degree of a randomly chosen node . also , since the sequences @xmath107 and @xmath2 need to have the same sum , we prefer to consider a sequence of bi - degree - sequences , i.e. , @xmath108 where @xmath109 , since otherwise the equal sum constraint would greatly restrict the type of sequences we can use ( e.g. , @xmath110 for all @xmath111 ) . the corresponding version of condition [ cond.regularityundirected ] is given below . [ cond.regularitydirected ] given a sequence of bi - degree - sequences @xmath108 satisfying @xmath112 let @xmath113 } , d^{[n]})$ ] denote the in - degree and out - degree of a randomly chosen node , i.e. , @xmath114 } , d^{[n ] } ) = ( i , j ) ) = \frac{1}{n } \sum_{k=1}^n 1(m_{nk } = i , d_{nk } = j).\ ] ] 1 . _ weak convergence . _ there exist finite random variables @xmath115 and @xmath116 taking values on the nonnegative integers and satisfying @xmath117 = e[\xi ] > 0 $ ] such that @xmath118 } , d^{[n ] } ) \rightarrow ( \gamma , \xi ) , \qquad n \to \infty.\ ] ] 2 . _ convergence of the first moments . _ @xmath119 } ] = e[\gamma ] \qquad \text{and } \qquad \lim_{n\to\infty } e[d^{[n ] } ] = e[\xi].\ ] ] 3 . _ convergence of the covariance . _ @xmath120 } d^{[n ] } ] = e[\gamma \xi].\ ] ] 4 . _ convergence of the second moments . _ @xmath121})^2 ] = e[\gamma^2 ] \qquad \text{and } \qquad \lim_{n\to\infty } e [ ( d^{[n]})^2 ] = e[\xi^2].\ ] ] we now state a result that says that the number of self - loops and the number of multiple edges produced by the random pairing converge jointly , as @xmath11 , to a pair of independent poisson random variables . as a corollary we obtain that the probability of the resulting graph being simple converges to a positive number , and is therefore bounded away from zero . the proof is an adaptation of the proof of proposition 7.9 in @xcite . consider the multigraph obtained through the directed configuration model from the bi - degree - sequence @xmath122 , and let @xmath123 be the number of self - loops and @xmath124 be the number of multiple edges in the same direction , that is , if there are @xmath125 ( directed ) edges from node @xmath20 to node @xmath106 , they contribute @xmath126 to @xmath124 . [ p.poissonlimit ] ( poisson limit of self - loops and multiple edges ) if @xmath108 satisfies condition [ cond.regularitydirected ] with @xmath117 = e[\xi ] = \mu > 0 $ ] , then @xmath127 as @xmath11 , where @xmath128 and @xmath129 are two independent poisson random variables with means @xmath130}{\mu } \qquad \text{and } \qquad \lambda_2 = \frac { e[\gamma(\gamma-1 ) ] e[\xi(\xi-1)]}{2 \mu^2},\ ] ] respectively . since the probability of the graph being simple is @xmath131 , we obtain as a consequence the following theorem . [ t.probsimple ] under the assumptions of proposition [ p.poissonlimit ] , @xmath132 it is clear from proposition [ p.mainconditions ] that condition [ cond.regularitydirected ] is satisfied by the bi - degree - sequence @xmath64 proposed in subsection [ ss.algorithm ] whenever @xmath0 and @xmath1 have finite variance . this implies that one way of obtaining a simple directed graph on @xmath7 nodes is by first sampling the bi - degree - sequence @xmath64 according to subsection [ ss.algorithm ] , then checking if it is graphical , and if it is , use the directed pairing model to draw a graph , discarding any realizations that are not simple . alternatively , since the probability of @xmath64 being graphical converges to one , then one could skip the verification of graphicality and re - sample @xmath64 each time the pairing needs to be repeated . the last thing we show in this section is that the degree distributions of the resulting simple graph will have with high probability the prescribed degree distributions @xmath0 and @xmath1 , as required . more specifically , if we let @xmath133 be the bi - degree - sequence of the final simple graph obtained through the repeated directed configuration model with bi - degree - sequence @xmath64 , then we will show that the joint distribution @xmath134 converges to @xmath135 , and the empirical distributions , @xmath136 converge in probability to @xmath137 and @xmath138 , respectively . the same result was shown in @xcite for the undirected case with i.i.d . degree sequence @xmath85 . [ p.repeateddistr ] for the repeated directed configuration model with bi - degree - sequence @xmath64 , as constructed in subsection [ ss.algorithm ] , we have : 1 . @xmath139 as @xmath11 , @xmath140 , and 2 . for all @xmath141 , @xmath142 note that by the continuous mapping theorem , ( a ) implies that the marginal distributions of the in - degrees and out - degrees , @xmath143 converge to @xmath144 and @xmath145 , respectively . the same arguments used in the proof also give that the joint empirical distribution converges to @xmath135 in probability . in this section we consider the erased directed configuration model , which is particularly useful when the probability of drawing a simple graph converges to zero as the number of nodes increases , which could happen , for example , when condition [ cond.regularitydirected ] ( d ) fails . given a bi - degree - sequence @xmath22 , the erased model consists in first obtaining a multigraph according to the directed configuration model and then erase all self - loops and merge multiple edges in the same direction into a single edge , with the result being a simple graph . note that the graph obtained through this process no longer has @xmath22 as its bi - degree - sequence . as for the repeated model , let @xmath146 be the bi - degree - sequence of the simple graph obtained through the erased directed configuration model with bi - degree - sequence @xmath64 . define the joint distribution @xmath147 and the empirical distributions , @xmath148 the following result is the analogue of proposition [ p.repeateddistr ] for the erased model . [ p.eraseddistr ] for the erased directed configuration model with bi - degree - sequence @xmath64 , as constructed in subsection [ ss.algorithm ] , we have : 1 . @xmath139 as @xmath11 , @xmath140 , and 2 . for all @xmath141 , @xmath142 in this section we give the proofs of all the results in the paper . we divide the proofs into two subsections , one containing those belonging to section [ s.degreesequences ] and those belonging to section [ s.configurationmodel ] . throughout the remainder of the paper we use the following notation : @xmath149 if @xmath150 , @xmath151 if @xmath152 , and @xmath153 if @xmath154 . suppose now that @xmath159 = \infty$ ] , which implies that @xmath160 $ ] . let @xmath161 , define @xmath162 , @xmath163 , and let @xmath164 be a sequence of i.i.d . random variables having distribution @xmath165 . then , @xmath166 \right\| + n\|e[\tilde z_1]\| > n^{1-\kappa+\delta_0 } \right ) + p\left ( \max_{1\leq i \leq n } \|z_i\| > t_n \right ) .\end{aligned}\ ] ] by the union bound , @xmath167 as @xmath11 , which converges to zero by basic properties of slowly varying functions ( see , e.g. , proposition 1.3.6 in @xcite ) . next , note that since @xmath168 = 0 $ ] , @xmath169\| = \frac{\| e [ z_1 1(\|z_1\| > t_n ) ] \|}{p(\|z_1\| \leq t_n ) } \leq \frac{e [ \|z_1\| 1(\|z_1\| > t_n ) ] } { p(\|z_1\| \leq t_n ) } \leq ( 1+o(1 ) ) \left ( t_n p(\|z_1\| > t_n ) + \int_{t_n}^\infty p(\|z_1\| > z ) dz \right).\ ] ] to estimate the integral note that @xmath170 where in the third step we used proposition 1.5.10 in @xcite . now note that @xmath171 from where it follows that @xmath169\| = o\left ( n^{-\kappa - \epsilon(\theta^{-1}-1 ) } ( l_f(t_n ) + l_g(t_n ) ) \right ) = o\left ( n^{-\kappa + \delta_0 } \right)\ ] ] as @xmath11 . in view of this , we can use chebyshev s inequality to obtain @xmath172 \right\| + n\|e[\tilde z_1]\| > n^{1-\kappa+\delta_0 } \right ) \leq \frac{{{\rm var } \mspace{1mu}}(\tilde z_1)}{n^{1 - 2(\kappa-\delta_0 ) } ( 1+o(1))}.\ ] ] finally , to see that this last bound converges to zero note that @xmath173 = \frac{1}{p(\|z_1\| \leq t_n ) } e [ z_1 ^ 2 1(\|z_1\| \leq t_n ) ] \leq ( 1+o(1 ) ) e\left [ \|z_1\|^{\theta^{-1}-\epsilon } \right ] t_n^{2-\theta^{-1}+\epsilon},\end{aligned}\ ] ] where @xmath174 < \infty$ ] by the remark following . we conclude that is of order @xmath175 as @xmath11 . this completes the proof . [ l.sumsorderstats ] let @xmath176 be an i.i.d . sequence of nonnegative random variables having distribution function @xmath23 , and let @xmath177 denote the @xmath68th order statistic . then , for any @xmath178 , @xmath179 \leq \int_0^\infty \min\left\ { n \overline{v}(x ) , k \right\ } dx.\ ] ] note that @xmath180 = \int_0^\infty p(x^{(i ) } > x ) \ , dx = \int_0^\infty \sum_{j = n - i+1}^n \binom{n}{j } \overline{v}(x)^j v(x)^{n - j } dx,\ ] ] from where it follows @xmath181 & = \sum_{i = n - k+1}^n \sum_{j = n - i+1}^n \binom{n}{j } \int_0^\infty \overline{v}(x)^j v(x)^{n - j } dx \\ & = \sum_{j=1}^{n } \min\{j , k\ } \binom{n}{j } \int_0^\infty \overline{v}(x)^{j } v(x)^{n - j } dx \\ & = \int_0^\infty e\left [ \min\ { b(n , \overline{v}(x ) ) , k \ } \right ] dx,\end{aligned}\ ] ] where @xmath182 is a binomial@xmath183 random variable . since the function @xmath184 is concave , jensen s inequality gives @xmath185 \leq \min\left\ { e [ b(n , \overline{v}(x ) ) ] , k \right\ } = \min\left\ { n\overline{v}(x ) , k \right\}.\ ] ] since by construction @xmath186 , it follows from theorem [ thm : graphical - condition ] that it suffices to show that @xmath187 fix @xmath188 and use the union bound to obtain @xmath189 by conditioning on how many of the @xmath67 are larger than @xmath190 we obtain that is bounded by @xmath191 where @xmath192 was defined in lemma [ l.differencelimit ] . now note that by the union bound we have @xmath193 as @xmath11 , where the last step follows from lemma [ l.differencelimit ] and basic properties of slowly varying functions ( see , e.g. , chapter 1 in @xcite ) . next , to analyze let @xmath194 and note that we can write it as @xmath195 where @xmath196 is the @xmath68th smallest of @xmath197 . now let @xmath198 = \overline{g}(0 ) > 0 $ ] and split the last probability as follows @xmath199 to bound use @xmath200 for all @xmath201 and chebyshev s inequality to obtain @xmath202 while the union bound gives that is bounded by @xmath203 where @xmath204 . for the second probability the union bound again gives @xmath205 as @xmath11 . finally , by markov s inequality and lemma [ l.sumsorderstats ] , @xmath206 \leq \frac{1}{b_n p(\mathcal{d}_n ) } \sum_{i = n - k_n+1}^n e [ \gamma^{(i ) } + 1 ] \\ & \leq \frac{1}{b_n p(\mathcal{d}_n ) } \left ( \int_0^\infty \min\left\ { n \overline{f}(x ) , k_n \right\ } dx + k_n \right ) \\ & = a_0^{-1 } ( 1 + o(1 ) ) \int_0^\infty \min\left\ { \overline{f}(x ) , n^{(1+\epsilon)/\beta - 1 } \right\ } dx + o(1 ) \\ & \leq a_0^{-1 } ( 1+o(1 ) ) \left ( n^{(1+\epsilon)/\beta-1 } + \int_1^\infty \min\left\ { k x^{-\alpha+\epsilon } , n^{(1+\epsilon)/\beta-1 } \right\ } dx \right ) + o(1 ) \\ & = o(1 ) + o\left ( \int_1^\infty \min\left\ { x^{-\alpha+\epsilon } , n^{(1+\epsilon)/\beta-1 } \right\ } dx \right)\end{aligned}\ ] ] as @xmath11 , where @xmath207 . since @xmath208 the proof is complete . let @xmath209 $ ] , @xmath210 , be a continuous bounded function , and let @xmath211 be defined as in lemma [ l.differencelimit ] . then , @xmath212 - e\left [ u(\eta_1 , \dots , \eta_r , \xi_1 , \dots , \xi_s ) \right ] \right\| \notag \\ & = \left\| e\left [ u(\gamma_{i_1}+\tau_{i_1 } , \dots , \gamma_{i_r}+\tau_{i_r } , \xi_{j_1 } + \chi_{j_1 } , \dots , \xi_{j_s } + \chi_{j_s } ) \| \mathcal{d}_n \right ] - e\left [ u(\gamma_{i_1 } , \dots , \gamma_{i_r } , \xi_{j_1 } , \dots , \xi_{j_s } ) \right ] \right\| \notag \\ & \leq \left\| e\left [ u(\gamma_{i_1}+\tau_{i_1 } , \dots , \gamma_{i_r}+\tau_{i_r } , \xi_{j_1 } + \chi_{j_1 } , \dots , \xi_{j_s } + \chi_{j_s } ) -u(\gamma_{i_1 } , \dots , \gamma_{i_r } , \xi_{j_1 } , \dots , \xi_{j_s } ) \| \mathcal{d}_n \right ] \right\| \label{eq : nochi } \\ & \hspace{5 mm } + \left\| e\left [ u(\gamma_{i_1 } , \dots , \gamma_{i_r } , \xi_{j_1 } , \dots , \xi_{j_s } ) \| \mathcal{d}_n \right ] - e\left [ u(\gamma_{1 } , \dots , \gamma_{r } , \xi_{1 } , \dots , \xi_{s } ) \right ] \right\| . \label{eq : conditional}\end{aligned}\ ] ] let @xmath213 . since @xmath214 is bounded then is smaller than or equal to @xmath215 \\ & \leq 2 m p\left ( \left . t \geq 1 \right\| \mathcal{d}_n \right ) \leq 2 m \left ( \sum_{t = 1}^r p ( \tau_{i_t } = 1 \| \mathcal{d}_n ) + \sum_{t=1}^s p(\chi_{j_t } = 1\| \mathcal{d}_n ) \right ) \\ & = \frac{2m}{p(\mathcal{d}_n ) } \left ( \sum_{t=1}^r e [ 1(\tau_{i_t } = 1 , \mathcal{d}_n ) ] + \sum_{t=1}^s e [ 1 ( \chi_{j_t } = 1 , \mathcal{d}_n ) ] \right).\end{aligned}\ ] ] to compute the last expectations let @xmath216 and note that @xmath217 = e [ 1(\mathcal{d}_n ) e[1(\chi_{j_t } = 1 ) \| \mathcal{f}_n ] ] = e\left [ 1(\mathcal{d}_n , \delta_n \geq 0 ) \frac{\binom{n-1}{\delta_n-1}}{\binom{n}{\delta_n } } \right ] = e\left [ 1(\mathcal{d}_n , \delta_n \geq 0 ) \frac{\delta_n}{n } \right],\ ] ] and symmetrically , @xmath218 = e\left [ 1(\mathcal{d}_n , \delta_n < 0 ) \frac{\|\delta_n\|}{n } \right],\ ] ] from where it follows that is bounded by @xmath219 + \sum_{t=1}^s e\left [ \left . \frac{\|\delta_n\|}{n } 1(\delta_n < 0 ) \right\| \mathcal{d}_n \right ] \right ) \leq 2 m ( r+s ) n^{-\kappa+\delta_0 } = o(1)\ ] ] as @xmath11 . to analyze we first note that by lemma [ l.differencelimit ] , @xmath220 as @xmath11 , hence @xmath221 & = \frac{1}{p(\mathcal{d}_n ) } e\left [ u(\gamma_{1 } , \dots , \gamma_r , \xi_{1 } , \dots , \xi_s ) 1(\mathcal{d}_n ) \right ] \\ & = e\left [ u(\gamma_{1 } , \dots , \gamma_r , \xi_{1 } , \dots , \xi_s ) 1(\mathcal{d}_n ) \right ] + o(1).\end{aligned}\ ] ] therefore , is equal to @xmath222 + o(1 ) \right\| \leq m p(\mathcal{d}_n^c ) + o(1 ) \to 0\end{aligned}\ ] ] as @xmath11 , which completes the proof . fix @xmath223 and let @xmath224 . for the first limit fix @xmath225 and note that by the union bound , @xmath226 where in the last step we used chebyshev s inequality . clearly , @xmath227 , and since by lemma [ l.differencelimit ] @xmath220 as @xmath11 , then the second term converges to zero . to analyze the first term note that at most one of @xmath228 or @xmath229 can be one , hence , @xmath230 as @xmath11 . next , for the average degrees we have @xmath231 \right\| > \epsilon \right ) & = p\left ( \left . \left\| \frac{1}{n } \sum_{i=1}^n ( \gamma_i+ \tau_i ) - e[\gamma_1 ] \right\| > \epsilon \right\| \mathcal{d}_n \right ) \notag \\ & \leq p\left ( \left . \left\| \frac{1}{n } \sum_{i=1}^n \gamma_i - e[\gamma_1 ] \right\| + \frac{\|\delta_n\|}{n } > \epsilon \right\| \mathcal{d}_n \right ) \notag \\ & \leq \frac{1}{p(\mathcal{d}_n ) } p\left ( \left\| \frac{1}{n } \sum_{i=1}^n \gamma_i - e[\gamma_1 ] \right\| + n^{-\kappa+\delta_0 } > \epsilon \right ) , \label{eq : in - degreewl}\end{aligned}\ ] ] symmetrically , @xmath232 \right\| > \epsilon \right ) & \leq \frac{1}{p(\mathcal{d}_n ) } p\left ( \left\| \frac{1}{n } \sum_{i=1}^n \xi_i - e[\xi_1 ] \right\| + n^{-\kappa+\delta_0 } > \epsilon \right ) , \label{eq : out - degreewl}\end{aligned}\ ] ] and since @xmath233 for all @xmath97 , @xmath234 \right\| > \epsilon \right ) & = p\left ( \left . \left\| \frac{1}{n } \sum_{i=1}^n ( \gamma_i \xi_i + \tau_i \xi_i + \gamma_i \chi_i - e[\gamma_1 \xi_1 ] \right\| > \epsilon \right\| \mathcal{d}_n \right ) \notag \\ & \leq p\left ( \left . \left\| \frac{1}{n } \sum_{i=1}^n \gamma_i \xi_i - e[\gamma_1 \xi_1 ] \right\| + \sum_{i=1}^n ( \tau_i \xi_i + \gamma_i \chi_i ) > \epsilon \right\| \mathcal{d}_n \right ) \notag \\ & \leq \frac{1}{p(\mathcal{d}_n ) } p\left ( \left\| \frac{1}{n } \sum_{i=1}^n \eta_i \xi_i - e[\gamma_1 \xi_1 ] \right\| + n^{-\kappa+\delta } > \epsilon \right ) \label{eq : mix - degreeswl } \\ & \hspace{5 mm } + p\left ( \left . \frac{1}{n } \sum_{i=1}^n ( \tau_i \xi_i + \gamma_i \chi_i ) > n^{-\kappa+\delta } \right\| \mathcal{d}_n \right ) , \label{eq : chiandeta}\end{aligned}\ ] ] for any @xmath235 . by lemma [ l.differencelimit ] , @xmath236 converges to one , and by the weak law of large numbers ( wlln ) we have that each of , and converges to zero as @xmath11 , as required . to see that converges to zero use markov s inequality to obtain @xmath237}{n^{-\kappa+\delta } } = \frac{e [ ( \tau_1 \xi_1 + \gamma_1 \chi_1 ) 1(\mathcal{d}_n)]}{p(\mathcal{d}_n ) n^{-\kappa+\delta}}. \label{eq : mixedproducts}\end{aligned}\ ] ] now let @xmath216 to compute @xmath238 = e [ ( \xi_1 e[\tau_1 \| \mathcal{f}_n ] + \gamma_1e[\chi_1\| \mathcal{f}_n ] ) 1(\mathcal{d}_n ) ] \leq e\left [ \left ( \xi_1 + \gamma_1 \right ) \frac{\|\delta_n\|}{n } 1(\mathcal{d}_n ) \right ] \leq 2\mu n^{-\kappa+\delta_0},\ ] ] which implies that converges to zero . finally , provided that @xmath93 < \infty$ ] , the wlln combined with the arguments used to bound give @xmath239 \right\| \right ) & \leq \frac{1}{p(\mathcal{d}_n ) } p\left ( \left\| \frac{1}{n } \sum_{i=1}^n \gamma_i^2 - e[\gamma_1 ^ 2 ] \right\| + \frac{1}{n } \sum_{i=1}^n ( 2\gamma_i \tau_i + \tau_i^2 ) > \epsilon , \mathcal{d}_n \right ) \\ & \leq \frac{1}{p(\mathcal{d}_n ) } p\left ( \left\| \frac{1}{n } \sum_{i=1}^n \gamma_i^2 - e[\gamma_1 ^ 2 ] \right\| + n^{-\kappa+\delta } > \epsilon \right ) \\ & \hspace{5 mm } + p\left ( \left . \frac{1}{n } \sum_{i=1}^n ( 2\gamma_i \tau_i + \tau_i^2 ) > n^{-\kappa+\delta } \right\| \mathcal{d}_n \right ) \\ & \leq o(1 ) + \frac{e [ ( 2\gamma_1 + 1)\tau_1\| \mathcal{d}_n]}{n^{-\kappa+\delta } } \\ & \leq o(1 ) + \frac{e[2\gamma_1 + 1]}{p(\mathcal{d}_n ) n^{\delta-\delta_0}},\end{aligned}\ ] ] and symmetrically , @xmath240 \right\| > \epsilon \right ) \to 0,\ ] ] as @xmath11 . this subsection contains the proofs of proposition [ p.uniformity ] , which establishes the uniformity of simple graphs , propositions [ p.poissonlimit ] and [ p.repeateddistr ] , which concern the repeated directed configuration model , and proposition [ p.eraseddistr ] which refers to the erased directed configuration model . suppose @xmath241 and @xmath98 have equal sum @xmath242 , and number the inbound and outbound half - edges by @xmath243 . the process of matching half edges in the configuration model is equivalent to a permutation @xmath244 of the numbers @xmath245 where we pair the @xmath68th inbound half - edge to the @xmath246th outbound half - edge , with all @xmath247 permutations being equally likely . note that different permutations can actually lead to the same graph , for example , if we switch the position of two outbound half - edges of the same node , so not all multigraphs have the same probability . nevertheless , a simple graph can only be produced by @xmath248 different permutations ; to see this note that for each node @xmath20 , @xmath249 , we can permute its @xmath17 inbound half - edges and its @xmath18 outbound half - edges without changing the graph . it follows that since the number of permutations leading to a simple graph is the same for all simple graphs , then conditional on the resulting graph being simple , it is uniformly chosen among all simple graphs having bi - degree - sequence @xmath22 . next , we give the proofs of the results related to the repeated directed configuration model . before proceeding with the proof of proposition [ p.poissonlimit ] we give the following preliminary lemma , which will be used to establish that under condition [ cond.regularitydirected ] the maximum in- and out - degrees can not grow too fast . define @xmath255 and note that @xmath0 and @xmath256 are both distribution functions with support on the nonnegative integers . define the pseudoinverse operator @xmath257 and let @xmath258 where @xmath259 is a uniform(0,1 ) random variable . it is easy to verify that @xmath260 and @xmath261 have distributions @xmath256 and @xmath0 , respectively . furthermore , the assumptions imply that @xmath262 as @xmath11 and @xmath263 = \sum_{j=0}^\infty j \frac{1}{n } \sum_{k=1}^n 1(a_{nk } = j ) = \frac{1}{n } \sum_{k=1}^n \sum_{j=0}^\infty j 1(a_{nk } = j ) = \frac{1}{n } \sum_{k=1}^n a_{nk } \to e[x]\ ] ] as @xmath11 , where the exchange of sums is justified by fubini s theorem . now note that by fatou s lemma , @xmath264 \geq e\left [ \liminf_{n \to \infty } x_n 1(x_n \leq \sqrt{n } ) \right ] = e[x],\ ] ] which implies that @xmath265 = 0.\ ] ] finally , @xmath266 = \sum_{j= \lfloor \sqrt{n } \rfloor + 1}^\infty j \frac{1}{n } \sum_{k=1}^n 1(a_{nk } = j ) = \frac{1}{n } \sum_{k=1}^n \sum_{j = \lfloor \sqrt{n } \rfloor + 1}^\infty j 1(a_{nk } = j ) = \frac{1}{n } \sum_{k=1}^n a_{nk } 1(a_{nk } > \sqrt{n}),\ ] ] from where it follows that @xmath267 which in turn implies that @xmath268 following the proof of proposition 7.9 in @xcite , we define the random variable @xmath269 to be the total number of pairs of multiple edges in the same direction , e.g. , if from node @xmath20 to node @xmath106 there are @xmath125 edges , their contribution to @xmath269 is @xmath270 . note that @xmath271 , with strict inequality whenever there is at least one pair of nodes having three or more multiple edges in the same direction . we claim that @xmath272 as @xmath11 , which implies that @xmath273 as @xmath11 . to prove the claim start by defining indicator random variables for each of the possible self - loops and multiple edges in the same direction that the multigraph can have . for the self - loops we use the notation @xmath274 to define @xmath275 and for the pairs of multiple edges in the same direction we use @xmath276 to define @xmath277 the sets of possible vectors @xmath278 and @xmath279 are given by @xmath280 respectively . it follows from this notation that @xmath281 we now proceed to prove that @xmath283 , where @xmath128 and @xmath129 are independent poisson random variables with means @xmath284 and @xmath285 , respectively . to do this we use theorem 2.6 in @xcite which says that if for any @xmath286 @xmath287 = \lambda_1^p \lambda_2^q,\ ] ] where @xmath288 , then @xmath289 as @xmath11 . to compute the expectation we use theorem 2.7 in @xcite , which gives @xmath290 = \sum_{{\bf u}_1 , \dots , { \bf u}_p \in \mathcal{i } } \sum_{{\bf w}_1 , \dots , { \bf w}_q \in \mathcal{j } } p\left(i _ { { \bf u}_{1 } } = \dots = i _ { { \bf u}_{p } } = j _ { { \bf w}_{1 } } = \dots = j _ { { \bf w}_{q } } = 1\right),\ ] ] where the sums are taken over all the @xmath75-permutations , respectively @xmath291-permutations , of the distinct indices in @xmath292 , respectively @xmath293 . next , by the fact that all stubs are uniformly paired , we have that @xmath294 unless there is a conflict in the attachment rules , i.e. , one stub is required to pair with two or more different stubs within the indices @xmath295 and @xmath296 , in which case @xmath297 therefore , from we obtain @xmath298 & \leq \sum_{{\bf u}_{1}, ... ,{\bf u}_{p}\in\mathcal{i } } \sum_{{\bf w}_{1}, ... ,{\bf w}_{q}\in\mathcal{j } } \frac{1}{\prod_{i=0}^{p+2q-1}(l_{n}-i ) } \notag \\ & = \frac{\|\mathcal{i}\| ( \|\mathcal{i}\| - 1 ) \cdots ( \|\mathcal{i}\| - p+1 ) \|\mathcal{j}\| ( \|\mathcal{j}\| -1 ) \cdots ( \|\mathcal{j}\| - q+1 ) } { l_{n}(l_{n}-1)\cdots(l_{n}-(p+2q-1 ) ) } , \label{eq : ub_factorialmoment}\end{aligned}\ ] ] where @xmath76 denotes the cardinality of set @xmath77 . now note that @xmath299 by lemma [ l.maximumgrowth ] and condition [ cond.regularitydirected ] we have @xmath300 as @xmath11 . hence , it follows from condition [ cond.regularitydirected ] that @xmath301+o(1 ) , \\ \frac{\|\mathcal{j}\|}{n^2 } & = \frac{1}{2 } e[\gamma(\gamma-1 ) ] e[\xi(\xi-1 ) ] + o(1 ) , \qquad \text{and } \\ \frac{n}{l_n } & = \frac{1}{\mu } + o(1)\end{aligned}\ ] ] as @xmath11 . since @xmath75 and @xmath291 remain fixed as @xmath11 , we have @xmath302 & = \left ( \lim_{n\rightarrow \infty } \frac{\|\mathcal{i}\|}{n } \right)^{p } \left ( \lim_{n\rightarrow \infty}\frac{\|\mathcal{j}\|}{n^2 } \right)^{q } \left ( \lim_{n\rightarrow \infty}\frac{n}{l_n } \right)^{p+2q}\\ & = \left({e[\gamma\xi]}\right)^{p } \left ( \frac{e[\gamma(\gamma-1 ) ] e[\xi(\xi-1)]}{2 } \right)^{q } \left ( \frac{1}{\mu } \right)^{p+2q } = \lambda_1^{p}\lambda_2^{q}.\end{aligned}\ ] ] to prove the matching lower bound , we note that occurs exactly when there is a conflict in the attachment rules . each time a conflict happens , the numerator of decreases by one . therefore , @xmath303 & = \frac{\|\mathcal{i}\| ( \|\mathcal{i}\| - 1 ) \cdots ( \|\mathcal{i}\| - p+1 ) \|\mathcal{j}\| ( \|\mathcal{j}\| -1 ) \cdots ( \|\mathcal{j}\| - q+1 ) } { l_{n}(l_{n}-1)\cdots(l_{n}-(p+2q-1 ) ) } \\ & \hspace{5 mm } - \sum_{{\bf u}_{1}, ... ,{\bf u}_{p}\in\mathcal{i } } \sum_{{\bf w}_{1}, ... ,{\bf w}_{q}\in\mathcal{j } } \frac{1({\bf u}_1 , \dots , { \bf u}_p , { \bf w}_1 , \dots , { \bf w}_q \text { have a conflict})}{\prod_{i=0}^{p+2q-1}(l_{n}-i ) } \\ & = \lambda_1^p \lambda_2^q - \frac{1}{(\mu n)^{p+2q } } \sum_{{\bf u}_{1}, ... ,{\bf u}_{p}\in\mathcal{i } } \sum_{{\bf w}_{1}, ... ,{\bf w}_{q}\in\mathcal{j } } 1({\bf u}_1 , \dots , { \bf u}_p , { \bf w}_1 , \dots , { \bf w}_q \text { have a conflict } ) + o(1)\end{aligned}\ ] ] as @xmath11 . to bound the total number of conflicts note that there are three possibilities : we now discuss each of the cases separately . for conflicts of type ( a ) suppose there is a conflict between the self - loops @xmath304 and @xmath305 ; the remaining @xmath306 self - loops and @xmath291 pairs of multiple edges can be chosen freely . then the number of such conflicts is bounded by @xmath307 , hence it suffices to show that the total number of conflicting pairs @xmath308 is @xmath309 as @xmath11 . now , to see that this is indeed the case , first choose the node @xmath20 where the conflicting pair is ; if the conflict is that an outbound stub is assigned to two different inbound stubs then we can choose the problematic outbound stub in @xmath310 ways and the two inbound stubs in @xmath311 ways , whereas if the conflict is that an inbound stub is assigned to two different outbound stubs then we can choose the problematic inbound stub in @xmath312 ways and the two outbound stubs in @xmath313 ways . thus , the total number of conflicting pairs is bounded by @xmath314 for conflicts of type ( b ) suppose there is a conflict between the self - loop @xmath304 and the pair of multiple edges @xmath315 ; choose the remaining @xmath316 self - loops and @xmath317 multiple edges freely . then , the number of such conflicts is bounded by @xmath318 , and it suffices to show that the number of conflicting pairs @xmath319 is @xmath320 as @xmath11 . similarly as in case ( a ) , an outbound stub of node @xmath20 can be paired to a self - loop and a multiple edge to node @xmath106 in @xmath321 ways , and an inbound stub of node @xmath20 can be paired to a self - loop and a multiple edge from node @xmath106 in @xmath322 ways , and so the total number of conflicting pairs is bounded by @xmath323 finally , for conflicts of type ( c ) we first fix @xmath324 and @xmath315 and choose freely the remaining @xmath75 self - loops and @xmath325 multiple edges , which can be done in less than @xmath326 ways . it then suffices to show that the number of conflicting pairs @xmath327 is @xmath328 as @xmath11 . a similar reasoning to that used in the previous cases gives that the total number of conflicting pairs is bounded by @xmath329 to prove part ( a ) note that for any @xmath140 , @xmath331 since the @xmath332 are identically distributed . now let @xmath333 and condition on @xmath334 to obtain @xmath335,\ ] ] from where it follows that @xmath336}{p(\mathcal{s}_n ) } \right\| + \left\| p(n_1 = i , d_1 = j ) - f_i g_j \right\| \\ & \leq e\left [ \left\| \frac{p(\mathcal{s}_n \| \mathcal{g}_n)}{p(\mathcal{s}_n ) } - 1 \right\| \right ] + \left\| p ( n_1 = i , d_1 = j ) - f_i g_j \right\|.\end{aligned}\ ] ] theorem [ t.weakconvergence ] gives that the second term converges to zero , and for the first term use theorem [ t.probsimple ] to obtain that both @xmath337 and @xmath338 converge to the same positive limit , so by dominated convergence , @xmath339 \leq e\left [ \lim_{n \to \infty } \left\| \frac{p(\mathcal{s}_n \| \mathcal{g}_n)}{p(\mathcal{s}_n ) } - 1 \right\| \right ] = 0.\ ] ] for part ( b ) we only show the proof for @xmath340 since the proof for @xmath341 is symmetrical . note that @xmath340 is a quantity defined on @xmath330 . fix @xmath223 and use the union bound to obtain @xmath342 by theorem [ t.probsimple ] and lemma [ l.differencelimit ] , @xmath337 and @xmath236 are bounded away from zero , so we only need to show that the numerators converge to zero . the arguments are the same as those used in the proof of proposition [ p.mainconditions ] ; for use chebyshev s inequality to obtain that @xmath343 as @xmath11 , and for @xmath344 which also converges to zero . this completes the proof . finally , the last result of the paper , which refers to the erased directed configuration model , is given below . since the technical part of the proof is to show that the probability that no in - degrees or out - degrees of a fixed node are removed during the erasing procedure , we split the proof of proposition [ p.eraseddistr ] into two parts . the following lemma contains the more delicate step . [ l.survivingstubs ] consider the graph obtained through the erased directed configuration model using as bi - degree - sequence @xmath64 , as constructed in subsection [ ss.algorithm ] . let @xmath345 and @xmath346 be the number of inbound stubs and outbound stubs , respectively , that have been removed from node @xmath347 during the erasing procedure . then , @xmath348 we only show the result for @xmath345 since the proof for @xmath346 is symmetric . define the set @xmath349 and note that in order for all the inbound stubs of node @xmath347 to survive the erasing procedure , it must have been that they were paired to outbound stubs of @xmath350 different nodes from @xmath351 . before we proceed it is helpful to recall some definitions from section [ s.degreesequences ] , @xmath352 , @xmath33 , @xmath353 , @xmath354 , and @xmath355 , where @xmath356 ; also , @xmath90 and @xmath87 are independent sequences of i.i.d . random variables having distributions @xmath0 and @xmath1 , respectively . now fix @xmath357 and let @xmath333 . then , since @xmath358 , @xmath359 \geq e\left [ \left . p\left ( e^+ = 0 \right\| \mathcal{g}_n \right ) 1(1\leq n_1 \leq n^\epsilon ) \right ] + p(n_1 = 0 ) \notag \\ & = e\left [ \frac{1(1 \leq n_1 \leq n^\epsilon ) } { l_n ! } \sum_{(i_1 , i_2 , \dots , i_{n_1 } ) \in \mathcal{p}^+_n } d_{i_1 } d_{i_2 } \cdots d_{i_{n_1 } } ( l_n - n_1 ) ! \right ] + p(n_1 = 0 ) \notag \\ & \geq e\left [ \left . \frac{1(1\leq \gamma_1+\tau_1 \leq n^\epsilon ) } { l_n ! } \sum_{(i_1 , i_2 , \dots , i_{(\gamma_1+\tau_1 ) } ) \in \mathcal{p}^+_n } \xi_{i_1 } \xi_{i_2 } \cdots \xi_{i_{(\gamma_1+\tau_1 ) } } ( l_n - \gamma_1 - \tau_1 ) ! \right\| \mathcal{d}_n \right ] \notag \\ & \hspace{5 mm } + p(n_1 = 0 ) \notag \\ & \geq e\left [ \left . \frac{1(1 \leq \gamma_1 \leq n^\epsilon ) 1(\tau_1 = 0)}{(l_n)^{\gamma_1 } } \sum_{(i_1 , i_2 , \dots , i_{\gamma_1 } ) \in \mathcal{p}^+_n } \xi_{i_1 } \xi_{i_2 } \cdots \xi_{i_{\xi_1 } } \right\| \mathcal{d}_n \right ] + p(n_1 = 0 ) . \label{eq : outboundsurvive}\end{aligned}\ ] ] next , condition on @xmath216 and note that @xmath360 it follows that the expectation in is equal to @xmath361 \\ & \geq e\left [ \left . \frac{\gamma_n}{\gamma_n+\|\delta_n\| } \cdot \frac{1(1 \leq \gamma_1 \leq n^\epsilon ) } { ( \gamma_n + \|\delta_n\|)^{\gamma_1 } } \sum_{(i_1 , i_2 , \dots , i_{\gamma_1 } ) \in \mathcal{p}^+_n } \xi_{i_1 } \xi_{i_2 } \cdots \xi_{i_{\gamma_1 } } \right\| \mathcal{d}_n \right ] \\ & \geq e\left [ \left . \frac{1(1 \leq \gamma_1 \leq n^\epsilon ) \gamma_n}{(\gamma_n + n^s)^{\gamma_1 + 1 } } \sum_{(i_1 , i_2 , \dots , i_{\gamma_1 } ) \in \mathcal{p}^+_n } \xi_{i_1 } \xi_{i_2 } \cdots \xi_{i_{\gamma_1 } } \right\| \mathcal{d}_n \right ] \end{aligned}\ ] ] @xmath362 \right ] \\ & = \frac{1}{p(\mathcal{d}_n ) } e\left [ 1(1 \leq \gamma_1 \leq n^\epsilon ) \frac{(n-1)!}{(n-1-\gamma_1 ) ! n^{\gamma_1 } } e\left [ \left . \frac { 1(\mathcal{d}_n ) \gamma_n n^{\gamma_1 } } { ( \gamma_n + n^s)^{\gamma_1 + 1 } } \cdot \xi_{1 } \xi_{2 } \cdots \xi_{\gamma_1 } \right\| \gamma_1 \right ] \right].\end{aligned}\ ] ] it follows by fatou s lemma , lemma [ l.differencelimit ] and theorem [ t.weakconvergence ] that @xmath363 \right ] + p(\gamma_1 = 0).\end{aligned}\ ] ] next , define the function @xmath364 as @xmath365 , \\\end{aligned}\ ] ] and note that it only remains to prove that for all @xmath366 , @xmath367 . now let @xmath368 and note that @xmath369 - p(\gamma_{n-1 } < an ) \\ & \hspace{5 mm } - e\left [ 1(\gamma_{n-1 } \geq an ) \left\| \frac{(\gamma_{n-1}+t ) n^t}{(\gamma_{n-1}+t + n^s)^{t+1 } } - \frac{1}{\mu^t } \right\| \xi_1 \xi_2 \cdots \xi_t \right ] . \end{aligned}\ ] ] the slln and bounded convergence give @xmath370 and @xmath371 \\ & \leq e\left [ \xi_1 \xi_2 \cdots \xi_t \limsup_{n \to \infty } \left\| \frac{(\gamma_{n-1}+t ) n^t}{(\gamma_{n-1}+t + n^s)^{t+1 } } - \frac{1}{\mu^t } \right\| \right ] = 0,\end{aligned}\ ] ] from where it follows that @xmath372 .\ ] ] the last step is to condition on @xmath373 and use fatou s lemma again to obtain @xmath374 \\ & = \liminf_{n\to \infty } e\left [ \frac { \xi_1 \xi_2 \cdots \xi_t } { \mu^t } p(\|\gamma_{n-1 } + t - \xi_{n } \| \leq n^s \| \xi_1 , \dots , \xi_t ) \right ] \\ & \geq e\left [ \frac { \xi_1 \xi_2 \cdots \xi_t } { \mu^t } \liminf_{n\to \infty } p ( \|\gamma_{n-1 } + t - \xi_{n}\| \leq n^s \| \xi_1 , \dots , \xi_t ) \right ] .\end{aligned}\ ] ] finally , by the same reasoning used in the proof of lemma [ l.differencelimit ] , we obtain @xmath375 since @xmath376 /\mu ^t = 1 $ ] , this completes the proof . to prove part ( a ) note that since the @xmath377 are identically distributed , then @xmath378 it follows that @xmath379 by theorem [ t.weakconvergence ] we have that @xmath380 , as @xmath11 , and for the remaining term note that @xmath381 \notag \\ & \leq e\left [ \left\| 1(d_1^{(e ) } = j ) - 1(d_1 = j ) \right\| \right ] + e \left [ \left\| 1(n_1^{(e ) } = i ) - 1(n_1 = i ) \right\| \right ] . \label{eq : eraseddifferences } \end{aligned}\ ] ] to bound the expectations in let @xmath345 and @xmath346 be the number of inbound stubs and outbound stubs , respectively , that have been removed from node @xmath347 during the erasing procedure . then , @xmath382 & \leq p\left ( e^- \geq 1 \right ) \quad \text{and } \\ e \left [ \left\| 1(n_1^{(e ) } = i ) - 1(n_1 = i ) \right\| \right ] & \leq p\left ( e^+ \geq 1 \right).\end{aligned}\ ] ] by lemma [ l.survivingstubs ] , @xmath383 which completes the proof of part ( a ) . for part ( b ) we only show the proof for @xmath340 , since the proof for @xmath341 is symmetrical . fix @xmath223 and use the triangle inequality and the union bound to obtain @xmath384 from the proof of proposition [ p.repeateddistr ] , we know that the second probability converges to zero as @xmath11 , and for the first one use markov s inequality to obtain @xmath385 \\ & \leq \frac{2}{\epsilon } p(e^- \geq 1 ) \to 0,\end{aligned}\ ] ] as @xmath11 , by lemma [ l.survivingstubs ] . j. kleinberg , r. kumar , r. raghavan , s. rajagopalan , and a. tomkins . , volume 1627 of _ lecture notes in computer science _ , chapter the web as a graph : measurements , models , and methods , pages 117 . springer - verlag , 1999 .
the statistical treatment of systems driven far from equilibrium presents exciting theoretical challenges @xcite . lacking the unified understanding afforded equilibrium phenomena by the work of boltzmann and gibbs , we are exploring unknown territory without recourse to an established theory . our physical intuition , developed in the context of equilibrium systems , can be misleading when faced with nonequilibrium problems . we therefore turn our attention to computational ` experiments ' in which a manifestly nonequilibrium state can be established and studied , seeking to identify key features shared by many nonequilibrium systems . while real physical systems are unquestionably important , complications and subtle details may obscure such common features . while motivated by realistic problems , microscopic rules and boundary conditions are chosen simple enough to facilitate a comprehensive computational study of the full parameter space . many of the questions relevant in equilibrium remain interesting , especially as concern the nature of phase transitions and the principle of universality . the richness of nonequilibrium phenomena is often surprising , as the relaxation of the detailed balance constraint allows a variety of unexpected possibilities : in contrast to equilibrium , the _ dynamics _ now affects the stationary ( long - time ) properties of the system . particularly dramatic effects have been observed in models where the violation of detailed balance is combined with spatial anisotropies and dynamic conservation laws @xcite . there , _ effective _ long - range interactions can be induced even if the microscopic rules are perfectly local in space and time @xcite . in this paper , we consider a model from this class , namely , a lattice gas of two species of particles and holes on a fully periodic lattice in two spatial dimensions . to drive the system out of equilibrium , we bias the hopping rates of the two species in opposite directions , reminiscent of an ` electric ' field , @xmath0 , acting on opposite ` charges ' ( though we stress that there is _ no coulomb interaction _ ) . a nonzero charge current signals the nonequilibrium steady state . the two species interact through an excluded volume constraint and nearest - neighbor attractions . we choose the interactions carefully , in order to unify three important models which appear as limiting cases of our more general theory . first , by letting all particles attract each other , irrespective of their identity ( charge ) , the non - driven limit corresponds to the familiar ising lattice gas @xcite . this well - known equilibrium model will serve as an anchor for our studies of driven systems . turning the bias on but removing all members of one species , we recover the _ driven _ ising lattice gas introduced by katz et al kls ( the kls model ) . the third limit , obtained by letting the interaction strength vanish , corresponds to a non - interacting two - species model first proposed by schmittmann et al @xcite ( the shz model ) . while the klsmodel phase - separates , via a continuous transition , into high- and low - density strips _ aligned _ with the field @xcite , the shz model orders into _ transverse , _ charge- _ _ _ _ and mass - segregated , strips @xcite , similar to jamming instabilities in traffic models @xcite . if a charge imbalance is imposed , these strips drift @xcite . our study will allow us to bridge the gap between these very different scenarios . to set the scene , we briefly survey some of the relevant previous work . as a single parameter modification of the ising lattice gas , the kls model is a _ minimal model _ for the study of nonequilibrium steady states ( ness ) . particle hops along one direction ( parallel to the @xmath1-axis ) occur at the normal equilibrium rate , as if in contact with a heat bath at temperature @xmath2 . particle hops in the other direction ( parallel to the @xmath3-axis ) are enhanced ( suppressed ) in the positive ( negative ) direction by coupling to an external field , @xmath0 . with periodic boundaries in the @xmath3-direction , a nonzero current is maintained , and the system settles into a ness . at half - filling , there remains a continuous transition , though with @xmath4 increasing monotonically with @xmath0 and saturating at @xmath5 @xcite . the transition falls into a novel universality class with exponents distinct from the ising ones @xcite . the critical behavior is strongly anisotropic , with distinct sets of exponents characterizing fluctuations _ perpendicular _ and _ parallel _ to @xmath0 . there has been some discussion regarding the nature of these fluctuations , with some authors disputing the original claim that the correct mesoscopic description is gaussian in the perpendicular direction @xcite . though the anisotropy makes numerical investigations of the critical behavior quite subtle and computationally intensive , recent high precision monte carlo studies compare the two mesoscopic descriptions . the results are in complete agreement with the predictions of the original field theory @xcite . as a final note , we mention that the combination of anisotropic dynamics and a conservation law introduces power law correlations at all @xmath6 eff - lr , a manifestation of the relaxation of the detailed balance constraint . these correlations are revealed by the structure factor , which has a discontinuity singularity at the origin . in this sense , even the ` disordered ' phase is quite non - trivial . turning to multispecies versions , the simplest ( ` shz ' ) model @xcite allows _ two _ different types of particles , distinguished only by their interaction with the external field . positive ( negative ) particles are biased to hop in the positive ( negative ) @xmath3-direction , and interact only via an excluded volume constraint . the temperature is absorbed into @xmath0 and the only parameters are @xmath0 , the overall mass density @xmath7 , and the overall charge density ( i.e. , the density difference of the two species ) , @xmath8 . here , the mechanism for ordering is the mutual volume exclusion of the particles , so that at sufficiently strong @xmath0 and large @xmath7 , the system locks into a high density strip _ perpendicular _ to @xmath0 , with positive and negative particles blocking each other . for non - zero charge density @xmath8 , this strip is found to drift in the direction of the minority species @xcite . depending on where the phase boundary is crossed , first - order or continuous transitions are observed @xcite . various other remarkable properties have been discovered . for a range of aspect ratios , configurations with non - zero winding number ( ` barber poles ' ) are quite frequently observed , _ in addition _ to the usual transverse strips , raising the possibility of bistability @xcite . power law correlations characterize the disordered phase , with _ directionality - dependent _ exponents @xcite . another subtle issue concerns the lower critical dimension : while an exact solution for a strictly one - dimensional model , characterized by a single ` lane ' parallel to @xmath0 , precludes a transition @xcite , monte carlo data for a ` two - lane ' model indicate the presence of a macroscopic cluster in finite systems @xcite . very subtle finite - size effects control the decay of this cluster in the thermodynamic limit @xcite . in this paper , we consider the two - species model at finite @xmath2 and @xmath0 , where interparticle interactions are expected to play an important role . by varying @xmath2 , @xmath0 , and @xmath8 , the fraction of the _ total _ population which are of the _ minority _ species , we can interpolate smoothly from the kls model to the ( non - interacting ) two - species shz model . hence , we expect a competition between the two types of ordered configurations parallel vs transverse strips favored by these two limits . as @xmath8 varies from @xmath9 ( kls model ) to @xmath10 ( equal numbers of each ) , there should be some critical @xmath8 where the preferred order switches . to explore these phenomena in more detail , we map out the phase diagram in @xmath11 @xmath8 and @xmath2 , for a range of system sizes . the energy scale is set by our choice of the interparticle attraction @xmath12 , and the overall mass density @xmath7 is fixed at @xmath10 so that the ising critical point remains accessible . many questions arise in connection with earlier work . how do nearest - neighbor attractions modify the two - species transition ? what will be the effect of a few ` impurities ' ( i.e. , minority particles ) on the kls transition ? at what concentration do the ` impurities ' become relevant and change the nature of the transition ? preliminary results , focusing on a restricted parameter space , were already reported in @xcite ; here , we explore a much wider parameter range , including several system sizes . we will be able , if not to answer these questions fully , then to at least suggest the character of their resolution sufficiently to guide further research . our main results are as follows . at fixed @xmath0 and sufficiently small @xmath8 , a line of continuous transitions emerges from the pure kls ( @xmath13 ) point , in the @xmath8-@xmath2 plane . this line separates the disordered phase from an ordered one , characterized by a particle - rich strip parallel to @xmath0 . as we increase @xmath8 , we encounter a bicritical point , where the transition line splits into a line of continuous order - disorder transitions , from disorder into a strip transverse to @xmath0 , and a line of first - order transitions along which transverse and parallel order coexist . if we fix @xmath8 and lower @xmath2 , we first observe the transition from disorder into the transverse strip , followed by a transition into parallel order . this topology persists at higher @xmath0 , except that all lines are shifted to slightly higher temperatures . the size - dependence of the phase diagram is subtle , since the main features are controlled by _ different _ scaling variables . on the one hand , the transition into the transverse strip is controlled by the _ effective _ drive @xmath14 where @xmath15 is the system size in the drive direction . on the other hand , the bicritical point appears to depend on the scaling variable @xmath16 which translates into the number of _ rows _ ( transverse to @xmath0 ) which can be filled with the minority species . finally , the pure kls point requires finite - size scaling at fixed shape factor @xmath17 @xcite , in two spatial dimensions . the remainder of the paper is organized as follows . we first describe in detail the microscopic model and the observables which are used to locate the different phases . we then present our simulation results , beginning with the structure of typical configurations in different parts of parameter space and their associated order parameters . by monitoring the signatures of first and second order transitions , we compile a cut through the phase diagram at fixed @xmath0 , with variable @xmath8 and @xmath2 . the phase boundaries and their dependence on system size are analyzed in some detail . to complete the picture , we present two cuts at different but fixed temperatures , crossing the phase boundaries by varying @xmath0 and @xmath8 . we conclude with a brief summary and a discussion of some open questions . we consider periodic square lattices of size @xmath18 , in two spatial dimensions , with @xmath0 parallel to the ( positive ) @xmath3-axis . a configuration is specified by the set of occupation variables , @xmath19 , where @xmath20 takes three values , @xmath21 denoting a positive ( negative ) particle or a hole at lattice site @xmath22 . often , we will only need to distinguish particles from holes , via @xmath23 . all lattices are half filled , i.e. , @xmath24 , so that the ising critical point remains accessible . an important parameter is the fraction of negative particles ( the ` minority species ' ) in the system : @xmath25 . clearly , we only need to consider the sector @xmath26 , from having no negative particles at all to equal numbers of each species . for later reference , we also introduce the charge density , @xmath27 . the nearest - neighbor attraction is modelled by the ising hamiltonian @xmath28we choose attractive interactions , @xmath29 , regardless of species . while many other choices are possible and interesting , ours provides maximum linkage to known cases : ising , kls and shz . the monte carlo dynamics conserves the number of each species and is specified as follows . an update attempt begins by picking a bond at random . if the bond connects a particle - hole pair , the contents are exchanged with the metropolis rate @xmath30 \right\ } $ ] @xcite . hence , at @xmath31 we recover the equilibrium ising model with conserved magnetization , coupled to a heat bath at temperature @xmath2 . we will set @xmath32 and measure @xmath0 in units of this ( arbitrary ) energy scale . @xmath2 will be quoted in units of the onsager temperature , @xmath33 . the change in @xmath3-coordinate , due to the proposed move , is denoted by @xmath34 , and @xmath35 is the associated change in internal energy . the term @xmath36 models the gain or loss of energy from the coupling to @xmath0 ; if @xmath37 is positive ( negative ) the move is favored ( unfavored ) . our model , in which @xmath0 and @xmath2 are varied independently , raises an interesting issue . if the ratio @xmath38 is quite large , it becomes almost impossible for particles to hop backwards . in a finite system , this implies that a relatively small fraction of the minority species provided the ` right ' fluctuation occurs is sufficient to form a stable blockage . even though such a fluctuation becomes less probable in a larger system , the dynamics nevertheless becomes nonergodic in the limit @xmath39 . in principle , this can be avoided by introducing , e.g. , a small probability for particles to exchange places @xcite . to limit the number of parameters , we circumvent these problems here by considering different initial configurations and a range of system sizes . the dynamics is diffusive , and therefore conserves both charge and mass density . though the _ local _ effect of the external field is analogous to the effect of an electrostatic potential on electric charges , the boundary conditions exclude the possibility of a _ global _ hamiltonian description . as overall density is conserved , we expect ordered configurations to be strips of higher density coexisting with strips of lower density . we therefore introduce the fourier transform of the local mass variable , @xmath40which is labelled by ( integer ) wavenumbers , @xmath41 , @xmath42 . the structure factor , @xmath43 then serves as a good order parameter , since it is sensitive to _ mass - segregated _ strip configurations . for example @xmath44 will be @xmath45 for a strip aligned with the field , characteristic of kls order ; similarly , @xmath46 will detect a strip transverse to @xmath0 which develops in the shz ( two - species ) model ; and both are normalized to @xmath47 for a disordered configuration . we also monitor a ` susceptibility ' , i.e. the fluctuations of the order parameter : @xmath48 \nonumber\end{aligned}\ ] ] we note that @xmath49 involves the fourier transform of the _ mass _ variable and is therefore not sensitive to any charge - segregated structures . replacing @xmath50 by @xmath51 in eq . ( [ s ] ) generates structure factors which respond to charge inhomogeneities . we have monitored these and their fluctuations throughout , and found that their behavior is consistent with the mass - based quantities . when @xmath52 is calculated , the average is taken over multiple steady - state configurations of a monte carlo run , with a typical run lasting @xmath53 m ( @xmath54 ) monte carlo steps ( mcs ) , and @xmath55 bond update attempts per mcs . data are collected every @xmath56 mcs ; fluctuations of observables indicate that this interval is sufficient to produce uncorrelated data in the largest @xmath57 systems considered . typically , the initial @xmath58 m mcs are discarded to ensure that data are taken from the steady state . near critical points and at low temperatures these numbers require modification , due to long correlation times and long - lived metastable states . in such cases the only recourse is a careful analysis of individual , very long runs . when that is necessary we will measure a quantity closely related to @xmath52 : @xmath59which measures the type of order present in a _ single _ configuration . we can then track @xmath60 for different @xmath61 s and @xmath62 s over the course of a run and see precisely how the averages are generated . finally , at each bond update we tally the quantity @xmath63 , which is then averaged over a run to give the charge current @xmath64 . however , it is not particularly illuminating to compare @xmath65 for different values of the effective drive @xmath38 . we are more interested in a quantity which is a property of the gross structure of the steady state , namely , the conductivity @xmath66 defined via @xmath67 . [ fig1 ] now that we have described the various quantities which will be used to probe the behavior of our model , we turn to the presentation of the data . in this section we seek the location and character of transitions by scanning in @xmath8 and @xmath2 at fixed drive @xmath0 . we choose @xmath68 since this intermediate value still allows for a significant fraction of backward jumps , thus avoiding the spurious metastable configurations discussed above . at the same time , it is large enough to induce measurable currents and other clear signatures of far - from - equilibrium behavior . the two order parameters @xmath44 and @xmath46 and their fluctuations are monitored in order to identify the different phases . large peaks in their fluctuations , or the presence of hysteresis , are used as indicators of continuous vs first order transitions , respectively . for clarity , we first present a quick overview of the topology of the phase diagram , and then turn to the details of the data which underlie this picture . [ fig3 ] fig . [ fig1 ] shows the phase diagram in the @xmath8-@xmath2 plane , at @xmath68 , for two different system sizes . three phases are found : a homogeneous , disordered phase ( do ) , a transverse strip ( ts ) phase as in the two - species model , and a parallel strip ( ps ) phase as in the kls model . the value of @xmath8 determines which phase is observed : at @xmath69 there is only one species of particles , and a single transition is observed from disorder into the parallel strip . as @xmath8 increases , this transition persists until the number of the minority species is sufficient to create a blockage and form the transverse strip . from here on , _ two _ transitions are observed : from disorder into the transverse strip , and at a lower temperature from the transverse strip into the parallel strip . upon increasing @xmath8 further , only the do - ts transition can be detected . although the ts - ps line can not have a critical endpoint for reasons of symmetry , for @xmath70 it occurs at such a low temperature that it can not be observed in simulations of a reasonable length . now that we have briefly discussed the phase diagram we can look in more detail at the phases and their boundaries . before presenting the data ( structure factors , their fluctuations , currents ) we will begin with some pictures of typical configurations at various points in the phase diagram . this way the reader can develop some intuition about the model . all pictures show @xmath71 lattices , with @xmath0 pointing up . white ( black ) pixels are positively ( negatively ) biased particles , blue - green ( gray in print ) pixels are holes . we will begin at the right side of the phase diagram and explore how typical configurations change as we move left , decreasing the fraction of the population which belongs to the minority species . [ fig5 ] in fig . [ fig3 ] we present four configurations at @xmath72 : the first picture is at @xmath73 , well above @xmath74 . unsurprisingly , this configuration lacks any visible structure since it falls deep into the homogeneous phase . this will be the only picture presented for the homogeneous phase , as the only _ visible _ feature which changes with @xmath8 is the ratio of white to black pixels . the next picture is at @xmath75 , and now we begin to see the two - species type phase separation . lowering the temperature further to @xmath76 , the horizontal strip appears quite clearly , though it remains diffuse at the boundaries , and there are many ` travelers ' ( isolated particles ) moving through the remainder of the system . at this temperature backward hops of particles are not too improbable , occuring with a rate @xmath77 , and thus a fair number of holes are able to enter the strip and allow particles to slip through the blockage . finally at @xmath78 we are deeply inside the ordered phase , and the boundaries of the strip are very sharp and travelers are few . at this temperature , holes are rarely able to penetrate into the interior of the strip . the irregular shape of the interface between the two species is a result of the quench from disorder ; a more gradual lowering of the temperature would result in a smoother interface . lowering @xmath8 to @xmath79 ( fig . [ fig5 ] ) , only @xmath80 rows of the minority species are left . the first frame shows a configuration just below criticality at @xmath81 . though a blockage can still form , the strip is no longer symmetric with respect to @xmath82 and @xmath83 . this leads to a drifting of the strip : occasionally , the rather thin blockage of the minority species is opened by backward hops , and the majority species pours through . these particles then travel quite rapidly around the periodic lattice and attach to the back of the majority blockage , the net result being an _ upward _ drift of the strip . lowering @xmath2 further to @xmath84 ( third frame ) it appears that interfaces _ parallel _ to @xmath0 are becoming favorable ; this type of configuration is common at these intermediate values of @xmath8 : here , parallel and transverse strips compete with each other . indeed the final frame ( @xmath85 ) shows the preferred low temperature configuration : a single strip of _ mixed _ charge parallel to @xmath0 , raising the possibility of a sequence of _ two _ transitions as a function of @xmath2 . now that we have developed a qualitative notion of the various regions in the phase diagram , we turn to a more quantitative analysis of the phases and their boundaries . as in the preceding section , we present the order parameters and their fluctuations as a function of @xmath2 , for a range of @xmath8 , though with one important difference : we will show data for at least _ two _ system sizes . while this can not replace a high - precision finite - size scaling analysis of the transitions , it is intended to provide a rough picture of how the fluctuations of the order parameter scale with system size . [ fig10 ] we begin as before at @xmath72 , with equal numbers of each type of particle . in figs . [ fig10 ] and [ fig11 ] , @xmath86 and its fluctuations are plotted as a function of @xmath2 for two system sizes ( @xmath71 and @xmath87 ) ; @xmath88 is not shown as the only transition here is from the homogeneous phase into the transverse strip . in both systems , @xmath86 goes smoothly to zero as @xmath2 is increased . a clean peak in @xmath89 is also observed in each system , increasing in amplitude with the system size . these two observations are consistent with a continuous transition into the transverse strip at @xmath72 , and we therefore use the location of the peak in @xmath89 to locate the phase boundary in fig . 1 , with @xmath90 and @xmath91 . reducing @xmath8 to just a few rows of the minority species introduces some new features into the data , which we describe briefly before continuing our analysis . here , we still observe the transition described above , with the same behavior of @xmath86 and @xmath89 . however , now that the strip is no longer symmetric , it drifts . the ordered phase therefore fluctuates significantly , and @xmath92 develops a shoulder at about @xmath93 of the value it reaches at the transition . we will see below that the fluctuations of the ordered phase can obscure the transition in smaller systems . [ fig11 ] [ fig14 ] upon reducing @xmath8 further we observe a sequence of _ two transitions _ as a function of @xmath2 . [ fig14 ] shows both order parameters , @xmath86 and @xmath88 for @xmath80 rows of the minority species : @xmath94 in the @xmath71 system . also shown in fig . [ fig15 ] is @xmath89 for @xmath80 rows of the minority species in both the @xmath71 and the @xmath87 system . we have omitted the @xmath87 data for the order parameters to keep the plot uncluttered . as before , @xmath86 and @xmath92 signal a continuous transition into the ts phase as @xmath2 is lowered , though the signal in @xmath89 is much more pronounced in the @xmath87 system . and also as before , there are significant fluctuations associated with this phase , due to strip drifting . specifically , in the @xmath87 system @xmath89 actually has a broad secondary peak in the ordered phase . the large magnitude of this signal is is quite unexpected and awaits a satisfactory explanation . for now , we only note that _ lowering _ @xmath2 _ increases _ the effective bias , @xmath38 , and therefore enhances fluctuations associated with the drive . lowering @xmath2 further we observe @xmath86 falling abruptly , while @xmath88 climbs rapidly , suggesting a discontinuous transition from the horizontal strip ( ts ) into the vertical strip ( ps ) . in the neighborhood of such a transition , one expects to see metastability of the unfavored phase , and this is indeed the case as shown in fig . [ fig16 ] . here we have plotted _ time traces _ ( as opposed to configurational averages ) of structure factors for individual configurations , @xmath95 and @xmath96 , defined in eqn . ( [ tt ] ) . clearly , when @xmath95 ( @xmath96 ) @xmath97 the configuration is a perfect vertical ( horizontal ) strip . sufficiently close to the transition , the time traces reveal the expected behavior , as the system switches between the two ordered phases . notice the length of the run shown : @xmath98 m mcs , which is a factor of @xmath98 longer than typical runs , indicating that the lifetimes of metastable configurations are already quite long even in the @xmath71 system , rendering such behavior nearly unobservable in the @xmath87 system . [ fig15 ] [ fig16 ] at smaller values of @xmath8 we are nearing the junction of the three phase boundaries , which considerably complicates the analysis of data from small systems in a couple of ways . the sequence of transitions ( do - ts followed by ts - ps ) becomes difficult to resolve , as they are quite close in temperature , and massive fluctuations from the first - order ts - ps line may wash out the signal in @xmath89 which locates the continuous do - ts line . and if the junction of the three lines is indeed a nonequilibrium bicritical point , we can expect unexplored finite - size effects to interfere with the analysis . we can , however , make some progress based on the assumption that the relevant control parameter near the bicritical point is the number of rows of the minority species . this hypothesis will be treated in more detail below , in the sections on scaling arguments . at precisely one row of the minority species it is no longer possible to accurately resolve the two transitions in the @xmath71 system . a weak transverse ordering is observed , with @xmath86 reaching at most @xmath99 of perfect order . in the vicinity of the ps - ts first - order transition , huge fluctuations associated with switching between metastable configurations are observed , which wash out the signal of the do - ps transition . however , it is interesting that these transitions can be resolved in larger systems at precisely one row of the minority species : there , these two transitions are sufficiently far apart in temperature since ( as we will see below ) @xmath100 increases with @xmath15 across the do - ts transition . this likely explains why the do - ts transition is not observed in the @xmath71 system . [ fig25 ] below one row of the minority species we no longer observe the transverse order , though we caution that this may be strictly correct only for the _ finite _ system . with @xmath8 just below a single row of the minority species , the high temperature phase is homogeneous and the low temperature phase is the parallel strip . in the vicinity of the transition , huge fluctuations are observed in @xmath46 and especially @xmath101 , where the fluctuations are an order of magnitude larger than the signal at the do - ts boundary . in this region neither @xmath102 nor @xmath46 posesses a well - defined average ; timetraces indicate vigorous competition between the two ordered phases . we conjecture that we are close to the bicritical point in the finite system , and are therefore unable to resolve the transition without some knowledge of scaling to guide the analysis . at smaller @xmath8 we are farther from the bicritical point , and the complications from the presence of the minority species are less severe . fig . [ fig25 ] shows @xmath44 for two system sizes at exactly one - half row of the minority species . we have not shown @xmath46 , as the signal has become insignificant . @xmath101 shows the low - temperature configuration to be a single parallel strip , the smooth approach to zero again suggesting a continuous transition . and indeed , the data for @xmath103 is consistent with this conjecture , with a sharp peak which increases with system size . the amplitude of the peak is smaller by a factor of five than the peaks observed near the bicritical point , suggesting that it is associated with fluctuations about a well - defined average , as opposed to transitions into and out of the ordered phase . timetraces support this conjecture . charge currents were also measured in order to look for signatures of the various phases and transitions . we note however that as we are changing @xmath2 , we are also changing the _ effective _ bias @xmath104 , and therefore a more appropriate quantity is the conductivity , @xmath105 . in fig . [ fig28 ] we plot @xmath66 as a function of @xmath106 for each @xmath8 discussed in this and the following sections ; we present only data for the @xmath71 system as there are no significant differences between the two system sizes . at temperatures below @xmath107 ( @xmath108 ) the conductivity vanishes in the @xmath72 systems ; as the temperature is raised the effective drive is reduced , and backward hops occasionally occur , allowing a small current to trickle through the blockage . upon raising the temperature further , the conductivity changes slope at @xmath109 in the @xmath71 system , which _ does not _ coincide with the phase transition . rather the maximal conductivity occurs in the disordered phase , but at a temperature which is not too large , so that backward hops are not so common as to begin reducing the current . the transition apparently corresponds instead to the _ inflection point _ ( @xmath110 ) , where the curvature changes sign . at @xmath111 , the conductivity has a slope discontinuity in the @xmath71 system at @xmath112 , slightly above the critical temperature @xmath113 . though there should be an inflection point in @xmath66 near the transition , our data are not precise enough to locate it . at @xmath94 and below , the conductivity drops smoothly to zero with @xmath2 , showing no indication of either transition . apparently , @xmath52 and @xmath114 are far more sensitive to the transitions in our model . [ fig28 ] we close this section on the @xmath8-@xmath2 phase diagram with a summary of the results . the picture at higher @xmath8 is clear : a clean continuous transition into the horizontal strip , with @xmath100 decreasing with @xmath8 . when @xmath8 is reduced to approximately three rows of the minority species , the signal of the transition remains clear , though it now sits atop a shoulder of fluctuations of the ordered phase . at yet smaller @xmath8 , a second transition appears between the two ordered phases at lower @xmath2 ; it has the characteristics of a first - order transition . at even lower @xmath8 , at approximately one row of the minority species , both fluctuations perpendicular and parallel to @xmath0 become so violent that the do - ts transition is only seen in larger systems , as the two transitions nearly overlap in the @xmath71 system . close to the @xmath69 point , the transition is once again clean and apparently continuous into a vertical strip of mixed charge . of course , all these statements are based on an analysis of _ finite _ systems . in order to draw robust conclusions , a more systematic analysis of larger samples is required . in the preceding section we studied a slice of the phase diagram at constant @xmath0 , varying the fraction of the minority species and the temperature . varying @xmath2 effectively varies both the strength of particle - particle attractions _ and _ the strength of the bias , since the relevant quantities in the rates are @xmath115 and @xmath38 . in this section we consider a different cut through the phase diagram . by varying @xmath0 and @xmath8 at fixed @xmath2 , the interparticle attractions are held constant while the strength of the bias is varied . in this way we can study directly the competition between the drive and the attractive interactions . in the following , we choose a value for @xmath2 , and then scan in @xmath0 for several values of @xmath8 . the temperatures are chosen by reference to the kls temperature : @xmath116 is above the critical temperature of the kls model at saturation , and @xmath117 is at the critical temperature for @xmath68 , studied in the previous section . at this stage , we have only data for @xmath71 systems , and are therefore as yet unable to speculate on results for larger systems . however , they cast a new light on the more detailed results of the previous sections . [ fig36 ] as before , we first survey the phase diagram with the help of some typical configurations . the qualitative picture will then be made more quantitative in the next section , by examining the behavior of currents and order parameters . [ fig36 ] shows a series of configurations at various @xmath8 for @xmath119 . the first frame clearly shows the transverse strip at @xmath72 , and the absence of travelers suggests that the strip is stationary . in the next frame we have reduced @xmath8 to @xmath120 , reducing the thickness of the minority species to exactly two rows . consequently we now see some travelers trickling through a break in the blockage . watching an animation in this region of the phase diagram reveals an interesting behavior : the strip is mostly quiescent , except for a few particles hopping back and forth at the particle - hole interface . aside from the different ratio of @xmath82 to @xmath83 , these configurations look similar to the @xmath72 strip . then , a sudden large fluctuation opens up a hole in the minority blockage : the @xmath82 particles pour through , and the strip fluctuates and drifts partway around the lattice , until the blockage is reestablished . reducing @xmath8 further to @xmath79 ( third frame ) we see a strip in the middle of one of these fluctuation events . in contrast to @xmath111 where such large fluctuations are relatively rare , the situation is now reversed ; i.e. , the quiescent periods become less frequent . in the fourth frame , we set @xmath121 , and while the strip is still clearly visible it now drifts continuously . the final two frames show @xmath122 and @xmath123 . now there is no longer any clear evidence of phase separation . this rough picture is consistent with our earlier investigation of the two - species transition , where we observed transverse order at and above a single row of the minority species . [ fig37 ] fig . [ fig37 ] presents the phase diagram at @xmath116 . the boundary separates a horizontal strip at high @xmath0 and @xmath8 from a homogeneous phase at small @xmath0 and @xmath8 . as we are above the critical temperature for the kls model at saturation bias , the vertical strip does not appear at any @xmath0 ; at low @xmath8 ( where we might otherwise expect to see such ordered configurations ) the system simply remains disordered for any @xmath0 and @xmath8 . [ fig38 ] and [ fig39 ] show @xmath46 and @xmath124 for several values of @xmath8 ; in the interest of clarity @xmath125 is plotted for only four @xmath8 s . as @xmath0 is increased the system [ fig38 ] [ fig39 ] orders into a transverse strip , with @xmath46 saturating at smaller values as @xmath8 is decreased . at @xmath121 ( exactly one row of the minority species ) @xmath52 saturates at only @xmath126 , indicating that the transverse ordering is rather weak , though comparison with the data for @xmath127 shows dramatically different behavior . here @xmath46 reaches a maximum of only @xmath128 , and the behavior can hardly be called ` saturation ' . @xmath124 also signals a transition , though the clean , sharp spike at @xmath72 becomes a broad bump at @xmath121 , and shows no signal at @xmath127 . the susceptibility also indicates a difference in the ordered phases at different @xmath8 : at large @xmath8 increasing @xmath0 suppresses fluctuations , while at smaller @xmath8 ( when the strip begins drifting ) increasing @xmath0 enhances fluctuations . it is important to note that the fluctuations at high @xmath0 are fluctuations about the ordered phase , as @xmath124 is always @xmath129 to @xmath130 orders of magnitude larger than @xmath103 . though the data is not included in the plots , we have checked the behavior of the ordered phase for @xmath0 as high as @xmath98 . the fluctuations for small @xmath8 ( @xmath131 ) saturate and bounce around a well - defined average , while for larger @xmath8 they are suppressed . this is true for @xmath121 and greater ; at @xmath127 the magnitude of fluctuations in either direction are comparable . we also examined the conductivity across the phase boundary . we do not present the data here , as it is qualitatively similar to the data in fig . [ fig28 ] . again , the transition apparently coincides with the inflection point of the conductivity . in the preceding section we found a simple phase diagram , with a single boundary separating two - species order from the homogeneous phase . since the temperature was chosen well above the maximum kls transition temperature , @xmath133 , the kls ordered phase could not be observed . now , we lower the temperature to @xmath134 and explore the corresponding ( @xmath8-@xmath0 ) slice of the phase diagram ( fig . [ fig41 ] ) . as [ fig41 ] long as we remain at @xmath70 , we observe a transition similar to the one at @xmath116 : from the homogeneous phase into two - species order . in contrast , for @xmath135 the kls transition is observed , since the minority species is too scarce to form a blockage and @xmath136 . between these two limiting values of @xmath8 we are again in the vicinity of the bicritical point , and the situation becomes complicated due to the competing types of order . [ fig42 ] shows typical configurations at two different values of @xmath8 for various @xmath0 . for @xmath137 these configurations look much as they did at [ fig42 ] @xmath116 , so they need not be included . the @xmath94 , @xmath119 configuration shows some very interesting structure , almost ` equal parts ' kls and two - species order , suggesting some serious competition between the two phases . we stress that this is a _ typical _ configuration . when @xmath8 is reduced to a single row ( @xmath121 ) this competition is reduced , and we see instead a kls phase with some local two - species order . this trend continues upon reducing @xmath8 to zero , where at high @xmath0 the kls order is observed . the other panels show @xmath94 and @xmath138 at smaller values of @xmath0 , @xmath139 and @xmath140 respectively . these @xmath0 values were chosen because they maximize the two - species order for these @xmath8 s . in each case the strip drifts rapidly around the lattice . interestingly , as the majority species is piled onto the back of the drifting strip it builds long fingers , leading to a very irregular interface . pictures at smaller @xmath8 s are not shown , as they do not form the transverse strip at smaller @xmath0 . in the next section we locate these boundaries and study the phases in greater detail by examining the behavior of @xmath52 , @xmath114 , and @xmath66 . in fig . [ fig44 ] we plot @xmath52 for the high-@xmath8 phases . for @xmath72 and @xmath120 , @xmath46 shows the system ordering into the two - species phase much as in the previous section . there is , however , one notable new feature for @xmath111 : we see a slight suppression of @xmath46 over an intermediate range of @xmath0 , from about @xmath141 to @xmath142 . in this range backward hops occur frequently enough to occasionally open a hole in the minority blockage , allowing the majority species to pour through until the blockage reforms . increasing @xmath0 reduces the likelihood of such events , until the strip is perfect except for some surface diffusion at the particle - hole interface . at @xmath94 the behavior changes dramatically . after maximizing the two - species order at @xmath143 , increasing @xmath0 further suppresses @xmath46 and enhances @xmath44 , until both saturate below @xmath58 . the system can hardly be said to have switched to kls order , though neither is it precisely disordered either . to understand this peculiar behavior , we recall that @xmath2 ( and therefore @xmath115 ) is held constant , so that increasing @xmath0 favors interfaces _ parallel _ to @xmath0 . yet , the number of minority particles is large enough to form bubbles of local two - species order , as one can discern from the top left panel of fig . [ fig42 ] . apparently the competition between the two phases is very balanced in this small system . it would be interesting to simulate larger systems and explore whether this type of ` phase competition ' persists , or whether the kls order eventually becomes stable . additional information is provided by the susceptibilities . observations of @xmath124 are consistent [ fig44 ] [ fig47 ] with a continuous transition from disorder into the transverse strip at small @xmath0 . seeking a signature of the kls phase , we show @xmath144 and @xmath124 for @xmath94 in fig . fig47 . here we observe a second peak in @xmath124 at @xmath145 , which corresponds to @xmath146 in fig . this peak is rather broad and its amplitude is more than twice that of the first peak ( associated with the do - ts transition ) . perhaps it suggests a first - order transition which would be observed in a larger system , separating the two - species phase from the kls phase . finally , at higher @xmath0 both @xmath124 and @xmath103 fluctuate around nonzero values , reflecting the fluctuations of the competing phases . we note no signal of a transition in @xmath103 . we now continue by looking at the data for @xmath121 , shown in fig . fig48 , where we have plotted @xmath52 for each type of order . we observe that [ fig48 ] the transverse strip only makes a brief appearance , with @xmath147 reaching a maximum of only @xmath148 . at higher @xmath0 , the parallel strip stabilizes and reaches a value slightly above @xmath149 . however , we observe no clear signature of a distinct transition between the two types of order . there is a vague remnant in @xmath125 of the double peak structure seen at @xmath94 , and then a broad , messy bump in @xmath103 before it settles down to capture the fluctuations of the ordered phase . only further investigation of larger systems can determine whether these weak signals in fact indicate a transition . it is clear , however , that the high @xmath0 phase here is not the ` phase competition ' seen previously at @xmath94 , since neither @xmath46 nor @xmath150 contain any trace of it . upon reducing @xmath8 below @xmath138 we find that the transverse strip has essentially disappeared . meanwhile , @xmath44 looks similar to fig . [ fig48 ] , saturating to about @xmath151 , indicating that the kls strip has formed . apparently the system is unable to completely order at any @xmath0 , since @xmath152 . we stress that this behavior is due to the proximity of the kls phase transition , not due to some residual competition with the transverse strip . finally , we note that the behavior of the susceptibility is consistent with a continuous transition into the parallel strip . the conductivities are shown in fig . [ fig52 ] . at @xmath72 and @xmath120 , @xmath66 has much the same form as seen before , decreasing monotonically with @xmath0 . this is no longer the case when @xmath8 is reduced to @xmath79 . now @xmath66 develops a broad minimum which coincides with the appearance of the [ fig52 ] two - species order , followed by a small peak at @xmath153 before falling off with @xmath0 . the peak coincides with the crossing of @xmath46 and @xmath44 , which suggests that the current is maximized by the ` phase competition ' . with the current maximized , increasing @xmath0 simply reduces @xmath66 ; indeed a glance at the raw data for the current shows that this is the case . the data for @xmath121 shows the same nonmonotonic structure as for the @xmath94 data . again , the peak corresponds to the crossing of the two order parameters , and then falls off with increasing @xmath0 . this similarity suggests that perhaps the high @xmath0 phase at @xmath94 is in fact the parallel strip , and such behavior would be observed in a larger system . for values of @xmath8 below @xmath138 we no longer observe the secondary peak , nor do we observe any phase competition . we must , however , leave questions about the nature of the high-@xmath0 phase open to further study . now that we have surveyed the phase diagram in some detail , we turn to look closely at some of the boundaries with the help of some scaling arguments . considering fig . [ fig1 ] again , we note two potentially troubling features . first , there is a shift in @xmath100 across the do - ts boundary of about @xmath154 between the two system sizes . we will characterize this shift using a mean - field scaling argument . second , we note that the bicritical point has shifted towards the @xmath69 axis in the larger system . it has been alluded to before that the number of rows of the minority species , rather than the fraction @xmath8 , might be the controlling variable . we will investigate this suggestion more carefully by considering some larger systems and rectangular geometries . finally we study the @xmath69 phase transition , using scaling arguments developed for the kls model . the shift in @xmath100 with system size is most pronounced at @xmath72 . previous work on the two - species model with @xmath155 treated the ordered phase in a mean - field approximation by solving equations of motion for the two different charge densities @xcite . * * * * it was found that the scaling functions depend on the combination @xmath156 , indicating that the effective bias @xmath38 introduces a new length scale . this scaling implies an infinite - volume limit in which @xmath38 @xmath157 as @xmath158 , while keeping @xmath156 fixed . earlier analyses of the _ ordered phase _ based on these ideas have worked quite well , so that we now attempt to extend this approach to analyze quantities near criticality and [ fig31 ] [ fig32 ] for @xmath159 . there is no reason to expect success a priori , as both critical fluctuations and nonzero @xmath12 may modify the scaling variables and the mean - field exponents . in figs . [ fig31 ] and [ fig32 ] we have plotted @xmath46 and @xmath124 for @xmath160 . ( we have divided @xmath124 , eqn . ( [ delta ] ) , by the volume in order to compare different system sizes more easily . ) rather than crossing the phase boundary by varying @xmath2 we have opted instead to vary @xmath0 since this allows us to vary the effective bias @xmath38 at _ constant _ interaction strength , @xmath115 . while the collapse of @xmath46 in fig . [ fig31 ] is not perfect , the mean - field scaling argument accounts for most of the shift in @xmath100 . there is a shift in the peak of @xmath161 of about @xmath162 between the largest and smallest systems ; if the same data were plotted without rescaling , the shift in @xmath163 is about @xmath164 . also of note is the extremely weak dependence on the transverse dimension , as predicted by the scaling argument . we also examined the same quantities for @xmath155 . interestingly , the data collapse for @xmath160 is better than in the @xmath155 case . this is somewhat surprising , as the original scaling argument was derived for @xmath155 , and we would expect the interactions to perhaps modify it . whatever the resolution of this puzzle , it is apparent that the scaling argument presented here explains the pronounced shift in @xmath100 seen in the phase diagram : increasing @xmath15 requires a corresponding decrease in the effective bias @xmath38 . another issue concerns the location of the junction of the three phase boundaries , shown in fig . it is clear that the junction moves toward the @xmath69 axis as the system size is increased . in fact , in both systems the boundaries merge just below a _ single row _ of the minority species , which naturally corresponds to a smaller @xmath8 in the larger system . in fig . [ fig33 ] we have replotted the data from fig . [ fig1 ] , [ fig33 ] replacing @xmath8 with @xmath165 , which is simply the _ number of rows _ of the minority species . near the junction of the three lines we have also included results from a few other system sizes with rectangular geometries . plotted vs. @xmath165 , the junctions of the boundaries coincide , within the error bars , for all system sizes , suggesting that the onset of the two - species order occurs , at least in relatively small finite systems , when there are sufficient minority particles to form a single row . the crucial question concerns the extrapolation of this result to an appropriate thermodynamic limit . if the system size goes to infinity in the most naive way , i.e. , @xmath166 at fixed aspect ratio @xmath167 , the particle density associated with a ` single row ' vanishes . it is possible that the do - ps transition exists in an infinite volume only at @xmath69 , and any _ finite density _ of ` disorder ' ( i.e. , the minority species ) induces the two - species order . preliminary studies @xcite indicate that the minority species does indeed constitute a relevant perturbation to the kls fixed point . we will have to leave discussion of this issue to future work and for now limit ourselves to statements about finite systems . at @xmath69 there is only one species , and we observe the kls transition at _ finite _ @xmath0 . though a great deal of study has been devoted to this transition at infinite @xmath0 , there has been no detailed work at finite @xmath0 . here we present a basic finite - size scaling ( fss ) analysis of this transition , in order to locate @xmath168 , and also to demonstrate the subtleties which can arise when studying phase transitions with anisotropic , nonequilibrium dynamics . field - theoretic studies of the kls model @xcite indicate that the critical behavior is strongly anisotropic , meaning that correlation lengths diverge with different exponents in the field direction and perpendicular to the field . specifically , the fluctuations perpendicular to the field are gaussian @xmath169 while those parallel to the field are not @xmath170 . correlations therefore grow faster in the parallel direction as @xmath171 , suggesting an analysis of rectangular samples such that the anisotropic aspect ratio @xmath172 is held fixed @xcite . while there is some discussion regarding the correct mesoscopic model @xcite , detailed numerical simulations show that the exponents cited above are the correct ones @xcite . in the following we will use only the phenomenological result of leung for the scaling of the order parameter at fixed @xmath173 : @xmath174where @xmath52 refers to @xmath44 . a detailed discussion of the subtleties of the fss analysis for the kls model and precision numerical results can be found in @xcite . [ fig34 ] presents our data for the scaled order parameter at @xmath175 ; the same data for saturation @xmath0 can be found elsewhere @xcite . this data is not intended as a test [ fig34 ] of the mesoscopic model , it merely is meant to indicate that the exponents at infinite @xmath0 are consistent with those at finite @xmath0 , and to determine @xmath176 . the data collapse is comparable to that seen at saturation @xmath0 , with the high temperature ( upper ) branch collapsing quite well and the low temperature ( lower ) branch showing small , but systematic deviations from scaling . these deviations remain unexplained . they are possibly due to corrections to scaling or perhaps the asymptotic region is only observed very close to @xmath100 . we have seen how subtle are the issues surrounding the transition at @xmath69 . perhaps a fruitful way to proceed when @xmath177 is to adopt the technique introduced by caracciolo _ et al_. @xcite , directly measuring finite volume correlation lengths for various geometries and volumes . in this way we may develop some understanding of how to approach the infinite volume limit in a simple way , minimizing corrections to scaling which would complicate an uninformed analysis . we have compiled a detailed phase diagram for a system of two species of particles , interacting via attractive ising interactions , and driven into opposite directions by an external `` electic '' field , @xmath0 . the purpose of our analysis was to unify previous studies which were restricted either to just one species or to having only excluded volume interactions . in the former case , particles order into a single strip aligned with the field direction while in the second model , the two oppositely driven species form a jam in the shape of a transverse strip . we monitor structure factors , their fluctuations , and , if necessary , their time traces to identify the location and character of the transitions . most of our data are taken at fixed @xmath0 , varying the fraction @xmath8 of the ` minority ' species and temperature @xmath2 . at high @xmath8 , we observe a continuous transition from a disordered phase into the transverse strip , as the temperature is lowered . noting a significant system - size dependence of the critical line , we invoke a mean - field scaling argument @xcite which suggests that @xmath156 is a good scaling variable . this is confirmed quite satisfactorily by our data . at smaller @xmath8 ( @xmath178 for a @xmath71 system ) we observe two transition as @xmath2 is lowered : first into the transverse strip ( continuous ) , and then into the parallel strip ( first order ) . and finally at the smallest @xmath8 a single , continuous transition is observed into the parallel strip . the junction of the three phases disorder , transverse and parallel strip appears to be a multicritical point . analyzing data for a range of system sizes suggests that its location scales with @xmath179 , rather than @xmath8 ; i.e. , the relevant quantity is the number of _ rows _ which can be formed by the minority particles . depending on how the thermodynamic limit is approached , the multicritical point can shift to @xmath69 . several projects suggest themselves to extend this work . first , an analysis of larger systems at fixed @xmath2 could clear up some questions , especially regarding the fate of the ` phase competition ' which is observed close to the multicritical point . many of the ` transitions ' in the @xmath71 system require an analysis of larger lattices in order to confirm their existence . second , a look at structure factors in the disordered phase is likely to reveal the presence of long - range correlations . since these are known to be quite distinct in the kls @xcite and shz @xcite models , it would be interesting to investigate how the crossover occurs . such a study would also yield considerable insight into the type of noise terms which would have to be added to the mean - field equations in order to capture fluctuations and critical properties accurately . these equations would then provide a reliable starting point for an analysis of the kls transition in the presence of a few minority charges , in order to understand the true nature of the kls critical point : does it mark the beginning of a critical line , or is it a multicritical point ? * acknowledgements . * we thank r.k.p . zia , i. georgiev , u.c . tuber , a. gambassi , m. gubinelli , g. korniss , and h.k . janssen for fruitful discussions . this work was supported in part by nsf grants dmr-0088451 , dmr-0414122 , and sbe-0244916 , as well as the jeffress memorial trust , grant no . j-594 . zhang , j - s . wang , j.l . lebowitz and j.l . valls , j. stat . phys . * 52 * , 1461 ( 1988 ) ; p.l . garrido , j.l . lebowitz , c. maes and h. spohn , phys . a * 42 * , 1954 ( 1990 ) ; g. grinstein , j. appl . phys . * 69 * , 5441 ( 1991 ) ; b. schmittmann and r.k.p . zia , j. stat . * 91 * , 525 ( 1998 ) .
in this pages there is an attempt to design and describe a naturalistic experiment on syntax acquisition . specifically , we want to build a _ syntactic network _ in order to study syntax with modern methods of complex network theory . the process is nor standard neither straightforward and deserves to be well described . there are interesting descriptive frameworks based on networks to study syntax . one of them is the so - called _ dependency grammar_@xcite . there are , also , theoretical approaches using graphs . a remarkable member is the _ word grammar_@xcite . the approach assumed here is closer to the word - grammar , despite we develop our own criteria , as well as we consider the graph representation as a linear projection of the constituency hierarchy . the paper is organized as follows : we firstly discuss the scope and validity of the conceptualization of syntactic relations within a network . the core of the work is devoted to the discussion of the ( descriptive ) structural criteria to tackle the problem of annotation in early grammars . finally , a brief compendium of network measures is shown , as well as an illustrating example . all analysis are performed over the peter corpora of childes database @xcite using the dga - annotator @xcite . every abstraction of a natural object implies a particular conception of it in order to answer a specific question . assuming that every abstraction implies a simplification , we have to explore , then , how different approaches can be complementary or whether some of these approaches are more fruitful than others -i.e . , what are the core questions leading to the understanding of such phenomena . focusing on language , research on syntax seeks to find the minimal set of rules that could generate all -and only- the potentially infinite set of sentences of a given language . thus , the question addressed by syntax is the problem of decidability or computability of the set of possible sentences of a given language . when dealing with language as a complex network , we have to note that statistical physics works from different perspectives : what are the global features of the dynamics of our system ? how the combinatorial space is filled ? what is -if any- the role of constraints ? thus , we do nt address questions concerning the structure of the inhabitants -sentences- of our system , but its global dynamics and organization . note that the questions are different than in the case of syntax : thus , the abstraction we are working in is also different . note , also , that we are not negating nor denying the particular features of sentence construction . simply , we work at other level of abstraction . we are confident that information from this different level of approach should be enlightening to questions addressed on grammar itself . if one wants to apply statistics on some syntactic phenomena , a word of caution is needed because there is a gap between the syntactic procedure and the statistical physics procedure : the former is focused on explaining almost _ every _ subtlety of sentence construction , while the latter works on averages over the largest possible set of data . thus , a compromise has to be assumed because it is not possible to deal with every syntactic phenomena but , also , the statistics has to be built on certain criteria . thus , the aim of this document is to present a set of descriptive criteria to identify _ structure _ in early child grammars . this is not a theoretical reflection about the concept _ structure _ or its evolution during the process of language acquisition . with these criteria , we want to build up the so - called syntactic networks from early child grammars . indeed , even though many features of the language acquisition process have been identified and well studied , there is a lack of a clear concept of what it is structured or not in early child grammars , namely , there is not a concept such as grammaticality @xcite or _ convergence _ @xcite , defined in adult grammars . if we take the adult - grammar concept of grammaticality , we will surely reject almost all of children s productions . but it will not be true that many of these rejected utterances are unstructured at all . in order to overcome these limitations , we developed a set of descriptive criteria to extract the syntactic network of different sets of the child s utterances belonging to successive time stages of the language acquisition process . as we discussed above , we present these criteria employed in the construction of the associated networks . formally speaking , a given language , @xmath0 is composed of an arbitrary large , but finite set of lexical items @xmath1 - or alphabet , in technical words- and of a restricted set of rules and axioms , @xmath2 . these rules describe how the elements from @xmath3 can be combined in order to _ 1_. obtain sentences of @xmath4 or to _ 2_. decide if a given sequence of elements from @xmath3 is a sentence of @xmath0 or not @xcite . syntax properties are , thus , indicators of grammatical complexity . if one intends to develop a syntax theory to decide whether a given sequence of words -namely , russian words- is a sentence of _ russian _ language , one needs to develop rules involving hierarchical and long range relations . moreover , the set of rules must be _ generative _ , in the sense that they should involve some recursive condition to grasp the potential infinity of sentences generated by russian grammar @xcite . thus , we can say , without any loose of generalization , that syntax works at the _ local level _ of language ] , i.e.it operates at the sentence level , no matter how long the sentence is . now , we wonder about the global profiles of syntactic relations . note that the question we want to address is not finding the specific rules needed to generate the possible sentences of @xmath0 , but we want to take a look at the system as a whole . this could seem bizarre when considered from the point of view of mainstream theories of syntax , but it is a common procedure in statistical physics . global profiles can provide information about general dynamics and constraints acting over the whole system as a complex entity . the unexpected profile given by zipf s law is an example of global behavior of language dynamics @xcite . a note must be added concerning the naturalistic character of this kind of experiments . syntax has been related with competence abilities . but statistical and naturalistic works are carried out over performance data . thus , we are inferring the global patterns of performance by assuming some competence abilities . networks revealed as an interesting abstraction to explore the global behavior and dynamics of complex real systems made from units and the associated relations between such units . let s explore such abstraction for syntax relations . a network @xmath5 is defined by the nodes @xmath6 and the links @xmath7 relating the nodes @xmath6 @xcite . these links can be directed or undirected ; we will use the directed ones , if the contrary is not indicated . to build a syntactic network , the mapping of @xmath8 onto a graph will be straightforward for the set @xmath9 i.e. , the set of lexical items of @xmath8 will be the set of nodes of @xmath10 . the mapping from @xmath2 to @xmath7 is not so obvious and needs further considerations . as we discussed above , the syntactic rules needed to generate any natural language revealed considerable degree of complexity . thus , it is clear that the statistical treatment employed here is an approximation . modern syntax is based on recursive operations of _ merge _ and _ move _ @xcite . such operations lead the syntactic derivations to display hierarchies and long range relations . these are features that can not be captured explicitly by a descriptive framework based on linear relations among lexical items -a network approach . but we are approaching the language structure from the point of view of statistical physics : we want to capture the global patterns of the system , thus we can not specify _ all _ the local properties . this is contrary to the procedure employed in the ising models of ferromagnetism , despite the success of this approach is universally acknowledged . thus we have to decide what is the most essential structure in a syntactic derivation . we assume that the most fundamental thing one can say from the syntactic point of view about a sentence is its _ constituent structure_. constituent structure can be captured by linear relations . in the following , we define an exact mapping from a hierarchical binary tree to a graph , an entity made of binary relations ( see figure ( [ grafets ] ) ) : 1 . find the basic syntactic structure of constituents without labels nor internal operations , with clear distinctions of complements and the head in every phrase . detect the verbs in finite forms . 2 . trace an arc from the complement to the head of the phrase . if the complement of a given phrase is also a phrase , trace and arc from the head of the internal phrase to the head of the external phrase . we want to recover the merging order . 3 . the head will be the semantically most relevant item . the verbs in finite forms are the head of the sentence . with the above criteria , we make an attempt to manage data with the less aggressive criteria . moreover , these assumptions do nt constrain us to one or other linguistic school and grasp reasonably with the observed syntactic development of children . with this method , we do not restrict our set of sentences to the one generated by finite combinatorics . we allow our sentence to be arbitrary long . thus , our model is only finite because real data is finite in nature , but it does nt negate the theoretical possibility of infinite generativity , a property expected for any approach of syntax @xcite . moreover , the relevance of the statistical physics properties generally is found in systems asymptotically large . chi : [ telephone go right here ] ( ... ) + chi : xxx [ need it ] [ my need it ] ( ... ) + chi : xxx ( ... ) + chi : [ put in there ] + + + + + + + + right@xmath11there@xmath11go@xmath12telephone+ it@xmath11need@xmath12my+in@xmath11there@xmath11put + + syntactic networks can be built by other procedures . dependency syntax @xcite has been used in other works @xcite . in such an approximation , syntactic networks have been built up by assuming syntactic relations as dependency relations among lexical items . dependency grammar generates a graph to describe the sentence structure and it is the reason why it is interesting to build networks . the mechanism to build large nets is straightforward . studies on language acquisition can be divided into two main types : _ experimental _ and _ naturalistic_. the experimental ones are focused on child s response to well - established situations in order to obtain data of some specific trait . naturalistic studies , at the other hand , are based on child s free speech corpora . these corpora can be extracted , for example , from a recorded session where the child speaks with adults spontaneously . @xcite . our study is clearly naturalistic , and this label takes here its whole meaning , because the procedures to build up _ biological _ networks , for example , are conceptually , the same . data has been extracted from the well - known _ childes database _ @xcite . the chosen corpus is the _ peter _ corpus , from bloom 1970 . we choose this data for many reasons : 1 ) time intervals are regular ( about 2 or 3 weeks ) . 2 ) extension of the corpora can be considered large enough to seize global properties , taking into account the intrinsic small size of the system . there is a little exception in corpus 2 , which is , by far , the smallest one . fortunately , this corpus does not seem to belong to a key stage in grammar evolution . 3 ) the acquisition stages of peter seem to be the standard ones observed in language acquisition . thus , it is reasonable to think that our results will not be biased to strange deviations of the particular case study . working data includes the 11th first corpora of peter s 20 corpora . the age period goes from 1 year and 9 months to 2 year and 4 months . as we said above , the aim of the study is to observe whether and to what extent syntactic networks can provide information on the process of language acquisition . the so - called _ syntactic spurt _ ( @xcite ) , which appears later than the _ lexical spurt _ , is clearly observable in the chosen corpora . thus , we manage data that begins when the lexical spurt has already taken place and ends when syntactic structures of child s productions are complex enough to be compared with the adult ones . this does not exclude the possibility of more abrupt changes in more advanced acquisition stages , but we stop our analysis here . .age of peter in successive corpora ( * years;months.days * ) . data from childes database @xcite [ cols="^,^,^,^,^",options="header " , ] @xmath13 the case concerning _ a _ , the schwa , will receive a particular attention below . some onomatopoeia appear together with its corresponding lexical item . to analyze them , we assume onomatopoeia to be nonexistent . take , for example : @xmath13 * peter 9_i want ta write the choo choo train @xmath14 i want ta write the train _ @xmath13 considerations related to other non - trivial interpretations , such as the role of _ ta _ , are extensively developed in the following lines . in addition , and following the enumeration of non - accepted productions , we find the general case where no structure is identified in a production . in this situation , we consider the utterance as a string of isolated lexical items . consequently , no links but only nodes corresponding to the lexical items are added to the graph . more attention has to be paid to imitations . the reason to consider imitations as unacceptable productions is that we have no confidence that such string is identified as a structured one or , simply , as a single lexical element . imitations are identified by analyzing the discursive context . some utterances of surprising complexity for its corresponding stage are produced after an _ untranscribed adult conversation _ : we cautiously removed from the graph such contributions . in peter 5 corpus , _ i ca nt see it _ is produced after an adult conversation and it is , by far , the most complex production of this corpus . it strongly suggests that this is an imitation from something said in such untranscribed adult conversation . as we stated above , structured productions and lexical items are taken into account . now we state another assumption : if in the whole utterance we can not find global structure but there are some structured strings , then we take these structured strings separately , ( see figure [ grafets ] ) . in the pre - functional stage ( identified in our data until corpus peter 6 ) there appear a lot of utterances where only thematic relations seem to be considered by the syntactic system of the child . thematic relations are fundamental at the syntactic level , and their appearance indicates presence of sub - categorization mechanisms in child grammar . no traces of more complex structure -like agreement- is found in this early stage of acquisition . we consider as syntactic relations the thematic relations between verb and arguments . moreover , subject elision is usual , due mainly to the facts that 1)utterances are in imperative mode or 2)there is no fixation yet of parametric variation associated to the explicit presence of subject in english . productions of this kind are : @xmath13 peter 5 * _ open box _ , instead of _ open the box _ , ( the determiner is missing . ) @xmath13 * peter 5 * _ wheel walk _ instead of _ the wheel walks _ , ( 3-singular english agreement is missing ) @xmath13 * peter 6 * _ two truck _ instead of _ two trucks _ , ( no plural agreement ) @xmath13 this leads to the logical conclusion that productions like _ open the _ will not be accepted . the reason is clear : if we assume thematic relations as the basic building blocks of child syntax , the non - presence of the semantically required argument but its determiner is not enough to define any relation . relations between verbal head and functional words are specifically considered in phrasal verbs . its isolated production is considered a structured utterance . several reasons support our choice : 1 ) their intrinsic complex nature , 2 ) we can not conclude that there are lexicalized imitations because , in adult speech , phrasal verbs usually are _ broken _ by a noun or determiner phrase : @xmath13 _ turn [ the wheel]@xmath15 out_. @xmath13 semantically vacuous predications ( those which involve the _ to be _ verb ) are often produced without realization the verb . we argued that missing arguments or lack of agreement in a production could be not the only reasons to conclude that there is not any structure in child utterances . this was justified because strong semantically items were present in discussed productions . the case of copulative constructions will be treated close to the ones involving missing functional words . in this case , no presence of the verb does not motivate the consideration of non - structured production . an interesting production is : @xmath13 * peter 5 * _ wheels mine _ instead of _ * the * wheels * are * mine _ @xmath13 in this case we have a predication _ mine _ from something _ wheels_. formally , _ are _ is a semantic link between predication and the element from which something is predicated @xcite . so the missing of the _ to be _ verb could be considered analogous to the missing of a functional particle . the same situation arises from : @xmath13 * peter 7 * _ that my pen _ @xmath13 usually , when inflectional morphology appears , some infinite forms are present without the finite form of the _ to be _ verb . this is the case of some present continuous utterances such as : @xmath13 * peter 8 * _ i writting _ @xmath13 this case should be treated as the above case : there is some predication with semantic structure . just the opposite is also found : presence of the _ to be _ verb with an infinitive or finite form : @xmath13 * peter 8 * _ i m write too _ @xmath13 in this case , we could assume that the child is acquiring inflectional morphology and that this utterance is a present continuous one without inflection . in the other hand , we could consider that _ m _ has not a role in the sentence and thus , this can be treated as a single finite sentence _ i write too_. analogously , @xmath13 * peter 7 * _ i m do it _ or * peter 6 * _ cars goes away _ @xmath13 are treated as single finite sentences : _ i do it _ and _ the cars go away _ some lexicalized phrases in adult language , such as _ back seat _ or _ thank you _ , have been considered as a complex structures . the reason is to be coherent : if we assume that _ fix it _ is clearly an imperative structured sentence , at this stages of acquisition there is no reason to think that _ back seat _ or _ thank you _ have to be considered differently . moreover , this interpretation is also coherent with the one developed for phrasal verbs . a special case of imitations involving the _ to be _ verb will be accepted . these imitations involve some _ adaptation _ of adult syntax to the syntax in which the child is competent . an example should be : @xmath13 * ( peter 6 ) * * adult * : _ is that a truck ? _ @xmath13 * child@xmath16 : _ that s a truck ? _ @xmath13 child@xmath17 : _ that a truck ? _ @xmath13 in this example , adult production involve an interrogative sentence with subject inversion . the first imitation ( _ child_@xmath16 ) retains all the lexical items but the sentence is translated as interrogative without subject inversion . in the second successive imitation ( _ child_@xmath17 ) the verb is missing . but the elements to define a predication are still at work -with a _ schwa _ as a determiner , suggesting that the child is entering into the functional stage . during the acquisition process , extension of meaning is subject to variations . to know which is the intrinsic nature of these changes is not our aim , but we have to manage such situations . thus , we find utterances where the child uses in the _ wrong _ way some lexical item that could be related semantically with the _ right _ lexical item . as an example : @xmath13 peter 5 * _ more screwdriver _ @xmath13 which could , checking the context , be properly replaced by constituents or lexical items with related meanings : @xmath13 _ another screwdriver _ @xmath13 or _ screw it again _ @xmath13 or _ screw it more _ ( or _ harder _ ... ) @xmath13 in the first case , we could consider that the child made some semantic extension of the word _ more _ and it has enough traces to define a syntactic relation . but context can lead us to a second or third interpretation . generally , if there is a great ambiguity we reject such utterances as structured ones . in this case , we should not consider any syntactic structure . thus , we do nt define any relation in productions such as : @xmath13 * peter 5 * _ screwdriver help _ @xmath13 * peter 5 * _ more [ fix it]@xmath18 _ ( we do nt define any relation between @xmath19_more _ and the sd _ fix it _ ) @xmath13 the semantics of the productions are intuitive , but is hard to justify clearly some kind of syntactic dependency . strings displaying mistakes in the use of personal pronouns and possessives have considered as structured . generally , we could associate such mistakes to the absence or weakness of case system . but many productions , as we reported above , have structure without any trace of case assignation . examples of this kind of utterances are : @xmath13 * peter 5 * _ my fix it_@xmath20instead of _ i fix it _ @xmath13 * peter 8 * _ me write_@xmath21 instead of _ i write _ @xmath13 this assumption is reinforced by realizing that , in some cases , a production with wrong pronoun is repeated correctly without any conversational pause : @xmath13 * peter 8 * _ me found it ( ... ) i find it _ @xmath13 this situation can not be confused with the missing of the _ to be _ verb such in the case of _ wheels mine_. this case has to be considered as above mentioned when dealing with missing _ to be _ verb structures . in early corpora ( 1 - 4 ) child productions display very poor structures . this is the so - called _ pre - functional _ stage , where no functional words appear in structured productions . beyond this point , some lexical elements -we are mainly talking about the @xmath22 , the _ schwa_- seem to act as a _ proto_functional particles . whether this schwa has a phonological or functional - syntactic character is an open question @xcite . some authors related the presence of these items as one step to combinatorial speech @xcite , but they realized that , in early stages , the role of these items is more related to phonological processes of language acquisition , without any functional or structural role , at the syntactic level . other authors such as veneziano @xmath23 sinclair `` _ linked these phenomena more specifically to the child s development of grammatical morphemes considering them as a sort of an intermediate form on the way to grammatical morphemes . _ '' @xcitepp 463 . roughly speaking , we can say that the core of this reasoning is rooted in the idea that the role of such items is dynamic , going at very first stages as _ filler syllabes _ without any syntactic role and acquiring grammatical features during the process to end as functional particles , with specific syntactic role . the lack of consensus around a topic that seems to be crucial in syntactic acquisition theorizations forces us to be really cautious in interpreting such items . furthermore , functional words such @xmath22 are strongly candidates to be the hubs in a fully developed syntactic network . hub are the most connected nodes on a network , being , thus core pieces in network organization . every candidate belonging to the set of functional particles is specially analyzed in order to discard simple phonological phenomena . thus , for every occurrence of such items there will be an individual decision , taking in account the context and with the framework defined by veneziano @xmath23 sinclair . sinclair s study , the chosen language is french , but we take as general some conclusions that seem to coincide with the observed phenomena in english acquisition ] specifically , we considered that sometimes the _ schwa _ plays a functional role . it is reasonable to assume , thus , that sometimes the _ schwa _ is substituting a specific functional particle . in this cases we assume that the _ schwa _ acts within the syntactic structure as the substituted particle . several examples can illustrate such reasoning : @xmath13 * peter 6 * _ light * a * hall _ @xmath13 * peter 6 * _ light in * a * hall _ @xmath13 * peter 6 * _ look * a * people _ @xmath13 in this case , it seems that _ a _ substitutes _ the_. _ * a * _ should be treated as a determiner . this is a very difficult choice , because purely phonological interpretation could be enough to justify the presence , specially in the third case . sometimes choice is really ambiguous . take for example : @xmath13 * peter 6 * _ there * a * new one _ @xmath13 such a case _ * a * _ could be easily interpreted as a pure phonological phenomena . but if we consider the vacuous semantic nature of the _ to be _ verb , we could understand these occurrences as protofunctionals . we removed these most ambiguous cases . we also rejected as unstructured utterances productions involving confuse sequences of functional particles as : @xmath13 * peter 6 * _ will an a in there _ @xmath13 any interpretation is really confusing . there are cases where the presence of the _ * a * _ is clearly purely phonological . for example : @xmath13 * peter 6 * _ more get * a * more _ @xmath13 * peter 6 * _ * a * ride a horsie _ @xmath13 * peter 5_a * this thumb _ @xmath13 * peter 7 * _ hmmm my a _ @xmath13 beyond the non - definition of personal pronouns due to the weakness of the case system , we find the pronoun _ i _ as an _ a _ : @xmath13 * peter 7 * _ a want milk _ @xmath13 * peter 7 * _ a want ta get out _ @xmath13 an interesting sequence of that reinforces our considerations is : @xmath13 * peter 7 * _ a put it on ( ... ) my put it on _ @xmath13 finally , it is interesting to note the presence of elements that are , to some extent , a mixing between _ a _ and _ to _ : _ ta_. the occurrence of this particle is rare and located explicitly at the very beginning of the funcional stage . the remarkable fact lies on the evidence that is located where it should be the preposition _ to_. this could imply that in fact there is a transition from a pure phonological role to a functional one . _ a @xmath14 ta @xmath14 to _ @xmath13 thus , we interpret _ ta _ as an intermediate stage but , due to its location within the sentence and the context , we assume it behaves as a preposition : @xmath13 * peter 6 * _ [ have [ ta [ screw it]]@xmath24 _ @xmath13 * peter 7 * _ [ have[ta [ screw it]]@xmath24 _ @xmath13 the emergence of english syntax is strongly tied to the emergence of functional particles . this is the reason why we decided to take into account this kind of lexical components : despite almost every utterance involving such items can be object of many considerations , there are enough motivation to try to define a descriptive criteria to deal with them . it is usual to find , at the beginning of the functional stage , that a verb that sub - categorizes , for example , a prepositional phrase , display two successive prepositions : @xmath13 * peter 6 * _ look * at in * there _ @xmath13 to manage this kind of productions we assumed , first , that these imply that the child conceives a syntactic structure that involves prepositional phrases . following this reasoning , we make the following structural description : @xmath13 _ look * at in * there _ @xmath14 _ [ look [ at there]@xmath24 _ @xmath13 in that case , _ in _ is interpreted as an independent lexical item . not only prepositions are involved in this duplication phenomena , but also determiners : @xmath13 * peter 9 * _ one that screwdriver _ @xmath13 interpretation rules out _ one _ as a member of any structure , leading the sd _ [ that screwdriver]@xmath15 _ alone . a situation analogous to famous one described by braine ( p.160 - 161 ) @xcite is : @xmath13 * peter 7 * _ get * another * one paper @xmath14 get another paper _ @xmath13 thus , as above , determiner duplication is not considered in the structural analysis . by this name , we designate the lexical items that are present in a conversational framework but can not be explicitly interpreted as members of some syntactic structure , such as _ hello _ , or_ok_. the reason to include these elements as connected to the network is due mainly because their are produced in non - arbitrary context . thus , we assume that they linked to the first element of the sentence they precede : @xmath13 * peter 5 * _ ok patsy _ @xmath13 obviously , previous reasons are at work when dealing with such items . thus , the conversational context has to be analyzed to interpret these items . for example , in the following situation , _ bye _ has not been considered as a member of any structure . the reason is that it is produced among analogous expressions , leading it to interpret more in a pragmatic sense : @xmath13 * peter 7 * _ see you , bye , see you _ @xmath13 sequences of numbers or other elements produced as a list are not considered as members of any structure : @xmath13 * peter 7 * _ one two three ... _ @xmath13 sometimes , personal nouns are produced by the child to demand attention from adult people . in these situations , we do not accept them as a members of structured sentences . some residual cases to be commented are the ones related with strings of nouns : @xmath13 * peter 9 * _ piece tape ... _ @xmath13 which are clearly unstructured , if conversational context does not conspire in the other way . sequences like @xmath13 * peter 9 * _ off on tv ... _ @xmath13 are ruled out as structured ones because any structure proposal leads us to a certainly bizarre sentence in terms of meaning and because there is a lack of many elements that can act as clues to find some structure . thus , they are considered as isolated lexical items : @xmath13 * peter 7 * _ * an * jenny _ @xmath13 when the functional stage is being consolidated , we find more complex structures . among others , interrogatives involving subject inversion or negation structures . negation structures sometimes imply the presence of the auxiliary _ to do _ are produced using the negative particle alone : @xmath13 * peter 7 * * no * _ put it here _ @xmath13 this context suggests us that _ no _ could be replacing the auxiliary form _ dont_. _ do nt put it here _ is its grammatical counterpart . nevertheless , we consider _ no _ as replacing _ do nt _ and , thus , as a member of a bigger syntactically structures utterance . obviously , as we said above , context has to rule out interpretations such as _ * no * , put it here_. analogous structures can be : @xmath13 * peter 7 * _ * no * ride a bike _ there are other situations where we considered suitable not to consider _ no _ as a member of any structure : @xmath13 * peter 8 * _ in the bag * no * _ @xmath13 we can not conclude that there is a syntactic relation among the negation operator and some other lexical item of the string . maybe a parametric @xcite hypotheses could save this production by suggesting that the location of the negation operator within the structure may be a parametric feature . despite interesting , we choose the rule these productions out for reliability purposes . furthermore , at the same time , there are productions like : @xmath13 * peter 9 * _ * no * in this box _ @xmath13 suggesting that the child _ knows _ the ordering of negation structures in english . in the latter case , as before , we considered not risky to identify _ no _ as a member of a bigger structures sentence . previous set of criteria enables us : 1 . to identify and characterize syntactic structures in early child language . 2 . to project them into word - word dependencies in order to annotate the corpus . the corpora is annotated _ by hand_. this enables us to be accurate and to manage ambiguous situations . the program used to perform the annotation is the so - called * dependency grammar annotator * ( dga annotator ) . this program was developed by marius popescu @xcite from the university of bucaresti and has a nice and easy interface . it works with xml files , whose internal structure will be described in the example of the last section . with these criteria in hand , we are ready to perform a first analysis of grammar complexity . such an analysis is closed to the classical mlu . what we can compute , now is the average size of syntactic structures . thus , in a production , we can , for example , find two syntactically unrelated structures@xmath26 and @xmath27 . the number of lexical items of these structures will be its sizes @xmath28 and @xmath29 . such an utterance will contribute to the computation of @xmath25 with two structures . for example , in @xmath13 _ look * at in * that _ @xmath13 we have two structures : @xmath30[look , [ at , that ] ] @xmath31 @xmath32 in @xmath33 @xmath34 ( note that single words are considered as size-@xmath35 structures ) to obtain @xmath25 the average is computed over all utterances . such a measure will provide us clues to decide whether the size of productions has information about grammatical complexity . its evolution can be related with working memory limitations . once we analyzed a conversation , we can build the network . the process is as follows . due to the nature of our analysis , we will have a collection of words , which define the set @xmath36 : @xmath37 this define the set of nodes of our network . if , during the conversation , we find some structure where two words @xmath38 are related syntactically -using the above criteria!- we say that @xmath39 and that there is a link @xmath40 . @xmath41 remark that : all the words and links only appear once a time . this enables us to separate -as far as possible- some contextual deviations from the specific conversations . also , there can be many isolated nodes . finally , we compute the adjacency matrix @xmath42 . this matrix is the representation of the graph and the abstract object where all computations of graph complexity are performed . if the child produced @xmath43 different words during the conversation , the size of this matrix will be , obviously @xmath44 . the adjacency matrix of the directed graph will be : @xmath45 if we consider the undirected version of this graph , @xmath46 will be defined as : @xmath47 note that @xmath46 is symmetrical , whereas @xmath42 it is not . now we are ready to perform an exhaustive analysis of network complexity . a first and fundamental question we find when dealing with such measures is whether the net is made of a large number or small , isolated graphs or if it displays a clearly differentiated giant connected component ( gcc ) that contains most of the connected words -i.e . words syntactically active in some production . the number of words contained on such a component or its relative size are interesting statistical indicators . strikingly , from the very beginning , child s syntactic graphs display a clear and very differentiated gcc . for mathematical purposes , we will use the matrix representation of the connectivity pattern of the gcc , the so - called _ adjacency matrix_. an element of such a matrix is @xmath48 if there exists a link among the words @xmath49 and @xmath50 and @xmath51 otherwise . if the contrary is not indicated , -we will compute the following measures over the gcc of our graphs . the number of links ( or _ degree _ ) @xmath52 of a given word @xmath53 gives a measure of the number of ( syntactic ) relations existing between a word and its neighbors . the simplest global measure that can be defined on @xmath54 is the average degree @xmath55 . for the @xmath56-th corpus , it will be defined as @xmath57 where @xmath58 indicates the number of words present in the @xmath56-th corpus . this number is known to increase through acquisition in a steady manner . this and other measures are computed on the largest component of the graph . beyond the average degree , two basic measures can be used to characterize the graph structure of the gcc of the @xmath56-th corpus . these are the average path length ( @xmath59 ) and the clustering coefficient ( @xmath60 ) . the first is defined as @xmath61 over all pairs @xmath62 , where @xmath63 indicates the length of the shortest path between two nodes . roughly speaking , a short path length means that it is easy to reach a given word @xmath64 starting from another arbitrary word @xmath65 . the second is defined as the probability that two vertices ( e.g. words ) that are neighbors of a given vertex are neighbors of each other . in order to compute the clustering , we define for each word @xmath66 a neighborhood @xmath67 . each word @xmath68 has been syntactically linked ( via the above defined projection ) at least once with @xmath66 in some sentence . the words in @xmath67 can also be linked among them , and it is what the clustering coefficient evaluates . the clustering @xmath69 of this set is defined as @xmath70 and the average clustering of the gcc concerning the @xmath56-th corpus is simply @xmath71 . the clustering @xmath72 provides a measure of the likelihood of having triangles in the graph . concerning the average path length , for random graphs with poissonian structure we have @xmath73\over \log \left [ z_2 /z_1\right ] } \ ] ] being @xmath74 the average number of neighbors at distance @xmath43 . for poissonian graphs , where @xmath75 and @xmath76 , we have the following approximation : @xmath77 it is said that a network is a _ small - world _ when @xmath78 ( and clearly @xmath79 ) . the key difference between a poissonian network and a real network is often @xmath80 @xcite . another quantity of interest is the degree of affinity among nodes with the same connectivity . in this way , the behavior of hubs is specially relevant , as well as they organize the overall structure of the net . a network is said to be _ assortative _ if hubs tend to be connected among them . at the other side , a network is said to be _ dissassortative _ if hubs tend to avoid connections among them . language networks at different scales display a high degree of dissassortativeness @xcite . to quantify the degree of assortativeness , we use the so - called pearson s coefficient for nets @xcite : @xmath81 where @xmath82 and @xmath83 are the degrees of the edges at the ends of the @xmath84th edge with @xmath85 , @xmath86 and being @xmath87 the number of edges . if @xmath88 the net is dissassortative , whereas if @xmath89 the net is assortative . below we have a fragment of the conversation transcribed in the corpus _ peter 7_. we will detail the analysis that we perform . firstly , we show the source corpus . we follow by selecting peter s productions . after that we select the structures and analyze this structures and we tag them . we finish by computing @xmath90 of this fraction of text and by showing the obtained net . pat : hey pete that s a nice new telephone looks like it must do@xmath13 @xmath91 everything it must ring and talk and .@xmath13 % mor : co\|hey n : prop\|pete pro : dem\|that v\|be @xmath23 3s det\|a adj\|nice adj\|new n\|telephone@xmath13 @xmath91 n\|look - pl v\|like pro\|it v : aux\|must v\|do pro : indef\|everything pro\|it v : aux\|must @xmath13 @xmath91 v\|ring conj : coo\|and n\|talk conj : coo\|and . @xmath13 % exp : peter has a new toy telephone on table next to him@xmath13 % com : < bef > untranscribed adult conversation@xmath13 * chi : xxx telephone go right there .@xmath13 % mor : unk\|xxx n\|telephone v\|go adv\|right adv : loc\|there . @xmath13 % act : < bef > reaches out to lift phone receiver , pointing to place where@xmath13 @xmath91 wire should connect receiver and telephone@xmath13 * mot : the wire .@xmath13 % mor : det\|the n\|wire . @xmath13 * pat : oh < the @xmath23 te > [ // ] the wire s gone ? @xmath13 % mor : co\|oh det\|the n\|wire v : aux\|be @xmath23 3s v\|go @xmath23 perf ? @xmath13 % com : < aft > untranscribed adult conversation@xmath13 * chi : xxx need it my need it xxx .@xmath13 % mor : unk\|xxx v\|need pro\|it pro : poss : det\|my n\|need pro\|it unk\|xxx .@xmath13 % act : < aft > goes to his room on mother s suggestion , returns with wire@xmath13 * chi : xxx .@xmath13 % mor : unk\|xxx . @xmath13 * pat : uhhuh .@xmath13 % mor : co\|uhhuh . @xmath13 * loi : why do nt you bring your telephone down here peter ? @xmath13 % mor : adv : wh\|why v : aux\|do neg\|not pro\|you v\|bring pro : poss : det\|your n\|telephone @xmath13 @xmath91 adv\|down adv : loc\|here n : prop\|peter ? @xmath13 * loi : why do nt you put it on the floor ? @xmath13 % mor : adv : wh\|why v : aux\|do neg\|not pro\|you v\|put @xmath23 zero pro\|it prep\|on det\|the n\|floor ? @xmath13 % act : < aft > peter puts it on floor < aft > peter is trying to attack `` wire''@xmath13 @xmath91 to phone and receiver@xmath13 % com : < aft > untranscribed adult conversation@xmath13 * loi : what re you doing ? @xmath13 % mor : pro : wh\|what v\|be @xmath23 pres pro\|you part\|do - prog ? @xmath13 * chi : 0 .@xmath13 % act : < aft > peter goes to hall closet , tries to open it@xmath13 * mot : what do you need ? % mor : pro : wh\|what v\|do pro\|you v\|need ? @xmath13 * chi : xxx .@xmath13 % mor : unk\|xxx .@xmath13 ( ... ) @xmath13 * chi : put in there .@xmath13 % mor : v\|put @xmath23 zero prep\|in adv : loc\|there . @xmath13 % act : attaching wire to phone@xmath13 * loi : ok it s all fixed oops it was out all fixed there .@xmath13 % mor : co\|ok pro\|it v\|be @xmath233s qn\|all part\|fix - perf co\|oops pro\|it v\|be @xmath23 past @xmath23 13s @xmath13 @xmath91 adv\|out qn\|all v\|fix - past adv : loc\|there . @xmath13 to work with the dga annotator , we need , firstly , to extract the child s productions . to do this , we programmed a routine in perl language able to extract child s productions . below there is a simple pseudocode as a sample : @xmath13 ` file = peter`@xmath92 ` for(i=5 ; i<=longfile ; i++ ) ` @xmath93 ` if(file[i]= /peter/ ) ` @xmath93 ` j = j+1 ; ` ` peter[j]= file[i ] ; ` @xmath94 @xmath94 @xmath13 ` for(i=0 ; i<=longfile ; i++ ) ` @xmath93 @xmath95 ` @peter[i]= tr/peter:/ / ; ` @xmath95 ` @peter[i]= tr/./ / ; ` @xmath95 ` @peter[i]= tr/,/ / ; ` @xmath95 ` @peter[i]= tr/;/ / ; ` @xmath95 ` @peter[i]= tr/:/ / ; ` @xmath95 ` @peter[i]= tr/!/ / ; ` @xmath95 ` @peter[i]= tr/</ / ; ` @xmath95 ` @peter[i]= tr/>/ / ; ` @xmath95 ` @peter[i]= tr/?/ / ; ` @xmath95 ` @peter[i]= tr// / ; ` @xmath95 ` @peter[i]= tr// / ; ` @xmath94 @xmath13 if we apply the above algorithm to the sample of text of the example , we obtain : @xmath13 xxx telephone go right there xxx need it my need it xxx xxx 0 xxx put in there further we need to provide the obtained strings of words with a xml format , in order to manage them with the dga annotator . below we have an example of the string _ put in there_. @xmath13 @xmath96`?xml version=1.0 encoding = iso-8859 - 1`@xmath97 @xmath96`!doctype dgadoc system dga.dtd`@xmath97 @xmath96`dgadoc`@xmath97 @xmath96`s`@xmath97 @xmath96`tok`@xmath97 @xmath98`orth`@xmath97`put`@xmath96`/orth`@xmath97 @xmath98`ordno`@xmath97`1`@xmath96`/ordno`@xmath97 @xmath96`/tok`@xmath97 @xmath96`tok`@xmath97 @xmath98`orth`@xmath97`in`@xmath96`/orth`@xmath97 @xmath98`ordno`@xmath97`2`@xmath96`/ordno`@xmath97 @xmath96`/tok`@xmath97 @xmath96`tok`@xmath97 @xmath98`orth`@xmath97`there`@xmath96`/orth`@xmath97 @xmath98`ordno`@xmath97`3`@xmath96`/ordno`@xmath97 @xmath96`/tok`@xmath97 @xmath96`/s`@xmath97 @xmath96`/dgadoc`@xmath97 @xmath13 we reject @xmath100 and @xmath101 as lexical items and proceed to annotate with the dga anotator . once the corpus is annotated ( with the criteria developed through the paper ! ) we generate the set of words . this can be done by sampling the xml file once annotated by using a routine close to the ones shown above ( perl or python are the ideal languages ) . to compute graph parameters and more mathematical artifacts , it is a good choice to use a _ stronger _ language , such as c or c++ . @xmath102 and the analysis is , roughly speaking : @xmath103_{pp}]_{vp}]_{tp}}\;\;\|s_1\|=4\ ] ] @xmath104_{vp}}\;\;\|s_2\|=2\ ] ] @xmath105_{vp}]_{tp}}\;\;s_3=3\ ] ] @xmath106_{pp}]_{vp}}\;\;s_4=3\ ] ] thus , we can compute @xmath107 : @xmath108 and , following the criteria developed above , we can define @xmath109 @xmath110 we have built the graph . my indebtedness for carlos rodrguez - caso for his ideas and his patient advising in teaching perl . i also want to acknowledge harold fellermann for his help in programming python routines , andreea munteanu for the careful reading of the manuscript and barbara soriano for its useful comments on language acquisition process . my indebtness , also , to ricard v. sol and sergi valverde as well as this work is a part of a bigger work where their ideas have been illuminating . finally , i d like to acknowledge ramon ferrer i cancho for his contributions at the beginning of the work . ferrer i cancho , r. ( 2005 ) the structure of syntactic dependency networks : insights from recent advances in network theory . in _ the problems of quantitative linguistics _ , altmann , g. , levickij , v. @xmath23 perebyinis , v. ( eds . ) . chernivtsi : ruta . pp . 60 - 75 .
the continuous - time random walks ( ctrws ) @xcite present sufficiently wide and general class of random walks , describing random motions of particles , when both the waiting ( sojourn ) time between successive jumps and the jump length ( including its direction ) are generally coupled random variables . the ctrws have been widely applied in building up models of anomalous diffusion and transport in physics @xcite ( to name a few ) and biology @xcite , and in economical problems @xcite . in their original form @xcite the ctrws describe random walks , or , more precisely , jumps , on lattices when only the waiting time is a random variable . in the general case of the uncoupled ctrws both waiting time and jump length are independent random variables , this leads to the jump model of the ctrws . the kinetic model of random walks , often called random flights , originates from the pearson rayleigh random walks @xcite . in this model the particle moves with constant velocity for a random time along straight lines between points where it changes randomly the direction of movement . the random flights can be considered both in the framework of the generalized linear transport ( kinetic ) equation @xcite and in the framework of the ctrws , when the waiting time and jump length are coupled random variables . the latter is the velocity or velocity - jump model of the ctrws @xcite . the jump model of the ctrws has gained high popularity due to its greater simplicity than the kinetic ( velocity - jump ) model . however , in the jump model the waiting time can generally be arbitrarily small or / and the jump length can be arbitrary long ( levy flights ) , and , therefore , a walking particle can move with infinite velocity . thus , the jump model may , in contrast to the kinetic one , violate the principle of causality . this is similar to the classic _ diffusion paradox _ @xcite related to the conventional diffusion equation , which is widely used for approximate macroscopic description of nonanomalous diffusion and brownian motion . the difference is that in classic einstein s model of brownian motion @xcite , leading to the diffusion equation , the particle moves with finite velocity . in fact einstein s model is nothing but the symmetric bernoulli random walk @xcite , which can be considered as a degenerate case of the ctrws in the frameworks of both the jump and kinetic models . the telegraph equation @xcite was proposed as an alternative to the diffusion equation . in contrast to the latter , which is parabolic , the telegraph equation is hyperbolic , providing the finite speed of signal propagation . two- and three - dimensional telegraph equations meet some formal problems since green s functions can become negative @xcite . though in the case of thermal conduction this can be fixed by imposing restrictions on the heat flux @xcite . yet the one - dimensional telegraph equation avoids the diffusion paradox and provides better model for nonanomalous diffusion than the one - dimensional diffusion equation . in ref.@xcite the authors discussed microscopic models of the telegraph equation . aside from the kinetic model they proposed another model based on uncoupled ctrws with the waiting time distributed according to the gamma law @xmath1 and the jump length having the finite second moment . the authors called the ctrws with the gamma - distributed waiting time delayed ctrws ( dctrws ) . similar arguments were used in ref.@xcite when deriving the telegraph equation with reaction . in ref.@xcite the telegraph equation with reaction was derived formally from the ctrw with exponentially distributed waiting time , and later , in ref.@xcite , from the ctrw with a general waiting time . earlier , the telegraph equation was derived in ref.@xcite from uncoupled ctrws with the waiting time having more general distribution than the gamma law . we call these ctrws for brevity also delayed ctrws ( dctrws ) . the conclusions that the dctrws simulates the telegraph equation at the microscopic level are strange . indeed , the telegraph equation describes phenomena with finite propagation speed , while the particle in the dctrws may move with infinite velocity , i.e. , the dctrws do not reflect the principal peculiarity of the telegraph equation . in spite of this discrepancy , one may believe that the telegraph equation still gives a precise approximation to the dctrws . in this paper we investigate how accurate are the approximations to the dctrws provided by the telegraph equation . we show that the diffusion equation , being the correct limit of the dctrws , gives better approximations to the dctrws than the telegraph equation . one can conclude , therefore , that the dctrws can not be a model for the telegraph equation at microscopic level . the rest of the paper is organized as follows . in section [ sec : ctrw ] we recall some facts concerning the uncoupled ctrws . section [ sec : teleapprox ] shows how the telegraph equation was derived from the dctrws in ref.@xcite . the asymptotic behaviour of the dctrws is discussed in section [ sec : asymptbehav ] . in section [ sec : numercompar ] we compare the telegraph and diffusion approximations to two ( discrete- and continuous - space ) dctrws , completely described in section [ sec : twoparticul ] , numerically , while asymptotic comparison of the approximations to the continuous - space dctrw at long times is performed in section [ sec : asymptcompar ] . section [ sec : concl ] contains some conclusive remarks . in this section we briefly formulate some facts concerning the ctrws for further reference . we adopt the one - dimensional model and consider uncoupled ctrws , when the waiting time and jump length are independent random variables , i.e. , the joint probability density @xmath2 for a particle to jump a distance @xmath3 after waiting a time @xmath4 is given by @xmath5 where @xmath6 and @xmath7 are probability densities for the waiting time and jump length , respectively , with obvious normalizations @xmath8 and @xmath9 . we consider here ctrws in a continuous space , however , they can be similarly described on a lattice , as it was originally proposed in @xcite , with obvious modifications . the ctrw is described by the density @xmath10 , which is a probability density that the particle is at the point @xmath3 at the time @xmath4 , so that @xmath11 . the density @xmath10 is given by the integral equation @xcite @xmath12 \shoveright{+ \int_0^t \int_{-\infty}^\infty \varphi(x - x',t - t ' ) \rho_{\mathrm{rw}}^{}(x',t ' ) { \mathop{}\!\mathrm{d}}{x ' } { \mathop{}\!\mathrm{d}}{t'}}\\[1ex ] \shoveleft{= \varpsi(t ) \delta(x)}\\[1ex ] + \int_0^t \psi(t - t ' ) \left [ \int_{-\infty}^\infty \lambda(x - x ' ) \rho_{\mathrm{rw}}^{}(x',t ' ) { \mathop{}\!\mathrm{d}}{x ' } \right ] { \mathop{}\!\mathrm{d}}{t'},\end{gathered}\ ] ] where @xmath13 is the dirac delta function , and @xmath14 is the survival probability , i.e. , the probability that a particle stays at the same position for the time @xmath4 . here one assumes that at the initial moment @xmath15 the particle was at the point @xmath16 , i.e. , @xmath17 . the fourier laplace transform of the equation , denoted by @xmath18 , implies the montroll weiss formula @xcite @xmath19}\\[1ex ] = \frac{1-{\mathcal{l}}\psi(s)}{s \,[1 - { \mathcal{f}}{\lambda}(k ) { \mathcal{l}}\psi(s)]},\end{gathered}\ ] ] where @xmath20 is the fourier transform of @xmath21 , @xmath22 is the laplace transform of @xmath23 . in the uncoupled ctrws the density @xmath24 can be expressed explicitly ( such a ctrw is the random walk subordinated to a renewal process ) @xcite : @xmath25 ( @xmath26 , with fixed @xmath4 , is actually a probability density of a random sum ) , where @xmath27 is the probability of @xmath28 jumps occurring up to the time @xmath4 , and @xmath29 is the probability density of a distance @xmath3 from the initial position reached by a particle after @xmath28 jumps . the probability density @xmath29 is given by @xmath30 where @xmath31 means the convolution , @xmath32 means the @xmath28-fold convolution of @xmath21 with itself , i.e. , @xmath33 , @xmath34 , and @xmath35 . equivalently , @xmath36^n.\ ] ] the probability @xmath27 is given by @xmath37 ( clearly @xmath38 and @xmath39 for @xmath40 ) , in particular , @xmath41 . equivalently , @xmath42^n.\ ] ] this completes the set of auxiliary propositions necessary for further considerations . consider a family of probability densities for waiting time @xmath4 @xcite @xmath43= \begin{cases } \dfrac{a}{\theta \sqrt{1-a } } { \mathop{\mathrm{e}}\nolimits}^{-t/\theta } \sinh \left ( \dfrac{t}{\theta } \sqrt{1-a } \right ) , & 0<a<1,\\[3.5ex ] \dfrac{t}{\theta^2 } { \mathop{\mathrm{e}}\nolimits}^{-t/\theta } , & a=1 , \end{cases}\end{gathered}\ ] ] ( obviously , @xmath44 ) . the laplace transform of the density @xmath45 is @xmath46 \equiv \frac{a}{\theta^2 } \frac{1}{(s + 1/\theta)^2 - ( 1-a)/\theta^2}.\end{gathered}\ ] ] the density @xmath47 was used in refs.@xcite ( see also ref.@xcite ) , and it belongs to the family of the gamma distributions . the density @xmath47 is a probability density of a sum of two independent exponentially distributed random variables with the probability density @xmath48 , and ctrws with such a distribution of waiting times were called in ref.@xcite the delayed ctrws ( dctrws ) . note that the mean waiting time for the exponential distribution is @xmath49 , and it is @xmath50 for the distribution @xmath47 . the mean waiting time for the distribution @xmath45 is @xmath51 , i.e. , the mean waiting time tends to infinity as @xmath52 . completely understanding the conditional character of this notion we call for brevity the ctrws with the probability densities @xmath45 of the waiting time also the dctrws . we denote densities describing the dctrws by @xmath53 . the montroll weiss formula with @xmath54 implies @xmath55 /\theta}.\ ] ] the straightforward calculations show that the density @xmath53 is a solution to the equation @xmath56 = \frac{a}{2 \theta } \left [ \int_{-\infty}^\infty \lambda(x - x ' ) \rho_{\mathrm{drw}}^{}(x',t ) { \mathop{}\!\mathrm{d}}{x ' } - \rho_{\mathrm{drw}}^{}(x , t ) \right]\end{gathered}\ ] ] with the initial value conditions @xmath57 consider a random walk on a regular one - dimensional lattice , where the distance between nearest - neighbour points is @xmath58 , and the coordinates of the lattice points are @xmath59 , @xmath60 . suppose that jumps are made to the nearest - neighbour points with equal probability . this discrete - space random walk is described by the probability density @xmath61/2 $ ] , and the equation becomes @xmath62 \shoveleft{= \frac{a}{2 \theta } \bigg [ \frac{\rho_{\mathrm{drw}}^{}(x_l - \sigma , t ) + \rho_{\mathrm{drw}}^{}(x_l + \sigma , t)}{2}}\\ - \rho_{\mathrm{drw}}^{}(x , t ) \bigg]\end{gathered}\ ] ] assuming that the density @xmath53 is smooth with respect to @xmath3 , one can expand it in the taylor series in @xmath3 . if @xmath58 is small enough then we neglect the derivatives of an order higher than two , and obtain the approximation for the density @xmath53 given by the solution of the telegraph equation ( te ) @xmath63 with the initial conditions @xmath64 the same result is obtained if jumps are made not only to the nearest - neighbour points , the only decisive condition is that the variance of the distribution @xmath65 is to be equal to @xmath66 . similar derivation of the telegraph equation from the discrete - space random walk with the probability density @xmath45 for waiting time was performed in ref.@xcite , where the authors wrote about `` the continuum limit '' as @xmath67 . note that both the equations and describe particles , moving with infinite velocity , while the telegraph equation describes phenomena with finite propagation speed . another derivation of the telegraph approximation is given in ref.@xcite . this is similar to the above one , but it is performed in the fourier laplace space . the derivation , given in ref.@xcite for @xmath68 , is in fact as follows . if the distribution of the jump length is symmetric and has the finite second moment with the variance equal to @xmath66 , then the fourier transform of the probability density of the jump length is approximately @xmath69 for @xmath70 . after substitution of this @xmath71 and @xmath72 into the montroll weiss formula one obtains that the fourier laplace transform of the density @xmath53 is approximately equal to @xmath73 with @xmath74 the inverse fourier laplace transform of @xmath75 implies that the density @xmath76 is a solution to the telegraph equation with the initial conditions . the authors of the paper @xcite call the telegraph equation `` the limit of small jumps ( diffusive approximation ) '' of the ctrw with the probability density @xmath47 for waiting time . the conclusion that the solution to the problem , approximates the dctrw is strange , since the telegraph equation describes phenomena with finite propagation speed . at the same time , the velocity of the motion of the particle in the dctrw is infinite , since waiting time can be arbitrarily small . moreover , the telegraph equation obviously is not `` the limit of small jumps '' of the dctrws . the asymptotic behaviour of various ctrws was studied in refs.@xcite . for convenience we briefly repeat the asymptotic analysis for the dctrws on the basis of refs.@xcite . to find the asymptotic behaviour of the dctrws the densities @xmath7 and @xmath77 are replaced by the scaled densities @xmath78 , @xmath79 , and @xmath80 , @xmath81 , respectively @xcite . the parameters @xmath82 and @xmath83 can be considered as the characteristic step length and waiting time , respectively . the fourier transform of @xmath84 is @xmath85 , and the laplace transform of @xmath86 is @xmath87 . we suppose that the distribution of the jump length is symmetric , @xmath88 , and has the finite second moment with the variance equal to @xmath66 , then the asymptotic behaviour for @xmath89 is @xmath90 as @xmath91 , @xmath92 @xcite . for the density @xmath86 one has the asymptotics @xmath93 as @xmath94 . the latter asymptotics are also valid in general case for @xmath95 , if the density @xmath23 has the finite expectation equal to @xmath51 @xcite . after substitution of the above asymptotics into the montroll weiss formula , one can see that the only possible nontrivial limit is @xmath96}\\[1ex ] \to { \mathcal{f}}{{\mathcal{l}}\rho}_{\mathrm{de}}^{}(k , s ) \quad\text{as}\enskip h \to 0 \enskip\text{and } \enskip \tau \to 0\end{gathered}\ ] ] under the scaling relation @xmath97 , where @xmath98 the inverse fourier laplace transform implies that the density @xmath99 is a solution of the diffusion equation ( de ) @xmath100 with the initial condition @xmath101 thus , the asymptotic behaviour of the dctrws is given by the diffusion equation . note that the diffusion equation is obtained from the telegraph equation by omitting the second time derivative . note also that @xmath102 and @xmath103 , or , equivalently , @xmath104 and @xmath105 for @xmath106 , i.e. , the law of conservation of particles is valid for both the telegraph and diffusion approximations . the other possible limits are @xmath107 as @xmath91 , and @xmath107 as @xmath108 and @xmath109 . the former is the limit of infinitesimal jumps with finite mean waiting time . this is the correct limit [ to be ] obtained in refs.@xcite . these asymptotics are trivial , since the inverse fourier laplace transform of @xmath110 gives @xmath111 , @xmath112 , which means that the particle does not leave an initial position . the diffusion limit , under the scaling relation @xmath113 , is valid for a variety of symmetric ( unbiased ) random walks , both continuous - time and discrete - time ones @xcite . in particular , it is valid for the symmetric bernoulli random walk , which is discrete both in time and space , when the waiting time is exactly @xmath83 and the jumps are exactly @xmath114 with equal probabilities , i.e. , the corresponding probability densities are @xmath115 \lambda_{\mathrm{d}}(x ) = \frac{1}{2 } [ \delta(x - h ) + \delta(x+h ) ] , \quad h>0.\end{gathered}\ ] ] however , in contrast to the ctrws , in which waiting time can be arbitrarily small and , hence , the particle moves with an infinite speed , the speed of the motion of the particle in the bernoulli random walk is finite . besides , it turns out that the solution of the telegraph equation approximates the density for the bernoulli random walk better than that of the diffusion equation @xcite . it is necessary to note here that the explanation , given in ref.@xcite , does not allow us to judge whether the telegraph equation approximates the random walk better than the diffusion equation in a wide time interval ? to check this conclusion we have performed calculations , in which the binomial distribution , being the distribution of the walking particle , was evaluated through the first two terms of stirling s asymptotic series for the gamma function @xcite @xmath116 $ ] as @xmath117 , @xmath118 , @xmath119 , @xmath120 , which is more accurate than conventional stirling s formula . these calculations have confirmed the conclusion up to time @xmath121 , which seems to be asymptotic . obviously , the solutions of both the diffusion equation and the telegraph one do not approximate the binomial distribution at short time . the reason , by which the solution of the telegraph equation approximates the bernoulli random walk better than that of the diffusion equation , seems to be clear : the bernoulli particle moves with finite velocity . however , the probability density for the waiting time @xmath122 , eq . , can be weakly approximated with arbitrary accuracy by the gamma distribution : @xmath123 \xrightarrow{\text{weakly } } \delta(t - \tau ) \quad \text{as } \enskip \mu \to + \infty,\end{gathered}\ ] ] where @xmath124 is the gamma function ; this can easily be derived with the laplace transform . in the random walk with the distribution @xmath125 of the waiting time the particle moves with infinite velocity , since the waiting time can generally be arbitrarily small . however , for sufficiently large @xmath126 the telegraph approximation to this random walk is clearly better than the diffusion one , since @xmath127 is very `` close '' to @xmath128 . thus , in spite of the diffusion asymptotic behaviour for the dctrws and the infinite speed of the motion of the particle , the question still remains : which of the telegraph or the diffusion equations gives better approximation to the dctrws ? to compare the telegraph and diffusion equations with the dctrws we consider two particular dctrws . one of them is a discrete - space random walk , the other is the continuous - space one . the _ discrete - space _ random walk takes place on a regular one - dimensional lattice , where the distance between nearest - neighbour points is @xmath58 , and the coordinates of the lattice points are @xmath59 , @xmath129 . jumps are made to the nearest - neighbour points with equal probability . this random walk is described by the probability density @xmath130.\ ] ] the subscript @xmath131 means bernoulli ( since this is the same distribution as in the bernoulli random walk ) . the variance of this distribution is equal to @xmath66 . the _ continuous - space _ random walk has the gaussian distribution of the jump length with the variance equal also to @xmath66 : @xmath132 its fourier transform is @xmath133 . the gaussian distribution of the jump length was used in ref.@xcite . the densities of the discrete- and continuous - space dctrws are given in appendix [ sec : twoparticul ] by eqs . and , respectively . the telegraph and diffusion approximations are given by eqs . and , respectively . all numerical results are obtained with the parameters @xmath134 and @xmath135 . figs.[fig : dens ] and [ fig : denslog ] show ( in cartesian and logarithmic scales ) the densities of the continuous- and discrete - space dctrws and their telegraph and diffusion approximations at intermediate values of time @xmath136 and @xmath137 with @xmath68 . figs.[fig : densdiscrepgauss ] and [ fig : densdiscrepbin ] show differences between each of the two approximations and the densities of the continuous- and discrete - space dctrws , respectively , obtained for the same values of time and the parameters . calculations show that in all the cases the @xmath0-discrepancy for the diffusion approximation is less than that for the telegraph one . in the case of the discrete - space random walk we calculated the @xmath138-discrepancy . note also that the maximum absolute value of the difference for the diffusion approximation is in all the cases less than that for the telegraph approximation . besides , it is important to emphasize that the telegraph approximation is incorrect with respect to the velocity of the motion of the particle , see figs.[fig : dens ] and [ fig : denslog ] for @xmath136 . the telegraph model gives the finite velocity , while it is infinite in the dctrws . the diffusion model is correct in this respect , however the diffusion paradox remains . and @xmath139 and the telegraph and diffusion approximations @xmath140 and @xmath141 at @xmath136 and @xmath137 with @xmath134 , @xmath68 and @xmath135 . the singular terms @xmath142 and @xmath143 are depicted by vertical arrows . ] , with logarithmic scale on the vertical axis . ] and @xmath144 at @xmath136 and @xmath137 with @xmath134 , @xmath68 and @xmath135 . note that the vertical scale unit is @xmath145 . ] and @xmath146 at @xmath136 and @xmath137 with @xmath134 , @xmath68 and @xmath135 . note that the vertical scale unit is @xmath145 . ] to assess the approximations at long times we have performed in appendix [ sec : asymptcompar ] asymptotic comparison of the telegraph and diffusion approximations to the continuous - space dctrw through their fourier transforms . fig.[fig : fourharmlongt ] shows graphs resulting from the fourier transforms @xmath147 , @xmath148 and @xmath149 at @xmath150 and @xmath151 with @xmath68 . these values correspond to the values of the small parameter @xmath152 and @xmath153 , respectively ( see appendix [ sec : asymptcompar ] ) . fig.[fig : fourharmlongtdiscrep2 ] shows the integrands of the integrals for the same values of time and parameters @xmath58 , @xmath154 and @xmath49 as in fig.[fig : fourharmlongt ] . note that the asymptotics and provide very good approximations ( not shown in the figure ) to the integrands . , @xmath155 and @xmath149 at @xmath150 and @xmath151 with @xmath134 , @xmath68 and @xmath135 , i.e. , @xmath156 and @xmath157 . ] and @xmath151 with @xmath134 , @xmath68 and @xmath135 , i.e. , @xmath156 and @xmath157 . note that the vertical scale unit is @xmath158 . ] for comparison we show in fig.[fig : densdiscrepbinlongt ] differences between each of the two approximations and the densities of the discrete - space dctrw ( not fourier transforms ) , obtained at the same long times @xmath150 and @xmath151 with @xmath68 . note that fig.[fig : densdiscrepbinlongt ] is qualitatively similar to fig.[fig : densdiscrepbin ] . in these cases the @xmath138-discrepancy for the diffusion approximation is also less than that for the telegraph one . we do not show differences between the two approximations and the densities of the continuous - space dctrw for @xmath150 and @xmath151 because fig.[fig : fourharmlongtdiscrep2 ] illustrates this via the fourier transforms . and @xmath146 at @xmath150 and @xmath151 , @xmath134 , @xmath68 and @xmath135 . note that the vertical scale unit is @xmath159 . with fig.[fig : densdiscrepbin ] . ] figs.[fig : densa01smallt ] and [ fig : densa01 ] show the densities of the continuous- and discrete - space dctrws and their telegraph and diffusion approximations with the parameter @xmath160 . fig.[fig : densa01smallt ] corresponds to the small value of time @xmath150 ( @xmath161 ) , which is equal to the mean waiting time . in this case the singular term @xmath142 is quite `` heavy '' : the probability that the particle stays at the initial point is @xmath162 . note that the telegraph and diffusion `` approximations '' essentially differ from the densities of the dctrws . at the same time they are very close to each other . fig.[fig : densa01 ] corresponds to intermediate and long time : @xmath151 , @xmath163 ( intermediate ) and @xmath164 ( long ) ( @xmath165 , @xmath166 and @xmath153 , respectively ) . for @xmath151 the singular term @xmath142 is still `` heavy '' enough : the probability that the particle stays at the initial point for the time @xmath151 is @xmath167 . it is necessary to emphasize that all the results are in agreement with the asymptotics ( even for intermediate and small values of time ) : the differences between the two approximations are much less than the differences between each of them , and the densities of the dctrws , while the diffusion approximation is slightly better than the telegraph one . and @xmath139 and the telegraph and diffusion approximations @xmath140 and @xmath141 at @xmath150 , @xmath134 , @xmath160 and @xmath135 ( @xmath161 ) . the telegraph and diffusion approximations are very close to each other . the singular terms @xmath142 and @xmath143 are depicted by vertical arrows . note that the singular term @xmath142 is quite `` heavy '' . ] and @xmath139 and the telegraph and diffusion approximations @xmath140 and @xmath141 at @xmath151 , @xmath163 and @xmath164 , @xmath134 , @xmath160 and @xmath135 ( @xmath165 , @xmath166 and @xmath153 ) . the telegraph and diffusion approximations are almost undistinguishable in the figure . the singular terms @xmath142 and @xmath143 ( only for @xmath151 ) are depicted by vertical arrows . cf . with fig.[fig : dens ] . ] the dctrws with small @xmath154 can be explained as follows . small values of @xmath154 mean that the particle mostly remains in rest than jumps , since the mean waiting time is long enough . small values of time can be described as lesser than or comparable with the mean waiting time @xmath51 , i.e. , such that @xmath168 . therefore , for small values of time both the diffusion and telegraph equations badly approximate the dctrw , see fig.[fig : densa01smallt ] . at the same time , if @xmath169 then @xmath170 , i.e. , the singular term of the telegraph approximation is negligible . in this case the solution of the telegraph equation behaves like the solution of the diffusion equation . at long times the telegraph and diffusion approximations are almost undistinguishable and both are good , while the diffusion one is slightly better . we should remind that the telegraph equation describes transport phenomena with finite propagation speed , while the velocity of the motion of the particle in the dctrws is infinite , i.e. , the dctrws do not simulate the most distinctive property of the telegraph equation . moreover , asymptotic analysis and computations performed in this paper show that the diffusion equation gives better approximations to the dctrws than the telegraph equation . in other words , _ the dctrws are always closer to their continuous limit than to the solutions of the telegraph equation_. this implies that , in contrast to the widespread opinion , the dctrws do not simulate the telegraph equation at the microscopic level . likewise , ctrws with the exponentially distributed and general waiting time do not simulate the telegraph equation at the microscopic level . summing up , we can conclude that , first , the dctrws do not provide any correct microscopic interpretation of the one - dimensional telegraph equation , and second , the kinetic ( exact ) model of the telegraph equation is different from the model based on the dctrws . an interesting question arises concerning discrete - space random walks with the distributions of the jump length @xmath65 , eq . , and the waiting time @xmath125 , eq .. for sufficiently large @xmath126 this walk is very close to the bernoulli random walk , eq .. for @xmath171 this is the discrete - space ctrw with the exponentially distributed waiting time @xmath172 . in the former case the telegraph approximation is better than the diffusion one , while in the latter case the diffusion approximation is better . being incompletely stated yet , the problem is the following one : is there @xmath173 such that for @xmath174 the diffusion equation gives better approximation to these discrete - space ctrws than the telegraph one , while for @xmath175 the telegraph approximation is better ? the support by the rfbr grants 08 - 01 - 00315-a , 10 - 01 - 00627-a , 11 - 01 - 00573-a , the ec collaborative project health - f5 - 2010 - 260429 and project no.14.740.11.0166 of the russian ministry of science and education is gratefully acknowledged . we thank anonymous referees for useful comments . the density of the dctrw with the distribution @xmath176 of the jump length results from the discrete - space variant of the formula . we denote the density by @xmath177 . it is given by @xmath178 where the probability density @xmath179 is given by the scaled binomial distribution @xmath180= \begin{cases } \dfrac{n!}{2^n \,[(n+\|l\|)/2 ] ! \,[(n-\|l\|)/2 ] ! } , \\[2ex ] \hspace{8.5em}\|l\| \leq n \enskip\text{and } & \hspace{-1ex}n+l \enskip\text{is even},\\[1.5ex ] 0 , \hspace{8.2em}n < \|l\| \enskip\text{or } & \hspace{-1ex}n+l \enskip\text{is odd } , \end{cases}\end{gathered}\ ] ] and @xmath181 stands for the probability @xmath182 , corresponding to the distribution @xmath45 . the laplace transform of @xmath181 is @xmath183^{n+1}}.\ ] ] therefore , the probability @xmath181 is given by @xmath184 , & 0<a<1,\\[3ex ] { \mathop{\mathrm{e}}\nolimits}^{-t/\theta } \bigg [ \dfrac{1}{(2n ) ! } \left(\dfrac{t}{\theta}\right)^{2n } \\[3ex ] \hspace{2.5em}+ \dfrac{1}{(2n+1 ) ! } \left(\dfrac{t}{\theta}\right)^{2n+1 } \bigg ] , & a=1 , \end{cases}\ ] ] where @xmath185 \times \left [ ( -1)^m { \mathop{\mathrm{e}}\nolimits}^{\sqrt{1-a } \,t/\theta } + ( -1)^{n+1 } { \mathop{\mathrm{e}}\nolimits}^{-\sqrt{1-a } \,t/\theta } \right]\end{gathered}\ ] ] with corresponding limit as @xmath186 . the density of the dctrw with the gaussian distribution @xmath187 of the jump length results from the formula . we denote the density by @xmath188 . it is given by @xmath189 where the singular term is @xmath190 with @xmath191 \hspace{1.5ex}+ \dfrac{1}{\sqrt{1-a } } \sinh \left ( \dfrac{t}{\theta } \sqrt{1-a } \right ) \bigg ] , & 0<a<1,\\[3ex ] { \mathop{\mathrm{e}}\nolimits}^{-t/\theta } \left ( 1 + \dfrac{t}{\theta } \right ) , & a=1 , \end{cases}\ ] ] and the regular term is @xmath192 the _ telegraph approximation _ is given by the solution to the initial - value problem , for the telegraph equation @xcite @xmath193 where the singular term is @xmath194,\ ] ] and the regular term is @xmath195 + t \left ( \sqrt{t^2 - \frac{x^2}{v^2 } } \right)^{-1 } \,i_1 \left ( \frac{1}{\theta } \sqrt{t^2 - \frac{x^2}{v^2 } } \right ) \bigg],\end{gathered}\ ] ] @xmath196 is the velocity , @xmath197 is the heaviside step function . the _ diffusion approximation _ is given by the solution to the initial - value problem , for the diffusion equation @xmath198 note that the significant difference between the density @xmath188 of the dctrw and the solution @xmath76 of the telegraph equation is that the support of @xmath142 is localized at the starting point of the walk ( @xmath16 ) , while the support of @xmath143 is localized at the moving front ( @xmath199 ) . the term @xmath142 is negligible for @xmath200 , while the term @xmath143 is negligible for @xmath170 , the former condition being stronger . note also that the support of @xmath201 is @xmath202 , while the support of @xmath203 is @xmath204 $ ] . however , the latter difference appears to be in general not so significant . indeed , according to the demoivre laplace and the central limit theorems @xcite the solution of the diffusion equation approximates the bernoulli random walk with @xmath205 and @xmath206 , @xmath207 in the expression being the number of jumps . the solution has the support @xmath202 , while the particle involved into the random walk has a finite velocity , nevertheless , the diffusion approximation is a classic and widely used one to the bernoulli random walk . in this section the long time means @xmath200 , i.e. , time is longer than the mean waiting time @xmath51 . we compare the telegraph and diffusion approximations to the continuous - space dctrw at long times by evaluation the @xmath208 norms of the differences @xmath209 and @xmath210 . it is more convenient to consider the @xmath208 norms of the fourier transforms of the differences . recall that the @xmath208 norms of a function @xmath211 and its fourier transform @xmath212 are related by @xmath213 . thus , we need to estimate the integrals @xmath214 \int_0^\infty \left\| { \mathcal{f}}\rho_{\mathrm{de}}^{}(k , t ) - { \mathcal{f}}\rho_{\mathrm{drwg}}^{\mathrm{r}}(k , t ) \right\|^2 { \mathop{}\!\mathrm{d}}{k}\end{gathered}\ ] ] at long times ( the integrals are taken over the interval @xmath215 , rather than @xmath216 , since the integrands are even with respect to @xmath217 ) . the fourier laplace transform of the density @xmath188 of the dctrw with the gaussian distribution @xmath187 of the jump length is the following one @xmath218 /\theta^2},\end{gathered}\ ] ] which results directly from the montroll weiss formula with @xmath219 and @xmath54 . this implies the fourier transform @xmath220 \shoveright{+ \frac{1}{\sqrt{1 - a r(k ) } } \sinh \left ( \frac{t}{\theta } \sqrt{1 - a r(k ) } \right ) \bigg],}\\[1ex ] r(k ) = 1 - { \mathop{\mathrm{e}}\nolimits}^{-\sigma^2 k^2 /2},\end{gathered}\ ] ] the fourier transform of the singular term being @xmath221 where @xmath222 is given by eq .. obviously @xmath223 as @xmath224 . the fourier laplace transform of the telegraph approximation can be rewritten as @xmath225 this implies the fourier transform @xmath226 + \left ( \sqrt{1 - a \frac{\sigma^2 k^2}{2 } } \right)^{-1 } \sinh \left ( \frac{t}{\theta } \sqrt{1 - a \frac{\sigma^2 k^2}{2 } } \right ) \bigg],\end{gathered}\ ] ] where we assume that if @xmath227 , then @xmath228 . the fourier transform of the singular term is @xmath229 obviously @xmath230 as @xmath224 . to estimate the integrals at long times it is convenient to introduce the small parameter @xmath235 and change the variable @xmath217 for @xmath236 . then we have in the new variable the fourier transform of the density of the continuous - space dctrw @xmath237\shoveright{+ \frac{1}{\sqrt{1 - a r(\zeta ) } } \sinh \left ( \frac{\sqrt{1 - a r(\zeta)}}{a\varepsilon } \right ) \bigg],}\\[1ex ] r(\zeta ) = 1 - { \mathop{\mathrm{e}}\nolimits}^{-2 \varepsilon \zeta^2},\end{gathered}\ ] ] the `` singular term '' being @xmath238 $ ] , the fourier transform of the telegraph approximation @xmath239 + \frac{1}{\sqrt{1 - 2 a \varepsilon \zeta^2 } } \sinh \left ( \frac{\sqrt{1 - 2 a \varepsilon \zeta^2}}{a\varepsilon } \right ) \bigg],\end{gathered}\ ] ] the `` singular term '' being @xmath240 , and the fourier transform of the diffusion approximation @xmath241 the integrals , within a factor of @xmath242 , become in the new variable @xmath243 \sqrt{\varepsilon } \int_0^\infty \left\| { \mathcal{f}}\rho_{\mathrm{de}}^{}(\zeta , t ) - { \mathcal{f}}\rho_{\mathrm{drwg}}^{\mathrm{r}}(\zeta , t ) \right\|^2 { \mathop{}\!\mathrm{d}}{\zeta}.\end{gathered}\ ] ] the `` regular term '' @xmath244 tends to zero monotonously and exponentially as @xmath245 , as well as the fourier transform @xmath246 . therefore , the second integral of the integrals can be approximated with arbitrary accuracy by the integral with the upper limit @xmath247 instead of @xmath248 , if @xmath247 is sufficiently large . for small @xmath249 the fourier transform @xmath250 tends to zero monotonously and exponentially on the interval @xmath251 $ ] , where the square root in eq . is nonnegative . the `` singular term '' @xmath233 can be neglected on this interval for small @xmath249 . asymptotic behaviour of the `` regular term '' is @xmath252 as @xmath245 . the reason for such a behaviour is that the regular term @xmath253 is discontinuous at @xmath254 . nevertheless , due to exponential decrease of @xmath255 for small @xmath249 the first integral of the integrals can be approximated with arbitrary accuracy by the integral with the upper limit @xmath247 instead of @xmath248 , if @xmath247 is sufficiently large . note that for any @xmath247 there exists sufficiently small @xmath249 such that @xmath256 , i.e. , @xmath257 \subset [ 0,1/\sqrt{2a\varepsilon}]$ ] . the straightforward calculations imply uniform asymptotics on the interval @xmath257 $ ] : @xmath258= \left\ { 1 + \left [ \frac{a}{2 } \left ( \zeta^2 - \zeta^4 \right ) + \zeta^4 \right ] \varepsilon \right\ } { \mathop{\mathrm{e}}\nolimits}^{-\zeta^2 } + o(\varepsilon^2)\\ \text{as}\quad \varepsilon \to 0\end{gathered}\ ] ] and @xmath259 \varepsilon \right\ } { \mathop{\mathrm{e}}\nolimits}^{-\zeta^2 } + o(\varepsilon^2)\\ \text{as}\quad \varepsilon \to 0\end{gathered}\ ] ] ( @xmath260 and @xmath250 have the same asymptotics ) . therefore , the asymptotics for the differences are @xmath261= - \zeta^4 { \mathop{\mathrm{e}}\nolimits}^{-\zeta^2 } \varepsilon + o(\varepsilon^2 ) \quad\text{as}\quad \varepsilon \to 0\end{gathered}\ ] ] and @xmath262= - \left [ \frac{a}{2 } \left ( \zeta^2 - \zeta^4 \right ) + \zeta^4 \right ] { \mathop{\mathrm{e}}\nolimits}^{-\zeta^2 } \varepsilon + o(\varepsilon^2 ) \quad\text{as}\quad \varepsilon \to 0.\end{gathered}\ ] ] therefore , at long times ( for small @xmath249 ) the integrals are @xmath263\simeq \varepsilon^{2.5 } \int_0^\infty \zeta^8 { \mathop{\mathrm{e}}\nolimits}^{-2\zeta^2 } { \mathop{}\!\mathrm{d}}\zeta = \sqrt{\frac{\pi}{2 } } \,\frac{7!!}{2 ^ 9 } \,\varepsilon^{2.5 } \\[1ex]\approx 0.257 \,\varepsilon^{2.5}\end{gathered}\ ] ] and @xmath264\simeq \varepsilon^{2.5 } \int_0^\infty \left [ \frac{a}{2 } \left ( \zeta^2 - \zeta^4 \right ) + \zeta^4 \right]^2 { \mathop{\mathrm{e}}\nolimits}^{-2\zeta^2 } { \mathop{}\!\mathrm{d}}\zeta\\[1ex ] \shoveleft{= \sqrt{\frac{\pi}{2 } } \bigg [ \left ( \frac{3!!}{2 ^ 7 } - \frac{5!!}{2 ^ 8 } + \frac{7!!}{2^{11 } } \right ) a^2 } \\[1ex ] \shoveright{+ \left ( \frac{5!!}{2 ^ 7 } - \frac{7!!}{2 ^ 9 } \right ) a + \frac{7!!}{2 ^ 9 } \bigg ] \varepsilon^{2.5 } } \\[1ex]\approx ( 0.020 \,a^2 - 0.110 \,a + 0.257 ) \,\varepsilon^{2.5}.\end{gathered}\ ] ] note that at long times the same integral for the difference @xmath265 is @xmath266\simeq \varepsilon^{2.5 } \int_0^\infty \left [ \frac{a}{2 } \left ( \zeta^2 - \zeta^4 \right ) \right]^2 { \mathop{\mathrm{e}}\nolimits}^{-2\zeta^2 } { \mathop{}\!\mathrm{d}}\zeta \approx 0.020 \,a^2 \,\varepsilon^{2.5}.\end{gathered}\ ] ] the asymptotic estimates and show that at long times the diffusion approximation to the density @xmath188 of the dctrw is better in the @xmath208 norm , than the telegraph approximation . at the same time , for small @xmath154 the telegraph approximation is almost as good as the diffusion one . e. w. montroll and m. f. shlesinger . on the wonderful world of random walks . in j. l. lebowitz and e. w. montroll , editors , _ nonequilibrium phenomena ii : from stochastics to hydrodynamics _ , volume 11 of _ studies in statistical mechanics _ , pages 1121 . north - holland , amsterdam , 1984 . b. i. henry , t. a. m. langlands , and s. l. wearne . anomalous diffusion with linear reaction dynamics : from continuous time random walks to fractional reaction - diffusion equations . , 74:031116 ( 15 pp ) , 2006 .
dynamics of a nonequilibrium system is one of the most stimulating topics in statistical mechanics . the dynamical property is determined by the interplay among various elements such as interaction , initial and boundary conditions , and so on . but little is known about how it depends on these elements and realizes universality in a situation where the elements are intertwined . the aim of this article is to approach such questions by analyzing the dynamics of the one - dimensional asymmetric simple exclusion process ( asep ) . the asep is one of the most typical models of interacting particle processes where particles diffuse to a preferred direction with hard core repulsive interaction @xcite . in spite of the simplicity of the model , it has been known that the asep shows various interesting phenomena caused by the collaboration between the diffusion and exclusion effect . furthermore , in the one - dimensional case , it has an integrable mathematical structure which allows us to analyze some physical quantities exactly . for example , in the open boundary condition a steady state can be constructed using the matrix product method and @xmath0-orthogonal polynomials , and the boundary - induced phase transition can be discussed @xcite . on the other hand , the dynamical ( non - stationary ) properties of the one - dimensional asep have been also analyzed recently . among various studies on the topic , we focus on the problem of the diffusion of a particular particle in the one - dimensional asep . ( we call the particle a _ tagged _ particle . ) this is a fundamental and elementary problem for the understanding of the dynamics of the asep . we are interested in how initial conditions and the exclusion effect through hard - core repulsive interaction affect the diffusion property and how it is different from the ( normal ) diffusion of the brownian particle . the scaling exponent of the diffusion provides insights into these problems . it was found that it depends on the initial conditions . in a steady state with a given density , the position fluctuation of a tagged particle grows as @xmath1 @xcite . the exponent 1/2 is the same as that of the brownian particle . on the other hand , for the fixed initial conditions where initially we create particles with a given density and fix the initial configuration , the exponent changes from 1/2 to 1/3 @xcite . such a diffusion is called an anomalous diffusion . this exponent 1/3 can be understood from the fact that the time evolution of the one - dimensional asep can be described by the one - dimensional kardar - parisi - zhang universality class @xcite . in the study of the current in the one - dimensional asep on a ring , the exponent 1/3 was calculated by the bethe ansatz technique @xcite . note that the two cases above have a common property that the density is invariant under both time and space translation and a tagged particle moves with a fixed velocity on average . on the other hand , let us now consider another initial condition where all sites on the left of a certain bond are occupied while all right sites are empty . we call this initial condition a _ step _ initial condition . this is the typical initial condition where the density and the average velocity of a tagged particle change with time . under the initial condition , we focus on the @xmath2th particle from the right as a tagged particle and consider the time evolution of the particle when @xmath2 is large ( but finite ) . in order to study this problem , we investigate the limiting behavior of the diffusion of the tagged particle as both @xmath2 and time go to infinity . in the totally asymmetric simple exclusion process ( tasep ) , where particles move only to the right , the average position of a tagged particle has been given in the scaling limit @xcite . on the other hand , we are interested in the limiting process itself which the tagged particle obeys , as well as the average position and the scaling exponent . such an attempt as getting more sophisticated information than the scaling exponent has been recently discussed in the study on the current fluctuations in the tasep . not only the exponent but also the scaling function were obtained and it was revealed that the function is equivalent to the largest eigenvalue distribution in the random matrix theory with various universality classes . this development is based on the results on the longest increasing subsequence in a random permutation @xcite . remarkable is that the scaling function detects a difference in initial conditions even if the scaling exponent does not . for instance , in the case of the step initial condition , the scaling function is equivalent to the gue tracy - widom distribution @xcite , which is the largest eigenvalue distribution of the gaussian unitary ensemble ( gue ) @xcite . in other initial conditions , on the other hand , these are described as the largest eigenvalue distributions with other universality classes in rmt such as gaussian orthogonal ensemble ( goe ) @xcite , two independent goe s which is denoted by goe@xmath3 @xcite , and so on @xcite . in addition , the equal - time multipoint distribution of the current fluctuations is obtained in the step and periodic initial conditions @xcite by use of the green function of the tasep @xcite . in this article , we calculate the multi - time distribution function of position fluctuations of a tagged particle in the tasep with the step initial condition based on the techniques of the directed polymer problem in a 01 matrix @xcite and stochastic growth of young diagram characterized by the schur process @xcite . we express the function as the fredholm determinant and discuss the scaling limit by the saddle point analysis . we remark that what will be considered in this paper is the pointwise convergence of the kernel . the convergence of the fredholm determinant itself is certainly expected to hold from previous works on related problems @xcite , but more elaborate asymptotic analysis is necessary for its proof . in the study on the asep , such correlations between different times have not been studied yet although the green function @xcite and the equal - time joint distributions of particle positions @xcite have been studied recently . in the one - time case , on the other hand , the position fluctuation is essentially the same as the current fluctuation which has already been studied in @xcite . but it has not been stated as the position fluctuation . furthermore , we also consider the dependence of the distribution function on defect particles located in front of the tagged particle . here defect particle means a particle with different hopping rate from other normal particles and we assume the number of them is finite . such a situation as the tasep with particle - dependent hopping rates has been recently discussed in @xcite . we clarify that when their hopping rate is smaller than that of normal particles , the limiting behavior of the multi - time distribution changes due to their presence although their number is finite . the paper is organized as follows . in the next section , we describe the definition of the model . the main result of this article is given in section [ main ] . in section [ ftts ] , we express the multi - time distribution of position fluctuations in the tasep as fredholm determinant . for this purpose , we first show that the directed polymer problem on a 01 matrix is related to the time evolution of a tagged particle in the tasep with the step initial condition . next , by applying the dual robinson - schensted - knuth algorithm to the 01 matrix , we map the tagged particle problem in the tasep to the stochastic growth of young diagram described by the schur process . in section [ asymp ] , we discuss the two types of scaling limit for the multi - time distribution function , the case where both time @xmath4 and the label of the tagged particle @xmath2 tend to infinity , and the one where @xmath4 goes to infinity with @xmath2 fixed . some discussions and concluding remark are given in section [ discussion ] and section [ conclusion ] respectively . in this section , we define the model and the quantity which we study in this article . let us consider the one - dimensional infinite lattice and particles as illustrated in fig . 1 . each site can be occupied by at most one particle . suppose all particles are labeled an integer @xmath5 from the right . a particle moves stochastically obeying the following rules . during each time step between @xmath6 and @xmath7 , the particle labeled @xmath5 hops to the right neighboring site with probability @xmath8 ( @xmath9 ) and stays at the same site with probability @xmath10 if the right neighboring site is empty . on the other hand , if the site is occupied by the particle labeled @xmath11 , the @xmath5th particle stays at the same site with probability 1 . this incorporates the exclusion effect which describes the hard core repulsive interaction . when hopping rates of all particles are the same , i.e. , @xmath12 , the system is the usual totally asymmetric simple exclusion process ( tasep ) with parallel update . we allow the particle - dependent hopping rates and study the effects of defect particles . in this article , we consider the step initial condition in which all sites on the left of a certain bond between two sites are occupied by particles and all sites on the right of the bond are empty . 1(b ) illustrates the initial condition . under these settings , we label the particles at @xmath13 from the right ( see fig . 1(b ) ) and consider the dynamics of the particle labeled @xmath2 . ( this particle which we focus on is called the tagged particle . ) for this purpose , we set the position coordinate such that the tagged particle is at the origin at time 0 as depicted in fig . let us define @xmath14 as the position of the @xmath2th particle at time @xmath4 . in other words , @xmath14 represents the distance travelled by the @xmath2th particle from time @xmath15 through @xmath4 . the quantity which we will investigate is the multi - time joint distribution function for @xmath14 defined as @xmath16 in particular , we are interested in the asymptotic behavior of as @xmath4 goes to infinity . in this article , two types of scaling limit will be discussed . the first one is the limit where both time @xmath4 and the label of the tagged particle @xmath2 go to infinity with the ratio @xmath17 fixed . in the second one , on the other hand , we take the @xmath18 limit with @xmath2 fixed . in order to see how the whole particles from the first to @xmath2th travel , we performed the numerical simulation of which the result is shown in fig . 2 . we considered the time evolution of the particles from the first to 100th and assumed the hopping rate of all particles is 0.9 except the first , 25th , 50th and 75th particles whose hopping rate is 0.8 . thus these four particles represent the defect particles . the typical example of the time evolution of the 100 particles from time 0 through 3000 is illustrated in fig . 2(a ) . in this figure , the vertical axis represents the position coordinate introduced in fig . 1(b ) , whereas the horizontal axis is time axis , and each particle is expressed as @xmath19 . from the macroscopic point of view such as fig . 2(a ) , we only find that they travel in a group . however , if we zoom the configuration of the particles in a certain time zone , we can see the microscopic pattern of the configuration . here we show two characteristic configurations which are depicted in fig . 2(b ) and 2(c ) . 2(b ) ( resp.(c ) ) represents the configuration of the particles when time is around 200 ( resp . 3000 ) . when the time zone which we focus on is early enough as in fig . 2(b ) , the particles move forming only one group and the four defect particles are included in the group . on the other hand , when time has passed sufficiently as in fig . 2(c ) , they form the four groups automatically and the top of each group is the defect particle . the similar platoon structure was also discussed in the study of the tasep with disordered hopping rates @xcite . in fact , as will be explained in the next section , the position fluctuation of a tagged particle also changes with time corresponding to the change of the configuration . our whole discussions in this article are based on a fact that the tagged particle problem for the step initial condition has a nice combinatorial structure which admits us to obtain a closed expression for the multi - time joint distribution in the form of the fredholm determinant . the results are summarized in the following theorem . the proof will be presented in section [ ftts ] . note that due to the rule of the tasep , the particle labeled @xmath2 can not move when @xmath20 . thus , throughout the article , we consider the multi - time function under the condition @xmath21 * theorem 1 * _ _ when holds , @xmath22 here @xmath23 is the fredholm determinant defined as @xmath24 where @xmath25 and the kernel @xmath26 is given by @xmath27 here @xmath28 in and denotes a contour enclosing the origin anticlockwise with radius @xmath29 and @xmath30 in satisfy the conditions @xmath31 and @xmath32 . using the results in theorem 1 , one can study the asymptotics of the joint distribution . in this subsection , we consider the scaling limit such that both @xmath4 and @xmath2 go to infinity with their ratio @xmath34 fixed . @xmath35 represents the scaled time . notice that we also take the limit of the label of the tagged particle @xmath2 . this may sound strange if the particle we are focusing on is varied as @xmath2 goes to infinity . rather we think that the results below give the asymptotic behaviors of the path statistics of the tagged particle for large but finite @xmath2 which is fixed . alternatively one might be able to interpret them as giving the asymptotic behaviors of the path statistics of the tagged particle near a large but finite @xmath4 . we also discuss the effect of defect particles . in particular , we consider the situation where there are finite @xmath36 defect particles with stay rates @xmath37 in front of the tagged particle . for this purpose , we set the stay rate @xmath10 of @xmath5th particle as follows . for some set @xmath38 with @xmath36 fixed , we assume @xmath39 note that the particle with the rate @xmath40 represents the defect particle whose label is @xmath41 whereas the remaining particles with the rate @xmath0 are the normal particles . the average position of the tagged particle divided by @xmath2 has a deterministic limit as @xmath33 . let us call it @xmath42 , @xmath43 we also define @xmath44 . ( note that @xmath45 is the stay rate of the slowest defect particle . ) if @xmath46 , which is the case where the slowest defect particle is slower than the normal particles , we obtain @xmath47 where @xmath48 and @xmath49 @xmath50 can be understood from theorem 1.1 . in @xcite . if @xmath51 , the average position is represented as only the first and second cases in . ( @xmath52 and @xmath53 do not appear in the case . ) note that @xmath50 does not depend on the stay rate @xmath45 of the slowest defect particle whereas @xmath53 does . @xmath42 is illustrated in fig . next we consider the position fluctuations of the tagged particle around the average position . there are four typical regions shown in fig . 3 . in each region , we can obtain the result on the position fluctuations by focusing on a point within the region and taking a proper scaling around the point . note that in fig . 3 , the regions 2 and 4 spread in the form of a line whereas the regions 1 and 3 are point - like . thus , in the region 1(resp . 3 ) , we always consider the fluctuation property around the point @xmath54 ( resp . @xmath55 ) . this is the region where the tagged particle just starts to move . this corresponds to @xmath56 . we set @xmath57 where @xmath58 . note that the time @xmath59 is approximately equal to the time on which some leftmost holes arrive at the site occupied by the tagged particle . one easily finds that the arrival time can be scaled as in which is the same scaling as the central limit theorem . applying the scaling to theorem 1 , we have the theorem as follows . the proof will be given in section [ pot21 ] . * theorem 2 - 1 * _ @xmath60 here the fredholm determinant in the right hand side is defined in where @xmath61 . @xmath62 is defined in . the kernel is given by @xmath63 here @xmath64 can be represented by use of the parabolic cylinder function @xmath65 in @xcite as @xmath66 where @xmath67 is taken to be positive , and @xmath68 where @xmath69 in represents the contour enclosing the origin anticlockwise with radius 1 , and @xmath70 is the hermite polynomial with degree @xmath36 @xcite . _ the special case where @xmath71 has appeared in the distribution of the height fluctuation in the `` critical regime '' of the oriented digital boiling model @xcite . recently the distribution in one time case ( @xmath72 ) has been also given in @xcite . our formula above gives a generalization to the multi - time version . in particular , the limiting distribution in the case @xmath72 can be described as @xmath73 where @xmath74 note that in the summation on @xmath75 where @xmath76 can be omitted . thus in this case , we can obtain exactly the limiting probability by calculating the determinants with finite rank . another expression of the kernel has recently appeared as the `` discrete hermite kernel '' in @xcite . in @xcite , the specific values of the probability are given in the case @xmath77 . this is the region where the effect of the defect particles does not affect the time evolution of the tagged particle and the dynamics of the ordinary tasep is dominant . the typical picture of the time evolution of the whole particles from 1st to @xmath2th in this region is illustrated in fig . 2(b ) . in this figure , we are interested in the position fluctuations of the bottommost particle . let us scale as @xmath79 where @xmath80 and @xmath81 . @xmath50 is defined in and @xmath82 note that the scaling exponent 1/3 in also appears in an anomalous diffusion of the tagged particle @xcite . the scaling exponents @xmath83 and @xmath84 in and are characteristic of the one - dimensional kpz universality class . they are expected to be universal for all models belonging to the kpz universality class . under the scaling defined above , we get the following theorem for the scaling function . this function is also universal but may depend on various elements such as initial conditions and the effect of the defect particles . for example , as we will show in theorem 2 - 3 , the scaling function in the region 3 is different from that in region 2 although we take the same scaling and in both regions . thus , the function provides more detailed viewpoint in the kpz universality than the scaling exponents . the proof of this theorem will be given in section [ pot223 ] . * theorem 2 - 2 * _ @xmath85 here the right hand side is the fredholm determinant defined as @xmath86 where @xmath87 are defined in terms of @xmath62 , as @xmath88 and the kernel @xmath89 is given by @xmath90 _ the kernel @xmath91 is called the extended airy kernel @xcite . the process characterized by the fredholm determinant with this kernel is called the airy process @xcite . this process appears as the limiting process of the largest eigenvalue in dyson s brownian motion model @xcite of the unitary class . the model is described as @xmath92 hermitian matrix where each independent element of @xmath93 obeys the ornstein - uhlenbeck process . the transition probability density @xmath94 from matrix @xmath95 to @xmath96 during @xmath97 is given by @xmath98 where @xmath99 is the normalization constant . if we choose the initial matrix @xmath100 to be gue random matrix , the joint density function of the probability that matrix @xmath96 is at time @xmath101 is represented as the hermitian multi - matrix model , @xmath102 where @xmath103 is the normalization constant . let @xmath104 be the largest eigenvalue of @xmath95 and we consider the quantity , @xmath105 when we set @xmath106 the limiting distribution is described as @xmath107 in the case @xmath72 , the distribution is the gue tracy - widom distribution @xcite . hence our theorem 2 - 2 says that in the appropriate scaling limit the dynamics of the tagged particle in this region is equivalent to the dynamics of the largest eigenvalue of dyson s brownian motion of the unitary type . in the context of the one dimensional kpz universality class , the airy process has already appeared in the study of the polynuclear growth ( png ) model . in @xcite , the process has first appeared as the equal - time multi - point height fluctuation . recently in @xcite , it has been shown that the process also describes the correlations among any `` space - like '' points in the space - time plane . our result is closely related to the latter situation . this region is the border between the region 2 where the normal particles in the tasep are dominant , and the region 4 where only the finite number of defect particles are dominant . first we consider the case where the stay rates @xmath109 of the defect particles are distinct . taking the same scaling as with @xmath108 , we obtain the following theorem . * theorem 2 - 3 * _ @xmath110 where the fredholm determinant and @xmath111 in the right hand side are defined in and respectively . the kernel is given by @xmath112 where @xmath91 is the extended airy kernel . _ this kernel has appeared in @xcite in the study of the height fluctuation property of the png model with external sources . when @xmath71 , it has been known that the fredholm determinant of this kernel is described as the distribution of the larger of the largest eigenvalues in two independent goes , which is denoted as goe@xmath3 @xcite . on the other hand , when @xmath113 , the fredholm determinant becomes the gue tracy - widom distribution . thus describes the transition of the largest eigenvalue distribution between gue and goe@xmath114 . next we consider the situation where the hopping rates @xmath115 are more or less the same . in addition to the scaling , we set the hopping rate of the defect particles as @xmath116 where @xmath117 is defined in . note that the parameters @xmath118 characterize the inhomogeneity of the hopping rates and the case @xmath119 corresponds to the situation where the hopping rates degenerate completely . under these settings , we obtain the following theorem . * theorem 2 - 3 * _ @xmath120 where the fredholm determinant and @xmath111 in the right hand side are defined in and respectively and the kernel is given by @xmath121 where the contour @xmath122 runs from @xmath123 to @xmath124 passing the down side of the points @xmath125 . _ in the case @xmath72 , this fredholm determinant has appeared in @xcite . note that if we set @xmath126 for some @xmath127 and @xmath128 for @xmath129 , we can realize the situation where the maximum of @xmath130 is unique such as the former case . indeed in the setting above , the kernel @xmath131 is reduced to @xmath132 . in that sense , the theorem is the generalization of theorem 2 - 3 . later we will give only the proof of the theorem in section [ pot223 ] . the fredholm determinant also describes the limiting largest eigenvalue distribution in the following multi - matrix model . let @xmath133 be the @xmath92 hermitian matrices . analogous to , the joint density function of the model is defined as @xmath134 here @xmath135 is the normalization constant and @xmath136 are parameters of the model such that @xmath137 gauges the strength of the connection between matrix @xmath138 and @xmath95 . @xmath139 represents the external source with rank @xmath36 of the model . under the scalings and @xmath140 the limiting joint distribution of the largest eigenvalues @xmath104 of @xmath141 is described as the fredholm determinant , @xmath142 in the case @xmath143 , was shown in @xcite . we can also show this equation for the general @xmath36 case in a similar manner . this is the region where the effect of the defect particles has become dominant for the dynamics of the tagged particle . as in the region 3 , we first consider the case @xmath145 . in this case , one can expect that the dynamics is effectively the same as that of the slowest defect particle with stay rate @xmath146 . to see nontrivial correlations , times @xmath147 should be macroscopically separated . we set @xmath148 where @xmath81 , @xmath149 is given in , and @xmath150 under this setting and introducing the parameter @xmath151 such that @xmath152 we get the theorem as follows . * theorem 2 - 4 * _ @xmath153 where the fredholm determinant and @xmath111 are defined in and respectively and @xmath154 _ this kernel represents the propagation of a brownian particle ( a particle obeying the ornstein - uhlenbeck process ) . actually the two point distribution is calculated as @xmath155 next we consider the case @xmath156 . the typical picture of the time evolution of the particles is depicted in fig . 2(c ) . in this figure , it seems that four groups of which each top particle is the defect particle are formed and each group behaves like one particle in the tasep . thus in the region , one can expect that the system is effectively the same as @xmath36 tasep particles . in addition to the scaling , we also set @xmath157 note that the scaling exponent 1/2 in is the same as that in the case of the central limit theorem . however the following theorem indicates that the limiting process of @xmath14 depends on the number @xmath36 of the defect particles . note that this is not the information which the scaling exponent can detect . * theorem 2 - 4 * _ @xmath158 the fredholm determinant and @xmath111 in the right hand side are defined in and respectively and the kernel is given by @xmath159 where @xmath160 in , @xmath161 represents the contour enclosing @xmath162 anticlockwise and @xmath163 represents the arbitrary path running from @xmath164 to @xmath165 . _ the kernel above has been given in @xcite . as in theorem 2 - 3 , it is reduced to @xmath166 if we set @xmath167 for some @xmath127 and @xmath168 for @xmath129 . thus , later we will give only the proof of this theorem in section [ pot24 ] . the fredholm determinant also describes the joint distribution function of the largest eigenvalue in the hermitian multi - matrix model , where we assume the rank of @xmath169 is @xmath36 and @xmath170 . the joint distribution of the largest eigenvalues @xmath104 of @xmath95 can be described as the fredholm determinant in the above theorem , @xmath171 note that in the equation above , we do not take the limit as the rank @xmath36 goes to infinity . here we consider another scaling limit such that @xmath2 is fixed and @xmath4 goes to infinity . we set @xmath172 and consider the asymptotic behavior as @xmath173 goes to infinity . we get the following theorem . the proof of the theorem will be given in section [ pot3 ] . * theorem 3 * _ @xmath174 where the right hand side is the same as that of with @xmath175 . _ in this subsection , we consider the continuous time version of the tasep which is also usually discussed in the study of the asep . in the case without defect particles , the rule is defined as follows . let @xmath176 be the time variable which can take any positive real number . between time @xmath176 and @xmath177 , a particle can hop to its right neighboring site with probability @xmath178 if the site is empty . if the site is occupied , the particle stays at the same site with probability 1 . we easily find that the continuous time version can be realized from the discrete time version defined in section [ model ] by taking the limit , @xmath179 with @xmath180 fixed . in addition , the scaling limit in section [ sclim1 ] can also be studied by taking the @xmath181 limit with their ratio @xmath182 fixed . applying and to the theorems in section [ sclim1 ] , we can rewrite the theorems for the continuous time tasep . here we show only the result which corresponds to theorem 2 - 2 . analogous to and , we scale time @xmath183 and the position @xmath184 as @xmath185 where @xmath186 and @xmath187 then we have @xmath188 here the right hand side is defined in . in this section , we discuss a relationship between the time evolution of a tagged particle in the tasep with the step initial condition described in section [ model ] and the stochastic growth of a young diagram @xcite which turns out to be a special case of the schur process @xcite . in the last part of this section , the proof of theorem 1 in section [ multitimed ] is obtained . in the study of the tasep , the technique using the enumeration of young diagrams was used in @xcite to study the current fluctuation . the discussion in @xcite was based on the relation of the tasep to the directed polymer problem on a matrix of which the elements obey the geometric distribution . in this section , on the other hand , we consider the correspondence between the directed polymer problem on a 01 matrix @xcite and the tasep , which is a similar but different mapping . for the analysis of the tagged particle problem , the mapping is more natural and convenient . let us consider an ensemble of @xmath189 01 matrices @xmath190 , whose matrix elements are either 0 or 1 . all matrix elements are independent random variables and each element obeys the bernoulli distribution , @xmath191 here @xmath192 is taken to be the same value as the hopping rate of the @xmath127th particle of the tasep defined in section [ model ] . each realization of the 01 matrix @xmath193 can be translated to a time evolution of the tasep as follows . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ @xmath194(resp . 1 ) : between time @xmath195 and @xmath196 , the @xmath127th particle tries to hop to the right neighboring site ( resp . stays at the same site ) . when the right neighboring site is occupied at time @xmath195 , particle can not move due to the exclusion effect even if @xmath194 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ since @xmath194 could mean both hopping or stay depending on whether the target site is occupied or empty , the mapping from a @xmath197 matrix to the time evolution of the tasep is not one - to - one . in fig . 4 , we depict an example of the time evolution in the tasep and the corresponding two 01 matrices . note that in the step initial condition , the @xmath127th particle can not move until time @xmath198 and that the time evolution of the @xmath127th particle at time step between @xmath5 and @xmath199 does not depend on that of the @xmath198th particle at the time step . thus one finds that the 01 matrix defined above has all informations needed for the time evolution of the @xmath2th particle from time 0 through @xmath200 . let us define a left - down path @xmath201 of a @xmath189 01 matrix @xmath190 by @xmath202 note that the row indices @xmath203 are _ strictly _ increasing while the column indices @xmath204 are _ weakly _ decreasing . in @xmath205 , the position of @xmath206 is always located on the down side or left - down side of the position of @xmath207 in the 01 matrix . we also define the quantity @xmath208 by @xmath209 where @xmath210 denotes the number of elements in the path @xmath205 . if we regard the 01 matrix as the random media in a plane , @xmath201 as a spatial configuration of a polymer chain , and @xmath211 as its energy , we can interpret as a statistical mechanical problem of a directed polymer . this kind of problem is studied in @xcite . now we find that this maximum length of the left - down path is directly related to the position of a tagged particle as @xmath212 where @xmath213 represents the number of times the @xmath2th particle stays at the same site from time 0 through @xmath214 . the equation ( [ prop1 ] ) enables us to study the dynamics of the tasep in terms of the 01 matrices and plays a fundamental role in our subsequent analysis . we can show in the following way . for the case @xmath215 ( the case where we tag the first particle ) , we can easily check this equation , since @xmath216 defined in is simply the total number of the value 1 in the sequence @xmath217 and it is clear that this quantity represents the number of times the particle stays from @xmath218 through @xmath219 . for the case @xmath220 , we can prove by mathematical induction about the row @xmath219 . here , we mainly consider only the case @xmath221 . for the case @xmath222 , we can check in a similar fashion to this case . when @xmath223 , we can check easily for all four cases ( @xmath224 and @xmath225 ) . we assume that holds for @xmath226 , and @xmath215 and @xmath227 , i.e. , @xmath228 under this assumption , we will show that @xmath229 as follows . from , one easily finds @xmath230 where @xmath231 , and @xmath232 and @xmath227 . considering these properties , we classify the problem into two cases , ( i ) : : @xmath233 ( ii ) : : @xmath234 the case ( i ) corresponds to the situation where the two particles occupy the neighboring sites at @xmath235 . thus the second particle ( the tagged particle ) must stay at time step between @xmath236 and @xmath237 . hence one finds @xmath238 regardless of the value of @xmath239 . on the other hand , due to , and , we have @xmath240 thus from , and , we have for the case ( i ) , @xmath241 the case ( ii ) corresponds to the situation where at @xmath235 , the distance between the first and the second particles is at least one site . note that in this case we do not need to consider the exclusion effect of the first particle and thus the situation is essentially the same as the case @xmath215 . we have @xmath242 on the other hand , from and we can also check that @xmath243 thus we find that holds also for the case ( ii ) and therefore holds for @xmath221 . in the case @xmath244 , we can also classify the situation into two cases as in ( i ) and ( ii ) for @xmath221 , i.e. , the case ( i) where all @xmath75 particles are packed and ( ii) where the particles form several groups which are separated by some successive empty sites . for the case ( i) , we can check in a similar manner to the case @xmath221 . for the case ( ii) , we can easily show that the case is essentially the same as the case ( i) for @xmath245 which is smaller than @xmath75 . the quantity we would like to discuss is the multi - time distribution of @xmath14 , @xmath246 where @xmath247 represents the probability measure of the tasep defined in section [ model ] . due to the result in section [ 01table](especially and ) , one finds @xmath248 where @xmath249 represents the probability measure of 01 table defined in and @xmath250 is defined in . in this section , we further restate our problem in terms of the combinatorics of young tableaux . for the definitions and basic properties of young tableaux and related subjects , we refer the readers to @xcite . there is a bijective mapping between an @xmath189 01 matrix and a pair @xmath251 where @xmath252 ( the transpose of @xmath253 ) and @xmath254 are the semistandard young tableaux ( ssyt ) with the condition that the shape of @xmath253 is the same as that of @xmath254 . the mapping is called the dual robinson - schensted - knuth ( rsk ) algorithm . in order to get @xmath251 , we first construct a two - line array ( or generalized permutation ) , @xmath255 where @xmath256 , from the 01 matrix by listing @xmath257 s for which @xmath258 . next , we construct @xmath253 ( resp . @xmath254 ) by arranging on a plane the figures in the second row ( resp . the first row ) . note that each positive integer constructing @xmath253(resp . @xmath254 ) is less than @xmath2(resp . @xmath219 ) . for details of the algorithm , see @xcite . the generalized permutation and @xmath251 corresponding to the 01 table in fig . 5(b ) are given in fig . now we express @xmath208 as the quantity related to a pair @xmath251 . let us denote the shape of @xmath253 ( or @xmath254 ) by the young diagram , @xmath259 where @xmath260 represents the length of @xmath5th row of the tableaux @xmath253 ( or @xmath254 ) . in the example of fig . 6(b ) , we see @xmath261 . then we have @xmath262 here @xmath263 means the transpose of the young diagram @xmath264 . thus @xmath265 is equal to the length of the first column of @xmath266 . . may be shown by examining the dual rsk algorithm directly but is also understood as follows . for the sequence of the numbers @xmath267 where @xmath268 , we define a nondecreasing ( resp . nonincreasing ) subsequence @xmath269 such that @xmath270 ( resp . @xmath271 ) . from the construction it is not difficult to see that @xmath208 is equivalent to the length of the longest nonincreasing subsequence of the second row @xmath272 in a generalized permutation . thus , @xmath208 is also regarded as the length of the longest nondecreasing subsequence of the opposite sequence @xmath273 of the second row in . we can apply the ( normal ) rsk algorithm @xcite to this reverse sequence to obtain a ssyt . let us denote the ssyt by @xmath29 . then ( [ dl ] ) is a consequence of the facts that the length of the longest nondecreasing sequence in the reverse sequence is equal to the length of the first row of @xmath29 ( this is known as a property of the normal rsk algorithm @xcite ) and the symmetry property , @xmath274 whose proof can be found in appendix a of @xcite . notice that in the example of fig . 5(a ) , @xmath275 thus @xmath276 while in fig . 6(b ) , @xmath277 . thus from , we get @xmath278 here @xmath279 in the right hand side represents the probability measure of the set of young diagrams obtained by the dual rsk algorithm from the 01 matrix defined in and @xmath280 means the young diagram obtained by applying the dual rsk algorithm to the @xmath281 01 submatrix @xmath282 of the @xmath189 01 matrix . from and , we managed to interpret the time evolution of a tagged particle in the tasep as the time evolution of @xmath283 , the length of the first column , regarding @xmath284 as a time parameter . in order to investigate the right hand side of , we consider the following probability , @xmath285 for a given set of young diagrams , @xmath286 . this equation describes the growth of a young diagram . . represents the joint distribution of them . 7 illustrates the growth of young diagram which corresponds to fig . notice that from and , the growth of @xmath287 describes the time evolution of @xmath2th particle in the tasep . let us denote the @xmath288 as the part of the @xmath254 tableau obtained by picking up only the symbols @xmath289 . from the rule of the dual rsk algorithm @xcite , we find that @xmath288 is equivalent to the @xmath254 tableau obtained by applying the dual rsk algorithm to the the submatrix @xmath290 . thus one has @xmath291 where @xmath292 denotes the shape of @xmath293 . hence the @xmath254 tableau records the growth of young diagram , that is , it one - to - one corresponds to the set of the young diagrams @xmath294 . compare the tableau @xmath254 in fig . 6(b ) with @xmath295 in fig . 7 . we can recognize that @xmath292 constructed from the tableau @xmath254 in fig . 6(b ) is equivalent to @xmath296 in fig . hence we have @xmath297 here @xmath298 represents the probability measure on the set of the pairs @xmath299 obtained by the dual rsk algorithm from the @xmath189 01 matrices defined in . hence it is found that the growth process of the young diagram is characterized by the @xmath254 tableau in the dual rsk correspondence . this type of growth process has been also discussed in @xcite . remembering @xmath253 is constructed from the second row of a generalized permutation , we find that the symbol @xmath127 in @xmath253 represents the column index of the element of the 01 matrix where @xmath193=1 . thus the total number of @xmath127s in @xmath253 is equal to the total number of the figure 1 in the @xmath127th column of the 01 matrix . hence , for a given @xmath300 , one finds @xmath301 where @xmath302 means the total number of the symbol @xmath5 in @xmath300 . applying the above equation to the combinatorial definition of the schur function , @xmath303 where the summation is taken to all ssyt @xmath304 with shape @xmath305 , we have @xmath306 where @xmath307 . the factor @xmath308 ensures that the probability is zero unless @xmath309 and @xmath310 have no two squares in the same column . thus combining with , we finally find @xmath311 this measure is the special case of the joint distribution function in the schur process @xcite . since the schur function is expressed as a determinant , the function is described as a product of determinants . furthermore , it is known that the process has a remarkable mathematical structure that the correlation function of @xmath312 can be expressed as a determinant and its scaling limit can be discussed exactly . the schur process also appears in other fields of physics and mathematics such as random growth process @xcite , melting problem of a three - dimensional crystal @xcite , random tiling model @xcite , and so on . from , and , we find @xmath313 here @xmath314 represents the probabilistic measure of the schur process . in general , the schur process is defined as follows . for the set of the young diagrams such that @xmath315 the following weight is assigned @xmath316 here @xmath317 is the schur function with the specialization of algebra @xmath318 . in our case , we notice only two cases @xmath319 where @xmath320 is defined in . substituting to , @xmath321 where @xmath322 and applying theorem 1 in @xcite or theorem 2.2 in @xcite to this case , we finally obtain the fredholm determinant representation(- ) of . in this section , we discuss the scaling limit of the multi - time distribution function by applying the saddle point method to the kernel - in theorem 1 . we consider two types of scaling limit . the first case is explained in section [ sclim1 ] . we focus on the situation where both time @xmath4 and the label of the tagged particle @xmath2 tend to infinity when there are @xmath36 defect particles with the stay rates @xmath323 in front of the tagged particle . the results on the limiting distributions are summarized as theorem 2 . their proofs are given in section 5.1 , 5.2 and 5.3 . in the second case , we take @xmath4 to be infinite with @xmath2 fixed . the limiting distribution is described in theorem 3 in section [ sclim2 ] , whose proof is given in section 5.4 . in this section , we consider the scaling limit in the region 1 in fig . we scale time @xmath59 as . in order to calculate the limiting distribution , we rewrite the kernel - with the condition as @xmath324 where @xmath325 and @xmath326 . @xmath327 denotes the contour with radius @xmath29 enclosing the origin anticlockwise and @xmath328 satisfy the condition , @xmath329 for @xmath330 and @xmath331 for @xmath332 . using and the relation @xmath333 one finds @xmath334 here the function @xmath335 and @xmath336 are given by @xmath337 where the radius @xmath338 of the contour @xmath339 in is taken such that @xmath340 . in the following discussion , we evaluate the asymptotics of the kernel by applying the saddle point method to the functions @xmath341 . first we consider the asymptotics of @xmath342 . substituting with , we set @xmath343 where @xmath344 from the equation @xmath345 , we find the critical point @xmath346 is given by @xmath347 scaling the variable @xmath348 around the critical point as @xmath349 where @xmath350 , we evaluate the asymptotics of the term @xmath351 in by the saddle point method , @xmath352 as @xmath353 . next we consider the asymptotics of other terms in . from , one obtains @xmath354 transforming @xmath355 into @xmath356 , @xmath357 and considering , one gets @xmath358 note that by the transformation , the function @xmath359 in the fredholm determinant is changed to @xmath360 . from - , we obtain the asymptotic form of @xmath342 , @xmath361 where @xmath362 . next we consider the asymptotics of @xmath363 . similar to @xmath342 , we set @xmath364 here @xmath365 is defined in and @xmath69 denotes the contour enclosing the origin anticlockwise with radius 1 . applying the saddle point method to this equation in the same way as @xmath342 , we get @xmath366 thus from , , and , we find @xmath367 where @xmath368 introducing the function @xmath369 by @xmath370 one finds that the function satisfies weber s equation @xcite , @xmath371 the initial condition was given in @xcite as @xmath372 where @xmath373 represents the gamma function . the function satisfying is known as the parabolic cylinder function @xcite . in addition , one easily finds @xmath374 can be represented as from the integral representation of the hermite polynomial @xmath375 @xcite , @xmath376 hence , noticing the prefactor @xmath377 does not contribute to the determinant in , we get the limiting kernel - . in this section we consider the scaling limit of the regions 2 and 3 in fig . we fix the scaled time @xmath35 such that @xmath378 for the region 2 and @xmath108 for the region 3 , where @xmath379 . we scale the time @xmath101 , the position @xmath184 of the tagged particle at @xmath101 , and the stay rates of the defect particles @xmath380 as , and respectively . first we analyze the asymptotics of @xmath381 in by the saddle point method . using the variable @xmath382 which will be fixed later , we set @xmath383 here @xmath384 , @xmath325 , and @xmath326 . the function @xmath385 is defined as @xmath386 we fix the value of @xmath382 in such a way that two saddle points of @xmath385 merge to one point . we have @xmath387 and the double critical point @xmath388 of @xmath385is obtained as @xmath389 note that @xmath390 is satisfied with @xmath382 fixed as . thanks to the property that @xmath391 when @xmath392 which is the condition of regions 2 and 3 , it is found that we can deform the contours @xmath393 around the double saddle point . thus we scale @xmath394 and @xmath395 as @xmath396 where @xmath117 is defined in . noting and the relation @xmath397 we have @xmath398 hence we had the asymptotic form of the term @xmath399 in . next we consider that of other terms in . combining with the relation @xmath400 we find @xmath401 furthermore , associated with , we scale @xmath402 as @xmath403 here @xmath50 is defined in . using , and , we get @xmath404 note that in , the asymptotic form is different between regions 2 and 3 . this leads to the difference of the limiting distribution between theorems 2 - 2 and 2 - 3. thus we obtain the asymptotic form of @xmath405 , @xmath406 here @xmath407 in , we used the integral representation of the airy function @xmath408 and the relation @xmath409 next we consider the asymptotics of @xmath410 . using @xmath385 and @xmath382 , the function is rewritten as for @xmath332 , @xmath411 let the variable @xmath348 scale as @xmath412 where @xmath388 defined in is the double saddle point of . due to , we have for @xmath413 @xmath414 from the above two equations , and , we get @xmath415 thus we obtain for @xmath332 , @xmath416 where @xmath417 is given in . from we finally find @xmath418 hence from and and noting that the term @xmath417 does not affect the determinant , we finally obtain the desired kernels for the regions 2 and 3 . in this section , we discuss the asymptotics of the region 4 in fig . 3 . in this region we take time @xmath101 in such a way that they are macroscopically separated and the scaled time @xmath81 is taken as @xmath419 . the position @xmath184 of the tagged particle at @xmath101 and the stay rates @xmath380 of the defect particles are also scaled as and respectively . deforming the contour @xmath339 in the kernel @xmath420 , we divide it into two parts , @xmath421 here for @xmath413 @xmath422 where @xmath325 and @xmath326 . in the equation above , the contour @xmath423 encloses @xmath424 anticlockwise while the contour @xmath425 is chosen in such a way that it encloses @xmath426 anticlockwise . first , we discuss the asymptotic form of @xmath427 . this is rewritten as @xmath428 where @xmath429 is a variable which will be fixed later and @xmath430 we choose the value of @xmath429 in such a way that the saddle point of @xmath431 is @xmath432 where @xmath433 . from the condition , @xmath434 we have @xmath435 scaling @xmath436 around the saddle point as @xmath437 and noting @xmath438 we can obtain the asymptotic form of @xmath439 in by the saddle point method , @xmath440 associated with , we set @xmath441 for @xmath413 . from this equation , and , we find @xmath442 where we used the parameter @xmath443 defined in . thus we eventually get @xmath444 here the contour @xmath161 encloses @xmath445 anticlockwise and @xmath163 is an arbitrary path running from @xmath164 to @xmath165 . next we consider the asymptotics of @xmath410 . by use of @xmath446 and @xmath429 , we set for @xmath332 , @xmath447 scaling @xmath348 as @xmath448 from this equation and , we get @xmath449 thus we obtain for @xmath332 , @xmath450 at last , we discuss that @xmath451 in does not contribute the asymptotic form of the kernel . the derivation can be done in the similar fashion to the one in theorem 3.1 . in @xcite . here we give only its outline . first we consider the case of equal time , @xmath452 . scaling @xmath436 as , we have @xmath453 where @xmath454 . noticing that @xmath455 for the region 4 ( @xmath456 ) and the asymptotic form of the airy function @xmath457 as @xmath458 , we find @xmath459 this indicates that under the scaling , the kernel @xmath460 vanishes as @xmath2 goes to infinity . next we consider the case for arbitrary @xmath461 . we scale @xmath394 around the double saddle point @xmath462 of @xmath463 and @xmath395 around the saddle point @xmath432 of @xmath464 as @xmath465 under these scalings , we find @xmath466 where @xmath467 considering the case @xmath468 in and the former result , one easily finds @xmath469 hence we finally get @xmath470 from , and and noting the factor @xmath471 does not affect the determinant , we obtain the desired expression for the limiting kernel . in this section we discuss the scaling limit explained in section [ sclim2 ] where time @xmath4 goes to infinity with @xmath2 fixed . in this limit , we scale the time @xmath101 , the particle position @xmath184 , and the stay rates @xmath472 as , , and respectively . first we consider the scaling form of @xmath381 . changing the variables @xmath436 to @xmath473 , we have @xmath474 where the contour @xmath475 encloses @xmath476 anticlockwise , and @xmath339 encloses the origin @xmath477 and @xmath476 . introducing @xmath478 , which will be chosen later , we set @xmath479 where the parameter @xmath97 is defined in and @xmath480 we choose @xmath481 in a way that @xmath482 has the saddle point at @xmath483 where @xmath0 is defined in . thus from the condition , @xmath484 , we have @xmath485 we scale the variable @xmath486 around the saddle point @xmath487 as @xmath488 considering , we also scale @xmath402 as @xmath489 from these equations and , we find @xmath490 from these equations , we eventually obtain @xmath491 here the contour @xmath161 encloses @xmath492 anticlockwise and @xmath163 is arbitrary path from @xmath164 to @xmath165 . next we consider the scaling limit of @xmath410 . changing the variable @xmath348 to @xmath493 , we set @xmath494 where @xmath482 and @xmath481 are given in and . we set @xmath495 from this equation and @xmath496 combining these equations , we finally find @xmath497 thus from and and noting that the prefactor @xmath498 does not affect the determinant , we get the desired expression for the kernel . in this section , we give the result of monte - carlo simulations about the position fluctuation of a tagged particle in the tasep in order to check our analysis discussed in the preceding sections . we performed the simulations in two situations , the case without defect particle and that with one defect particle . 8 shows the data of the scaled position of a tagged particle obtained by the monte - carlo simulation and the probability distribution functions which must fit them . 8(a ) corresponds to the time region @xmath499 where @xmath0 ( resp . @xmath45 ) is the stay rate of the normal particle ( resp . the defect particle ) . note that in the region , the both situations belong to the region 2 and due to theorem 2 - 2 , the gue tracy - widom distribution must fit the data . 8(b ) represents the case @xmath108 . in this case the first situation comes under the region 2 . on the other hand , the second one belongs to the region 3 where the position fluctuation are described by the limiting largest eigenvalue distribution of goe@xmath3 as explained in theorem 2 - 3 . in fig . 8(c ) , only the data for the second situation are shown for the case @xmath500 . this case is classified as the region 4 and from theorem 2 - 4 , the fluctuation is supposed to be gaussian . in all figures , we see a good agreement between the data and the distribution functions which they must obey . in this article , we have discussed the multi - time distribution of the tagged particle . in order to analyze it , we have introduced the directed polymer problem of the 01 matrix in section [ ftts ] . here we consider the directed polymer problem of another random matrix where each element is geometric random variable . by this analysis we can discuss the correlations of other quantities in the tasep with the step initial condition . let @xmath501 be the @xmath189 matrix and the element @xmath502 is the geometric random variable , @xmath503 where @xmath504 is the parameter of the geometric distribution and we identified this with the stay rate of the @xmath127th particle of the tasep with the step initial condition defined in section [ model ] . for the matrix , we introduce the quantity @xmath505 analogous to @xmath208 as @xmath506 here @xmath507 is the set of right / down paths from @xmath508 to @xmath509 , @xmath510 in this setting , we consider the following quantity , @xmath511 it is known that @xmath505 can be interpreted as a quantity in the tasep as follows @xcite , _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ @xmath512 : in the tasep with the step initial condition , the time until which the @xmath2th particle moves @xmath219 sites to its right is @xmath4 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ thus the quantity represents the correlations of the arrival times of the @xmath2th particle at the site @xmath513 . ( note that we set the site coordinate as in fig . furthermore we easily find that @xmath514 here @xmath515 represents the number of particles which passed the site @xmath219 until time @xmath4 . thus the probability also means the correlations of currents between different times and different sites . applying the similar technique in section [ ftts ] to , we can also represent it as the growth process of young diagram characterized by the schur process . the result is as follows , @xmath516 where @xmath280 is the young diagram obtained by applying the normal rsk algorithm to the submatrix @xmath517 and @xmath518 is the length of its first row . the probability measure in the right hand side of this equation is characterized by the following joint distribution function , @xmath519 where @xmath520 is the schur function . this is also a special case of the schur process and we can get the fredholm determinant representation of . although in one point case , the asymptotics of this equation was discussed in @xcite , the multi - point distribution has not been discussed yet . in this article , we have studied the multi - time distribution function of position fluctuations of a tagged particle in the tasep with the step initial condition . the main results are summarized as theorems 1 , 2 and 3 in section [ main ] . first , we have obtained the fredholm determinant expression of in theorem 1 . for this purpose , we have mapped the time evolution of a tagged particle in the tasep to the growth process of young diagram which is related to the special case of the schur process . next , using the fredholm determinant in theorem 1 , we have studied the two types of scaling limit . the first one is the case where both time @xmath4 and the label of a tagged particle @xmath2 go to infinity . in the second one , we take the @xmath18 limit with @xmath2 fixed . the results for the first and second ones are shown in theorems 2 and 3 respectively in section [ main ] . in the first scaling limit , if the hopping rates of all particles are the same , we can divide the scaled time into two characteristic regions according to the limiting behavior of , the region 1 where a tagged particle begins to move and the region 2 which is after the region 1 . in the region 1 , the limiting process of a tagged particle converges to the spatially discrete process which is described in theorem 2 - 1 . the process reflects on the discreteness of the model . in the region 2 , on the other hand , we have shown in theorem 2 - 2 that the limiting process becomes the airy process , which is characteristic of the one - dimensional kpz universality class . if there are @xmath36 defect particles with small hoping rates in front of a tagged particle , the limiting distribution changes at the scaled time @xmath52 which is determined by the hopping rate of the slowest defect particle . the limiting process around @xmath52 ( region 3 ) is equivalent to that of the largest eigenvalue in gue dyson s brownian motion with rank one external source as described in theorem 2 - 3 . theorem 2 - 3 indicates that the rank becomes @xmath36 if the hopping rates of the defect particles are the same . when the scaled time is after @xmath52 ( region 4 ) , we have found in theorem 2 - 4 that the process is equivalent to the one dimensional brownian motion . if the hopping rates are degenerate , it is equivalent to the process of the largest eigenvalue of @xmath521 gue dyson s brownian motion model as described in theorem 2 - 4. this indicates that in the region 4 , the effect of the defect particles is dominant whereas that of an infinite number of normal particles is irrelevant . theorem 3 shows the result for the second scaling limit . the limiting distribution is also described as the fredholm determinant whose kernel is the same as that in theorem 2 - 4. the authors would like to thank a. borodin for drawing out attention to the reference @xcite . the work of t.i . is supported by core research for evolutional science and technology of japan science and technology agency . the work of t.s . is supported by the grant - in - aid for young scientists ( b ) , the ministry of education , culture , sports , science and technology , japan . fig . 1 : definition of the tasep with the step initial condition . ( a ) during each time step , the @xmath5th particle can hop to the right with probability @xmath8 . however , @xmath522th particle can not hop to the right neighboring site since the site is occupied by @xmath199th particle . ( b ) step initial condition . fig . 2 : typical time evolution of the particles from the first to 100th in the tasep with the step initial condition . the hopping rate of all particles is 0.9 ( @xmath523 ) except the four defect particles ( the first , 25th , 50th and 75th particles ) whose hopping rate is 0.8 ( @xmath524 ) . ( a ) illustrates the whole time evolution from @xmath218 through 3000 . its closeup around @xmath525 and @xmath526 is shown in ( b ) and ( c ) respectively . fig . 3 : average position of the tagged particle . the thick line shows . we divide this line into four regions , where @xmath35 is near @xmath54 , @xmath378 , @xmath35 is near @xmath52 , and @xmath456 . we denote them as region 1 , 2 , 3 , and 4 respectively . fig . 4 : example of the time evolution of the tasep and corresponding 01 matrices . in this case , two tables is assigned for one example of the time evolution . since the second particle can not hop to its right neighboring site between @xmath527 and @xmath528 , there are two possibilities of assigning both 0 and 1 in the @xmath529 element of the 01 matrix . fig . 7 : growth of young diagram obtained from fig . 5(b ) by the dual rsk algorithm . the diagrams @xmath531 are obtained from the submatrices @xmath532 of fig . the set of the diagrams @xmath533 corresponds to the @xmath254 tableaux in fig . fig . 8 : probability distributions of the scaled position of the 100th particle from the right ( @xmath534 ) for @xmath525 ( a ) , @xmath535 ( b ) and @xmath526 ( c ) . in these figures , @xmath536 represents the data for the case without defect particles and we set the stay rate @xmath523 . on the other hand , @xmath19 corresponds to the case where the first particle is a defect particle with @xmath524 while remaining particles are normal ones with @xmath523 . the number of samples are 10000 for each case . in ( a ) , both cases belong to the region 2 and are fitted into the gue tracy - widom distribution shown as the dashed line . in ( b ) , the second case belongs to the region 3 since @xmath537 where @xmath538 and it is described by the distribution denoted as goe@xmath3 while the first case remains in region 2 . in ( c ) , the second cases comes under the region 4 where the distribution is described by gaussian ( dashed line ) . ( 400,171 ) ( 0,160)(a ) ( 65,171)@xmath218 ( 160,171)@xmath539 ( 257,171)@xmath527 ( 65,147)@xmath540 ( 160,147)@xmath543 ( 257,147)@xmath544 ( 65,123)@xmath545 ( 160,123)@xmath546 ( 257,123)@xmath547 ( 65,98)@xmath548 ( 50,85 ) ( 0,70)(b ) ( 70,55)@xmath232 ( 70,40)@xmath541 ( 70,26)@xmath549 ( 70,11)@xmath550 ( 70,-3)@xmath551 ( 70,-18)@xmath552 ( 82,70)@xmath542 ( 118,70)@xmath227 ( 133,70)@xmath528 ( 147,70)@xmath553 ( 100,-20 )
the central parts of our galaxy were prospected by baade ( 1946 ) , in order to detect its nucleus , and to have an indication of the morphological type of our galaxy . the so - called baade s window was revealed , and from variable stars identified in the field , the bulge stellar population was identified to be similar to that defined in baade ( 1944 ) as population ii . in the early 60s the notion of a galactic bulge was already established ( e.g. courtes & cruvellier 1960 ) . mcclure ( 1969 ) found that the bulge stars were super metal - rich compared to the stars nearby the sun . whitford and rich ( 1983 ) confirmed the high metal content of bulge stars from individual star spectroscopy . catalogues prepared along the decades show a steady growth of overall samples of globular clusters ( hereafter referred to as gcs ) , together with photometric and spectroscopic information . cannon ( 1929 ) used the harvard plate spectra to give integrated spectral types of 45 globular clusters . mayall ( 1946 ) measured radial velocities of 50 globular clusters , and the integrated spectral types of 40 of them . kinman ( 1959 ) listed 32 gcs with indication of metallicity from their integrated spectra ; morgan ( 1959 ) directed efforts on a relatively bright sample of 13 bulge gcs , now known to be metal - rich . surveys with schmidt plates since the 60s provided the palomar and eso star clusters , among others . terzan ( 1968 and references therein ) significantly increased the number of central gcs by reporting new faint ones in the bulge direction . these discoveries provided an important sample of gcs in the bulge . other studies of gc overall samples were presented by van den bergh ( 1967 ) , zinn ( 1985 and references therein ) , bica & pastoriza ( 1983 ) . webbink ( 1985 ) provided a catalogue of 154 gcs and candidates , with an important impact on subsequent observational efforts . more recently , harris ( 1996 , updated in 2010 , hereafter harris10 ) reports properties for 157 gcs . it is interesting to see how the notion of gcs pertaining to the bulge and their spatial distribution evolved in the last decades . frenk & white ( 1982 ) found evidence that metal - rich gcs formed a bar - like structure . based on metallicity , scale height , and rotational velocities available at that time , armandroff ( 1989 and references therein ) interpreted a sample of low galactic latitude metal - rich gcs as belonging to a disk system . minniti ( 1995 ) instead , from metallicity and kinematics of gcs in the central 3 kpc ( about 20@xmath1 ) from the galactic center , found evidence for these gcs to belong to the bulge . ct ( 1999 ) derived metallicities and radial velocities from high resolution spectroscopy for a significant sample of gcs within 4 kpc from the galactic center . he concluded that metal - rich gcs are associated with the bulge / bar rather than the thick disk . an important step in the understanding of the nature of the bulge gcs was presented by ortolani et al . ( 1995 ) , where the metal - rich gcs ngc 6528 and ngc 6553 were found to have an old age , identical to the bulk of the bulge field stars and comparable to that of the halo clusters . barbuy et al . ( 1998 ) derived new results and summarized the properties of 17 gcs in the bulge projected within 5@xmath1 of the galactic center . they concluded that these clusters shared comparable properties with the bulge field stars , including not only metal - rich gcs but also intermediate metallicity ones . they also found that there are no clusters in a strip 2.8@xmath1 wide centred at about 0.5@xmath1 south of the galactic plane . several comprehensive recent reviews have addressed the stellar populations , both field and gcs , in the galactic bulge ( e.g. harris 2001 ; rich 2013 ; gonzalez & gadotti 2015 ) . these reviews address also comparisons of the galactic bulge with external galaxies . in the following we describe recent advances on detailed chemical abundances , distances , kinematical properties and hints on possible association with subsystems in the central region of the galaxy . we also prepare for the future by defining a bulge gc sample , which includes suggestions for future studies , in terms of unstudied objects . in particular it is clear that proper motion derivation is still needed for most gcs . in the past , angular distances from the galactic center were the basis for selecting bulge gcs in the central parts of the galaxy ( e.g. zinn 1985 ; barbuy et al . 1998 ) . as a first approach we selected clusters with angular distances below 20@xmath1 . our next step was to use the galactocentric distances of the clusters , given that now they are more reliable - see harris10 . we adopted a distance to the galactic center of 8 kpc ( bobylev et al . 2014 ; reid et al . 2014 ) . for selecting the clusters we have experimented different distances from the galactic center of 6 , 5 , 4 and 3 kpc . in each case , we checked for bulge clusters and halo intruders . we conclude that a cutoff of 3 kpc is best in terms of isolating a bona - fide bulge cluster sample , with little contamination . we finally applied a metallicity filter : zoccali et al . ( 2008 ) , hill et al . ( 2011 ) , ness et al . ( 2013 ) , rojas - arriagada et al . ( 2014 ) have shown that the lower metallicity end of the bulge is around [ fe / h]=-1.3 . this is confirmed with the findings by pietrukowicz et al . 2012 , 2015 ) , dkny et al . ( 2013 ) and lee ( 1992 ) , all of them having demonstrated that there is a peak of rr lyrae with [ fe / h]=-1.0 , centrally concentrated and spheroidal ( except for pietrukowicz et al . 2015 that found it to be elongated ) , corresponding to an old bulge . we verified that gcs with [ fe / h ] @xmath2 -1.5 corresponded to well - known halo clusters in most cases . criteria using the space velocity v , as proposed by dinescu et al . ( 1997 , 1999a , b , 2003 ) , and casetti - dinescu et al . ( 2007 , 2010 , 2013 ) , are only feasible if proper motions are available , besides radial velocities . in particular dinescu et al . ( 2003 ) verified the classification of clusters as members of different galaxy components in terms of kinematics . the use of this criterion is possible for about a third of the gcs , at the present stage . gaia results in a few years will bring higher precision , and new data for a number of gcs . a few groups are applying proper motion cleaning to bulge clusters ( e.g. dinescu et al . 2003 ; rossi et al . therefore carretta et al . ( 2010 ) criterion of space velocity v values to select stellar populations , is not possible presently for having only the radial velocity in many cases , but may be applicable in a few years . in table [ tab1 ] we present the selected bulge clusters , following the criteria explained above . djorgovski 1 is added to the list , because it was in the barbuy et al . ( 1998 ) s list , and the distance available in harris10 , based on infrared data , is probably overestimated . the clusters are ordered by right ascension , to be compatible with harris10 . the distances from the sun are from harris10 mainly based on the visual magnitude of the hb , while valenti et al . ( 2007 ) give distances based on isochrone fitting in jhk . we also report distances derived from the galactic center , as given by harris10 . for kronberger 49 , not listed in harris10 , data are from ortolani et al . the distance derivation is crucial in order to get the position of the clusters relative to the galactic center and the bar . in most cases , for a cluster located inside the bar , its orbit is trapped , given the high mass of the bar . in fact more than half the bulge mass is in the bar , and for this reason , the bulge clusters that show low kinematics are trapped ( rossi et al . 2015a ) . given the old age of the clusters it is likely that they were formed before the bar , and later trapped . the dynamical behaviour of such clusters was illustrated for hp 1 ( ortolani et al . 2011 ) , and for 9 gcs by rossi et al . ( 2015b ) , all of them showing a low maximum height . in table [ tab2 ] we list possible intruders to our main selection , as well as limiting cases , and we collected a few more gcs with other distance and metallicity criteria . subsamples are classified as follows : \a ) probable halo intruders with [ fe / h]@xmath2 - 1.5 : we found 6 gcs with distances to the galactic center smaller than 3 kpc ( our selected bulge volume ) , but with metallicity lower than [ fe / h]@xmath2 - 1.5 , and besides with very high spatial velocities ( dinescu et al . 1999a , 2010 ) , all pointing to perigalactic locations of halo gcs . \b ) outer bulge shell with distances 3@xmath2r@xmath24.5 kpc : this surrounding bulge shell does probably contain true bulge gcs , that should be near apogalacticon . interestingly , the few space velocities availabe ( dinescu et al . 1999a , cudworth & hanson 1993 ) for this sample are rather low . \c ) outer bulge shell intruders : five gcs are located in this shell b ) , but showing low metallicities , as for example m22 , having also a high velocity typical of halo , like in group a ) . \d ) metal - rich gcs ( [ fe / h]@xmath3 - 1.0 ) beyond r@xmath34.5kpc : this sample corresponds to the ones that conveyed the idea of `` disk globular clusters '' ( e.g. armandroff 1989 ) . many of these key and well - known gcs do not have space velocities derived so far . this would be of great interest to constrain the possibility to link them to the thick disk . \e ) intruders to d ) . data for fsr 1767 are from bonatto et al . ( 2007 , 2009 ) . \f ) little studied gcs without enough parameters : probably key objects for future analysis ( sect . [ sect7 ] ) terzan 5 was identified to have at least two stellar populations and given its high mass , it was proposed to be a stripped dwarf galaxy ( ferraro et al . 2009 ; massari et al . saracino et al . ( 2015 ) concluded that liller 1 is as massive as terzan 5 or @xmath4 centauri . these massive clusters have absolute magnitudes around m@xmath0@xmath5 - 10 . on the other hand , the faintest bulge gcs are as faint as m@xmath0@xmath5 - 4 , such as terzan 9 and al 3 . these magnitudes are comparable to those of the ultra - faint galaxies ( e.g. mcconnachie 2012 ; bechtol et al . 2015 ) . by multiple stellar populations we refer to different metallicities [ fe / h ] in a same cluster . for most clusters there are hints of two populations , but no difference in metallicity . a comprehensive discussion on na - o anticorrelation was given by gratton et al . ( 2015 ) , where differences between red and blue horizontal branch stars allow to derive some important conclusions . see further discussions on multiple populations revealed by na - o anticorrelations in sect . since this effect appears to be present for most clusters , this is incorporated in the definition of globular clusters , and it is not a concern in the present work . distances of bulge gcs are mainly based on the horizontal branch ( hb ) luminosity level . the calculation of distances depends on 3 basic inputs : ( 1 ) the hb absolute magnitude , ( 2 ) the reddening , and ( 3 ) the reddening law . these three datasets slightly depend on the metallicity . the hb absolute magnitude may also depend on the he abundance . the current statistics is based mostly on the distances reported in harris10 , where the hb level is given by the relation : m(hb)=0.16[fe / h]+0.84 . in harris10 if the hb level is not available , specific references are used . this relation is somewhat different from that used by barbuy et al . ( 1998 ) in their catalogue of 17 inner bulge clusters , adopted from jones et al . ( 1992 ) : m(hb)=0.16[fe / h]+0.98 . this latter relation gives an hb level of about 0.14 magnitudes fainter , producing as well smaller distance moduli , but this difference is negligible when compared to other uncertainties . it is also important to point out that the main difference between harris10 and barbuy et al . ( 1998 ) values , is that the latter was based on optical cmds only . the list of distances also makes use of the jhk colour - magnitude diagrams ( cmds ) when it is considered more reliable . finally the catalogue by bica et al . ( 2006 ) uses basically the same assumptions as harris10 . _ reddening and reddening law : _ the reddening is a key parameter in the derivation of the distances based on the photometric technique . for most of the low latitude bulge gcs , an error of 5% in reddening produces a typical error of @xmath615% in the absorption , which propagates with the same fraction to the distance modulus . in harris10 the reddening has been derived from webbink ( 1985 ) , zinn ( 1985 ) and reid et al . the standard reddening law ( r@xmath0=3.1 ) has been used to derive the visual absorption . bica et al . ( 2006 ) uses basically the same input values , but they converted e(b - v ) into a@xmath0 using r@xmath0=3.1 for clusters with [ fe / h]@xmath2 - 1.0 and r@xmath0=3.6 for [ fe / h]@xmath3 - 1.0 following grebel and roberts ( 1995 ) , and r@xmath0 values have been interpolated in the intermediate metallicity interval . in principle this choice should produce shorter distances than harris10 for high metallicity clusters . a different computation was performed for most of the inner bulge clusters presented in barbuy et al . a1 of dean et al . ( 1978 ) was used to convert e(v - i ) to e(b - v ) and then the reddening dependence of r@xmath0 on e(b - v ) as given in olson ( 1975 ) has been used to convert e(b - v ) into a@xmath0 : r@xmath0=3.1 + 0.05([fe / h ] ) . independent distances and reddening values have been obtained by valenti et al . ( 2007 ) for 37 bulge clusters , using infrared jhk photometry . the main advantage of the jhk derived distances is that the reddening vs. absorption ratio is almost constant even if r@xmath0 varies ( fitzpatrick 1999 ; cardelli et al . the comparison of the optical vs. ir distances give a chance to probe the reddening law , in the optical regime , in the direction of the galactic bulge . figure [ fig2 ] shows the distance differences between different r@xmath0 values versus reddening . the absolute magnitude of the horizontal branch m(hb ) is assumed to be of m(hb)=0.68 for [ fe / h]=-1.0 , at e(b - v)=0 . therefore at the distance of the galactic center , assumed here to be of 8 kpc , the distance modulus is m - m=14.5 , and m(hb)=14.5 + 0.68=15.18 . so for the zero point we assume : e(b - v)=0 , m(hb)=15.18 . we use 12 clusters in common with valenti et al . ( 2007 ) , and adding terzan 9 ( ortolani et al . 1999 ) . from this sample , an average distance difference of d(ir)-d(optical ) = 0.65 kpc has been derived . in the infrared sample we have also 3 clusters ( terzan 4 , terzan 5 and ngc 6528 ) with data from hst / nicmos ( ortolani et al . 2007 ) . in these cases we adopted an average value between nicmos and valenti et al . ( 2007 ) . at an average distance of 8 kpc , the 0.65 kpc difference is equivalent to about 0.3 mag . in distance modulus . conclusions from fig . 2 are : a ) it is clear that the optical data produce shorter distances , in particular at high reddening ; b ) in order to get comparable infrared and optical distances , we have to adopt an average total - to - selective absorption r@xmath0=3.2 . this average value is slightly lower than that adopted in barbuy et al . ( 1998 ) of r@xmath0=3.39 ( see their table 4 ) , but it is still higher than the standard r@xmath0=3.1 value , and it is in agreement with recent studies of the reddening law in different conditions of reddening , metallicity and intrinsic colours of the stars ( hendricks et al . 2012 ; mccall 2004 ) . a further test can be performed following racine & harris ( 1989 ) and barbuy et al . ( 1998 ) , plotting v(hb ) as a function of reddening . assuming that , on the average , the inner bulge clusters ( within 3 kpc from the galactic center ) are concentrated around this distance , the hb level should be related to the reddening by means of a slope r@xmath0 . this is plotted in fig . [ fig2 ] , which is the updated version of fig . 5 in barbuy et al . this analysis includes 12 clusters plus an arbitrary point at e(b - v)=0 and v(hb)=15.18 , which should correspond to the v(hb ) at 8.0 kpc from the sun , with [ fe / h]=-1.0 , m@xmath0=0.8 , as explained above . therefore the best fit is 3.3 @xmath2 r@xmath0 @xmath23.1 . the higher slope of r@xmath0=3.6 seems too steep . this confirms that a choice of an average reddening value of r@xmath0=3.2 is adequate and that the baricentric distance from the sun of the considered clusters should be around 8.0 kpc . finally , we address a few comments on the distance of uks 1 , since there is disagreement between different authors . we employed 4 sets of data : a ) nicmos data ( ortolani et al . 2001 , 2007 ) , measured relative to ngc 6528 , assumed to be at a distance of d@xmath7 = 7.7 kpc , results in ( m - m)@xmath8= 14.43 . assuming that both clusters have a similar metallicity , and the hst / nicmos reddening law , we get d@xmath7 = 15.8 kpc for uks 1 ; b ) nicmos data measured relative to liller 1 , assuming for liller 1 a distance of d@xmath7 = 8.1 kpc ( from ( m - m)@xmath8=14.55 , saracino et al . 2015 ) ; given a @xmath9(m - m)@xmath8=1.02 , we get ( m - m)@xmath8=15.57 and a distance of d@xmath7 = 13.0 for uks 1 ; c ) assuming minniti et al . ( 2011 ) absolute reference for the hb colour and magnitude m@xmath10=-1.65 , j - k=0.71 ( from a study of red clump calibrated stars with hipparcos data ) , and using nicmos calibrated data , we obtain j@xmath11=17.96 . from m(hb)@xmath12 = -0.94 and a@xmath12=2.78 , obtained from the comparison with liller 1 ( saracino et al . 2015 ) , we get ( m - m)@xmath8=18.63 - 2.78=15.85 , and d@xmath7 = 14.8 kpc ; d ) from liller 1 ( saracino et al . 2015 ) , but with a recalculated absolute distance from minniti et al . ( 2011 ) values of m(hb)@xmath10 and j - k(hb ) , we have : ( m - m)@xmath8(liller 1)=14.85(d@xmath7 = 9.33 kpc ) , and ( m - m)@xmath8=14.85@xmath131.02=15.87 and d@xmath7 = 14.9 kpc for uks 1 . all these methods give an average of d@xmath7 = 14.6 kpc for uks 1 , as reported in table [ tab1 ] . the metallicity distribution of bulge clusters as given in table [ tab1 ] is shown in fig . it shows a peak around [ fe / h]@xmath5 - 1.0 , suggesting that this population is intrinsically significant . that such an old bulge stellar population is important , is confirmed by studies of rr lyrae by lee ( 1992 ) , dkny et al . ( 2013 ) , pietrukowicz et al . ( 2012 , 2015 ) . also , from the metallicity distribution function of field stars by zoccali et al . ( 2008 ) , ness et al . ( 2013 ) , and rojas - arriagada et al . ( 2014 ) , this moderate metallicity corresponds to the lower end of the metallicity distribution function ( mdf ) of the bulge bulk field stellar population . table [ tab3 ] shows the chemical abundances for a subset of gcs from table [ tab1 ] , that have available high spectral resolution abundance analyses . _ carbon and nitrogen : _ c and n show the expected anticorrelation due to cno processing along the red giant branch . two clusters with low n abundances should be further studied since this is not expected . _ odd - z elements na , al _ a crucial issue concerns possible na - o anticorrelation , which would indicate the presence of a second stellar generation . as a matter of fact , most gcs are presently being found to have at least two stellar generations , except possibly the least massive clusters . a threshold in mass for a second generation not to occur is presently estimated to be at 3 - 4@xmath1410@xmath15 m@xmath8 ( r.g . gratton , private communication ) , or in other words , essentially all gcs should show the na - o anticorrelation . the second generation is detected via na - o , mg - al anticorrelations , and also by the presence of both cn - strong and cn - weak stars ( carretta et al . the origin of these chemical anomalies is probably hot bottom burning ( hbb ) ( ventura et al . 2013 ) in massive asymptotic giant branch ( agb ) first generation stars , with yields ejected in the internal cluster medium , and incorporated by the second generation stars , the latter showing these anomalies . a thorough discussion on the origin of multiple stellar populations in globular clusters is given in renzini et al . ( 2015 ) . from table [ tab3 ] we see that very few of the bulge clusters were investigated in terms of na and al . in fact , four stars analysed by barbuy et al . ( 2014 ) show no na - o anticorrelation ( see their fig . 6 ) , however more stars have to be analysed for a firm conclusion . an important investigation on this matter was carried out by gratton et al . ( 2015 ) , where 17 bhb and 30 rhb stars of ngc 6723 were analysed . it was found that rhb and intermediate - bhb stars appear to belong to a first generation , showing o - rich and na - poor abundances , whereas the bluest of the bhb stars show lower oxygen and higher na ( four bhb stars show [ o / fe]=+0.23 and [ na / fe ] = + 0.11 ) , in contrast with the mean values given in table [ tab3 ] . gratton et al . ( 2015 ) consider that extended blue hb stars might correspond to a second generation of lower mass stars . _ alpha - elements o , mg , si , ca , ti _ in fig . [ fig7 ] are plotted the abundances of these alpha elements vs. [ fe / h ] , including the abundances for the bulge sample , as given in table [ tab3 ] , compared with abundances for 57 field bulge giant stars from lecureur et al . ( 2007 ) , gonzalez et al . ( 2011 ) and barbuy et al . ( 2015 ) , and 58 bulge dwarf stars from bensby et al . the oxygen abundances are as given in barbuy et al . ( 2015 ) , where they were revised with respect to lecureur et al . some discrepancy between the oxygen abundances from barbuy et al . ( 2015 ) and bensby et al . ( 2013 ) can be explained by the facts that : barbuy et al . analysed red giants , using the forbidden [ oi]6300.31 line and bensby et al . ( 2013 ) analysed dwarfs , using the permitted triplet oi lines at 7771.94 , 7774.16 , 7775.39 lines . solar oxygen abundances adopted were respectively @xmath16(o ) = 8.87 and 8.85 for barbuy et al . and bensby et al . , which would tend to invert the small shift between the two sets results . all in all , given that the permitted oi lines are well - known to be subject to strong non - lte effects , and tend to overestimate the oxygen abundances , we can consider that there is a good agreement between the two sets of oxygen abundances . the panels in fig [ fig7 ] indicate therefore that the alpha - elements o , mg , si and ca are overabundant by about [ o , mg , si , ca , ti / fe]@xmath50.3 to 0.4 dex for the more metal - poor stars . the same is found for the field ( see also alves - brito et al . 2010 ; bensby et al . 2013 ) . this implies early fast enrichment by core collapse supernovae snii , which in turn give a short timescale for bulge formation , where even the metal - rich clusters are old ( ortolani et al . 1995 ) . _ heavy elements _ very few heavy element abundance derivation is available for individual stars of bulge gcs . the heavy elements of first peak y , sr , zr , together with a few elements of the second peak ba , la , and the r - element eu , can reveal the nature of the first stars , or else to reveal if agbs were acting as important chemical contributors . a threshold of about [ ba / eu]=0.6 would be indicative if ba was produced by r- or s - process . if [ ba / eu]@xmath30.6 , this could be a hint of early enrichment by massive spinstars , and the same applies to the ratios y / ba , sr / ba , and zr / ba ( frischknecht et al . 2015 ) . these ratios were studied for the globular cluster ngc 6522 , by barbuy et al . ( 2009 ) , chiappini et al . ( 2011 ) , and ness et al . ( 2014 ) using giraffe spectra from the survey by zoccali et al . barbuy et al . ( 2014 ) used new uves spectra observed in 2012 for four of the same stars , and these new results are reported in table [ tab3 ] . sr lines are unreliable as shown by barbuy et al . ( 2014 ) , and ness et al . ( 2014 ) , superseding abundance values given in chiappini et al . zr is difficult to derive : in table [ tab3 ] it is not reported given that it is only available for ngc 6522 , but it is worth mentioning that cantelli et al . ( 2015 , in preparation ) seems to detect a clear variation in zr abundances among 12 member stars of ngc 6522 observed with giraffe in 2012 . the ratios from first to second peak of heavy elements can also be explained by production of s - process heavy elements in massive agb stars , and predicted relative ratios between different heavy elements was presented in bisterzo et al . ( 2010 , 2013 ) . finally , it is possible that all heavy elements in old stars were produced by the r - process only , as first suggested by truran ( 1981 ) . . [ fig1 ] [ cols= " < , > , > , > , > , > , > , > , > , > , > , > , > , > , > , > , > , > , > " , ] [ tab4 ] in this review we provide a state - of - the - art list of bulge globular clusters and their properties . we tried to constrain their properties including their kinematics when available . kinematics is becoming a key tool to identify their nature , and the link with the field stellar populations , within the complex substructures of the bulge . in recent years good progress in the knowledge of the globular clusters in the central parts of the galaxy has been achieved . for spectroscopy the multi - object spectrographs in 8m - class telescopes have made it possible to derive chemical abundances for considerable numbers of stars . progress in instrumentation thanks to e.g. imaging with mcao in the infrared ( mad / vlt , gems / gemini ) and hst / nicmos , made possible long time baseline of ccd data with excellent quality , allowing proper motion cleaning of cmds . finally , surveys with larger apertures such as the 4 m vista telescope , provided the discovery of new objects . the vvv survey in particular has provided several new bulge globular clusters and candidates , to be explored in coming years . much work is still needed as concerns bulge globular clusters , such as the monitoring of variable stars , in particular rr lyrae , to obtain deep cmds allowing age derivation , and the derivation of metallicities and chemical abundances from high - resolution spectroscopy . the derivation of na and o abundances will be crucial to define if these very old clusters have a unique or multiple stellar generations . these results should allow to better compare the bulge globular clusters with outer bulge , inner halo and outer halo ones . alves - brito , a. , barbuy , b. , zoccali , m. et al . 2006 , a&a , 460 , 269 alves - brito , a. , melndez , j. , asplund , m. , ramrez , i. , yong , d. 2010 , a&a , 513 , a35 armandroff , t. 1989 , aj , 97 , 375 athanassoula , e. 2005 , mnras , 358 , 1477 baade , w. 1944 , apj , 100 , 137 baade , w. 1946 , pasp , 343 , 249 babusiaux , c. , gmez , a. , hill , v. et al . 2010 , a&a , 519 , 77 babusiaux , c. , katz , d. , hill , v. et al . 2014 , a&a , 563 , 15 barbuy , b. , bica , e. , ortolani , s. 1998 , a&a , 333 , 117 barbuy , b. , zoccali , m. , ortolani , s. , et al . 2006 , a&a , 449 , 349 barbuy , b. , zoccali , m. , ortolani , s. , et al . 2007 , aj , 134 , 1613 barbuy , b. , zoccali , m. , ortolani , s. , et al . 2009 , a&a , 507 , 405 barbuy , b. , chiappini , c. , cantelli , e. et al . 2014 , a&a , 570 , 76 barbuy , b. , friaa , a. , da silveira , c.r . 2015 , a&a , 580 , 40 bechtol , k. , drlica - wagner , a. , balbinot , e. et al . 2015 , apj , 807 , 50 bensby , t. , yee , j.c . , feltzing , s. et al . 2013 , a&a , 549 , 147 bica , e. , pastoriza , m. 1983 , apssci , 91 , 99 bica , e. , bonatto , c. , barbuy , b. , ortolani , s. 2006 , a&a , 450 , 105 bisterzo , s. , gallino , r. , straniero , o. et al . 2010 , mnras , 404 , 1529 bisterzo , s. , travaglio , c. , wiescher , m. et al . 2013 , arxiv : 1311.5381 blitz , l. , spergel , d.n . 1991 , apj , 379 , 631 bobylev v. v. , mosenkov , a.v . , bajkova , a.t . , gontcharov , g.a . 2014 , astl , 40 , 86 bonatto , c. , bica , e. , ortolani , s. , barbuy , b. 2007 , mnras , 381 , l45 bonatto , c. , bica , e. , ortolani , s. , barbuy , b. : 2009 , mnras , 397 , 1032 borissova , j. , bonatto , c. , kurtev , r. et al . 2011 , a&a , 532 , 131 borissova , j. , chen , a. n. , ramirez alegria , s. , et al . 2014 , a&a , 569 , 24 bureau , m. , aronica , g. , athanassoula , e. et al . 2006 , mnras , 370 , 753 cannon , a.j . 1929 , bharo , 868 , 1 cardelli , j.a . , clayton , g.c . , mathis , j.s . 1989 , apj , 345 , 245 carretta , e. , cohen , j.g . , gratton , r.g . , behr , b.b . 2001 , aj , 122 , 1469 carretta , e. , bragaglia , a. , gratton , r.g . 2009 , a&a , 505 , 117 carretta , e. , bragaglia , a. , gratton , r.g . 2010 , a&a , 516 , 55 casetti - dinescu , d.i . , girard , t.m . , herrrera , d. et al . 2007 , aj , 134 , 195 casetti - dinescu , d.i . , girard , t.m . , korchagin , v.i . 2010 , aj , 140 , 1282 casetti - dinescu , d.i . , girard , t.m . , jlkov , l. et al . 2013 , aj , 146 , 33 clarkson , w. , sahu , k. , anderson j. et al . 2008 , apj , 684 , 1110 clarkson , w. , sahu , k. , anderson , j. et al . 2011 , apj , 735 , 37 chiappini , c. , frischknecht , u. , meynet , g. et al . 2011 , nature , 472 , 454 cohen , j.g . , gratton , r.g . , behr , b.b . , carretta , e. 1999 , apj , 523 , 739 ct , p. 1999 , aj , 118 , 406 courtes , g. , cruvellier , p. 1960 , pohp , 5 , 3 cudworth , k.m . , hanson , r.b . 1993 , aj , 105 , 168 dean , j.f . , warpen , p.r . , cousins , a.j . 1978 , mnras , 183 , 569 dkny , i. , minniti , d. , catelan , m. et al . 2013 , apj , 776 , l19 dinescu , d.i . , girard , t.m . , van altena , w.f . 1997 , aj , 114 , 1014 dinescu , d.i . , van altena , w.f . , girard , t.m . 1999a , aj , 117 , 277 dinescu , d.i . , girard , t.m . , van altena , w.f . 1999b , aj , 117 , 1792 dinescu , d.i . , girard , t.m . , van altena , w.f . , lpez , c.e . 2003 , aj , 125 , 1373 feltzing , s. , johnson , r.a . 2002 , a&a , 385 , 67 ferraro , f.r . , dalessandro , e. , mucciarelli , a. et al . 2009 , nature , 462 , 483 fitzpatrick e.l . 1999 , pasp , 111 , 63 frenk , c.s . , white , s.d.m . 1982 , mnras , 198 , 173 frischknecht , u. , hirschi , r. , pignatari , m. et al . 2015 , mnras , submitted gratton , r.g . , lucatello , s. , sollima , a. et al . 2015 , a&a , 573 , a92 grebel , e.k . , roberts , w.j . 1995 , a&as , 109 , 293 gonzalez , o.a . , rejkuba , m. , zoccali , m. et al . 2011 , a&a , 530 , 54 gonzalez , o.a . , gadotti , d.a . 2015 , arxiv:1503.07252 , to appear in _ galactic bulges _ , eds . laurikainen e. , peletier r. , gadotti d. harris , w.e . 1996 , aj , 112 , 1487 harris , w.e . 2001 , star clusters , saas - fee advanced courses , volume 28 . isbn 978 - 3 - 540 - 67646 - 1 . springer - verlag berlin heidelberg , 2001 , p. 223 hendricks , b. , stetson , p.b . , vandenberg , d.a . , dallora , m. 2012 , apj , 144 , 25 hill , v. , lecureur , a. , gmez , a. et al . 2011 , a&a , 534 , 80 ivanov , v. d. ; kurtev , r. ; borissova , j. 2005 , a&a , 442 , 195 jones , r.v . , carney , b.w . , storm , j. , latham , d.w . 1992 , apj , 386 , 646 kinman , t.d . 1959 , mnras , 119 , 538 lecureur , a. , hill , v. , zoccali , m. et al . 2007 , a&a , 465 , 799 lee , j .- w . , carney , b.w . , balachandran , s.c . 2004 , aj , 128 , 2388 lee , y .- w . 1992 , aj , 104 , 1780 lee , j.w . 2007 , revmexaa , 28 , 120 lee , y .- w . , joo , s .- chung , c. 2015 , arxiv:1508.05942 martinez - valpuesta , i. , shlosman , i. , heller , c. 2006 , apj , 637 , 214 massari , d. , mucciarelli , a. , ferraro , f.r . 2014 , apj , 795 , 22 mayall , m. u. 1946 , apj , 104 , 290 mccall , m. 2004 , aj , 128 , 2144 mcclure , r.d . 1969 , aj , 74 , 50 mcconnachie , a.w . 2012 , aj , 144 , 4 mcwilliam , a. , zoccali , m. 2010 , apj , 724 , 1491 melndez , j. , barbuy , b. , bica , e. et al . 2003 , a&a , 411 , 417 minniti , d. 1995 , aj , 109 , 1663 minniti , d. , zoccali , m. 2008 , iau symp . 245 , p. 323 minniti , d. , hempel , m. , toledo , i. et al . 2011 , a&a , 527 , 81 moni bidin , c. , mauro , f. , geisler , d. et al . 2011 , a&a , 535 , 33 morgan , w.w . 1959 , aj , 64 , 432 muoz , r. r. , geha , m. , ct , p. et al . 2012 , apj , 753 , l15 nataf , d.m . , udalski , a. , gould , a. , fouqu , p. , stanek , k.z . : 2010 , apj , 721 , l28 ness , m. , freeman , k. , athanassoula , e. et al . 2013 , mnras , 432 , 2092 ness , m. , asplund , m. , casey , a.r . 2014 , mnras , 445 , 2994 ness , m. , freeman , k. , athanassoula , e. et al . 2013 , mnras , 430 , 836 olson , b.i . 1975 , pasp , 87 , 349 origlia , l. , rich , r.m . , castro , s. 2002 , aj , 123 , 1559 origlia , l. , rich , r.m . 2004 , aj , 127 , 3422 origlia , l. , valenti , e. , rich , r.m . 2005b , mnras , 356 , 1276 origlia , l. , valenti , e. , rich , r.m . , ferraro , f. 2005b , mnras , 363 , 897 origlia , l. , valenti , e. , rich , r.m . 2008 , mnras , 388 , 1419 origlia , l. , rich , r.m . , ferraro , f. et al . 2011 , apj , 726 , l20 origlia , l. , massari , d. , rich , r.m . et al . 2013 , apj , 779 , l5 ortolani , s. , renzini , a. , gilmozzi , r. et al . 1995 , nature , 377 , 701 ortolani , s. , bica , e. , barbuy , b. 1999 , a&as , 138 , 267 ortolani , s. , barbuy , b. , bica , e. et al . 2001 , a&a , 376,878 ortolani , s. , barbuy , b. , bica , e. , zoccali , m. , renzini , a. 2007 , a&a , 470 , 1043 ortolani , s. , bonatto , c. , bica , e. , barbuy , b. : 2009 , aj , 138 , 889 ortolani , s. , barbuy , b. , momany , y. et al . 2011 , apj , 737 , 31 ortolani , s. , bonatto , c. , bica , e. et al . 2012 , aj , 144 , 147 pealoza , f. , pessev , p. , vsquez , s. et al . 2015 , pasp , 127 , 329 pfenniger , d. 1984 , a&a , 134 , 373 pietrukowicz , p. , udalski , a. , sosynski , i. et al . 2012 , apj , 750 , 169 pietrukowicz , p. , kozlowski , s. , skowron , j. et al . 2015 , arxiv:1412.4121 racine , r. harris , w.e . 1989 , aj , 98 , 1609 reid , m.j . , menten , k.m . , brunthaler , a. et al . , 2014 , apj , 783,130 renzini , a. , dantona , f. , cassisi , s. et al . 2015 , mnras , submitted rich , r.m . 2013 , planets , stars and stellar systems vol . 5 , by oswalt , terry d. ; gilmore , gerard , isbn 978 - 94 - 007 - 5611 - 3 . springer science+business media dordrecht , 2013 , p. 271 rojas - arriagada , a. , recio - blanco , a. , hill , v. et al . 2014 , a&a , 569 , 103 rossi , l. , ortolani , s. , bica , e. , barbuy , b. , bonfanti , a. 2015a , mnras , 450 , 3270 rossi , l. , ortolani , s. , casotto , s. , barbuy , b. , bica , e. , 2015b , mnras , in preparation saracino , s. , dalessandro , e. , ferraro , f.r . 2015 , apj , 806 , 152 smith , h.a . , wehlau , a. 1985 , apj , 298 , 572 terzan , a. 1968 , pohp , 10 , 4 truran , j. 1981 , a&a , 97 , 391 valenti , e. , ferraro , f. , origlia , l. 2007 , aj , 133 , 1287 valenti , e. , origlia , l. , rich , r.m . 2011 , mnras , 414 , 2690 valenti , e. , origlia , l. , mucciarelli , a. , rich , r.m . 2015 , a&a , 574 , 80 van den bergh , s. 1967 , aj , 72 , 70 vsquez , s. , zoccali , m. , hill , v. et al . 2013 , a&a , 555 , 91 ventura , p. , di criscienzo , m. , carini , r. , dantona , f. 2013 , mnras , 431 , 3642 webbink , r.f . 1985 , iau symposium 113 , p. 541 wegg , c. , gerhard , o. , portail , m. 2015 , mnras , 450 , 4050 whitford , a.e . , rich , r.m . 1983 , apj , 274 , 723 zinn , r. 1985 , apj , 293 , 424 zoccali , m. , renzini , a. , ortolani , s. et al . 2001 , aj , 121 , 2638 zoccali , m. , barbuy , b. , hill , v. et al . 2004 , a&a , 423 , 507 zoccali , m. , renzini , a. , ortolani , s. et al . 2003 , a&a , 399 , 931 zoccali , m. , hill , v. , lecureur , a. et al . 2008 , a&a , 486 , 177
the abundance - metallicity relation of low mass extremely metal - poor ( emp ) stars ( @xmath4 \la -3 $ ] ; @xmath5 \equiv \log_{10 } ( n_{\rm x}/n_{\rm y } ) - \log_{10 } ( n_{\rm x}/n_{\rm y})_\odot$ ] where @xmath6 and @xmath7 are the abundances of elements x and y , respectively ) and very metal - poor ( vmp ) stars ( @xmath8 \la -2 $ ] ) has been clarified by observations of very high quality spectra ( e.g. , * ? ? ? * ; * ? ? ? this relation is expected to provide the information of supernovae ( sne ) of population ( pop ) iii ( the first generation ) massive stars ( e.g. , * ? ? ? * ) and the first stage of galactic chemical evolution ( gce ) ( e.g. , * ? ? ? the emp stars are considered to have suffered the injection of heavy elements from one or a few sne ( e.g. , * ? ? ? a halo interstellar medium becomes gradually homogeneous with metallicity @xcite . although recent studies of the star formation in the first stage of gce have been in progress , the characteristics of first generation stars , such as initial mass function , have not been clarified . in order to clarify such characteristics , we should investigate nucleosynthesis in the candidates of the first generation stars as well as the observed variations of the abundance distributions of the emp and vmp stars . recent observations have indicated that the abundance ratios [ x / fe ] averaged in the metal - poor stars with @xmath4 \la -2 $ ] are between @xmath9 and 0.5 for most of observed elements @xcite . their scatter around the average values deduced from the new observation is much smaller than found in earlier studies . the elemental abundance distribution of the metal - poor stars is not so different from that of the solar - system composition . therefore , the small scatter is considered to suggest primordial burst of high - mass stars or the very rapid mixing of matter from different bursts of early star formation . from theoretical viewpoint , @xcite compared the pop iii hypernova ( hn ) yields with the abundance pattern of emp stars with the metallicity range @xmath10 < -3.5 $ ] in @xcite . they adopted the mixing - fallback model for the hne . the hn yields give good agreement with the observed abundances for c , na , mg , al , si , ca , ni , and zn . they also compared the pop iii sn and hn yields integrated over the salpeter initial mass function ( imf ) with the abundances of vmp stars in the metallicity range @xmath11 < -2.0 $ ] in @xcite . they reproduced the observed trends of abundance ratios for c , o , na , mg , al , si , ca , cr , ni , and zn . the sn and hn yields of n , k , sc , ti , mn , and co are short for explaining the observed yield ratios , although sc and ti may be improved in the high - entropy explosion models . the yields of mn and co may be reproduced by the modification of electron fraction @xmath12 , of which value is still uncertain in the innermost region of sne and hne . additional nucleosynthesis processes also remain open for these elements . during a sn explosion , the collapsed core becomes extremely high temperature and high density , and neutrinos are produced through pair creation . the neutrinos are almost thermalized by neutrino - nucleus scattering at the center . they pass away through the neutrino sphere with carrying the gravitational binding energy of the core . the neutrino irradiation lasts for about @xmath13 s. the total number of neutrinos emitted in a sn is very huge , i.e. , about @xmath14 . the neutrinos interact with nuclei in the surrounding stellar material to produce new species of nuclei . this is called the @xmath0-process . the @xmath0-process is expected to be an important production process for some elements in pop iii sn explosions . the @xmath0-process is important for the synthesis of light elements , li and b @xcite , f ( e.g. , * ? ? ? * ; * ? ? ? * ) , and some heavy neutron - deficient nuclei such as @xmath15la and @xmath16ta @xcite . the @xmath0-process products are synthesized through the spallations by neutrinos from abundant seed nuclei . it also produces protons and neutrons and their capture reactions also enhance the abundances of some elements . for pop iii sne , the seed nuclei are @xmath17-nuclei abundantly produced in complete and incomplete si burning during the explosions . thus , the @xmath0-process in pop iii sne will produce additional some odd-@xmath1 elements . we note that most of the nucleosynthesis calculations including the @xmath0-process have been carried out using spherical explosion models @xcite . although the total neutrino energy should be close to the gravitational binding energy released from the collapsing core , properties of the neutrinos are still uncertain . the properties strongly depend on explosion mechanism but details of the mechanism have not been clarified . on the other hand , aspherical explosions have been investigated recently ( e.g. * ? ? ? * ; * ? ? ? in such a case , aspherical structure of neutrino sphere can be formed and the properties of the neutrinos are also still uncertain . therefore , the dependence of the abundances of pop iii sne and hne on neutrino properties should be investigated . in this study , we focus on the effect of the @xmath0-process on the production of odd-@xmath1 iron - peak elements , sc , mn , and co , during pop iii sn and hn explosions . we calculate the sn nucleosynthesis including the @xmath0-process . then , we investigate the dependence of the abundances of sc , mn , and co in the sn ejecta on the energy of neutrinos , the stellar mass , and the explosion energy . sc , mn , and co are expected to be at least partly synthesized through the @xmath0-process . we also indicate whether the abundance ratios of sc , mn , and co to fe reproduce those observed in the emp stars when we take into account the @xmath0-process . in 2 we explain stellar evolution models and the explosion models of sne and a hn . we also describe the nucleosynthesis models including the @xmath0-process and the neutrino irradiation models . in 3 we show the abundance distributions of sc , mn , and co in the ejecta of the sn and hn models . we explain additional production of these elements through the @xmath0-process . in 4 we explain the effect of the @xmath0-process on the abundance ratios of sc , mn , and co to fe and the dependence on the neutrino irradiation strength . we also compare our results with the abundance ratios observed in the emp stars . in 5 we discuss other production processes of sc , mn , and co proposed in recent studies and the effect of the @xmath0-process . finally , we summarize this study in 6 . we calculate the evolution of population iii stars for initial masses of 15 @xmath2 and 25 @xmath2 from zero - age main - sequence to the onset of the core collapse . we use a henyey - type stellar evolution code including a nuclear reaction network with about 300 isotopes to calculate detailed nucleosynthesis and energy generation . the input physics is the same as the models used in @xcite . sn explosions are calculated using a hydrodynamical code of spherically symmetrical piecewise parabolic method @xcite . we set the explosion energy of the 15 and 25 @xmath2 sn models to be @xmath18 , where @xmath19 is the explosion energy in units of @xmath20 ergs . we also calculate the explosion of a hn for the 25 @xmath2 star model with @xmath21 . the location of the mass cut of the sn explosions is determined as follows . for the 15 @xmath2 star model , we set the location of the mass cut at @xmath22 ( @xmath23 being the mass coordinate of a star ) to obtain the ejected @xmath24ni mass of 0.07 @xmath2 . in the case of 25 @xmath2 sn , we set the mass cut location at @xmath25 , so that large ( co , zn)/fe ratios are obtained . this location corresponds to the bottom of the layer where electron fraction @xmath12 is close to 0.5 @xcite . a large amount of @xmath24ni is produced in the region of @xmath26 ( see 3.2 for details ) . since the location of the mass cut is deep , this model ejects the @xmath24ni mass of 0.36 @xmath2 . this model corresponds to a sn with the explosion energy of @xmath27 and yielding a large @xmath24ni amount ( @xmath28(@xmath24ni)@xmath29 ) such as sn 2005bf @xcite . note that the mixing - fallback model is not applied to these cases . in the case of the 25 @xmath2 hn model , we apply the mixing - fallback model @xcite . this model approximates aspherical effects of hn explosions @xcite . during aspherical explosion , the central core of a star first collapses . then , jet - like explosion ejects the surrounding stellar materials . some of the materials in an inner region may fallback to the central core from off - axis direction of the jet . there are three parameters in the mixing - fallback model ; the initial mass cut @xmath30(ini ) , the outer boundary of the mixing region @xmath31(out ) , and the ejection factor @xmath32 . the initial mass cut corresponds to the mass of the initially collapsing core . the outer boundary of the mixing region corresponds to the location where the surrounding materials infall to the core during the explosion . the ejection fraction is the ejected material fraction in the mixing region . details of the mixing - fallback model have been described in @xcite . the values of the mixing - fallback parameters should be related to the explosion features such as asphericity and energy deposition rate @xcite . however , the explosion mechanism has not been clarified . therefore , we determine these parameter values similar to case a in @xcite . the location of the initial mass cut is set as @xmath33 , which is the bottom of the @xmath34 layer . this condition deduces large zn / fe ratio @xcite . the outer boundary of the mixing region is determined as @xmath35 where the mass fraction of @xmath24ni decreases to @xmath36 . the ejection factor is determined as @xmath37 in order to yield [ o / fe ] = 0.5 . the electron fraction in complete and incomplete si burning regions is not modified in this study . the @xmath24ni mass of each explosion model is listed in table 1 . a huge amount of neutrinos are emitted during sn explosions . for spherically symmetrical sn explosions , neutrinos are emitted isotropically from the neutrino sphere and carry the majority of gravitational energy of a proto - neutron star . on the other hand , aspherical sne and jet - like explosion would emit neutrinos anisotropically . in this case neutrino irradiation may be very strong in a specific direction . since we assume spherically symmetrical sn explosions , we assume isotropic neutrino emission . the strength of the neutrino flux is determined by the total neutrino energy . the adopted values of the total neutrino energy are @xmath38 ergs , @xmath39 ergs , and @xmath40 ergs . the first value corresponds to normal neutrino irradiation and is almost equal to the gravitational binding energy of a 1.4 @xmath2 neutron star @xcite . we assume that the neutrino luminosity decrease exponentially with time . the decay time of the neutrino luminosity is set to be 3.0 s. the spectra of the sn neutrinos are assumed to obey fermi distribution with zero chemical potentials . the temperatures of @xmath41 , @xmath42 and @xmath43 , @xmath44 are set to be @xmath45 mev/@xmath46 and 4 mev/@xmath46 , respectively . this set of the neutrino temperatures is the same as in @xcite and @xcite . we note that @xcite provided the sn yields adopting the @xmath41 and @xmath42 temperature of @xmath47 mev and @xmath48 ergs . therefore , we briefly mention the yields obtained using this parameter set . detailed nucleosynthesis during the sn explosions is calculated by postprocessing using a code of @xcite to which the @xmath0-process reactions in @xcite were added . we should note that the evolution of neutrino luminosity and energy spectra of hne may be different from that of sne . the neutrino emission from rapidly accreting disks surrounding black holes during gamma ray bursts and hne has been investigated ( e.g. , * ? ? ? they showed that @xmath49 and smaller amount of @xmath50 are mainly emitted from the accretion disks with slightly hard energy spectra compared with those from proto - neutron stars . however , the evolution of neutrino emission during hn explosions is still largely uncertain . on the other hand , the main contribution of the @xmath0-process is neutral - current interactions , which do not depend on neutrino flavors @xcite . furthermore , the yields of the @xmath0-process products depend on both neutrino energy spectra and the total neutrino energy . roughly speaking , a larger total neutrino energy corresponds to harder neutrino energy spectra @xcite . therefore , we set the total neutrino energy as a parameter . the value of total neutrino energy @xmath51 ergs may be too large even in the hn model . if an element is not produced abundantly through the @xmath0-process with @xmath51 ergs , the @xmath0-process contribution should be very small for the production of the element . ccccc no neutrinos & @xmath52 & @xmath53 & @xmath54 & @xmath55 + @xmath56 & @xmath52 & @xmath57 & @xmath58 & @xmath59 + @xmath39 & @xmath60 & @xmath61 & @xmath62 & @xmath63 + @xmath40 & @xmath64 & @xmath65 & @xmath66 & @xmath67 + @xmath56 ( @xmath68=8 mev ) & @xmath52 & @xmath69 & @xmath70 & @xmath71 + no neutrinos & @xmath72 & @xmath73 & @xmath74 & @xmath75 + @xmath56 & @xmath76 & @xmath77 & @xmath78 & @xmath79 + @xmath39 & @xmath80 & @xmath81 & @xmath82 & @xmath83 + @xmath40 & @xmath84 & @xmath58 & @xmath85 & @xmath86 + @xmath56 ( @xmath68=8 mev ) & @xmath76 & @xmath87 & @xmath63 & @xmath88 + no neutrinos & @xmath89 & @xmath90 & @xmath91 & @xmath92 + @xmath56 & @xmath89 & @xmath93 & @xmath94 & @xmath95 + @xmath39 & @xmath89 & @xmath96 & @xmath97 & @xmath98 + @xmath40 & @xmath99 & @xmath100 & @xmath101 & @xmath102 + @xmath56 ( @xmath68=8 mev ) & @xmath89 & @xmath103 & @xmath104 & @xmath83 + we explain the mass fractions of sc , mn , and co as well as the contribution of the @xmath0-process for their production . figure 1@xmath105 shows the distributions of the mass fractions of sc , mn , and co in the 15 @xmath2 sn model without taking into account the @xmath0-process . in this figure , the main product in the range of @xmath106 is @xmath24ni . in this region a large amount of iron - peak elements are produced through complete si burning . we call this region the ni layer . in the region outside the ni layer , incomplete si burning occurs during the explosion and a large amount of si as well as a smaller amount of iron - peak elements are produced . we call this region ( @xmath107 ) the si / s layer . the characteristics of the mass fractions and main production sites of sc , mn , and co without the @xmath0-process are explained as follows . 1 . small mass fraction of sc is produced at @xmath108 . the mass fraction is about @xmath109 at the maximum . most of sc is produced through incomplete si burning . most of mn is produced in the si / s layer through incomplete si burning . the mass fraction at the maximum is about @xmath110 , which is much larger than that of sc . a part of mn is also produced in the ni layer but the mass fraction is much smaller . co is also produced in the ni layer . the mass fraction of co is about @xmath111 throughout this layer . hereafter we consider the effect of the @xmath0-process on the production of sc , mn , and co. figures 1@xmath112 and 1@xmath113 show the distributions of the mass fractions taking into account the @xmath0-process with @xmath114 ergs and @xmath40 ergs , respectively . [ [ sc ] ] sc ^^ we see that additional sc is produced in the si / s layer ( @xmath115 ) and in the ni layer . the mass fraction of sc is less than @xmath109 . most of the sc is first produced as @xmath116ti and @xmath116v . in the ni layer , a large amount of @xmath117ti is produced through complete si burning . at the same time , @xmath116ti and @xmath116v are partly produced through neutron- and proton - captures of @xmath117ti . without the @xmath0-process , however , the produced @xmath116ti and @xmath116v are smoothly captured or decomposed again during the explosion . with the @xmath0-process , additional protons are produced through the @xmath0-process , so that some @xmath116v is produced again by the proton capture . [ [ mn ] ] mn ^^ we see the increase in the mass fraction of mn in the ni layer by about three orders of magnitude compared with the case without the @xmath0-process . the mass fraction in the si / s layer also increases by about a factor of 1.4 . during the sn explosion , mn is produced as @xmath118co . the amount of @xmath118mn originally produced in the sn is much smaller . in the si / s layer , @xmath118co is produced by @xmath119fe(@xmath120co or photodisintegration from @xmath24ni . when the @xmath0-process is considered , @xmath118co is also produced through @xmath24ni(@xmath121co after the production of @xmath24ni through complete and incomplete si burning . since the decay time of the neutrino luminosity is longer than the time scale of si burning , the neutrino irradiation proceeds after ceasing si burning . [ [ co ] ] co ^^ when we consider the @xmath0-process , the mass fraction of co increases by about a factor of three throughout the ni layer . the main production region of co is not changed when we consider the @xmath0-process . most of co is produced as @xmath122cu and about @xmath123 of @xmath122co is produced as @xmath122ni . we see increase in the mass fractions of @xmath122cu and @xmath122ni similarly in the ni layer . most of @xmath122cu and @xmath122ni are produced through the captures of protons and neutrons by @xmath124ni . the protons and neutrons are produced through the @xmath0-process even after the cease of complete si burning . therefore , the additional production raises the mass fraction of co. we also consider the case of larger neutrino irradiation with @xmath56 ergs . the distributions of the mass fractions are shown in fig . we see that the mass fractions of sc , mn , and co in the ni layer are larger by a factor of @xmath125 than those in the case of @xmath48 ergs . the dependence of the obtained mass fractions on the total neutrino energy is slightly smaller than the linear one . in this case all three elements are mainly produced in the ni layer although mn is mainly produced in the si / s layer in the case of the former cases . the mass fractions of the 25 @xmath2 sn model without the @xmath0-process is shown in fig . 2@xmath105 . in this figure , the ni layer where complete si burning occurs is in the range between @xmath126 and @xmath127 in the mass coordinate . the si / s layer which suffered incomplete si burning is between @xmath127 and @xmath128 . the ejected masses of these layers are larger than the corresponding layers in the 15 @xmath2 sn model . the distributions of the mass fractions of sc , mn , and co in the 25 @xmath2 sn model are roughly similar to those of the 15 @xmath2 sn model . most of sc is produced in outer part of the si / s layer ( @xmath129 ) . mn is mainly produced in the si / s layer , too . most of co is produced in the ni layer . the mass fractions of sc , mn , and co with the @xmath0-process of @xmath48 ergs are shown in fig . when we consider the @xmath0-process , their increase is seen in the ni layer . for sc , the mass fraction reaches about @xmath130 . it is about @xmath131 for mn but the main mn production is still in the si / s layer . increase in the abundance by the @xmath0-process is seen in the si / s layer . for co , the abundance increases by about a factor of two . the mass fractions of sc , mn , and co in the case of a larger neutrino irradiation ( @xmath51 ergs ) are shown in fig . we obtain larger ones for sc , mn , and co. the mass fraction of sc in the ni layer is about @xmath132 , which is larger than that of the 15 @xmath2 sn . for mn and co , the abundances in the ni layer are close to those of the 15 @xmath2 sn . the increase in the mn abundance in the si / s layer is also seen . we explain effects of the @xmath0-process on the production in the 25 @xmath2 hn model with @xmath133 . figure 3@xmath105 shows the distribution of the mass fractions in the 25 @xmath2 hn model without the @xmath0-process . the mass coordinate ranges of the ni and si / s layers are @xmath134 and @xmath135 , respectively . the mass contained in each of the two layers for the 25 @xmath2 hn is larger than that of the corresponding layer of the 25 @xmath2 sn model . we see that the mass fractions of sc , mn , and co in each burning region are not so different from those in the 25 @xmath2 sn ; most of sc and mn are produced in the si / s layer and co is produced in the ni layer . the average value of the co mass fraction in the ni layer is larger than that in the 25 @xmath2 sn . the distribution of the mass fractions in the 25 @xmath2 hn model with the @xmath0-process ( @xmath48 ergs ) is shown in fig . 3@xmath112 . increase in the mass fraction by the @xmath0-process is mainly seen for sc and mn . as shown in the cases of the sne , additional sc is produced in the ni layer . in addition , some mn is also produced in there . since the range of the ni layer is large , the contribution from the ni layer is larger than that from the si / s layer in this case . on the other hand , clear effect of the @xmath0-process is not seen for co. figure 3@xmath113 shows the mass fractions in the 25 @xmath2 hn model with larger neutrino irradiation ( @xmath51 ergs ) . as same as the sn case , the sc abundance in the ni layer is larger by about a factor of ten than that in the case of the normal neutrino irradiation ( @xmath48 ergs ) . the mass fraction of mn is similarly larger . in the si / s layer , the mn amount also increases . on the other hand , the mass fraction of co does not so change even with large neutrino irradiation . co is produced through complete si burning in the ni layer and most of it is not decomposed during the explosion . since hne produce large amount of co even without neutrino irradiation @xcite , the additional production through the @xmath0-process hardly affects the co production . we have shown that some amounts of sc , mn , and co are produced through the @xmath0-process and the following capture reactions of neutrons and protons . this additional production increases the total abundances of these elements in pop iii sn and hn explosions . hereafter we show the distributions of abundance ratios in the 15 @xmath2 and 25 @xmath2 sn and the 25 @xmath2 hn models . we also compare our results with the observed abundances of low mass emp stars . as a reference of the observed abundances , we use the abundance ratios to fe averaged in the data of 22 low mass halo stars with [ fe / h ] @xmath136 in @xcite . we show the abundance ratios in the 15 @xmath2 sn model in fig . the ratios of [ sc / fe ] , [ mn / fe ] , and [ co / fe ] are listed in table 1 . the abundance ratios of the three elements are much smaller than the corresponding solar ratios . for other elements , the abundance ratios of odd-@xmath1 elements are smaller than those of the even-@xmath1 elements in the neighbors . when we consider the @xmath0-process , the ratios [ sc / fe ] , [ mn / fe ] , and [ co / fe ] increase by @xmath137 dex ( see table 1 ) . thus , the @xmath0-process raises the abundances of these elements . since the fe abundance is not altered by the @xmath0-process , the increase in the abundance ratios is due to the increase in the abundances by the @xmath0-process . larger irradiation of neutrinos enhances the abundances of these elements . in the case of @xmath40 ergs , the abundance ratios for mn and co are larger than the corresponding solar ratios ( see table 1 and fig . the abundance ratios of v , mn , and co are larger than those of the even-@xmath1 elements ti , cr , and fe , respectively . we see that [ sc / fe ] , [ mn / fe ] , and [ co / fe ] for @xmath138 ergs and @xmath40 ergs increase by @xmath139 dex and @xmath140 dex , respectively , compared with those with @xmath48 ergs . the total neutrino energies with @xmath141 ergs and @xmath40 ergs are three times and ten times larger than that with @xmath142 ergs . simple consideration leads to that the abundances of odd-@xmath1 elements produced through the @xmath0-process are proportional to the neutrino flux , i.e. , the total neutrino energy . numerical simulations of the @xmath0-process show that this property is approximately correct . the reaction rates of the @xmath0-process are proportional to the total neutrino energy . at the same time , production rates of protons and neutrons produced through the @xmath0-process are also proportional to the total neutrino energy . the increase in the production rates brings about the increase in the production of odd-@xmath1 elements through proton- and neutron - captures . however , decomposition rates of the odd-@xmath1 elements through proton- and neutron - captures also increase . thus , the abundances produced through the @xmath0-process increase less than proportional to the total neutrino energy . we compare the obtained [ sc / fe ] , [ mn / fe ] , and [ co / fe ] with the corresponding observed results . the observed [ sc / fe ] averaged in 22 low mass emp stars is 0.06 . this value is slightly larger than the solar ratio and is still much larger than the calculated abundance ratio even in the case of the largest neutrino irradiation . for mn , the observed [ mn / fe ] is @xmath143 . this abundance ratio is larger than the ratio without the @xmath0-process and is reproduced by the result with the @xmath0-process of the normal neutrino irradiation ( @xmath48 ergs ) . on the other hand , the mn / fe ratio in the case of @xmath51 ergs is much larger than the largest value in the observed ratios . the observed [ co / fe ] is 0.31 , which is larger than the solar ratio . it is larger than the value in the sn model even with the largest neutrino irradiation . figure 5 shows the abundance ratios of the 25 @xmath2 sn model . when we do not consider the @xmath0-process , the abundance ratios of sc , mn , and co are much smaller than the corresponding solar ratios ( see also table 1 ) . the @xmath0-process also enhances the production of these elements . when we consider the @xmath0-process , the sc / fe ratio in this model is larger than that in the 15 @xmath2 sn model . this is because the seed nucleus of @xmath116v , i.e. , @xmath117ti , is abundantly produced in complete si burning for the 25 @xmath2 sn . however , the sc / fe ratio is smaller than the observed ratio even in the case of the largest neutrino irradiation . on the other hand , the mn / fe ratio in this model well reproduces the observed ratio in the case with the @xmath0-process of normal neutrino irradiation . the model with the largest neutrino irradiation ( @xmath51 ergs ) overproduces mn . the observed co / fe ratio is reproduced by the 25 @xmath2 sn model with the largest neutrino irradiation @xmath51 ergs ; large neutrino irradiation is required similarly to the 15 @xmath2 sn case . figure 6 shows the distribution of the abundance ratios of the 25 @xmath2 hn model with @xmath144 . in this model we considered mixing - fallback as described in 2.1 . when we do not consider the @xmath0-process , the [ sc / fe ] and [ mn / fe ] are @xmath145 and @xmath146 , respectively ( see also table 1 ) . the mn / fe ratio is much smaller than the corresponding ratios in the other two models . mn is mainly produced in the si / s layer without the @xmath0-process ( see fig . 3@xmath105 ) , whereas the range of the ni layer in the hn is larger than in the 25 @xmath2 sn . on the other hand , the co / fe is larger than those in the other sn models ( see also table 1 ) . as shown in fig . 3@xmath105 , the hn provides a large amount of co through complete si burning . the mass fraction of co in the ni layer is of order @xmath147 for the 15 @xmath2 and 25 @xmath2 sn models , and @xmath148 for the 25 @xmath2 hn . when we consider the @xmath0-process , the abundance ratios of sc and mn increase with the total neutrino energy . however , the co abundance scarcely changes even taking into account the @xmath0-process . the @xmath0-process raises the abundances of sc and mn in the ni layer . although it also enhances the co abundance there , the contribution from the @xmath0-process is hindered by large co production through complete si burning . we also compare our results with the observed ones . the obtained sc / fe ratio is smaller than the observed ratio . even in the case of the largest neutrino irradiation , the abundance ratio is smaller by about 0.5 dex . the mn / fe ratio is marginally consistent with the observed ratio in the largest neutrino irradiation case . at the same time , the co / fe ratio is also marginally consistent with the observed ratio despite the @xmath0-process ; the smallest co / fe ratio is reproduced . the hn model reproduces the observed mn / fe ratio when the @xmath0-process with large neutrino irradiation is taken into account . it reproduces the observed co / fe ratio even without the @xmath0-process . when we do not consider the @xmath0-process , the abundance ratios of sc / fe , mn / fe , and co / fe in the pop iii sne are smaller than the corresponding observed ratios . on the other hand , the observed mn / fe ratio is reproduced by the 15 @xmath2 and 25 @xmath2 sn models with the @xmath0-process . the observed co / fe ratio is marginally reproduced by the 25 @xmath2 hn model even without neutrino irradiation . thus , we can say that the @xmath0-process in the pop iii sne produce most of mn observed in low mass emp stars . the @xmath0-process also produces sc but the abundance is still short of the reproduction of the observed sc / fe ratios . hne may provide co being abundant enough to reproduce the observed co / fe ratios . we briefly mention the abundance ratios in the case of the @xmath0-process with @xmath47 mev and @xmath48 ergs , which has been adopted in @xcite . the abundance patterns in the 15 and 25 @xmath2 sn and the 25 @xmath2 hn models are shown as dotted lines in figs . the values of [ sc / fe ] , [ mn / fe ] , and [ co / fe ] are tabulated in table 1 . the yields of sc , mn , and co in this neutrino parameter set are larger by about 0.1 dex than those in the case of @xmath45 mev . they are still smaller than those in the case of @xmath45 mev and @xmath150 ergs . therefore , the dependence on the neutrino temperature is not strong for the sc , mn , and co yields . the abundance pattern of vmp stars are considered to reflect partly mixed interstellar materials in the early universe . recently , the averaged abundance pattern of vmp stars with @xmath11 < -2.0 $ ] in @xcite has been reproduced by the pop iii sn and hn yields integrated over the salpeter imf function @xcite . however , the evaluated [ sc / fe ] and [ mn / fe ] are still short of the observed ratio . we discuss the enhancement of these ratios by the @xmath0-process . @xcite evaluated the sc / fe ratio as [ sc / fe ] = @xmath151 whereas the observed value is [ sc / fe ] = @xmath152 . in this study , we showed that the @xmath0-process raises the sc yield by 1.1 dex at the maximum . however , this enhancement is still short for reproducing the observed value . production sites other than the @xmath0-process is needed to reproduce it ( see 5.2 ) . the [ mn / fe ] ratio observed in vmp stars is @xmath153 . the evaluated value of [ mn / fe ] is @xmath154 . when we take into account the @xmath0-process , this value increases by @xmath155 dex with @xmath48 ergs and @xmath156 dex with @xmath150 ergs . the increase by the @xmath0-process with @xmath157 ergs will provide better fitting of the pop iii sn and hn yields to [ mn / fe ] in vmp stars . the pop iii sn and hn models in @xcite indicate [ co / fe ] close to the lowest ratio in vmp stars ; [ co / fe ] = @xmath158 . taking into account the @xmath0-process with @xmath48 ergs , [ co / fe ] increases by @xmath159 dex in sne and it does not in hne . this enhancement is enough to reproduce the average [ co / fe ] ratio in vmp stars . the @xmath0-process of the pop iii sne and hne is one of the preferable sites of the mn and co in vmp stars . we have shown that sc , mn , and co are produced through the @xmath0-process in the pop iii sne and hne . for other than mn , the @xmath0-process in the sne and hne does not produce large enough abundances to reproduce the abundances observed in low mass emp stars . so , we will discuss other synthesis processes proposed in recent studies . [ [ sc-1 ] ] sc ^^ _ aspherical explosions _ as shown in previous studies ( e.g. , * ? ? ? * ) , spherically symmetrical sn and hn models do not provide enough amount of sc to reproduce the observed abundance of low mass emp stars . on the other hand , aspherical explosion and jet - like explosion produce larger amount of sc @xcite . in the case of spherical explosions , low - density structure enables to supply larger amount of sc @xcite . thus , higher temperature and low density , which are realized by aspherical explosions , are one of the favorable environments to produce sc . the amount of sc produced in sne and hne is also sensitive to @xmath12 in complete si burning region . large @xmath12 value ( @xmath160 ) brings about a large amount of sc @xcite . _ nucleosynthesis in innermost region of sne _ detailed thermal evolution of the innermost region of sn ejecta , i.e. , the neighbor of the `` mass cut has not been solved strictly . however , recent progress of hydrodynamical calculations has gradually revealed thermal evolution in such a deep region . at the same time , detailed nucleosynthesis in such a deep region has also been investigated . @xcite investigated the nucleosynthesis in convective bubbles in the innermost region of sn ejecta and in early stage of neutrino - driven winds . they used results of two - dimensional hydrodynamical calculations by @xcite . @xcite examined the simulations of one - dimensional self - consistent treatment of core - collapse sne with modified neutrino cross sections to explode . they also investigated the nucleosynthesis in the innermost region just above the mass cut produced self - consistently . both of the studies showed that the electron fraction @xmath12 exceeds 0.5 in such regions , which enhances the production of sc . in the latter study , they took into account all charged - current weak - interactions that changes @xmath12 in the ejecta . they obtained that the amount of sc with @xmath161 is produced in the innermost region with @xmath160 . the sc / fe ratio evaluated in their study is slightly smaller than the averaged abundances observed in low mass emp stars @xcite . thus , the sc / fe ratio is larger than our result . they also showed that a significant amount of nuclei with mass of @xmath162 in this region , especially light @xmath163-nuclei such as @xmath164mo and @xmath165ru ( the @xmath166-process ; * ? ? ? * ) . it is noted that the region of @xmath160 is convectively unstable so that the region will mix with outer smaller @xmath12 region in dynamical time scale . however , they expected that the @xmath12 remains high in an average sense . indeed , the former study took into account the convection in two - dimensional simulations and obtained @xmath160 . we also note that neutral - current @xmath0-process reactions were not included in @xcite . they also pointed out that neutrino - induced spallation reactions can change the final abundances of some nuclei . we expect that these reactions enhance the abundance of sc and that the sc / fe ratio becomes close to the observed ratio . it is also noted that materials blown off in neutrino - driven winds in early time after a sn explosion can achieve large @xmath12 value . accordingly , the @xmath167-process can take place in the materials . this synthetic process is strongly preferable for the production of the light @xmath163-nuclei @xcite . however , the contribution of the @xmath167-process in neutrino - driven winds to sc production is small . [ [ mn-1 ] ] mn ^^ the amount of mn produced in sne and hne is strongly sensitive to electron fraction @xmath12 in incomplete si burning region @xcite . their hn models modifying @xmath12 value to @xmath168 produce enough amount of mn to reproduce the observed one in low mass emp stars . however , hn models , whose @xmath12 in incomplete si burning region is not modified , indicated that the amount of mn is less than the observed value . aspherical explosions suppress mn production @xcite . we have shown that the mn production during sn explosions is mainly due to the @xmath0-process rather than to the incomplete si burning . the @xmath0-process with normal neutrino irradiation ( @xmath48 ergs ) in sne reproduces the mn / fe ratio averaged in low mass emp stars even without @xmath12 modifications . in the case of @xmath150 ergs , the mn amount marginally reproduces the upper limit of the observed mn abundance . in the case of @xmath51 ergs , however , the sn models overproduce mn . thus , the @xmath0-process in the pop iii sne is one of the main production processes of mn observed in low mass emp stars . the observed abundance constrains the total neutrino energy @xmath169 in sne . the observational constraint is @xmath170 ergs . [ [ co-1 ] ] co ^^ a large amount of co can be produced through complete si burning in aspherical explosion @xcite . for spherical explosions , @xcite pointed out that the co abundance is significantly enhanced for @xmath160 in the case of hne . however , the co / fe ratio in @xcite is still smaller than the observed ratio in @xcite . their model corresponds to a normal sn explosion . therefore , hne ( @xmath171 ) with @xmath172 produce co of which amount reproduces observed co / fe ratio in low mass emp stars . neutral - current @xmath0-process reactions may help additional production in the deep region just above the mass cut . however , in the case of hne , large amount of protons are also produced through @xmath17-rich freeze out . this effect may hinder the proton production by the @xmath0-process . the nucleosynthesis in pop iii sne have been conducted in several groups . here , we briefly compare with the abundances of sc , mn , and co of 15 @xmath2 and 25 @xmath2 pop iii sn models in @xcite , hereafter abbreviated to ww95 , ( z15a and z25b models ) and in @xcite , abbreviated to cl04 . the sn nucleosynthesis models in ww95 include the @xmath0-process . the temperature of @xmath41 in ww95 is assumed to be 8 mev , which is larger than the one adopted in this study . the sn nucleosynthesis with updated physics inputs of ww95 has been shown in @xcite but the pop iii sn nucleosynthesis has not been calculated . the nucleosynthesis models in cl04 do not include the @xmath0-process . the explosion models in cl04 were set to be the ejected mass of @xmath24ni equal to 0.1 @xmath2 . @xcite indicate [ sc / fe]=@xmath98 for @xmath173 model ; the sc / fe is much larger than our result . on the other hand , cl04 indicates sc / fe slightly larger than our results of sn models without the @xmath0-process ( [ sc / fe]=@xmath174 for 25 @xmath2 model ) . although the sc / fe in ww95 is much larger than the other models , it is still smaller than the observed ratio . therefore , other synthesis processes such as discussed above are necessary for sc production . the mn / fe shown in ww95 is smaller than the observed ratio ( [ mn / fe]@xmath175 ) . the 15 @xmath2 model in cl04 also indicates the mn / fe smaller than the observed one ( [ mn / fe]@xmath176 ) . the 25 @xmath2 model in cl04 indicates larger mn / fe ratio , which is comparable to the observed one ( [ mn / fe]@xmath177 ) . we showed that the @xmath0-process increases the mn abundance , but difference of stellar evolution models may affect the abundance . the 15 @xmath2 models of ww95 and cl04 show [ co / fe ] of @xmath178 and @xmath179 , which are larger than the ratio of our sn models . their 25 @xmath2 models indicate the co / fe similar to our results without the @xmath0-process . the evaluated co / fe ratios are still short of the observed ratio , so that hne will have an important role for co production in the early universe . among li , be , and b , @xmath180li and @xmath181b are mainly produced through the @xmath0-process . when the @xmath0-process is not considered , the li / fe and b / fe ratios are much smaller than the corresponding solar ratio . sn explosions with the neutrinos of @xmath182 ergs bring about li yield ratio of [ li / fe ] @xmath183 . they also produce b of which yield ratio to fe is larger than the solar ratio . therefore , sn explosions may contribute to the b production in the early universe . the @xmath180li is mainly produced as @xmath180be in the he - rich region . the @xmath181b is produced as @xmath181c and @xmath181b in c - enriched oxygen layer and c - enriched he layer . the li and b production in hne is less effective than that in sne . higher maximum temperature in the he layer decomposes @xmath180be and @xmath181c produced through the @xmath0-process and the following @xmath17-captures . in addition , strong explosion makes the region of c - rich oxygen layer small . [ [ f ] ] f ^ the @xmath0-process is a main f production process in sne @xcite . the f / fe ratio including the @xmath0-process is much larger than the one without the @xmath0-process . the seed nucleus of f in the @xmath0-process is @xmath184ne . most of f is produced in o and ne - enriched region . stellar mass dependence of the f / fe ratio is small in the sn models . on the other hand , the hn model shows smaller f / fe ratio . the hn explosion brings about stronger explosive ne burning , so that the o and ne - enriched region becomes smaller . figures @xmath149 show that the @xmath0-process scarcely affects the yields of na and al . furthermore , the yield ratios are much smaller than the observed ratios even in the case of the largest neutrino irradiation . although the @xmath0-process enhances the abundances of na and al in incomplete si burning region , na and al are mainly produced in carbon and neon shells . the yields produced in the outer two shells strongly depend on the nucleosynthesis in stellar evolution . @xcite successfully reproduced relatively large [ na / fe ] and [ al / fe ] observed in two hyper - metal - poor stars . they considered that the effect of overshooting during presn evolution and extensive matter mixing and fallback during sn explosions with small explosion energies . [ [ v ] ] v ^ vanadium is mainly produced as @xmath185mn in incomplete si burning region in sne . in the case of the hn model , it is mainly produced in complete si burning region . additional production through the @xmath0-process from @xmath186fe in incomplete si burning region contributes to the enhancement of v in sne . low metallicity stars with @xmath187 } \la -2 $ ] indicate the v abundance ratios of @xmath188 } \la 0.6 $ ] ( e.g. , * ? ? ? * ; * ? ? ? the average abundance ratio is reproduced by the 15 and 25 @xmath2 sn models with the @xmath0-process of @xmath189 ergs . in order to reproduce the v / fe ratios , strong neutrino irradiation in the sne is favored . in the case of the hn model , v is also produced through the @xmath0-process . however , the produced abundance is much smaller than that produced in complete si burning region . the chemical compositions of low mass emp stars are expected to be injected from one or a few sn or hn explosions evolved from population iii massive stars . the abundance ratios of odd-@xmath1 elements to fe observed in these stars are close to the corresponding solar ratios . on the other hand , the abundances produced in complete and incomplete si burning in sn explosion models are still in short . in this study , we investigated the @xmath0-process of odd-@xmath1 iron - peak elements , sc , mn , and co , in the sn and hn explosions of 15 @xmath2 and 25 @xmath2 pop iii stars . then , we compared the abundance ratios of these elements in the sn and hn models with those observed in the emp stars . the obtained results are summarized as follows . 1 . sc , mn , and co are produced through the @xmath0-process and the following capture reactions of protons and neutrons in complete and incomplete si burning region of pop iii sne . the produced amounts of these elements are roughly proportional to the total neutrino energy . 2 . the observed mn / fe ratio averaged in low mass emp stars is reproduced by the 15 @xmath2 and 25 @xmath2 pop iii sn models with the @xmath0-process . therefore , the @xmath0-process in pop iii sne is one of the main synthesis processes of mn observed in low mass emp stars . the observed mn abundance constrains the total neutrino energy to @xmath170 ergs . this @xmath169 value is roughly consistent with the gravitational binding energy of a typical neutron star . in the case of the 25 @xmath2 pop iii hn model , large neutrino irradiation enables to reproduce the observed mn / fe ratio . 3 . the observed co / fe ratio is reproduced by the sn models . however , required total neutrino energy is @xmath190 ergs , which is larger than typical value of the gravitational binding energy of a proto - neutron star . complete si burning in pop iii hne should be favorable to co production rather than the @xmath0-process of sne and hne . although larger amount of sc is produced through the @xmath0-process , the sc / fe ratio is still smaller than the ratio observed in low mass emp stars . explosive nucleosynthesis during aspherical explosions and in a deep region just above the mass cut would be favorable to produce sc and co. the @xmath0-process and the capture reactions of protons and neutrons may also enhance the production of sc , mn , and co in these environments . we would like to thank anonymous referee for valuable comments . numerical computations were in part carried out on general common use computer system at astronomical data analysis center ( adac ) of national astronomical observatory of japan . this work has been supported in part by the ministry of education , culture , sports , science and technology , grants - in - aid for young scientist ( b ) ( 17740130 ) , and scientific research ( s ) ( 18104003 ) , ( c ) ( 16540229 , 18540231 ) . t.y . was supported by the 21st century coe program `` exploring new science by bridging particle - matter hierarchy in graduate school of science , tohoku university . argast , d. , samland , m. , gerhard , o. e. , & thielemann , f .- k . 2004 , , 356 , 873 cayrel , r. , et al . 2004 , , 416 , 1117 chieffi , a. & limongi , m. 2004 , , 608 , 405 colella , p. & woodward , p. r. 1984 , j. comput . phys . , 54 , 174 domogatsky , g. v. , eramzhyan , r. a. , & nadyozhin , d. k. 1978 , in neutrino physics and neutrino astrophysics , ed . m. a. markov et al . ( moscow : nauka ) , 115 frhlich , c. , et al . 2006a , , 637 , 415 frhlich , c. , martnez - pinedo , g. , liebendrfer , m. , thielemann , f .- k . , bravo , e. , hix , w. r. , langanke , k. , & zinner , n. t. 2006b , , 96 , 142502 goriely , s. , arnould , m. , borzov , i. , & rayet , m. 2001 , , 375 , l35 heger , a. , kolbe , e. , haxton , w. c. , langanke , k. , martnez - pinedo , g. , & woosley , s. e. 2005 , phys . b , 606 , 258 hix , w. r. & thielemann , f .- k . 1996 , , 460 , 869 honda , s. , aoki , w. , kajino , t. , ando , h. , beers , t. c. , izumiura , h. sadakane , k. , & takada - hidai , m. 2004 , , 607 , 474 iwamoto , n. , umeda , h. , tominaga , n. , nomoto , k. , maeda , k. 2005 , science , 309 , 451 iwamoto , n. , umeda , h. , nomoto , k. , tominaga , n. , thielemann , f .- k . , & hix , w. r. 2006 , in international symposium on the origin of matter and evolution of galaxies 2005 , ed . s. kubono et al . ( san francisco : aip ) , 409 janka , h .- th . , buras , r. , & rampp , m. 2003 , nucl . phys . a , 718 , 269 kobayashi , c. , umeda , h. , nomoto , k. , tominaga , n. , & ohkubo , t. 2006 , , 653 , 1145 kotake , k. , sawai , h. , yamada , s. , & sato , k. 2004 , , 608 , 391 lattimer , j. m. & prakash , m. 2001 , , 550 , 426 maeda , k. & nomoto , k. 2003 , , 598 , 1163 mclaughlin , g. c. & surman , r. 2007 , , 75 , 023005 nomoto , k. , tominaga , n. , umeda , h. , kobayashi , c. , & maeda , k. 2006 , nucl . a777 , 424 norris , j. e. , ryan , s. g. , & beers , t. c. 2001 , , 561 , 1034 pruet , j. , hoffman , r. d. , woosley , s. e. , janka , h .- buras , r. 2006 , , 644 , 1028 pruet , j. , woosley , s. e. , buras , r. , janka , h .- t . , & hoffman , r. d. 2005 , , 623 , 325 rauscher , t. , heger , a. , hoffman , r. d. , & woosley , s. e. 2002 , , 576 , 323 scheck , l. , plewa , t. , janka , h .- th . , kifonidis , k. , & mller , e. 2004 , , 92 , 011103 shigeyama , t. & tsujimoto , t. 1998 , , 507 , l135 tominaga , n. , maeda , k. , umeda , h. , nomoto , k. , tanaka , m. , iwamoto , n. , suzuki , t. , & mazzali , p. 2007a , , 657 , l77 tominaga , n. , umeda , h. , & nomoto , k. 2007b , , 660 , 516 tominaga , n. , et al . 2005 , , 633 , l97 umeda , h. & nomoto , k. 2002 , , 565 , 385 umeda , h. & nomoto , k. 2005 , , 619 , 427 wanajo , s. 2006 , , 647 , 1323 woosley , s. e. , hartmann , d. h. , hoffman , r. d. , & haxton , w. c. 1990 , , 356 , 272 woosley , s. e. & weaver , t. a. 1995 , , 101 , 181 yoshida , t. , kajino , t. , & hartmann , d. h. 2005a , , 94 , 231101 yoshida , t. , kajino , t. , yokomakura , h. , kimura , k. , takamura , a. , & hartmann , d. h. 2006a , , 649 , 319 yoshida , t. , kajino , t. , yokomakura , h. , kimura , k. , takamura , a. , & hartmann , d. h. 2006b , , 96 , 091101 yoshida , t. , terasawa , m. , kajino , t. , & sumiyoshi , k. 2004 , , 600 , 204 yoshida , t. , umeda , h. , & nomoto , k. 2005b , , 631 , 1039
open clusters are considered excellent laboratories for studies of stellar evolution and the dynamics of stellar systems . a large fraction of galactic open clusters are situated close to the galactic plane , thus potentially making them appropriate objects for probing the structure and evolution of the galactic disc although the presence of dust in the galactic plane restricts their usefulness in this sense . one of many important aspects in open cluster studies is related to the effects of mass segregation ; that is , the more massive member stars in the cluster will be more centrally concentrated and/or have a distinct distribution in velocity space compared to the lower - mass members . the key observational questions of interest in open star cluster research include whether or not mass segregation exists in a given star cluster and , if so , what the extent of this effect is in relation to the cluster s age . for cluster samples as a whole , the similarities and differences of the effects of mass segregation for clusters of various ages , abundances and galactocentric distances are of great interest . all of these properties are related to the formation process as well as the dynamical evolution of clusters . they are also ( sometimes significantly ) influenced by the galactic tidal fields , in particular in the outer regions of a given cluster . young open clusters are important stellar systems for our understanding of the star - formation process . in very young clusters , most of the low - mass stars will still be in their pre - main - sequence ( pms ) evolutionary stage , and one may expect to derive global stellar initial mass functions ( imfs ) that are not yet significantly affected by the effects of stellar and dynamical evolution . however , if a significant amount of `` primordial '' mass segregation ( i.e. , mass segregation intrinsic to the star - formation process itself ) is present , this would seriously complicate the interpretation of an observed stellar luminosity function ( lf ) at a given position within a star cluster in terms of its imf @xcite . in this paper , we use photographic plate material with longer time baselines than published previously , in order to determine the proper motions and membership probabilities of stars in the regions of two very young galactic open clusters ngc 2244 and ngc 6530 . we aim to quantify the possible effects of mass segregation and place these objects in the context of their expected future evolution . ngc 2244 ( also referred to as ngc 2239 ) is the core ob cluster of the monoceros ob2 ( mon ob2 ) association , embedded in the rosette nebula ( ngc 2237/2246 ) , a well - developed `` blister '' hii region . this system is located in the northwest quadrant of the rosette molecular cloud complex , which itself is one of the most massive giant molecular cloud ( gmc ) complexes in the milky way ( see @xcite ) ; such gmc complexes are particularly conducive to the formation of massive ob stars . because of the strong ionizing radiation from the massive ob stars in the core of ngc 2244 ( see , e.g. , @xcite for a census of the ob stars in the cluster ) , the cluster has excavated the gmc complex , allowing us a deeper look into the cloud than otherwise possible . ngc 2244 represents the youngest of two or three subgroups of ob stars and stellar aggregates in the mon ob2 association @xcite . since ngc 2244 is one of the youngest star - forming open clusters in the milky way , and relatively nearby , it has been the subject of a number of detailed studies . using the webda open cluster database @xcite as our guide , the equatorial coordinates of the cluster center are @xmath3 , and the galactic coordinates @xmath4 , i.e. , the cluster is located quite close to the direction of the galactic anticenter . @xcite , using near - infrared imaging from the two micron all - sky survey ( 2mass ) , estimated a mean visual extinction , @xmath5 mag within a radius of 10 arcmin centered on the apparent center of the hii region . they pointed out that this independent low near - infrared extinction estimate was consistent with previous estimates in the optical , which is also supported by results from co and c@xmath6o maps ( e.g. , @xcite ; @xcite ) . in the optical , the most comprehensive extinction analysis for the cluster was published by @xcite , who reported a mean reddening of @xmath7 mag , with a non - negligible amount of differential reddening across the cluster : their reddening values for individual cluster stars range from @xmath8 to 0.98 mag . this is fully consistent with the results of @xcite ; and references therein ) , who found @xmath9 mag from the individual reddening values for 28 cluster member stars brighter than @xmath10 mag , and a smaller amount of differential reddening , ranging from @xmath11 to 0.56 mag . they also found that ngc 2244 is most likely characterized by a `` standard '' reddening law , with a total - to - selective extinction ratio of @xmath12 . as such , we adopt a foregound reddening of @xmath7 mag in this paper . the distance to ngc 2244 is reasonably well established , at @xmath13 kpc , and generally consistent among different distance indicators used in the literature ( see also @xcite for a review ) . for instance , @xcite obtained @xmath14 pc from the distance moduli of the individual ob - type stars in the cluster . more recently , @xcite derived a distance of @xmath15 kpc to the cluster based on spectroscopic observations of one of the cluster member stars , the eclipsing binary v578 mon ( hd259135 ) . @xcite derived a distance modulus of @xmath16 based on @xmath17 photometry , corresponding to @xmath18 kpc . this larger distance value seems to be favoured by many authors in the field ( e.g. , @xcite ) , and as such we will adopt it here as well . for the age of ngc 2244 , most authors agree on a value of 24 myr for the main sequence ( ms ) turn - off age ( @xcite and references therein ) . in addition , recently an age estimate of @xmath19 myr was derived from the age of the eclipsing binary v578 mon @xcite . based on high spatial resolution multi - filter photometry of ngc 2244 , @xcite concluded that while most of the cluster stars are aged up to 3 myr , significantly younger low - mass stars ( 5% or more of the cluster stars ) exist in the cluster as well , indicating that star formation is still in progress in the rosette nebula / ngc 2244 region ( see also @xcite , the latter authors found a ms turn - off age of 1.9 myr , but a pre - ms age spread of about 6 myr ) . however , they also note that up to 45% of the cluster stars may be older than 3 myr , thus suggesting an earlier star - forming episode in the region . moreover , the dynamical age of the nebula is just @xmath20 myr @xcite , while the dynamical age of the hi shell is as old as 4 myr ( see @xcite ) . the first proper motion study of this cluster @xcite identified 22 probable members . subsequently , @xcite performed more precise proper motion measurements of 287 stars down to @xmath21 mag and selected 89 stars with membership probabilities , @xmath22 , and data quality , @xmath23 , in an area of @xmath24 arcmin@xmath25 . however , @xcite cautioned that their `` member '' sample may contain several non - cluster members of the mon ob2 association because of the large rectangular field of view employed by the @xcite study compared to the cluster s angular size of @xmath26 arcmin and the roughly spherical shape described by @xcite . using the @xcite astrometric membership determinations , combined with spectroscopically determined stellar mk classifications of 193 stars by @xcite , @xcite redetermined the membership probabilities of the brightest cluster stars ( 30 ob stars and 21 pre - ms members and candidates brighter than @xmath27 mag ) . @xcite and @xcite redetermined the cluster s stellar membership probabilities based on the early astrometric data first analyzed by @xcite . they confirmed 92 ( out of 202 ) stars as physical members of the cluster , with @xmath28 . in this paper , we will analyze high - quality , wide - field photographic plates of shanghai astronomical observatory , with an maximum likelihood algorithm involving careful assessment of the distributions of both the proper motions and the positions , thus allowing us to arrive at a valuable , independent estimation of membership probabilities . ngc 6530 is another prime example of an extremely young open cluster that has been the subject of a number of previous studies . as the core cluster of the sgr ob1 association , it is projected onto the eastern part of the lagoon nebula ( m8 ) , which is one of the brightest nearby nebulae and hii regions @xcite . the lagoon nebula is the illuminated part of the gmc from which the cluster formed by generating a `` blister '' hii region due to the ionizing radiation from its hot massive stars ( e.g. , @xcite . once again , this allows us to probe into the depths of this gmc region much better than would have been possible otherwise . the equatorial coordinates of ngc 6530 are @xmath29 and the galactic coordinates @xmath30 , i.e. , the cluster is located in the direction not far from that of the galactic center . this implies that one has to be very careful in disentangling cluster members from field interlopers ( see also @xcite . as a case in point , in early work on this cluster , @xcite obtained photo - electric magnitudes and colors for 118 stars in the ngc 6530 region and proposed that a band of its low - mass members above the ms in color - magnitude space may be pms stars that are still in the stage of gravitational contraction . it has since been shown ( e.g. , @xcite ) that many of these pms candidates are , in fact , foregound or background field stars ( see the discussions in @xcite , and @xcite ) . subsequently , the properties of this very young cluster have been studied extensively by different authors , leading to an age estimate of 1.53.0 myr @xcite , and the cluster s distance to the sun has been estimated to be in the range from 0.62.0 kpc @xcite . the most recent median age estimate for the cluster , based on ground - based @xmath31 photometry down to @xmath32 mag @xcite is 2.3 myr , i.e. , well inside the range quoted by previous authors , although the same team also derived a much younger median age of 0.8 myr ( @xcite ; see also @xcite for the pre - ms stars ; @xcite ) . both estimates are not necessarily in conflict with each other , because the cluster appears to host stars formed over a significant length of time . while @xcite argued that star formation in the cluster may have been ongoing since a few @xmath33 ago , until the present , @xcite based on their detailed study of low - mass pms stars argue that the time - scale for star formation in ngc 6530 must have been much shorter , starting only a few @xmath34 yr ago . nevertheless , this clearly implies that the cluster environment has been an active star - forming environment for a significant period ( see also @xcite ) . more specifically , @xcite conclude , based on a detailed analysis of the distribution of optical / infrared - excess stars throughout the region , that star formation has proceeded rather undisturbed and for a longer period in the northern cluster region than near the cluster center . this is in support of suggestions by @xcite and @xcite that the star - formation activity may have proceeded from north to south in the cluster region , which is further supported by a spatial age segregation analysis by @xcite . the most recent distance estimate to ngc 6530 , based on @xmath31 photometry , was derived by @xcite , @xmath35 pc , corresponding to a distance modulus @xmath36 mag . this is towards the bottom end of the range in possible distances quoted by a variety of authors , @xmath37 pc ( @xcite , their table 1 ) , generally obtained photometrically from the brightest cluster member stars . in contrast , it is significantly larger than the distance obtained from hipparcos parallaxes of seven member stars , @xmath38 pc @xcite . in other recent studies , both @xcite and @xcite determined a best distance estimate to ngc 6530 of @xmath39 kpc [ @xmath40 mag , based on optical photometry and a detailed consideration of the extinction across the cluster ( see also @xcite ) . in this paper , we opt to use the @xcite distance determination , @xmath35 pc . following the webda compilation of cluster parameters , we adopt a mean reddening @xmath41 mag ( @xcite ; see also @xcite ) , in addition to a foreground reddening of @xmath42 to 0.20 mag @xcite . there is significant variation in reddening across the ngc 6530 region , as is evident from both the distribution of stars in color - magnitude space and from the spatial variation of stellar density ( @xcite , and references therein ) . in fact , the amount of differential reddening ranges from @xmath43 to 0.50 mag across the cluster ( @xcite ; also @xcite , but see @xcite ) , although @xcite also uncovered evidence for the presence of a large number of heavily reddened objects , characterized by @xmath44 mag . @xcite were the first to determine the membership probabilities of 363 stars in a @xmath45 region centered on ngc 6530 from relative proper motion data with a time baseline of 34 years . they found 76 stars with membership probabilities , @xmath46 , greater than 0.5 with @xmath47 mag . more recent studies resulted in firm membership determinations for 451 stars with @xmath48 @xcite , and 237 stars @xcite , the latter study based on spectroscopic observations around the lithium and h@xmath49 lines . @xcite has suggested that the cluster may contain @xmath50 members . it should be noted that a reliable stellar membership determination is crucial for resolved star cluster studies . the shanghai astronomical observatory has in its archives wide - field photographic plates with a long time baseline of these two open clusters , thus providing valuable , independent observational materials for high - accuracy measurements of stellar proper motions , and for the determination of cluster membership . this constitutes a strong starting point for further detailed cluster studies , including those dealing with the effects of mass segregation , as we show in this paper . in this paper , we use photographic plate material with longer time baselines or higher quality than published previously , in order to determine the proper motions and membership probabilities of stars in the regions of both clusters ( sect . [ obs.sec ] ) . we present a new , detailed investigation of these two clusters based on our membership determination ( sect . [ members.sec ] ) , of their stellar luminosity functions and the observed effects of mass segregation ( sect . [ results.sec ] ) . in sect . [ discussion.sec ] we discuss the clusters in the context of their evolutionary state and summarize the main results obtained in this paper . for the open clusters ngc 2244 and ngc 6530 , all photographic plates used for the proper motion reduction were taken by the 40-cm refractor telescope ( with a focal length of 6895 mm ) at the sheshan ( z s ) station of the shanghai astronomical observatory , operated by the chinese academy of sciences . for ngc 2244 , four plates for each of the first and second epochs were taken in 19631964 , and 19981999 , respectively , spanning a time baseline of 35 years . for ngc 6530 , two first epoch plates and four second plates were taken in 1912 and 1999 , respectively , spanning 87 years . the field of view achieved is @xmath51 and @xmath52 , for the first and second epoch plates , respectively . no filter was used to obtain these plates ; their sensitive wavelength region is close to the @xmath53 band and the 30 min exposure time resulted in a brightness limit of about 16 mag . the stellar coordinates on the plates of both clusters were measured with the photometrics data systems ( pds ) microdensitometer @xcite at the dominion astrophysical observatory ( dao ) in canada . for each plate , raster scanning was performed over the full @xmath54 central regions of the clusters . the scanning diaphragm used had a size of @xmath55 m squared , with both a step length and line separation of @xmath56 m ; a speed of 15 mm s@xmath2 was used . in the region where we performed the full scan , stellar images were extracted and their rectangular coordinates , @xmath57 , obtained using the processing software provided by dao . after the preliminary identification , the measured coordinates on each plate were transformed linearly to a unified system using six plate constants , and those with residual errors @xmath58 m were excluded . the cross - identification resulted in a final set of 498 stars for ngc 2244 and 365 stars for ngc 6530 , respectively , with each star appearing on at least one plate of both the early _ and _ later epochs . in the reduction of our proper motion data , an iterative central - overlap technique @xcite was used . for ngc 2244 , 78 tycho-2 stars were selected as reference stars ; their reference positions and proper motions were adopted based on the tycho-2 catalogue @xcite . during the iterative solution , three stars with extraordinarily large residual errors were removed . the final solution provided us with robust position and proper motion data of 495 stars in the cluster region . similarly , 79 tycho-2 catalogue stars were used as reference stars for ngc 6530 , and only 1 star with an extremely large residual error in the iterative solution was removed . this resulted in a total of 364 stars with final position and proper motion values . the internal levels of uncertainty on the position and proper motion of the individual stars are given in table 1 . estimating the membership probability of the observed cluster stars involves a careful assessment of the distributions of both the proper motions and the positions . we assume that in position space the surface number density of the field stars follows a uniform distribution , @xmath59 , while the cluster members approximately follow a gaussian profile ( see , e.g. , @xcite , for a preliminary analysis of the stars in the ngc 6530 region with membership probabilities @xmath60 ) , @xmath61 , with a central density , @xmath62 , and a dispersion , @xmath49 . then , @xmath63 where @xmath64 denotes the distance of the @xmath65 star to the cluster center . let @xmath66 and @xmath67 be the normalized @xmath59 and @xmath61 , respectively , and @xmath68 . then we have @xmath69 ( by definition ) , and @xmath70 both the field stars and the cluster members have independent proper motion and position distributions . we assume that the field star proper motion is characterized by an elliptical gaussian distribution , and that of the cluster stars is a circular gaussian function . the overall distribution function for the field stars , @xmath71 , and for the cluster stars , @xmath72 , can then be written as : @xmath73\},\end{aligned}\ ] ] and @xmath74\}\end{aligned}\ ] ] here , @xmath75 is the reduced proper motion of the @xmath65 star and @xmath76 are the estimated observational errors . @xmath77 and @xmath78 are the distribution centers of the field stars and cluster members in the vector point diagram ( vpd ) , respectively , whereas @xmath79 and @xmath80 are the intrinsic proper motion dispersions of the field stars and the cluster members respectively , and @xmath81 is the correlation coefficient . together with the parameters @xmath49 and @xmath82 , there are 10 parameters awaiting determination . these parameters were estimated by means of a maximum likelihood method . when searching for the maximum of the likelihood function , a bipartition algorithm was adopted @xcite , and the ( internal ) uncertainty of the parameter estimation was evaluated based on the second derivatives of the likelihood function @xcite . in order to get a handle on the external uncertainty of our proper motion determinations , we compared our derived proper motion values to those derived from the tycho-2 catalogue , for the 78 stars in ngc 2244 and the 79 stars in ngc 6530 in common in both data sets . this resulted in a mean deviation of @xmath83 mas yr@xmath2 , with a dispersion of @xmath84 mas yr@xmath2 , in very good agreement with the internal level of accuracy of the tycho-2 catalogue . after the distribution parameters are determined , the membership probability of the @xmath65 star can be calculated as : @xmath85 table 2 gives the maximum likelihood estimates of the distribution parameters , and the corresponding uncertainties , for both clusters . we used the distribution parameters to calculate the membership probabilities of the individual stars in the ngc 2244 and ngc 6530 regions . figure 1 shows the vpd for both ngc 2244 and ngc 6530 ( see also the preliminary analysis by @xcite ; the solid points represent the stars with membership probabilities , @xmath60 , whereas the crosses indicate the @xmath86 . the open circles have @xmath46 values in between these boundaries . the vpds already show that our membership determination appears to be quite effective . to underscore this point , fig . 2 shows the histograms of the membership probabilities ; the left - hand panel is for ngc 2244 and the right - hand panel for ngc 6530 . returning to the clusters vpds , both @xcite and @xcite published vpds for ngc 2244 based on independent proper motion analyses . although their samples of cluster members are much smaller than ours ( of order 100 member stars ) , the vpds are qualitatively similar in extent as well as morphology . we note , however , that in particular the @xcite vpd appears somewhat more elongated in the @xmath87 direction than our vpd in fig . 1 . we will return to this issue below . from fig . 2 , one can deduce _ quantitatively _ that the discrimination of membership for ngc 2244 is quite effective . of the total of 495 stars , about 16% , or 78 stars , have probabilities in the range of @xmath88 . as an aside , we note that a comparison of the membership histograms in @xcite and fig . 2 shows that our membership determination is more efficient than theirs . meanwhile , for ngc 6530 , the separation of members from the field stars is even more effective , with only a few stars having probabilities around @xmath89 , i.e. , only 4.4% ( or 16 stars ) have @xmath88 ; 451 stars have @xmath60 ( see also @xcite for a discussion of the results in relation to the earlier work by @xcite ) . we note that in the context of ngc 6530 , the @xcite and @xcite studies cover a much smaller field of view ( diameter @xmath90 arcmin ) than our observations presented here ( sect . [ obs.sec ] ) , although at a higher spatial resolution . secondly , the @xcite sample , which in essence only covers the cluster core , is fainter ( @xmath91 mag ) than ours ( @xmath92 mag ) . as such there is virtually no overlap between both data sets . star counts allow us to statistically determine the properties of clusters with respect to the surrounding stellar background . the cluster radius itself is one of the most important cluster parameters , and particularly useful for dynamical studies . from our results , the member stars of both ngc 2244 and ngc 6530 show an obvious concentration on the sky , while this is clearly not the case for field stars . in essence , this supports the result for ngc 2244 by @xcite based on the cluster s vpd . they determined robust membership probabilities for 92 stars in the area using both the cluster s vpd and its membership probability histogram , akin to our fig . 2 . in the vpd they clearly showed the ( elliptical ) distribution of the proper motions of the cluster members around a centroid , and the very small dispersion of stars with @xmath93 , clearly setting them apart from the field stars , with @xmath94 ( see also @xcite ) . as regards ngc 6530 , @xcite and @xcite provided compelling evidence for the superiority of the membership determinations in the present data set compared to that presented by @xcite . in order to derive the radial stellar surface density , we first chose the cluster center as the median coordinates of all the member stars with @xmath95 . the adopted centers are at the ( j2000.0 ) equatorial coordinates of @xmath96 for ngc 2244 and @xmath97 for ngc 6530 , respectively . these updated center positions are slightly different from the values provided by webda , although the displacements between our new center coordinates and those provided by webda are well inside the respective cluster core radii ( see below ) . for comparison , @xcite determined the 2mass center position of ngc 2244 as being very close to the centroid determined from our proper motions , resulting in j2000.0 equatorial coordinates of @xmath98 . the radial number density profile , shown in fig . 3 , was constructed by performing star counts inside increasing concentric annuli around the cluster center and normalized by the surface areas covered by the respective annuli . figure 3 shows the radial surface number density profiles of the cluster members @xmath99 and the field stars @xmath100 in our observational fields of view , on the left - hand side for ngc 2244 and on the right - hand side for ngc 6530 . in each panel , the profiles for the cluster members and the field stars are completely different , with the field star distributions essentially flat , whereas the cluster members exhibit obvious concentrations towards the cluster centers . this result shows once again that our membership determinations are reasonably robust and reliable . the basic structural parameters were derived by fitting the two - parameter surface density @xcite profile to our observational data in a least - squares sense . we used fitting radii of 32 and 33 arcmin for ngc 2244 and ngc 6530 , respectively . the core radii thus obtained are @xmath101 arcmin for ngc 2244 and @xmath102 arcmin for ngc 6530 , corresponding to @xmath103 and @xmath104 pc , respectively . meanwhile , the half - number radii ( i.e. , the radii on either side of which half of the member stars are located ) of the clusters were also determined , as 14 arcmin for ngc 2244 and 21 arcmin for ngc 6530 , or @xmath105 and @xmath106 pc , respectively . our main reason for using the clusters half - number radii instead of their half - light radii is practical : the position and proper motion information , eventually leading to membership probabilities , is based on the results from the pds scans . during this process , some of the brightest stars had to be avoided because of saturation effects on the photographic plates ( which would lead to unacceptably large positional uncertainties for our analysis ) . in other words , our sample of cluster members is incomplete at bright magnitudes , and the membership probabilities of the ( few ) brightest stars are unknown . the half - number radii are not significantly affected by the omission of these few stars , which would have an extraordinarily large contribution to the clusters total fluxes , however . although @xcite found that the radial profile in the near - infrared of the ngc 2244 member stars disappears into the background field star population at a radius of about 20 arcmin , our radial profile in fig . 3 clearly shows that we can follow the cluster s profile out to almost 30 arcmin . this is in keeping with the cluster radius of 11.8 pc derived by @xcite , which at an assumed distance of 1.4 kpc corresponds to about 29 arcmin . our size determination for ngc 6530 is less clear - cut , as fig . 3 shows : the radial profile composed by its member stars appears to merge with the field star population at a radius of around 20 arcmin , although this value is rather uncertain . we believe , therefore , that our result is in reasonable agreement with the diameter of roughly 35 arcmin quoted by @xcite . for both clusters , the model profiles tend to roughly follow the observational data points in the outer regions , but not in detail . this suggests that both clusters do not have well - defined outer radii , but are instead characterized by a significant amount of substructure . there is , in fact , strong evidence that ngc 2244 possesses distinct substructures . @xcite , using 2mass near - infrared imaging , found that the cluster is resolved into an extended distribution of stellar sources associated with its compact core in the south center of the blister hii region , but also into a clear elongated density enhancement in the west . 6530 also exhibit distinct substructures : in a region towards the northwest of the actual cluster center there appears to be a density enhancement of slightly older stars than in the rest of the cluster ( e.g. , @xcite ) . in addition , the vpds of both clusters ( fig . 1 ; see also @xcite , in particular ) appear to have axial ratios significantly deviating from unity . finally , @xcite and @xcite point out that instead of using a @xcite profile , a more appropriate radial density profile might be represented by a @xmath107 profile , following suggestions by @xcite regarding the radial density distributions of embedded star clusters . in fig . 3 , we show the best - fitting @xmath107 profiles for both clusters . for ngc 2244 , the best - fitting profile follows @xmath108 , while the equivalent profile for ngc 6530 is @xmath109 . we note that whereas the fit to the ngc 6530 radial profile is statistically good , the ngc 2244 distribution is better approximated by a king profile . @xcite provided their @xmath110 observations centered on ngc 2244 with the kpno 0.9 m telescope and a tektronix 2048@xmath1112048 ccd , covering a 2800 arcmin@xmath25 area . this resulted in accurate photometric data for a total of 773 stars in the cluster region , down to @xmath112 mag . by matching our photographic plate observations to the @xcite sample using a conservative matching tolerance of 0.3 arcsec , 323 stars in common were identified . furthermore , using a quadratic function , our initial instrumental magnitudes , @xmath113 , were converted to the @xmath53 magnitudes in the @xcite system . as regards ngc 6530 , @xcite performed @xmath31 observations of the ngc 6530 region using the eso 2.2 m telescope at la silla , characterized by a field of view of @xmath114 arcmin@xmath25 . the limiting magnitude of their work is @xmath115 mag . we compared our @xmath113 instrumental magnitudes with the @xmath53 magnitudes of 62 stars in common with @xcite s sample . except for a few stars , we found an obvious linear correlation between the magnitudes , which we used to convert our @xmath113 photometry to the b magnitudes in the @xmath110 system . figure 4 shows the comparison between our instrumental @xmath113 photometry and the @xmath53 magnitudes of @xmath110 system , for both clusters . the photometric uncertainties in both our own data and those taken from the literature are small : @xcite state that their uncertainties are @xmath116 mag for stars brighter than @xmath117 mag , and `` do not become appreciable until @xmath118 mag '' . the photometric uncertainties in the @xcite data have a mean value of 0.02 mag , with a maximum error @xmath119 mag . for member stars in a given open cluster , all at roughly the same distance , the range in visual magnitudes corresponds to that in luminosity . in addition , when only main - sequence stars are considered , the stellar luminosity is a proxy for the stellar mass . in order to assess the effects of mass segregation , we first examined the luminosity functions of the member stars in different ( radial ) ranges of the clusters . figures 5 and 6 show the @xmath120 luminosity functions of stars with membership probabilities @xmath121 in the two cluster regions . figures 5a ( top ) and b ( middle ) show the luminosity functions of the ngc 2244 cluster members in the inner @xmath122 and outer @xmath123 areas of ngc 2244 , where @xmath124 arcmin is the half - number radius derived from our full sample of 216 member stars . 5c ( bottom ) shows the cumulative distributions for both radial regions . figures 6a ( top ) , b ( middle ) , and c ( bottom ) show the equivalent results for ngc 6530 ( see also @xcite ) . here , we derived a half - number radius of @xmath125 arcmin from our total sample of 250 member stars . from these two figures , the effects of mass segregation are evident : the relative fraction of brighter stars ( or , equivalently , more massive stars ) is higher in the inner than in the outer regions . as an alternative approach to investigate these mass segregation effects , we calculated the half - number radii for member stars in different absolute magnitude ranges , i.e. , for @xmath126 and @xmath127 mag . the results of this exercise are listed in table 5 ( see @xcite for a complementary discussion related to ngc 6530 ; in particular their table 6 ) . it can be seen that for both clusters , the brighter members are characterized by smaller half - number radii another indication of the existence of mass segregation in these clusters . to further examine the mass segregation effects , we also investigated the radial density distributions of member stars ( @xmath95 ) in different luminosity ranges in the clusters , which are shown in figs . 7 ( for ngc 2244 ) and 8 ( for ngc 6530 ) . in each cluster , a position - dependent lf is found . in essence , the fainter ( i.e. , lower - mass ) stars are spread throughout the entire cluster in both objects , while the brighter ( massive ) stars are predominantly located in the inner region . if the mass segregation effects solely result from the process of two - body relaxation and the associated energy equipartition , one should find some evidence for this effect not only in the spatial distribution of the cluster members , but also in their velocity distribution : the more massive members will have a smaller velocity dispersion than the less massive stars . accordingly , for true cluster members ( with @xmath128 ) in different magnitude ranges , the intrinsic proper motion dispersions @xmath129 as derived from the observed dispersions @xmath130 , and the mean errors intrinsic to the stellar proper motions were investigated . the final results are shown in figs . it can be seen from these figures that the intrinsic proper motion dispersions are not significantly different for various magnitude groups . thus , our data do not support any significant velocity ( dispersion)-mass correlation , neither in ngc 2244 , nor in ngc 6530 ( see also @xcite ) . the traditional explanation for ( position dependent ) mass segregation in an open cluster is that it results from a situation in which all cluster stars approach energy equipartition . however , this is unlikely the case for ngc 2244 and ngc 6530 , since both clusters are very young , low - density objects ( but see @xcite for counterarguments ) . with ages of only around 23 myr , their ages are comparable to the time - scale of about only one or two crossing times . theoretically , the nature and degree of `` primordial '' mass segregation , i.e. , mass segregation intrinsic to the star - formation process itself , is presumably determined by the properties of the interactions of protostellar material during the star - forming episode in a cluster . more massive stars are subject to more mergers , hence accrete even more mass ( cf . @xcite and references therein ) , and therefore dissipate more kinetic energy . in addition , they tend to form near the cluster center , in the highest - density region , where the encounter - rate is highest @xcite . this will lead to an observed position - dependent stellar mass function containing more low - mass stars at larger radii compared to the mass function in the cluster center ( although low - mass stars are still present at small radii , see also the discussions in @xcite ; but see @xcite for arguments in favor of excluding primordial mass segregation to explain recent observations of the arches cluster near the galactic center ) . from an observational point of view , the study of very young star clusters still embedded in the molecular clouds from which they originated might give us a handle to constrain the degree of primordial mass segregation . in the milky way , in three such young star clusters mass segregation effects have been studied in great detail in the past two decades . while @xcite suggested , based on ground - based observations , that the brighter stars in ngc 2024 ( and the mon r2 complex at large ) seem to be more centrally concentrated than the fainter cluster members , this evidence was deemed inconclusive by @xcite . they argued that this result was based on an incomplete sample of cluster members , although mass segregation might be limited to the massive , bright ob stars forming in the very center . these same authors argued that for masses below @xmath131 , mass segregation effects in the mon r2 complex amount to only a @xmath132 result . secondly , a combination of both ground - based ( e.g. , @xcite and _ hubble space telescope ( hst ) _ observations ( e.g. , @xcite ) of the orion nebula cluster ( onc ) , and in particular of its very core , the trapezium stars , have presented clear evidence for mass segregation for the @xmath133 component , with some evidence for general mass segregation down to @xmath134@xmath135 ( @xcite ; see also the review by @xcite ) . finally , r136 , the central cluster in 30 doradus in the large magellanic cloud ( lmc ; age @xmath136 34 myr ; cf . @xcite ) , has been studied extensively , both from the ground and with the _ a variety of techniques have revealed a significant overabundance of high - mass stars in its very center , thus supporting strong mass segregation ( e.g. , @xcite ) . thus , in most of the ( still partially embedded ) young star clusters that can be resolved in individual stars , mass segregation effects are observed , although to varying degrees . this underlines the importance of our understanding of the physical processes involved in the formation and evolution of star clusters , and in particular of the imf , which will ultimately determine the time - scale on which a young star cluster will eventually be destroyed . we may conclude from the above discussion that the observed spatial mass segregation in the young open clusters ngc 2244 and ngc 6530 might have resulted from a combination of both initial conditions in the early stages of evolution and , possibly , two - body relaxation process . the latter process will lead to a manifestation of cluster - wide mass segregation on the half - mass relaxation time - scale . however , it will proceed much faster in higher - density regions in the cluster , and among the more massive stars both of these conditions favor an decreased time - scale in the cluster center . finally , we will consider the long - term fate of the two young clusters discussed in this paper in relation to their ( observed ) imf and velocity dispersion . both clusters have been scrutinized with the aim of determining their stellar make - up . because of their very young ages , the stellar content in essence reflects the mass distribution at the time of their birth , as insufficient time has passed for any significant number of stars to have undergone a full cycle of stellar evolution ( and neither for significant dynamical evolution to have occurred , as argued above ) . @xcite determined the imf of ngc 2244 in the mass range @xmath137 . they found a flat imf slope , with @xmath138 , where the equivalent slope of the @xcite imf would be @xmath139 . their result is in close agreement both with that of @xcite in the range from @xmath140 to @xmath141 , and with that of @xcite for stars with masses in excess of @xmath142 , @xmath143 . such a flat imf implies that the cluster s stellar population contains too many massive stars with respect to their lower - mass counterparts for it to survive the dynamical effects leading to cluster dissolution for any significant amount of time ( see , e.g. , @xcite in the context of the young massive cluster m82-f ; and @xcite for theoretical arguments in support of this notion ) . the precise dissolution time - scale depends sensitively on the imf mass range ( kouwenhoven et al . , in prep . ) this conclusion is supported by arguments related to the cluster s velocity dispersion . its intrinsic overall velocity dispersion is @xmath144 mas yr@xmath2 ; at the adopted distance , @xmath18 kpc , this is equivalent to @xmath145 km s@xmath2 . myr by @xcite . ] this large linear velocity dispersion is significantly in excess of the `` typical '' velocity dispersions older open clusters in the milky way characterized by a salpeter - type imf , or equivalent ( de grijs et al . , in prep . ) . thus , the combination of these arguments leads us to conclude that ngc 2244 will most likely dissolve on a short time - scale ( see also @xcite for a more conservative back - of - the - envelope calculation supporting our conclusion ) . if not due to internal two - body relaxation effects , then most likely because of external shocks expected to operate near its location in , or close to , the galactic disk ( although we note that these effects may not be very significant in the galactic anticenter direction ) ngc 6530 , on the other hand , exhibits a power - law imf slope of @xmath146 for stellar masses between 0.6 and @xmath147 @xcite , fully consistent with the @xcite imf . they note that the cluster imf peaks towards smaller masses and then declines . this is in keeping with the detailed photometric study of @xcite , who quote an imf slope of @xmath148 , largely independent of the mass to luminosity conversion employed . its intrinsic velocity dispersion is @xmath149 mas yr@xmath2 ; at the adopted distance , @xmath35 pc , this is equivalent to @xmath150 km s@xmath2 , i.e. , much closer to the dynamical state required for longer - term survival , although still fairly large ( de grijs et al . , in prep . ) . in the absence of sizeable external perturbations , we speculate , therefore , that ngc 6530 has the potential to survive for a significant length of time , possibly of order a few @xmath151 yr . however , we also note that the cluster is situated very close to the galactic plane in the direction of the galactic center , such that external perturbations are unlikely to remain insignificant for any length of time . finally , we point out that the large velocity dispersions obtained for both young clusters imply that these objects are unlikely to survive for any significant length of time . the velocity dispersions of ` classical ' old ( @xmath152 yr ) bound open clusters are of order @xmath153 km s@xmath2 as shown by , e.g. , @xcite in what is probably the most comprehensive study of open cluster velocity dispersions available to date . a small subset of these measurements have since been confirmed in more recent studies of individual objects ( de grijs et al . , in prep . ) . for instance , the velocity dispersion of the @xmath154 myr - old @xcite central cluster in coma berenices is @xmath155 km s@xmath2 ( @xcite ; cf . @xmath156 km s@xmath2 measured in the cluster core by @xcite ) ; that of m67 ( at an age of @xmath157 gyr ; @xcite ) ranges from @xmath158 km s@xmath2 , with a best estimate of @xmath159 km s@xmath2 ( see * ? ? ? * de grijs et al . , in prep . ) , while the velocity dispersion of praesepe ( m44 ) is @xmath160 km s@xmath2 @xcite , at an age of @xmath161 myr @xcite . m35 ( @xmath162 ; @xcite ) is characterized by @xmath163 from @xmath164 km s@xmath2 @xcite to @xmath165 km s@xmath2 @xcite . although we refer the reader for further details on the kinematic structure and long - term stability of galactic open clusters to a forthcoming paper ( de grijs et al . , in prep . ) , these examples show that in order for an open cluster to survive for a few @xmath151 yr , its velocity dispersion needs to be sufficiently small so as to prevent early dissolution of the cluster due to internal kinematic effects . neither of the clusters discussed in this paper satisfies this empirical condition . the only old ( @xmath166 ; @xcite ) galactic open cluster with a measured velocity dispersion in excess of the range indicated above is m11 , with @xmath163 from @xmath167 km s@xmath2 @xcite to 2.9 km s@xmath2 @xcite ; the latter measurement was later reduced to 2.0 km s@xmath2 @xcite . this high velocity dispersion , for an ostensibly stable _ massive _ open cluster , is still more than a factor of two below that of ngc 6530 , which is both significantly less massive and less compact than m11 , and thus less stable to kinematic dissolution . 1 . from the photographic plate data of the shanghai astronomical observatory , the proper motions of stars in the region of the open clusters ngc 2244 and ngc 6530 were reduced by means of a central - overlapping technique , and the distribution parameters of the clusters as well as the membership probabilities of the individual stars in the cluster region were determined using a maximum likelihood principle . both clusters show clear evidence of mass segregation , but there is no definite evidence for a velocity ( dispersion)-mass ( or , equivalently , a velocity - luminosity ) dependence in these clusters . the observed mass segregation might be due to a combination of both initial conditions and relaxation processes . based on arguments related to the clusters imfs , in conjunction with our new measurements of their internal velocity dispersions , we provide strong arguments in favor of the dissolution of ngc 2244 on very short time - scales , while we speculate that ngc 6530 may be more stable against the effects of internal two - body relaxation . rdg was partially supported by an `` international joint project '' grant , jointly funded by the royal society in the uk and the chinese national science foundation ( nsfc ) . lc & jlz want to thank dr . peter stetson of dao for his kind help in measuring plates with the pds machine . lc and jlz are supported by the nsfc ( grant nos . 10333050 , 10333020).this research has made use of the webda database , operated at the institute for astronomy of the university of vienna . baba , d. , et al . 2004 , apj , 614 , 818 berghfer , t . w. , & christian , d. j. 2002 , a&a , 384 , 890 blitz , l. , & stark , a . a. 1986 , apj , 300 , l89 bonnell , i. a. , bate , m. r. , clarke , c. j. , & pringle , j . e. 1997 , mnras , 285 , 201 bonnell , i. a. , & davies , m. b. 1998 , mnras , 295 , 691 bonnell , i. a. , bate , m. r. , & zinnecker , h. 1998,mnras , 298 , 93 bonnell , i. a. , bate , m. r.,clarke , c. j. , & pringle , j. e. 2001a , mnras , 323 , 785 bonnell , i. a. , clarke , c. j. , bate , m. r. , & pringle , j . e. 2001b , mnras , 324 , 573 brandl , b. , sams , b. j. , bertoldi , f. , eckart , a.,genzel , r. , drapatz , s. , hofmann , r. , lwe , m. , & quirrenbach , a. 1996 , apj , 466 , 254 campbell , b. , et al . 1992 , aj , 104 , 1721 carpenter , j. m. , meyer , m. r. , dougados , c. , strom , s . e. , & hillenbrand , l. a. 1997 , aj , 114 , 198 chen , w. p. , chiang , p.s . & 2004 , chjaa , 4 , 153 chernoff , d. f. , & shapiro , s. l. 1987 , apj , 322 , 113 chernoff , d. f. , & weinberg , m. d. 1990 , apj , 351 , 121 chini , r. , & neckel , t. 1981 , a&a , 102 , 171 cudworth , k. 1976 , aj , 81 , 519 damiani , f. , et al . 2004 , , 608 , 781 damiani , f. , prisinzano , l. , micela , g. , & sciortino , s. 2006 , a&a , 459 , 477 de grijs , r. , gilmore , g. f. , johnson , r. a. , & mackey , a. d. 2002a , , 331 , 245 de grijs , r. , gilmore , g. f. , mackey , a. d. , wilkinson , m. i. , beaulieu , s. f. , johnson , r. a. , & santiago , b. x. 2002b , , 337 , 597 girard , t. m. , grundy , w. m. , lpez , c. e. , & van altena , w. f. 1989 , aj , 98 , 227 goodwin , s. p. 1997 , mnras , 284 , 785 hensberge , h. , pavlovski , k. , & verschueren , w. 2000 , a&a , 358 , 553 higuera g. , m. a. , uribe , a. , & barrera , r. s. 2002 , rev . . de conf . ) , 14 , 33 hillenbrand , l. a. 1997 , aj , 113 , 1733 hillenbrand , l. a. , & hartmann , l. e. 1998 , apj , 492 , 540 hg e. , fabricius c. , makarov v.v . , urban s. , corbin t. , wycoff g. , bastian u. , schwekendiek p. & wicenec a. 2000 , a&a , 355 , l27 hunter , d. a. , shaya , e. j. , holtzman , j. a. , light , r. m. , oneil , e. j. , & lynds , r. 1995 , apj , 448 , 179 jones , b. 1971 , aj , 76 , 470 kalirai , j. s. , fahlman , g. g. , richer , h. b. , & ventura , p. 2003 , aj , 126 , 1402 king , i. 1966 , aj , 71 , 64 lada , c. j. , gottlieb , c. a. , gottlieb , e. w. , & gull , t. r. 1976 , apj , 203 , 159 lada , e. a. , depoy , d. l. , evans , n. j. , & gatley , i. 1991 , apj , 371 , 171 larson , r. b. 1991 , in : fragmentation of molecular clouds and star formation , iau symp . 147 , falgarone , e. , boulanger , f. , & duvert , g. , eds . , dordrecht : kluwer , p. 261 larson , r. b. 1993 , in : the globular cluster galaxy connection , smith , g. h. , & brodie , j. p. , eds . , asp conf . 48 , ( asp : san francisco ) , p. 675 larson , r. b. 2003 , asp conf . ser . , 287 , 65 li , j. z. 2005 , apj , 625 , 242 li , j. z. , & smith , m. d. 2005a , a&a , 431 , 925 li , j. z. , & smith , m. d. 2005b , aj , 130 , 721 lohman , w. 1972 , astron . nachr . , 293 , 259 loktin , a. v. , & beshenov , g. v. 2001 , astr . , 27 , 386 malumuth , e. m. , & heap , s. r. 1994 , aj , 107 , 1054 marschall , l. a. , van altena , w. f. , & chiu , l .- t . 1982 , aj , 87 , 1497 massey , p. , johnson , k. e. , & degioia - eastwood , k. 1995 , apj , 454 , 151 mathieu , r. d. 1984 , apj , 284 , 643 matthews , w. g. 1967 , apj , 147 , 965 mccall , m. l. , richer , m. g. , & visvanathan , n. 1990 , apj , 357 , 502 mcnamara , b. j. , & sanders , w. l. 1977 , a&a , 54 , 569 mcnamara , b. j. , & sekiguchi , k. 1986 , aj , 310 , 613 mermilliod , j .- c . , & paunzen , e. 2003 , a&a , 410 , 511 odenkirchen , m. , soubiran , c. , & colin , j. 1998 , newa , 3 , 583 ogura , k. , & ishida , k. 1981 , pasj , 33 , 149 park , b .- , & sung , h. , 2002 , aj , 123 , 892 paunzen , e. , & netopil , m. 2006 , mnras , 371 , 1641 prez , m. r. , the , p. s. , & westerlund , b. e. , 1987 , pasp , 99 , 1050 prez , m. r. , jones , m. d. , the , p. s. , & westerlund , b. e. 1989 , pasp , 101,195 prez , m. r. 1991 , rev . mex . a&a , 22 , 99 portegies zwart , s. f. , gaburov , e. , chen , h .- c . , & grkan , m. a. 2007 , mnras ( letters ) , in press ( astro - ph/0702693 ) prisinzano , l. , damiani , f. , micela , g. , & sciortino , s. 2005 , a&a , 430 , 941 prisinzano , l. , damiani , f. , micela , g. , & pillitteri , i. 2007 , a&a , 462 , 123 rauw , g. , naz , y. , gosset , e. , stevens , i. r. , blomme , r. , corcoran , m. f. , pittard , j. m. , & runacres , m. c. 2002 , a&a , 395 , 499 russell , j. l. 1976 , ph.d . thesis , university of pittsburgh sabogal - martnez , b. e. , garca - varela , j. a. , higuera g. , m. a. , uribe , a. , & brieva , e. 2001 , rev . a&a , 37 , 105 sagar , r. , & joshi , u. c. 1978 , mnras , 184 , 467 salpeter , e. e. 1955 , apj , 121 , 161 smith , l. j. , & gallagher , j. s. iii 2001 , mnras , 326 , 1027 sterzik , m. f. , alcala , j. m. , neuhuser , r. , & schmitt , j. h. m. m. 1995 , a&a , 297 , 418 stilburn , j. r. , stetson , p. b. , & fisher , w. a. 1992 , jrasc , 86 , 140 sung , h. , bessell , m. s. , lee h .- w . , kang , y. h. , & lee , s .- w . 1999 , mnras , 310 , 982 sung , h. , chun , m .- y . , & bessell , m. s. 2000 , aj , 120 , 333 the , p .- s . 1960 , apj , 132 , 40 townsley , l. k. , feigelson , e. d. , montmerle , t. , broos , p. s. , chu , y .- h . , & garmire , g. p. 2003 , apj , 593 , 874 turner , d. g. 1976 , apj , 210 , 65 van altena , w. f. , & jones , b. f. 1972 , a&a , 20 , 425 van den ancker , m. e. , the , p. s. , feinstein , a. , vzquez , r. a. , de winter , d. , & prez , m. r. 1997 , a&as , 123 , 63 van schewick , h. 1958 , veroff . sternwarte , no . 51 verschueren , w. 1991 , ph.d . thesis , vrije univ . brussels ( vub ) walker , m. f. 1957 , apj , 125 , 636 wang , j. j. , chen , l. , & zhao , j. h. 1996 , acta astr . sin . , 37 , 68 wang , j. j. 1997 , annals of shanghai observatory , acad . sinica , 18 , 45 wen , w. , zhao , j .- l . , & chen , l. 2006 , chin . a&a , 30 , 274 williams , j. p. , blitz , l. , & stark , a. a. 1995 , apj , 451 , 252 zhao , j .- l . , chen , l. , & wen w. 2006 , chja&a , 4 , 435 zhao , j. l. , & he , y. p. 1987 , acta astr . sin . , 28 , 374 cccccc [ tab1 ] @xmath49 & @xmath168 & @xmath169 & & @xmath168 & @xmath170 + @xmath171 & @xmath172 & @xmath173 & & @xmath174 & @xmath175 + @xmath176 & 1.39 mas yr@xmath2 & 8.88 mas yr@xmath2 & & 0.95 mas yr@xmath2 & 2.74 mas yr@xmath2 + @xmath177 & 1.56 mas yr@xmath2 & 9.54 mas yr@xmath2 & & 1.21 mas yr@xmath2 & 4.05 mas yr@xmath2 + cccc @xmath178(mas yr@xmath2 ) & @xmath179 & & @xmath180 + @xmath181(mas yr@xmath2 ) & @xmath182 & & @xmath183 + @xmath184(mas yr@xmath2 ) & @xmath185 & & @xmath186 + @xmath187(mas yr@xmath2 ) & @xmath188 & & @xmath189 + @xmath190(mas yr@xmath2 ) & @xmath191 & & @xmath192 + @xmath193(mas yr@xmath2 ) & @xmath194 & & @xmath195 + @xmath196(mas yr@xmath2 ) & @xmath197 & & @xmath198 + @xmath199 & @xmath200 & & @xmath201 + @xmath202 & @xmath203 & & @xmath204 + @xmath205 & @xmath206 & & @xmath207 + rrrrrrrrr 1 & 06 31 35.826 & + 04 30 51.70 & @xmath2086.142 & 0.919 & 4.685 & 0.405 & 8.720 & 0.62 + 2 & 06 32 10.470 & + 04 57 59.75 & @xmath2081.600 & 1.486 & @xmath2080.421 & 2.184 & 8.350 & 0.92 + 3 & 06 31 36.322 & + 04 55 59.53 & 0.336 & 0.464 & @xmath2084.710 & 1.008 & 8.830 & 0.89 + 4 & 06 31 38.396 & + 05 01 36.38 & @xmath2081.673 & 0.639 & @xmath2080.232 & 1.303 & 8.110 & 0.89 + 5 & 06 31 16.839 & + 04 37 25.92 & @xmath2080.074 & 0.425 & @xmath2083.398 & 1.127 & 9.039 & 0.73 + rrrrrrrrr 1 & 18 04 25.848 & @xmath20824 23 08.27 & 6.469 & 0.815 & @xmath2086.376 & 2.249 & 8.55 & 0.94 + 2 & 18 06 04.702 & @xmath20824 11 43.84 & 5.989 & 0.591 & @xmath2088.649 & 1.670 & 8.64 & 0.83 + 3 & 18 04 15.222 & @xmath20824 11 00.07 & 2.958 & 0.877 & @xmath2082.713 & 1.067 & 9.12 & 0.99 + 4 & 18 02 39.798 & @xmath20824 14 47.40 & 7.732 & 0.405 & 0.474 & 3.538 & 9.22 & 0.74 + 5 & 18 04 14.538 & @xmath20824 14 36.31 & 8.511 & 0.853 & 9.241 & 1.665 & 9.35 & 0.00 +
we define a _ sociotechnical system _ as one involving interactions between _ autonomous _ social entities such as people and organizations mediated by technical components . by emphasizing the autonomous nature of social entities , our definition generalizes over , yet more precisely captures , the traditional connotation of the interaction between humans and societal infrastructure . our definition contrasts with one of the conventional uses of the term , which covers any interaction between people and computers . researchers in software engineering ( se ) picked up the theme of sociotechnical systems in two major directions : ( 1 ) how to model a sociotechnical system as a combination of social and software components , as in @xcite ; and ( 2 ) how to elicit , model , and manage the requirements of the social components so that suitable software components may be designed @xcite . often , there is substantial overlap between the two directions , for example , as in i * @xcite . baxter and sommerville conceptualize a sociotechnical system as one that _ actively pursues goals _ in an organizational setting ; that is , a sociotechnical system is one actor : a single locus of control . therefore , we refer to such systems as being _ conceptually centralized_. the centralized conception is shared by current se approaches . for clarity , we reserve the term _ principal _ to refer to an autonomous social party who participates in a system at runtime and the term _ stakeholder _ to refer to one who originates some of the requirements . in general , any principal would also be a stakeholder in the system ( suitably abstracted via the notion of _ roles _ ) . we restrict our attention to systems whose membership and structure can change dynamically . open _ sociotechnical systems are properly viewed as _ societies _ of principals : we can not build a `` complete '' system but can only specify how its members may _ interact _ , leaving as out of our scope the engineering of the members themselves . thus whether the members _ comply _ with the specified interactions is crucial . because of its emphasis on interaction , we term our approach interaction - oriented software engineering or _ iose_. we claim that iose is necessary and show that the centralized conception is not applicable in sociotechnical systems . common settings such as business services and social computing realize open sociotechnical systems . consider schield et al.s description of a healthcare system in the united states , including its key stakeholders and their ( presumed ) objectives . * us society : the collective public , private , and personal interests of the united states citizens . * regulatory bodies : public and private that make and enforce policies . * mco : managed care organizations ( i.e. , licensed insurance organizations ) , who offer and administer managed care plans . * institutional providers : hospitals and laboratories . * clinical or professional providers : individual doctors and communities of practice . * employers : fully insured and self - funded organizations who offer managed care plans to their employees . * consumers : the enrollees of managed care plans . * medicaid or medicare beneficiaries . schield et al . find that the stakeholders objectives often conflict . for example , payers must keep costs down whereas providers must maximize revenue ; insurers want to shift risk to providers and consumers for matters of cost control whereas both providers and consumers want protection from potentially catastrophic costs . consumers want shorter waiting periods for appointments with their physicians whereas physicians want to increase their panel of patients . further , schield et al . point out how one principal can , in pursuing its own objectives , compromise the objectives of others . for example , making a patient wait longer may compromise the patient s health , thereby increasing the cost of care and decreasing patient satisfaction . regulatory bodies have the objective of making sure mcos and providers play by the rules and ensuring that cost , quality , and access criteria are met but that conflicts with society s goals of lower - cost healthcare . assume that the stakeholders for a healthcare system manage to resolve their differences and specify a system that meets their stated collective requirements . now , individual _ principals _ such as specific mcos , hospitals , physicians , employees , and consumers can join or leave the managed healthcare system of their own volition . the principals act according to their own private business policies , some of which may not have been envisaged by the stakeholders . for example , an mco may have a private objective to acquire an independent call center . further , principals would generally employ their own internal information systems to support their interactions with other principals . for example , providers may set up appointment systems to handle appointment - related communications with consumers . hospitals and mcos may use complex information systems that help process payments from each other and the consumers . mcos may employ complex actuarial and decision - support processes in handling claims . recognizing and addressing conflicts among stakeholders requirements has been a longstanding area of research in software engineering . once the conflicts are addressed , through whichever means , engineers would seek to model and realize a system that meets those requirements . this paper treats conflict analysis as out of scope and instead focuses on identifying modeling assumptions and criteria that capture sociotechnical systems more faithfully . this paper makes two main contributions . first , it shows via conceptual analysis why iose is ( 1 ) a novel paradigm and ( 2 ) well - suited for the software engineering of open sociotechnical systems . our argument proceeds as follows . one , we show that the foundational models of traditional se are conceptually centralized . this is because traditional se concerns itself with the engineering of what is conceptually a single machine ( section [ sec : machine - orientation ] ) . two , we present the conceptual model of a system in iose ( section [ sec : iose ] ) and discuss how it accommodates multiple perspectives . three , based on the characteristics of sociotechnical systems , namely , autonomy , accountability , and loose coupling , we formulate criteria that any se approach for them should meet and show that whereas machine - orientation , and therefore traditional se , fails the criteria , iose meets them ( section [ sec : eval - machine - iose ] ) . the second contribution of this paper lies in reinterpreting the broad se principles of modularity , encapsulation , abstraction , and the separation of concerns in accordance with the above criteria ( section [ sec : principles - iose ] ) . the reinterpretation provides the elements of a methodology for iose . further , we discuss in detail well - known software methodologies that are motivated either by sociotechnical systems or the modeling of interactions among principals and show how they violate one or more of these principles ( section [ sec : prominent - approaches ] ) . section [ sec : literature ] discusses additional literature that is relevant to iose . section [ sec : discussion ] concludes the paper with pointers to future directions . figure [ fig : re - conceptual - model ] ( from van lamsweerde ) shows the traditional conceptual model for se . the _ system as is _ represents the system with identified problems , inefficiencies , and limitations . the _ system to be _ , whose _ objectives _ are to avoid those deficiencies , is to be engineered . the idea is to understand the problem domain in order to come up with a set of _ services _ , _ constraints _ , and _ assumptions _ under which the stated objectives would be met . some of these would be met by the _ software to be _ as part of the system to be ; the rest would be assigned as responsibilities to components in the _ environment _ , namely , _ people _ , _ devices _ , and _ existing software_. = [ blue!40!black , font= * * ] = [ font= ] = [ align = center ] * * = [ draw , rectangle , align = center , minimum height=5ex , text width=8ex ] ( why ) at ( 0,5.5 ) why ? ; ( what ) at ( 0,3.2 ) what ? ; ( who ) at ( 0,0 ) who ? ; ( cloud ) at ( 7.0,5.75 ) objectives ; ( core ) at ( 2.4,5.55 ) [ cols="^ " , ] ; background ( lbl.north-\| e.east ) rectangle ( p.south-\| p.west ) ; even more generally , given a set of requirements , the essential idea is to come up with the specification of a _ machine _ ( equivalently , the software to be above ) that along with reasonable domain assumptions satisfies the requirements of the stakeholders @xcite . the _ system as is _ is the system as it exists without the machine and helps us understand the environment in which the machine will be introduced . the system to be is what the system will be when the machine is introduced . if all goes well in the system to be , the stakeholders requirements are satisfied . it is important to understand the nature of the machine . figure [ fig : machine - environment ] ( from @xcite but based on @xcite ) illustrates an operational conceptualization of the machine - environment interface . here , the _ software to be _ represents the machine . a machine is a _ controller _ that maps inputs from its environment ( by monitoring certain variables ) to outputs or effects in the environment ( by setting certain variables ) . in other words , it processes inputs to produce outputs . to its users ( in the environment ) , a machine provides computational services ( functionality ) . in its very conception , a machine corresponds to a single locus of control . indeed , it is common in software engineering to talk of machines as acting in pursuit of goals @xcite . traditional se is machine - oriented it concerns itself with the specification of a machine ( even if implemented via distributed components ) that would meet stakeholder requirements . thus traditional se reflects a _ conceptually centralized _ perspective , namely , of the stakeholders , considered collectively . = [ thick , draw = none , text width=15ex , align = center ] = [ box , fill = blue!20!gray!40 ] = [ box , fill = black!10 ] = [ box , draw = black , thin , fill = white , text = black ] = [ blue!40!black , draw = none , text width=10ex , align = center ] = [ font= ] = [ align = center ] ( machine_envt ) [ draw = none , row sep=1 mm , column sep=3 mm ] ( empty - tl ) ; & ( input ) input devices + ( sensors ) ; & ( empty - tr ) ; + ( environment ) environment ; & & ( software ) software to be ; + ( empty - bl ) ; & ( output ) output devices + ( actuators ) ; & ( empty - br ) ; + ; ( environment ) .. controls ( empty - tl ) .. node [ edge_label , above ] monitored + variables ( input ) ; ( input ) .. controls ( empty - tr ) .. node [ edge_label , above ] input + data ( software ) ; ( output ) .. controls ( empty - bl ) .. node [ edge_label , below ] controlled + variables ( environment ) ; ( software ) .. controls ( empty - br ) .. node [ edge_label , below ] output + data ( output ) ; the above account of machine - orientation , taken from leading writings on se , is agnostic to particular software methodologies . machine - orientation manifests itself in specific modeling notations and methodologies , some of which have been highly influential in se . many existing requirements - based approaches , although differing in their details , instantiate the same concepts at their core : machine , environment , system as is , and system to be . tropos @xcite and i * @xcite refer to a model of the system as is and the system to be as the early and late requirements models , respectively . in their terminology , the environment is a set of _ actors _ ; the machine itself is the _ system actor_. for example , tropos would create a system actor for an entire healthcare system . this actor would capture ( a consistent subset of ) the goals of all stakeholders , thereby functioning as a logically centralized machine . following kaos @xcite , one would create a set of _ agents _ with designer - assigned individual goals . however , the set of agents is conceptually a single machine because there is a single locus of design . this follows from the fact that goals are assigned to agents and , therefore , at a high - level the agents have already been specified . indeed , referring specifically to kaos ( and to feather s work , which provides the conceptual basis for kaos ) , zave and jackson observe that even though kaos supports multiple agents , it is only a refinement of their more basic single - machine framework . let us apply the above conceptual model of systems to the healthcare scenario . imagine that a healthcare machine were built to a specified set of stakeholder requirements . the machine would control various medical devices and other equipment . users such as consumers , providers , mcos , and so via on would interface with the machine via a web interface . consider the example of scheduling an appointment with a physician . the physician would have configured his or her preferences in the machine : his or her daily schedule , how long a time period to allot each patient , and so on . those are the physician s inputs to the machine , which the machine uses in scheduling patient appointments . the patient s interface would display the available slots and enable him or her to choose from among those slots . the machine records each selection , and produces as output a confirmation of the appointment , perhaps also by email . further , the machine grays out the slot for future scheduling . the machine additionally supports the processing of claims and payments among hospitals and insurers . the machine would encode the insurance company s process of queuing claims above $ 10,000 separately for detailed examination . the insurance company through its interface can obtain the claims from the appropriate queues . the machine would implement the appropriate access control so that users can access the appropriate functionality . traditional se , in spite of its conceptually centralized perspective , may be applied to yield a physically distributed machine . web applications suggest physical distribution and applying kaos would result in a physically distributed set of agents . in general , techniques from software architecture may be applied to internally decompose a machine into distributed components . rapanotti et al . employ such an approach in the decomposition of machines in the problem frames approach . we define a _ protocol _ of a sociotechnical system as a specification of how interactions among its roles would proceed . in business settings , the protocol can take the form of a business contract that principals would enter into upon adopting different roles . for example , in texas , if medco wishes to play the role of mco , it must enter into a _ uniform managed care _ contract with the state health commission ( which alone plays the regulatory body role ) @xcite . at any point during the interaction , the set of social expectations represents the _ social state _ of the system . a protocol serves three purposes . one , it makes public the social expectations of the participants while giving them the flexibility to follow their individual objectives . for example , public funding agencies may expect that mcos , upon notification , refund payments made in error within 30 days . two , it identifies which participant is accountable to which participant for what expectation . for instance , if medco , who plays mco , does not refund an erroneous payment in time , the funding agency may hold medco accountable . three , it frees participants to implement their information systems as they please as long as they satisfy the given protocol . for example , medco might employ any representation it chooses and may apply its policies in deciding whether to provide a refund early or late within the 30-day window . = [ thick , draw = none , text width=11ex , align = center ] = [ box , fill = green!10!blue!30!gray!60 ] = [ box , fill = black!10 ] = [ box , fill = green!10!blue!30!gray!40 ] = [ box , draw = black , thin , fill = white , text = black ] = [ blue!40!black , draw = none , text width=10ex , align = center ] = [ font= ] = [ align = center ] ( machine_envt ) [ draw = none , row sep=1 mm , column sep=3 mm ] ( environment - l ) environment ; & & & ( environment - r ) environment ; + & ( empty - top - l ) ; & ( empty - top - r ) ; + ( empty - principal - l ) ; & ( principal - l ) principal ; & ( principal - r ) principal ; & ( empty - principal - r ) ; + ; + ( empty - ml ) ; & ( role - l ) role ; & ( role - r ) role ; & ( empty - mr ) ; + ; + & ( empty - bcl ) ; & ( empty - bcr ) ; + ; ( sociotechnical system ) sociotechnical system as protocol ; ( protocol ) + meanings ; ( principal-l.south ) .. controls ( protocol.south ) .. ( principal-r.south ) ; ( environment-l.south ) .. controls ( empty - principal - l ) .. ( principal-l.west ) ; ( environment - r ) .. controls ( empty - principal - r ) .. ( principal - r ) ; background ( empty - top - l.north-\| principal-r.east ) rectangle ( protocol.south-\| principal-l.west ) ; accordingly , the objective of iose is to specify a _ sociotechnical system _ as a protocol that defines two or more _ roles_. figure [ fig : system ] describes how iose applies . for the sociotechnical system to be instantiated , principals adopt roles in the protocol . each principal continues to interact with its environment even after it joins the system . the principals communicate with each other within the scope of the system , which means that they are subject to the meanings defined in the protocol that specifies the system . in engineering terms , the main artifact produced by iose is the protocol which , through its roles , itself serves as a requirement for the principals who would adopt those roles . whereas traditional se seeks to specify a machine , iose seeks to specify a protocol . whereas a machine captures a single locus of control , iose naturally supports multiple perspectives as roles in a protocol . in iose , each principal , who adopts one or more roles , is a locus of control a reasoner and a sender and receiver of communications . we introduce commitment protocols @xcite as an example iose approach . a ( social ) commitment captures an elementary business relationship between a debtor and a creditor @xcite . specifically , the commitment @xmath0 says that the debtor @xmath1 commits to the creditor @xmath2 that if the antecedent @xmath3 comes to hold , then it will bring the consequent @xmath4 . the elementary operations on commitments are create , release , cancel , delegate , and assign . the social nature of commitments owes to the fact that they progress due to interactions among principals , not due to the internal reasoning of any principal . commitment protocols exploit this connection by specifying the meanings of _ messages _ in terms of how they affect commitments , thus enabling principals to interact flexibly . other abstractions could potentially be used in addition to commitments but , for simplicity , we confine the present discussion to commitments . table [ table : scheduling - protocol ] shows a partial appointment scheduling protocol . the protocol involves two roles , phy ( for physician ) and pat ( for patient ) . assume that via the enactment of some other protocol , ( any physician playing ) phy is committed to ( any patient playing ) pat to provide the latter with a list of available appointment slots upon the latter s request for appointment ( this commitment , for instance , could be set up when a patient registers with a physician for the first time ) . we represent this commitment as @xmath5 . the availableslots message from phy to pat conveys the list of available slots to the pat : it means that phy commits to pat that if pat commits to show up for one of those slots , then phy will commit to showing up for that slot as well . the select message from pat to phy commits pat to a selected slot . the confirmslot message from phy to pat for a particular slot commits phy to the slot . a complete and effective meeting scheduling protocol would additionally include messages that deal with meeting cancellation and rescheduling . figure [ fig : ms - progression ] depicts the progression of social state according to the meanings in table [ table : scheduling - protocol ] and using the commitments introduced in table [ table : scheduling - protocol - commitments ] . alessia plays phy and bianca plays pat . upon bianca sending the requestappointment message to alessia , the antecedent of @xmath6 is satisfied . this makes alessia unconditionally committed to sending her the available slots ( @xmath7 ) . when alessia sends the list of available slots ( at 1400 and 1600 hours ) , she discharges @xmath7 and creates @xmath8 . bianca communicates her selection of 1400 , which makes alessia unconditionally committed to confirming it ; in other words , bianca creates @xmath9 , which creates @xmath10 ( the unconditional version of @xmath8 ) . finally , alessia confirms the selected slot , thus creating @xmath11 and discharging @xmath10 . in the final state of figure [ fig : ms - progression ] , both alessia and bianca are committed to each other for showing up at the selected time . = [ thick , draw = none , text width=7ex , minimum height=7ex , align = center ] = [ box , fill = green!10!blue!30!gray!60 ] = [ blue!40!black , auto = true , font= ] = [ edge_label , fill = white , auto = false ] ( soc-1 ) [ draw = blue!20!black , row sep=25 , column sep=0 ] ( alessia-1 ) alessia + mi ; + ( bianca-1 ) bianca + mp ; + ; ( soc-2 ) [ draw = blue!20!black , row sep=25 , column sep=0,right of = soc-1 ] ( alessia-2 ) alessia + mi ; + ( bianca-2 ) bianca + mp ; + ; ( soc-3 ) [ draw = blue!20!black , row sep=25 , column sep=0,right of = soc-2 ] ( alessia-3 ) alessia + mi ; + ( bianca-3 ) bianca + mp ; + ; ( soc-4 ) [ draw = blue!20!black , row sep=25 , column sep=0,below of = soc-3,left of = soc-3,node distance=90 ] ( alessia-4 ) alessia + mi ; + ( bianca-4 ) bianca + mp ; + ; ( soc-5 ) [ draw = blue!20!black , row sep=25 , column sep=0,left of = soc-4 ] ( alessia-5 ) alessia + mi ; + ( bianca-5 ) bianca + mp ; + ; ( alessia-1.south ) to node [ edge_label_in ] @xmath12 ( bianca-1.north ) ; ( alessia-2.south ) to node [ edge_label_in ] @xmath13 ( bianca-2.north ) ; ( alessia-3.south ) to node [ edge_label_in ] @xmath14 ( bianca-3.north ) ; ( alessia-4.240 ) to node [ edge_label_in ] @xmath15 ( bianca-4.120 ) ; ( bianca-4.60 ) to node [ edge_label_in ] @xmath16 ( alessia-4.300 ) ; ( alessia-5.240 ) to node [ edge_label_in ] @xmath17 ( bianca-5.120 ) ; ( bianca-5.60 ) to node [ edge_label_in ] @xmath16 ( alessia-5.300 ) ; ( soc-1.east ) to node [ pos=0.5,edge_label_in ] requestappointment ( ) ( soc-2.west ) ; ( soc-2.east ) to node [ pos=0.5,edge_label_in ] availableslots(1400,1600 ) ( soc-3.west ) ; ( soc-3.south ) to node [ pos=0.5,edge_label_in , xshift=10 ] selectslot(1400 ) ( soc-4.east ) ; ( soc-4.west ) to node [ pos=0.5,edge_label_in ] confirmslot(1400 ) ( soc-5.east ) ; protocols for handling claims between providers and mco would be more complex than those for appointment scheduling . desai et al . specify protocols for automobile insurance and show how to compose protocols . we focus on appointment scheduling because ( 1 ) it illustrates the essential concepts of iose , and ( 2 ) it bears similarity to meeting scheduling , an application that has been studied extensively in the literature , and which we discuss in detail in section [ sec : prominent - approaches ] . figure [ fig : system - types ] describes three takes on a sociotechnical system . figure [ fig : old ] shows the most traditional setting , which even predates it . here , one can imagine a sociotechnical system as realized by two or more principals , each with its own ledgers , and interacting via the postal service or foot messenger . the principals are autonomous . the same situation holds in the case of digital communication , where we can think of the messaging system as providing the necessary message delivery functionality . in this case too , the messages the principals send and receive are opaque to the communication network . in either case , there is no computational support for the social state . indeed , there is no computational support for anything beyond message transport and , possibly operational constraints such as message ordering . each principal maintains its expectations and commitments in its ledgers and acts according to its local policies . for example , a physician may send a request to a laboratory to find a patient s cholesterol and the laboratory may send back the results along with an invoice . = [ thin , draw , align = center , anchor = center , minimum height=20,inner sep=2pt , font= ] = [ box , fill = blue!20!gray!20,draw = none , minimum height=6ex , rounded corners ] = [ box , draw = none , circle , fill = green!10!blue!30!gray!60,minimum width=6 ] = [ shape = cloud , align = center , thick , aspect=4,fill = green!10!blue!30!gray!40 ] = [ blue!40!black , draw = none ] = [ annotation , midway , anchor = south , align = center ] = [ align = center ] = [ anchor = base east ] = [ anchor = base east ] 0.3 ( trust ) [ draw = none , matrix of nodes , row sep=10,column sep=0 ] ( xx ) ; & [ 35]&[35 ] ( yy ) ; + ( empty - tl ) ; & & ( empty - tr ) ; + ( empty - ml ) ; & & ( empty - mr ) ; + ( empty - bl ) ; & & ( empty - br ) ; + ; let 1 = ( xx.west ) , 2 = ( yy.east ) , 3 = ( empty-bl.south west ) , 4 = ( empty - ml ) in node ( bl ) at ( 1 + 0,3 ) node ( br ) at ( 2 - 0,3 ) node ( tl ) at ( 1 + 0,4 - 10 ) node ( tr ) at ( 2 - 0,4 - 10 ) ; node [ activity , fit= ( bl ) ( br ) ( tl ) ( tr ) ( empty-bl.south ) ( empty-br.south),inner sep=0,minimum height=40 ] ( communication ) ; let 1 = ( communication.west ) , 2 = ( communication.east ) , 3 = ( communication.north ) in node [ font=,draw = none , align = center , minimum height=40,anchor = north ] at ( 3,3 - 0 ) + ; ( yy.south ) \|-(yy.south\|-communication.center)-\| ( xx.south ) ; 0.3 ( trust ) [ draw = none , matrix of nodes , row sep=10,column sep=0 ] ( xx ) ; & [ 35]&[35 ] ( yy ) ; + ( empty - tl ) ; & ( empty - tr ) ; + ( empty - ml ) ; & ( empty - mr ) ; + ( empty - bl ) ; & & ( empty - br ) ; + ; let 1 = ( xx.west ) , 2 = ( yy.east ) , 3 = ( empty-bl.south west ) , 4 = ( empty - ml ) in node ( bl ) at ( 1 + 0,3 ) node ( br ) at ( 2 - 0,3 ) node ( tl ) at ( 1 + 0,4 - 10 ) node ( tr ) at ( 2 - 0,4 - 10 ) ; node [ activity , fit= ( bl ) ( br ) ( tl ) ( tr ) ( empty-bl.south ) ( empty-br.south),inner sep=0,minimum height=40 ] ( communication ) ; let 1 = ( communication.west ) , 2 = ( communication.east ) , 3 = ( communication.north ) in node [ font=,draw = none , align = center , minimum height=40,anchor = north ] at ( 3,3 - 0 ) + ; ( xx.south ) to node [ edge_label , right ] api ( xx.south\|-communication.north ) ; ( yy.south ) to node [ edge_label , left ] api ( yy.south\|-communication.north ) ; 0.3 ( trust ) [ draw = none , matrix of nodes , row sep=10,column sep=0 ] ; & ( social ) ; & ; + ( xx ) ; & & ( yy ) ; + ( empty - tl ) ; & [ 70 ] ( empty - tr ) ; + ( empty - ml ) ; & & ( empty - mr ) ; + ( empty - bl ) ; & & ( empty - br ) ; + ; let 1 = ( xx.west ) , 2 = ( yy.east ) , 3 = ( empty-bl.south west ) , 4 = ( empty - ml ) in node ( bl ) at ( 1 + 0,3 ) node ( br ) at ( 2 - 0,3 ) node ( tl ) at ( 1 + 0,4 - 10 ) node ( tr ) at ( 2 - 0,4 - 10 ) ; node [ activity , fit= ( bl ) ( br ) ( tl ) ( tr ) ( empty-bl.south ) ( empty-br.south),inner sep=0,minimum height=40 ] ( communication ) ; let 1 = ( communication.west ) , 2 = ( communication.east ) , 3 = ( communication.north ) in node [ font=,draw = none , align = center , minimum height=40,anchor = north ] at ( 3,3 - 0 ) + ; \(xx ) to node [ edge_label , anchor = north ] protocol ( yy ) ; ( yy.south ) \|-(yy.south\|-communication.center)-\| ( xx.south ) ; ( social.south ) \|-(social.south\|-xx.east ) ; figure [ fig : machine ] shows a more sophisticated , but traditional , approach wherein a machine mediates all interactions among the principals . the machine offers an api to the principals through which they can request changes in the social state . such conceptually centralized machines are not common for healthcare today , so we use ebay as an example and then return to healthcare . the auctioneer ebay offers a machine for conducting auctions that provides an api ( requiring little more than a web browser ) through which sellers can list items for sale and buyers can bid for listed items . the machine determines the social state : whether something is on sale , its reserve price , the time an auction closes , whether a bid is legitimate , the winning bid , and so on . the principals can represent their local view in their ledgers but what matters is the ebay machine s view . the machine happens to usually respect the requests of buyers and sellers but it does not need to . returning to the healthcare machine of section [ sec : machine - orientation ] , the patient may attempt to schedule an appointment with the physician , but the machine determines whether the appointment holds , i.e. , whether the patient and physician are each committed to showing up at the appointed time . the patient and the physician may maintain their local views of the social state but what the machine maintains is definitive . the same situation arises between the hospital and medco regarding their payments . in this setting , the only way the parties can interoperate is if they maintain their local views tightly synchronized with the central machine . in contrast , figure [ fig : iose ] shows how the principals enact a protocol with one another . the enactment takes advantage of infrastructure such as for communication . however , the social state is determined not by the infrastructure but through the enactment of the protocol . that is , the protocol specifies how the social state progresses . here too the principals can maintain their individual ledgers but what the protocol specifies is definitive . the important challenge of formalizing protocols to ensure that the local views remain sufficiently consistent with the protocol - specified `` public '' state without synchronization is out of our present scope . chopra and singh present a relevant approach . let us use the healthcare example of section [ sec : healthcare ] to motivate the key criteria that any software engineering methodology for sociotechnical systems must support . accommodate autonomy the autonomy of a principal ( a human , organization , or other social entity ) means that it is an independent locus of control with its own goals and policies , which reflect its business motivations and are normally private . thus even as a participant in a system , each principal is free in principle to act in pursuit of its own goals , without any consideration of others . for example , a hospital may decide not to entertain patients with a particular insurance provider , or physicians may increase the panel of patients they treat at the expense of reducing availability . figure [ fig : old ] is compatible with autonomy only to the extent that it does not model social state . figure [ fig : machine ] violates autonomy because the machine not only determines the definitive social state but also controls transitions on it . the principals have no direct relationship with each other . for example , a bid may be declined and a seller may not be able to change his mind and accept a bid lower than the reserve price . contrast the above with protocols . a protocol is not a computationally active entity and can not provide a service . a protocol simply specifies the correctness criteria for interaction among the participants . figure [ fig : iose ] expresses a protocol that specifies the appropriate social relationships among potential principals , capturing their legitimate expectations of each other . further , principals would enter into these relationships of their own volition ; no relationship is forced upon a principal , not even to ensure compliance . support accountability accountability is the flip side of autonomy . a participant is free to act as it pleases ; however , it is accountable to those that have legitimate social expectations of it . for example , a laboratory may legitimately expect that the mco transfer funds for services provided within a certain time after submission of claims . this means that the mco is accountable to the laboratory for transferring funds in time ; if it does not , it violates the laboratory s expectation . figure [ fig : old ] provides no support for accountability because it does not represent the social state . figure [ fig : machine ] holds each principal accountable to the central authority and , if it permits , to the other principals . for example , a buyer is accountable to ebay to pay in time . but if she persuades ebay to grant her an extension , the seller has no recourse within the ebay system . figure [ fig : iose ] bases accountability on a protocol and thus each principal is accountable to any other principal in whom it has created a legitimate expectation . in particular , through the protocol , each principal ought to be able to judge the compliance with its expectations of each principal with whom it interacts . accommodate loose coupling loose coupling captures the idea that each principal is an independent locus of design . it has full control over the design of its own information systems but no control over the design of other principals information systems . for example , a hospital may store information about a surgeon s willingness to work the night shift whereas the insurance company may capture the total charges for procedures led by that surgeon . and , hospitals in different jurisdictions may capture different patient monitoring data during a surgery . figure [ fig : old ] is compatible with loose coupling insofar as it elides the treatment of social state . in practice , it forces an ad hoc design of the interaction ( e.g. , message formats ) . because the message meanings are hidden , the principals who adapt their information systems risk failing interoperability . figure [ fig : machine ] requires that the principals adopt the same representation of the social state as the machine . the meanings of the messages sent over the api could be explicit but they usually are hardcoded in the machine , and sometimes in the principals information systems , if any . this produces a tighter coupling than desirable between the principals and the machine ( both considered as endpoints ) : when the machine changes , the principals must potentially change their own behaviors and any information systems that support them accordingly . by contrast , figure [ fig : iose ] employs a protocol with explicit meanings ( not hardcoded in any principals information system ) for the interactions . the iose approach , in effect , makes public the extent of the coupling . the principals may adapt their information systems without consideration of others as long as they follow the protocol . we introduce the core principles of iose . specifically , we contrast how the key principles of se : modularity and encapsulation @xcite , separation of concerns @xcite , and abstraction are manifested in current modeling approaches with how they are manifested in iose . we find that although the intuitions behind the principles hold , the technical details are completely different and , in many cases , antithetical to what we encounter in conventional se . modularity refers to the functional , most commonly hierarchical , decomposition of systems @xcite . the benefit of appropriately modularizing systems is composability . in the extreme , modules from different vendors may be composed as needed . as figure [ fig : machine - environment ] shows , the first - level application of modularity in se is the decomposition of the system into the environment and the machine . rapanotti et al . apply architectural patterns to decompose a problem frame into its constituent problem frames . kaos , i * , and tropos , being agent - oriented , support finer grained units of modularity . the agents therein may have responsibilities or provide services , as captured by the dependencies among them . in iose , a principal as an autonomous social entity is the natural unit of modularity . with autonomy comes accountability , as motivated by mamdani and pitt . without the ability to check compliance , though , accountability would be meaningless . _ example . _ although bianca is free to not show for the appointment once she commits to a slot , she is accountable for her commitment . _ benefit . _ promotes autonomy by not unduly restricting a principal s course of action . promotes accountability by providing a basis for ensuring _ correctness _ : a principal who does not comply may face sanctions from principals to whom it is accountable for the given expectation each principal autonomously becomes the debtor of any commitments @xcite . that is , the debtor must have initiated an interaction ( sent a message ) that leads to it being committed . in some cases , the message could itself create the commitment . in other cases , the debtor may have created some commitment ( as debtor ) whereby actions by other parties could lead to the creation of the given commitment . because commitments are created ultimately due to the communications of the debtor , the debtor is accountable for them . demands placed on a principal other than as the debtor of a commitment have no bearing on compliance or enforcement . abstraction refers to the level of the concepts used in a specification . the ideal abstraction is sufficiently high - level to hide details and reduce complexity , yet sufficiently low - level to support drawing the necessary conclusions . tropos and i * offer high - level abstractions such as goals , capabilities , and goal dependencies . sommerville et al . apply high - level notions of responsibility and delegation to requirements modeling . using high - level abstractions places requirements away from low - level notions such as tasks , plans , and state machines . in contrast , iose emphasizes abstractions that capture the meaning of an interaction . explicit social meaning make all social expectations explicit in the system specification . the meanings of individual communications must be explicitly formalized in terms of what they _ count as _ in the society being designed . in general , the meaning involves the creation or other manipulation of the commitments among the parties involved . _ example . _ table [ table : scheduling - protocol ] specifies the message meanings in an appointment scheduling protocol . _ benefit . _ explicit social meaning promotes loose coupling . as figure [ fig : ms - progression ] illustrates , the meaning captures how the principals social state progresses . the true social state progresses even if we have no computational representation of it . but lacking an explicit meaning , each principal could interpret messages in incompatible ways . for example , in the healthcare setting , a laboratory may interpret the messages with the claim information as leading to an unconditional commitment from the mco to honor the claim . the mco , however , may interpret the commitment as being conditional on the claim being valid . such misalignments could have serious repercussions for the principals ( e.g. , in producing their balance sheets ) and may lead to legal disputes . if the principals negotiate the meanings of the messages and hard - code them in their information systems , they would produce hidden couplings among themselves : changes in how one principal handles messages would need to be propagated to the others . additionally , explicit social meaning promotes accountability because if the meanings are public , then principals can potentially check the compliance of themselves and others . solely social meaning a system specification must specify _ only _ the possible communications and their meanings and nothing else . further , the meaning must be expressed in terms of social abstractions such as commitments . specifications of any operational details that have significance at the social level ( e.g. , a convention to pay first ) , must be captured via social abstractions ( e.g. , commitments ) . specifications that capture ordering and occurrence constraints separately from the meaning violate this principle . for example , one could specify in the appointment scheduling protocol that availableslots follow requestappointment . but here no one is accountable for the constraint : is the physician at fault for not delaying sending availableslots or is the patient at fault for not sending requestappointment ? instead , if this constraint is necessary as a social requirement , one or more of the principals should commit to enforcing it , e.g. , adopting marengo et al.s approach and placing temporal constraints in commitments . further , specifying the technical infrastructure does not capture a sociotechnical system . we might either model the social actor who provides the infrastructure and engage it via commitments or omit such technical constraints altogether since they apply at a lower level of abstraction . _ example . _ the above ordering constraint can be expressed @xcite as @xmath18 where the dot ( ` @xmath19 ' ) means _ occurs before_. _ benefit . _ promotes autonomy and accountability . no central controller enforces constraints in a sociotechnical system . every constraint logically ought to be some principal s responsibility . expressing a constraint as a hard constraint to be enforced magically by the environment either under - specifies the functioning of the system or ( in most current thinking ) postulates a central entity that is the sole autonomous principal and can impose its will upon all the other participants . separation of concerns refers to the treatment of each aspect of a problem independently of _ yet _ in relation to others . it refers to the sorting out of the different threads from what would otherwise be a tangled mess . often , the invocation of this principle is implicit . for example , zave and jackson s identification of domain assumptions , machine requirements , and user requirements as the separate but essential categories for re is at its heart an application of this principle . dardenne et al . and yu express early requirements in the form of goal models , thus separating the exploration of the problem space from the solution space . mylopoulos et al . separate nonfunctional from functional requirements . finkelstein et al . separate concerns explicitly based on stakeholders , acknowledging the fact that , in general , each stakeholder has different concerns and may employ different representations for expressing them . for sociotechnical systems , we must separate social and technical considerations . principals such as people and organizations must be distinguished from technical entities such as resources , legacy systems , software components , devices , communication infrastructure , and other technical objects in the environment . this is because social relationships are meaningful only among principals . a principal may only bear a control relationship ( e.g. , ownership , invocation , or access ) toward a technical entity as may one such entity toward another . only principals are autonomous and accountable : a patient can not sue an operating table but can sue a surgeon or a hospital . _ example . _ as figure [ fig : ms - progression ] shows , alessia and bianca maintain a social relationship . however , each of them maintains and controls an information system , which is not socially visible . _ benefit . _ separating the social and technical entities makes clear the kinds of relationships that would make sense among them . promotes accountability by making clear only principals are accountable to each other in the social sense . enables engineers of sociotechnical systems to focus solely on the social aspects . encapsulation refers to the principle that a module reveal no more information than is necessary to effectively use it , in particular , that it reveal no implementation details . figure [ fig : machine - environment ] highlights the interface between the machine and the environment but does not bind the machine to any particular internal implementation . zave and jackson characterize re as the process by which you arrive at the machine - environment interface ; anything more would amount to prematurely determining an implementation . in i * and tropos , dependencies among actors correspond to their interfaces . the idea of encapsulation , namely , to avoid examining the internals of a component , remains appropriate in iose . a direct consequence of this principle is that a sociotechnical system can not be specified in terms of mental abstractions such as beliefs , goals , intentions , and so on neither of its stakeholders nor of the principals who may participate in it . in particular , roles can not be specified in terms of mental abstractions . in iose , each role in a protocol refers only to the social commitments resulting from the communications that a principal adopting it would be involved in . _ example . _ neither the phy role nor the pat role has goal of scheduling appointment ; neither do they have a shared ( joint ) goal to schedule appointments . _ benefit . _ promotes loose coupling by hiding details not relevant to the interaction . also promotes accountability : a key reason we can not use mental concepts to specify a sociotechnical system is that they make determining compliance impossible @xcite . as mentioned earlier , accountability is meaningless if we can not check compliance . table [ table : principles ] presents the principles and their benefits at a glance . we now evaluate iose with some prominent approaches from se . we choose these either because ( 1 ) they are representative of broad classes of approaches for modeling sociotechnical systems ( i * , tropos , kaos ) , or ( 2 ) they give emphasis to interaction and protocols ( gaia and choreographies ) . lacking a treatment of healthcare in i * and tropos approaches , we study yu s treatment of meeting scheduling in i * @xcite , which is similar in spirit to the appointment scheduling protocol discussed earlier . despite its simplicity , meeting scheduling provides sufficient subtlety to demonstrate various modeling approaches @xcite and is extensively used in the literature . the main requirement in this scenario is to automate and make efficient some aspects of meeting scheduling so that the meeting initiator s burden is reduced . figure [ fig : sd - system - to - be ] shows the _ system to be _ for a meeting scheduling system in the i * notation . the circles represent the actors in the system : meeting initiator ( mi ) , meeting participant ( mp ) , and meeting scheduler ( ms ) . the directed links between the actors represent _ dependencies_. for example , mi depends on mp for achieving the goal that _ the participant attend the meeting _ ( ) . in the i * terminology , figure [ fig : sd - system - to - be ] is a _ strategic dependency _ ( sd ) diagram . according to yu , in _ early requirements engineering _ , i * helps come up with an initial set of requirements . specifically , i * helps identify the goals of the various stakeholders and their dependencies in achieving those goals . a new system actor is introduced unto which some of these goals are _ delegated_. the system actor represents the machine to be designed to meet stakeholder goals . this _ goal - modeling _ phase helps refine the system actor s goals and its dependencies with the other actors . in figure [ fig : sd - system - to - be ] , the meeting scheduler ms is the system actor and is responsible for many tasks that mi was responsible for in the system as is ( not shown ) , such as obtaining availability information from participants and choosing a date that suits all participants . = [ draw , rectangle , rounded corners=4pt , inner sep=2pt , text width=8ex , align = center ] = [ draw , ellipse , text width=6ex , align = center ] = [ draw , rectangle , text width=8ex , align = center ] = [ draw , circle , text width=6ex , align = center ] = [ myarrow data=0.5 ] = [ font= ] ( as - is ) [ draw = none , row sep=0 mm , column sep=2mm , align = center ] ( empty - tl ) ; & [ 0 mm ] & ( attends ) attends + meeting ; & [ 4 mm ] ( empty - tr ) ; + & ( scheduled ) meeting scheduled ; & & ( empty - two - r ) ; + & ( empty - three - l ) ; & ( available ) available dates ; & + ( initiator ) meeting + initiator + ( mi ) ; & ( scheduler ) meeting + scheduler + ( ms ) ; & ( proposed ) proposed dates ; & ( participant ) meeting + participant + ( mp ) ; + & & ( agreed ) agreement ; + & ( range ) date range ; & ( subgoal ) attends + meeting ; + ( empty - bl ) ; & ( assured ) assured ; & & ( important ) important + participant + ( ip ) ; + ; ( important ) to ( participant ) ; ( initiator ) .. controls ( empty - tl ) .. ( attends ) ; ( attends ) .. controls ( empty - tr ) .. ( participant ) ; ( initiator ) to [ bend left ] ( scheduled.west ) ; ( scheduled ) to ( scheduler ) ; ( initiator ) to [ bend right ] ( range ) ; ( range ) to ( scheduler ) ; ( scheduler ) .. controls ( empty - three - l ) .. ( available.west ) ; ( available.east ) to [ bend left ] ( participant ) ; ( scheduler ) to ( proposed ) ; ( proposed ) to ( participant ) ; ( scheduler ) to ( agreed.west ) ; ( agreed.east ) to ( participant ) ; ( initiator ) .. controls ( empty - bl ) .. ( assured ) ; ( assured ) to ( important ) ; ( initiator.south ) to ( subgoal ) ; ( subgoal ) to ( important ) ; tropos is an agent - oriented software engineering ( aose ) methodology that builds on the i * metamodel @xcite . tropos follows i * s goal modeling phase with an architectural phase that maps the system actor to new actors that are interconnected through appropriate data and control dependencies . the detailed design phase models each of these actors as one or more belief - desire - intention agents . the implementation phase realizes the agents , thereby completely implementing the system actor . both i * and tropos would create a system actor for an entire healthcare system . this actor would capture ( a consistent subset of ) the goals of all stakeholders , thereby functioning as a logically centralized machine . the comments below apply equally to i * and tropos . * _ accountability modularity : _ violated . there is no discrete principal behind the system actor and it is not meaningful to talk of its accountability . for example , ms , the system actor in figure [ fig : sd - system - to - be ] is not accountable to the stakeholders . in general , dependencies in i * do not support accountability just because an actor depends on another for a goal does not make the latter accountable for it . for example , figure [ fig : sd - system - to - be ] shows that the mi depends on the mp for attending the meeting ; however it does not automatically follow ( without reasoning about communication and the resulting commitments , as in iose ) that the mp is accountable for showing up . * _ explicit social meaning : _ partially fulfilled . the purpose behind the notion of dependencies in i * was to capture interactor relations in terms of high - level abstractions . however , i * dependencies refer to actors mental states , and are not social , as further explained under encapsulation below . * _ solely social meaning : _ partially fulfilled . dependencies are the only interactor abstraction . however , as we mentioned above , i * dependencies are not social abstractions . * _ separating social from technical : _ partially fulfilled . i * treats social and technical actors alike with the same kinds of dependencies between any pair of actors . for example , ms is a technical actor whereas mp and mi are stakeholders but ms s dependencies with mi and mp are of the same nature as those between mi and mp . however , to its credit , in the early requirements modeling phase , all the actors are stakeholders and only in the later phases ( starting from late requirements modeling ) are technical actors introduced . * _ no principal internals : _ violated . an i * goal dependency between two actors @xmath1 and @xmath2 for some goal @xmath20 means that @xmath1 has a goal @xmath20 , and @xmath2 is _ able _ to achieve @xmath20 and in addition _ intends _ to deliver @xmath20 . for example , as figure [ fig : sd - system - to - be ] shows , mi depends on ms for its goal . this dependency means that ( 1 ) mi has a goal to have the meeting ; ( 2 ) ms is able to schedule the meeting ; and ( 3 ) ms intends to schedule the meeting . in other words , a goal dependency expresses a shared goal , indicating _ joint intentionality _ , among the actors . ( soft goal , task , and resource dependencies are interpreted analogously . ) thus the connection expressed between actors is rooted in their internals . indeed , yu states this : `` the strategic dependency model aims to present a picture of agents by explicitly modeling only their external intentional relationships with each other . the semantics of these external relationships , however , are characterized in terms of some _ presumed internal intentional features _ of the agent '' ( * ? ? ? * , emphasis added ) . in summary , the good points about i * and tropos are that they model stakeholders and stakeholder relations explicitly as actors and dependencies . their shortcomings include the machine - orientation underlying the notion of system actor ; mentalist dependencies ; and violation of encapsulation . that is , i * and tropos can help design a machine in a cooperative setting , where autonomy is not a consideration , but do not adequately address the engineering of a sociotechnical system . dardenne et al.s kaos resembles tropos in its goal - orientation . kaos first elicits stakeholders goals and represents them as and - or graphs . next , it selects a particular variant of the graph as the basis for further engineering . domain analysis reveals the possible sets of agents in the system to be along with their capabilities . kaos s strength lies in the methodological details of deriving operational constraints from goals and assigning them to particular agents as responsibilities depending on their capabilities . for uniformity , we consider examples from van lamsweerde et al.s study of the meeting scheduler . the notion of _ agents _ in kaos is fundamentally different from that of principals in iose . agents in kaos may be social ( human ) or technical , e.g. , devices and software programs . further , an agent is considered a performer of actions , which are themselves defined in terms of input , output , precondition , and postcondition . for example , van lamsweerde et al . specify a agent that has the capability to perform the action , whose preconditions are that the meeting be requested but not scheduled and whose postconditions are ( 1 ) if the meeting is feasible , it is scheduled and ( 2 ) if it is not , the scheduling attempt fails . other agents in van lamsweerde et al.s meeting scheduling system are and , which are analogously specified . in kaos , a system specification captures adequately refined goals ( operational in nature ) and their assignment to agents with the requisite capabilities . for example , the goal of notifying participants is assignable either to the or because they are both capable of notifying participants ; the goal of maintaining an up - to - date agenda that the could consult is the responsibility of every ; and so on . the idea is that if the agents are appropriately specified and the goals are appropriately assigned , then overall system goals would be met . kaos specifications capture neither communications among the agents nor explicit relationships among them . instead kaos conceptualized agents as monitoring and setting the appropriate system - level ( meaning global ) variables and predicates . kaos would model the healthcare system via multiple agents but assign the goals of each agent , thereby creating a conceptually centralized machine . the contrast with iose principles is stark : * _ accountability modularity : _ violated . kaos s notion of composite system is exactly that of figure [ fig : machine - environment ] : the two components are the automated system and the environment @xcite . the users are considered part of the environment . the automated system consists of software agents . thus kaos supports the specification of a distributed machine ( for example , above , we discussed the kaos specification of agents for the meeting scheduler system ) . the automated system represents the realization of the stakeholder requirements ; hence it is meaningless to talk its accountability to anyone . we note here that the notion of responsibility in kaos in not a social one : a responsibility is simply an operational constraint that is assigned ( and therefore , designed ) into an agent . * _ explicit social meaning : _ violated . kaos does not model communication nor employ any communication - related social abstractions . further , unlike tropos , it does not have abstractions to capture any interagent relations . all interaction between agents is captured only indirectly through the setting of the appropriate variables and predicates in the environment . * _ solely social meaning : _ violated . since kaos does not support social meaning . * _ separating social from technical : _ violated . kaos models a sociotechnical system in a purely technical way . neither stakeholders nor principals are accommodated in the system model . in fact , kaos does not even model the goals as being of particular stakeholders : there is just one goal tree , which is progressively refined until the leafs can be assigned to agents . the names of the agents in the meeting scheduling system ( , and so on ) may suggest that kaos accommodates principals . however , that would be misleading : agent specifications together with goal assignments are essentially abstract programs . the entire automated system is a collection of programs interacting via variables . * _ no principal internals : _ violated . the system is ultimately a collection of agents whose assigned responsibilities determine their implementations . since the system s correct functioning depends on the goals assigned to its member agents , kaos breaks encapsulation . like tropos , kaos applies in designing a technical system in a cooperative setting , but its machine - orientation precludes engineering an open sociotechnical system . zambonelli et al.s gaia is one of the earliest aose methodologies . later aose methodologies bear many conceptual similarities with gaia . hence , it deserves an in - depth discussion . gaia takes an organization - based approach in which agents may adopt roles . zambonelli et al . consider a conference management system in which agents may adopt the appropriate roles ( such as reviewer , pc member , author , and so on ) . a role defines the permissions and responsibilities of a prospective agent . permissions describe what an agent could do with resources in the environment . responsibilities are algebraic expressions over the protocols and internal activities that an agent must perform . for reviewer , the permissions are and and the responsibilities are first ( a protocol ) , second ( an internal activity ) , and third ( a protocol ) . below we discuss gaia with respect to the iose principles . * _ accountability modularity : _ violated . in gaia , technical components such as mail clients and active databases are agents . it is meaningless to talk about their accountability . even if the agents represented only stakeholders , gaia says nothing about to whom they are accountable . for example , if a reviewer may not send a review form , would it be responsible to the overseeing pc member ? gaia says nothing about that . + zambonelli et al . conceptualize the multiagent system as an organization that `` can exploit , control or consume when it is working towards the achievement of the organizational goal '' ( p. 328 ) . further , gaia specifies _ organizational rules _ `` that the organization should respect and enforce in its global behavior '' ( p. 335 ) . both these points seem to hint toward a conceptually central entity in the system . although gaia never explicitly mentions it , perhaps all agents that adopt roles are accountable only to this central entity . however , this resembles a conceptually centralized perspective and makes the system conception in gaia similar to the one in figure [ fig : machine ] . * _ explicit social meaning : _ partially fulfilled . whereas gaia supports specifying interaction among roles , it does not specify the meaning of the communications itself . gaia has the notions of permissions and responsibilities at the role level . the intent behind these notions seems to have been to capture social aspects of organizations ; however , gaia falters in important details . as stated above , permissions and responsibilities in gaia are not interagent relationships . for example , the reviewer has the permissions and . but this seems to assume a computational intermediary ( left unspecified in gaia ) with resources ( the papers ) where these activities must be performed . ( in iose , would be modeled as a communication from the reviewer to the program chair or the overseeing pc member and its body would contain the updated review . ) responsibilities specify what an agent adopting the role ought to do but again this seems to assume an unspecified entity to whom the agent would be responsible . * _ solely social meaning : _ violated . as we noted above , gaia does not have any true social abstractions . * _ separating social from technical : _ partially fulfilled . gaia aspires to social - level modeling by modeling systems as organizations and having agents adopt roles in organizations . gaia , to its credit , distinguishes between open systems and closed systems . it explicitly mentions that agents could represent different stakeholders . however , unlike a principal in iose , an agent in gaia is anything that has its own thread of control . specifically , even in open systems , `` active '' components such as active databases , are modeled as agents . * _ no principal internals : _ violated . gaia makes internal agent behavior ( e.g. , ) explicit by placing activities in role specifications , thereby exposing an agent s internal decision - making . ( would not appear in an iose protocol because it does not involve communication it is an internal activity . ) in summary , gaia aspires to modeling open systems via roles and protocols , but falters in important details . although it is not explicitly machine - oriented , it betrays a conceptually centralized mindset . as such , it is not adequate for the modeling of open sociotechnical systems . a choreography specifies the schemas of the messages exchanged as well as constraints on their ordering and occurrence . a choreography is conceptually decentralized and involves roles that principals may potentially adopt . interestingly , choreographies fit in the paradigm of figure [ fig : old ] . principals adopting roles in a choreography would potentially maintain their own ledgers and interact via a digital messaging system that carries their messages to each other . however , there is no computational support for the social state : choreographies do not specify the social meanings of messages and as a result do not capture the social expectations interacting principals may have of each other . choreography description languages , e.g. , ws - cdl @xcite , support specifying the internal activities of principals in the choreography itself . for instance , the ws - cdl notion of _ workunits _ may be used to specify conditional actions by a principal . likewise , mendling and hafner use workunits to specify the internal compliance checks that a tax adviser would make in handling a client s annual statement ( p. 532 ) , thereby violating encapsulation . * _ accountability modularity : _ partially fulfilled . choreographies support compliance at the technical level : principals must send messages in the prescribed order , otherwise they are not compliant with the choreography . however , lacking representation of social meaning and social state , a violation of a choreography does not necessarily amount to a violation at the social level . * _ explicit social meaning : _ violated . lack a representation of social meaning . * _ solely social meaning : _ violated . follows from the lack of a representation of social meaning . * _ separating social from technical : _ fulfilled . choreographies specify interactions with reference to roles that principals may adopt . * _ no principal internals : _ partially fulfilled . in principle , choreographies seek to specify interactions ; however , in practice , they also specify the internal activities of principals . choreographies seek to model interactions ; however , they do so in terms of low - level control and data flow abstractions and often specify aspects of principals internal behavior . therefore , they also do not fare well with respect to the iose principles . table [ table : principles - evaluation ] contrasts traditional approaches with commitment protocols , explained in section [ sec : principles - iose ] as an exemplar of iose . i * , tropos , and kaos are all machine - oriented since their primary specification artifact is that of a machine ( the system actor in i * and tropos and a collection of agents in kaos ) that would meet stakeholder requirements . nonetheless , i * and tropos perform better than kaos in our evaluation because unlike kaos , they model stakeholders and their relations explicitly . in the foregoing , we have extensively discussed the literature most pertinent to our claims . here , we discuss other relevant literature . commitments . commitments are recognized in the literature as a key social abstraction . singh introduced social commitments as a way of formalizing agent communication that was suited to open systems . fornara and colombetti gave an operational semantics for commitments . yolum and singh specified commitment protocols in the event calculus . newer work has tended more toward se themes : methodologies for composing commitment protocols @xcite , patterns @xcite , the relationship of commitments with the notion of goals in tropos @xcite , and monitoring requirements via commitments @xcite . young and antn apply commitments to identify software requirements from regulatory policies and paja et al . to extract security requirements for organizations . baldoni et al . present commitment protocols that support temporal properties . the value we add to this body of work is making explicit connections between the ontology and principles implicit in commitment - based approaches with those implicit in traditional se . iose , as a new paradigm for the sociotechnical systems , provides a natural home for commitments . agent - oriented software engineering despite the apparent similarity between the notion of agents and principals and talk of interaction in many aose methodologies , they ( as exemplified by gaia and tropos ) violate key iose principles . some aose approaches are logically centralized , e.g. , geared toward efficient problem - solving . gaia and other aose methodologies acknowledge the distinction between open and closed systems and emphasize interactions . they correctly classify problem solving systems that distribute a problem across agents as closed . however , aose methodologies betray , if not centralized mindsets , at least considerable conceptual difficulties . vsquez - salceda et al . model the objectives of social systems as goals and the systems as controllers ( p. 338 ) : `` facilitation roles are usually provided by agents controlled by the society , and follow a trivial contract . '' but in a sociotechnical system , there is no unitary multiagent system or society that controls resources . and an organization has no goals that overarch the goals of other participants that is , an organization would be one of many participants in a sociotechnical system . van riemsdijk et al . too are motivated by the modeling of open systems but conceptualize a multiagent system as having collective goals . wagner s agent - object - relationship ( aor ) ontology proposes two kinds of models to support the design of a multiagent system : _ external _ ( observer s perspective , including interaction models ) and _ internal _ ( agent s perspective ) . newer work by guizzardi et al . is based on the aor ontology . wagner s internal - external distinction aligns well with iose . unfortunately , aor subverts the internal - external distinction by including in external models aspects of the internal organization of agents , such as their beliefs ( in _ agent diagrams _ ) as well as reactive rules that specify how an agent would respond to events . singh introduced social commitment as a concept distinct from the mentalist concept of _ internal commitment _ @xcite . aor , tropos , and i * employ internal commitments . serrano and leite map i * concepts to agent programming concepts such as beliefs , desires , and intentions . they map dependencies between two agents to the joint desires of those agents . using their methodology , one specifies the internals of one or more agents . thus even though serrano and leite do nt involve an explicit system actor , the approach is conceptually centralized . how can we reconcile the facts that many aose methodologies are conceptually centralized yet model interactions ? aose methodologies yield physically distributed systems of agents that pursue centrally assigned goals . interactions in aose are focused on the technical means to achieve such goals whereas interactions in iose have social standing . information systems modeling mccarthy proposed rea as a generalized model for accounting that is based on _ resources _ , _ events _ , and _ agents _ and relations among them . rea is finding applications in services and information systems modeling , e.g. , @xcite . rea seeks to capture the fundamental economic model underlying accounting . our endeavor is analogous : to develop a general model of sociotechnical systems , based on social abstractions . the key distinction between iose and rea is that rea models ultimately represent an organization s internal perspective , not its interactions with others . this is evident from rea s inside - outside distinction . one of rea s _ accountability _ relations identifies the responsible agent for an event _ inside _ an organization ; another is between an outside agent and an event . iose commitments are more general : each commitment identifies the accountable party and the party to whom it is accountable for what . the second distinction stems from the _ duality _ of events and the corresponding _ reciprocity _ of commitments reflecting the roots of rea in accounting . duality in rea relates two events that together characterize the give - and - take of an economic exchange ( e.g. , and ) . these events _ execute _ ( i.e. , fulfill ) the corresponding reciprocal commitments . iose does not give any special place to either duality or reciprocity . no iose commitment is necessarily reciprocal : if @xmath1 is committed to @xmath2 , there is no requirement that @xmath2 be committed to @xmath1 ( though it is allowed ) . gordijn and akkermans propose the @xmath21-value ontology and associated methodology in order to specify and analyze so - called business models to evaluate their profitability for all stakeholders involved . a business model represents stakeholders as actors ( e.g. , customer ) . other important concepts are _ value objects _ ( e.g , online articles ) , and _ value exchanges _ among actors ( e.g. , with an article provider for articles and their payment ) . commitment protocols could be used to capture the notion of value exchange among actors ; further commitment protocols could be analyzed for feasibility @xcite . however , unlike commitment protocols and in violation of iose principles , @xmath21-value models also depict the internal flow of values in petri net - like notation . human - computer interaction ( hci ) following developments in hci , the term `` social '' is often associated in re with the interplay between social factors and the design of technical artifacts ( as in the _ design of work _ , e.g. , see @xcite ) , especially as concerns user interface and experience ( e.g. , see @xcite ) . the iose treatment of `` social '' is more general and includes not just people but entities such as enterprises and organizations that may be conceptualized as social actors which enables iose to tackle sociotechnical systems as we define them . compliance compliance with service - level agreements , regulations , and business contracts is gaining interest in software engineering . organizations are increasingly concerned with the question of compliance with regulatory frameworks such as hipaa @xcite and sarbanes - oxley . the challenges here are threefold : one , how to model regulations ; two , how to determine the compliance of a principal s behavior with regulations ; and , three , how to design a principal s information systems such that it is ( likely to be ) compliant with the regulations . siena et al . develop a tropos - inspired modeling language to capture and reason about such regulations . iose can potentially support a more flexible notion of compliance geared toward open systems . although high - level concepts such as obligations , permissions , and so on are often employed in software engineering , they are used to capture behavioral constraints abstractly but are not grounded in communication as commitments are . a more detailed discussion is out of the scope of this paper but interested readers are referred to @xcite for an in - depth discussion . our work began from the recognition that open sociotechnical systems are conceptually decentralized and their principals are autonomous social entities . specifying a sociotechnical system amounts to specifying a protocol , i.e. , a specification of the interactions among its potential participants . we showed that existing approaches take a fundamentally centralized view of sociotechnical systems design : they are geared toward the specification of a machine , not a protocol . we presented iose as a new approach for the engineering of sociotechnical systems . iose is driven by key technical demands : autonomy , accountability , and loose coupling . to illustrate iose , we introduced protocols in which the meanings of the messages are specified in terms of commitments as a natural way to address these technical demands . we formulated the iose principles inspired by , but distinct from , the traditional se principles . we undertook a detailed analysis of some leading relevant se approaches to show which iose principles they violate . traditional se falls short not because the abstractions it employs are low level , but because they deemphasize interaction . to support this claim , we consider well - known requirements modeling approaches , which emphasize high - level abstractions such as goals and dependencies @xcite but do not accord first - class status to interactions . even in the simple appointment scheduling setting , the benefits of iose are obvious : what was previously seen as the problem of specifying a meeting scheduler , under iose turns into two independent problems of ( 1 ) specifying the meeting scheduling protocol , and ( 2 ) specifying principals to participate in the protocol . this reflects parnas conception of modularity as the _ division of labor_. our evaluation of the traditional approaches is not a blanket criticism but rather a critical examination that highlights the differences in their assumptions from the principles of iose . further , our evaluation was limited to the system models or specifications that existing approaches advocate . we did not , for instance , dwell upon the modeling of functional versus nonfunctional requirements or the elicitation of requirements from stakeholders , areas in which current sociotechnical systems approaches ( including i * , tropos , and kaos ) have made substantial progress . although we carefully differentiate iose from traditional se , the two complement each other . iose deals with systems ( viewed as protocols ) whereas traditional re deals with components or principals behaviors ( viewed as machines ) . the former is typified by commitments and other social constructs ; the latter by goals and other mental constructs . recent works @xcite have begun to relate these views from the standpoint of building agents who can function effectively in sociotechnical systems characterized by commitments . we have recently @xcite , sought to apply tropos - like models towards the specification of sociotechnical systems . these works replace intentional dependencies with commitments and do not consider a system actor . further , each actor is understood as a role , i.e. , inert and lacking any goals . goals are applied in the modeling the principals who may potentially adopt roles in the sociotechnical system . we defer the task of systematically laying out the methodological elements of iose , as they overlay the main concepts and principles we introduced above . we expect to go beyond methodologies for specifying and composing protocols , e.g. , @xcite , by incorporating the key re considerations of stakeholders and requirements . the nature of requirements for protocols appears quite different from requirements for machines . for example , a requirement such as `` a meeting is scheduled within three hours of the request '' makes sense for a machine but not a protocol though it could be negotiated between the principals playing the mi and mp roles . a requirement that `` a principal can observe whether another principal is complying with some commitment '' makes sense for a protocol but less so for a machine . the bases from which to engineer protocols is a timely and pertinent direction of research . we imagine that existing re methodologies would provide a useful starting point for investigations into a methodology for iose . amit chopra was supported by a marie curie trentino cofund grant and the erc advanced grant 267856 . we thank fabiano dalpiaz , paolo giorgini , michael huhns , michael jackson , john mylopoulos , and erik simmons for helpful discussions . , dalpiaz , f. , giorgini , p. , and mylopoulos , j. 2010a . modeling and reasoning about service - oriented applications via goals and commitments . in _ proceedings of the 22nd international conference on advanced information systems engineering ( caise)_. lncs series , vol . springer , 113128 . , dalpiaz , f. , giorgini , p. , and mylopoulos , j. 2010b . reasoning about agents and protocols via goals and commitments . in _ proceedings of the ninth international conference on autonomous agents and multiagent systems ( aamas)_. 457464 . , narendra , n. c. , and singh , m. p. 2008 . checking correctness of business contracts via commitments . in _ proceedings of the 7th international conference on autonomous agents and multiagent systems ( aamas)_. ifaamas , estoril , portugal , 787794 . \2002 . operational specification of a commitment - based agent communication language . in _ proceedings of the 1st international joint conference on autonomous agents and multiagent systems ( aamas)_. acm press , 535542 . requirements engineering as the reconciliation of technical and social issues . in _ requirements engineering : social and technical issues _ , m. jirotka and j. goguen , eds . academic press , 165200 . , baldoni , m. , baroglio , c. , chopra , a. k. , patti , v. , , and singh , m. p. 2011 . commitment with regulations : reasoning about safety and control in regula . in _ proceedings of the 10th international conference on autonomous agents and multiagent systems ( aamas)_. 467474 . , dalpiaz , f. , poggianella , m. , roberti , p. , and giorgini , p. 2012 . : using commitments to specify socio - technical security requirements . in _ proceedings of the er 2012 workshops_. lncs series , vol . springer , 396399 . development of agent - driven systems : from i * architectural models to intentional agents code . in _ proceedings of the 5th international i * workshop_. ceur workshop proceedings series , vol . 766 . ceur-ws.org , 5560 . , armellin , g. , mameli , g. , mylopoulos , j. , perini , a. , and susi , a. 2010 . establishing regulatory compliance for information system requirements : an experience report from the health care domain . in _ proceedings of the 29th international conference on conceptual modeling_. lncs series , vol . springer , 90103 . commitments in multiagent systems : some history , some confusions , some controversies , some prospects . in _ the goals of cognition : essays in honor of cristiano castelfranchi _ , f. paglieri , l. tummolini , r. falcone , and m. miceli , eds . college publications , london , chapter 31 , 591616 . in press ; available at http://www.csc.ncsu.edu/faculty/mpsingh/papers . , lock , r. , storer , t. , and dobson , j. 2009 . deriving information requirements from responsibility models . in _ proceedings of the 21st international conference on advanced information systems engineering_. amsterdam , 515529 . \2009 . enhancing tropos with commitments : a business metamodel and methodology . in _ conceptual modeling : foundations and applications _ , a. borgida , v. chaudhri , p. giorgini , and e. yu , eds . number 5600 in lncs . springer , 417435 . , yorke - smith , n. , and singh , m. p. 2012 . relating goal and commitment semantics . in _ proceedings of the 9th international workshop on programming multiagent systems ( promas 2011)_. lncs series , vol springer , taipei , 2237 . , darimont , r. , and massonet , p. 1995 . elaboration of requirements for a meeting scheduler : problems and lessons learnt . in _ proceedings of the second ieee international symposium on requirements engineering_. 194203 . , hindriks , k. , and jonker , c. 2009 . programming organization - aware agents : a research agenda . in _ engineering societies in the agents world x _ , h. aldewereld , v. dignum , and g. picard , eds . lncs series , vol . springer , 98112 . , johannesson , p. , andersson , b. , arachchige , j. j. , and bergholtz , m. 2011 . management services a framework for design . in _ proceedings of the 23rd international conference on advanced information systems engineering ( caise)_. lncs series , vol . 6741 . springer , 582596 . flexible protocol specification and execution : applying event calculus planning using commitments . in _ proceedings of the 1st international joint conference on autonomous agents and multiagent systems_. acm press , 527534 .
many efficient sources of quasi - monochromatic hard radiation exploit transmission of ultrarelativistic electrons through periodic structures ( crystals or undulators ) @xcite . for such so - called coherent sources , high radiation brightness is relatively easy to achieve by increasing the periodic structure length . the price to pay , however , is that as the photon emission probability reaches the order of unity , the description of the source performance must regard the possibility of creation of a few photons and electron - positron pairs per passing electron , i.e. , essentially electromagnetic cascading @xcite . the proper procedure for calculation of electromagnetic multiple particle production at ultra - relativistic energies is via a system of kinetic equations for sequential photon and @xmath0 pair creation , allowing for energy redistribution at each branching . in a non - trivial field of the radiator , this complete procedure is involved , and generally requires numerical simulation . fortunately , in a quite typical case when typical photon energies are inferior to the incident electron energy , the calculations may appreciably simplify . first of all , the probability of pair production is not enhanced so strongly as that of radiation @xcite , and thus can be neglected in the first approximation . secondly , the negligibility of photon recoils allows treating the electron current as fully determined by the electron s initial conditions , entailing statistical independence of multiple photon emission acts @xcite . altogether , that opens the possibility for semiclassical description of the cascading process . however , an extra impediment is that when emitted photon energies belong to gamma - range and @xcite ) . ] , the measured radiation spectra begin to depend on the photon detection method . if it was feasible experimentally , in spite of high radiation intensity , to distinguish arrivals of individual photons , the detector signal under soft photon emission condition would be described merely by formula from classical electrodynamics for the radiation energy spectrum . but narrow beaming of the radiation from ultra - relativistic electrons causes pileup of reactions from different photons in the downstream detector volume , thus affecting the detector signal . spectrometers for gamma - quanta @xcite today are predominantly based on calorimetry ( measurement of the total energy deposition in a detector ) @xcite , but to capture most of the energy of electromagnetic shower created by the hard photon , the calorimeter transverse size must be in excess of the typical shower lateral spread ( molire radius ) , which amounts a few centimeters @xcite . that in turn impedes angular resolution of the radiation within its lorentz - contracted emission cone , which might otherwise be employed to eliminate photon pileups . scheme of gamma - radiation spectrum measurement yielding a multiphoton spectrum . the spent electron is deflected by a magnet , so only the forward - flying photons hit the calorimeter . ] in view of appreciable difficulties with eliminating photon pileup effects , a common attitude at laboratory measurements of gamma - radiation spectra is not to pursue the objective of photon counting at all . then , a single electromagnetic calorimeter is used to measure the _ total _ energy deposited by _ all _ the @xmath1-quanta per event of electron passage through the radiator ( see fig . [ fig : calorim - scheme ] ) . clearly , the latter method is equivalent to measuring the electron s radiative energy loss spectrum . furthermore , sometimes a simplified procedure of radiation spectrum measurement is adopted , when spent electron energies are measured , which then truly yields the spectrum of full energy losses by electrons in the radiator . even though the calorimetric method for spectrum measurement hinders applicability of formulas of classical electrodynamics , those can still be useful for the spectrum calculation . indeed , the theory predicts that multiple soft photon emission probabilities factorize , expressing through the same classically calculated spectral intensity . to obtain the multiphoton spectrum , one thus needs to resum all @xmath2-photon probabilities while holding the total irradiated energy at a preselected value . technically , the problem is similar to that of fast charged particle energy loss under statistically independent atom ionization acts in the medium @xcite , and some other problems @xcite , whose solutions were attained by integral transform methods . for generic energy loss spectrum by a fast particle in the medium , a laplace transform solution was known since 1944 @xcite . in application to radiation , it was later utilized in @xcite , with the purpose of investigating soft ( photon - dominated ) electromagnetic showers in crystals and polycrystals , and checking the possibility of the shower length reduction due to coherent radiation effects . relatively recently , in @xcite this method was used for studying the shape of the soft part of the radiation spectrum from electrons in an amorphous medium . model solutions for shapes of synchrotron - like radiation spectra in crystals were discussed in @xcite . there were also several investigations of shapes of radiation spectra in crystals , based on numerical simulation ( usually by monte - carlo ) , with @xcite or without @xcite neglect of @xmath0 pair production as a first approximation . notwithstanding the pretty long history of the problem , there still remain many open questions concerning multiphoton effects on coherent radiation spectrum shapes . for example , since coherent radiation spectra are typically sharply peaked and discontinuous , the prime task is to investigate how do such features modify under multiphoton emission conditions . in so doing , it is also important to incorporate an incoherent bremsstrahlung component , which may require some special treatment , because of its divergent total probability . next , it is important to give proper study for the behavior of the resummed coherent radiation spectrum in the high intensity ( or photon multiplicity ) limit . there , one generally expects the spectrum to tend to a gaussian form , but corrections to this form may be significant at practice , and demand evaluation . besides the abovementioned most urgent problems , there is a number of developments which are desirable to make in the theory . first , it is expedient to supplement the notion of the resummed spectrum by that of photon multiplicity spectrum , which was proven measurable in experiments at cern @xcite . secondly , it would be valuable for various purposes to find a possibility to reconstruct the single - photon spectrum from a calorimetrically measured one . other issues include correspondence with the perturbation theory , separation of energetically ordered and anti - ordered contributions , etc . the goal of the present paper is to provide a comprehensive formulation for the theory of resummed soft coherent radiation spectra , and with its aid , to tackle problems raised above . although conditions of coherent radiation in different sources may vary , we concentrate on one of the most typical cases dipole radiation ( for overview of its conditions see , e.g. , @xcite ) , under the simplifying assumption of approximately harmonic transverse oscillation of the radiating electron ( the neglect of higher harmonics ) . we incorporate also an incoherent radiation component , which , fortunately , has a simple structure , free from model assumptions . the interference between coherent and incoherent spectral components may be neglected in the first approximation . therewith , there emerges a fairly model - independent description of radiation , which can embrace diverse practical situations . this paper is laid out as follows . in sec . [ sec:2 ] we establish main relations of the theory of soft multiphoton spectra . along with the resummed spectrum itself , the photon multiplicity spectrum is defined . in sec . [ sec : coh+incoh ] , the developed generic techniques are applied to the study of coherent and incoherent contributions to the radiation . modification of various spectral features due to multiphoton effects , as well as interplay of coherent and incoherent radiation components , is investigated . [ sec : high - intensity ] explores the high - intensity limit , benefiting from the analogy between the multiphoton energy loss and a random walk . having calculated the multiphoton spectrum in the central region , we also examine the spectrum behavior beyond it , and assess long - range effects due to the incoherent bremsstrahlung component . [ sec : reconstr ] works out the principles of reconstruction of the single - photon radiation spectrum from the calorimetrically measured one . the summary and some outlook is given in sec . [ sec : summary ] . a few technical issues are relegated to the appendices . as was mentioned in the introduction , for radiation emitted by an ultra - relativistic electron ( or positron ) , the radiation angles are typically small ( inversely proportional to the electron s lorentz factor ) , wherefore they often remain experimentally unresolved . in the latter case , only the radiation spectrum is measured . for coherent radiation , the spectrum is concentrated on a characteristic energy scale @xmath3 , which is often much smaller than the initial electron energy @xmath4 : @xmath5 under those circumstances , the radiation recoils may be neglected , and the radiation process be treated as statistically independent emissions of photons by a classically moving charged particle . the description of the multiphoton spectrum then proceeds as follows . ] . let the classically evaluated radiation energy spectrum , integrated over relativistically small emission angles , equal @xmath6 . the corresponding photon number density could be derived as @xmath7 that might as well be interpreted as the probability distribution for emission of a photon , but only provided its integral @xmath8 which is to correspond to the total emission probability , is @xmath9 . sometimes the latter condition is fulfilled thanks to the smallness of electron coupling with the electromagnetic field , @xmath10 . on the other hand , in a long radiator , @xmath11 can accumulate and get arbitrarily large . as it becomes sizable , proper probabilistic treatment must take into account emission of an arbitrary number of photons . assuming that the radiation process remains soft and intrinsically semiclassical , photon emission acts can be proven to be statistically independent @xcite , thereby obeying poisson statisticsa word of caution must be sound that poisson statistics holds only for a completely prescribed classical electromagnetic current . at practice , the radiation spectrum can yet depend on the electron impact parameters and momentum dispersion within the initial beam , fluctuations of the intra - crystal field due to thermal vibrations and lattice defects , etc . therefore , the poisson distributions must be averaged over the beam and target ensembles , but only at the last step of the calculation ( cf . , e.g. , @xcite ) . in the present paper , we will refrain from implementation of the averaging procedures , just indicating which of them are most relevant for specific coherent radiation sources . fortunately , in a number of cases we consider , the ensemble averaging effects are negligible . ] : @xmath12 here factors @xmath13 and @xmath14 are consequences of probability conservation during photon generation is conventionally attributed to photon equivalence . ] . from the normalization of the total probability to unity , @xmath15 one readily derives a relation of @xmath13 with the single - photon spectrum : @xmath16 since @xmath13 may be regarded as the 0-th term of the series , it is to be interpreted as the probability of photon non - emission . quantity @xmath11 also admits an independent physical meaning , being equal to the mean number of emitted photons ( the photon multiplicity ) : @xmath17 from the experimental point of view , completely exclusive spectrum @xmath18 is arguably too detailed . if we ( idealistically ) limit ourselves to a single - inclusive spectrum , where the energy of only one of the photons is measured , the corresponding observable would equal @xmath19 thus reproducing the photon number spectrum , connected with the classical energy spectrum via eq . ( [ primary ] ) . for the case of calorimetric measurement , however , the situation is different . there , the probability of emission of an arbitrary number of photons with definite aggregate energy @xmath20 is measured , taking the form - functions in eqs . ( [ class ] ) or ( [ generic - sum - int ] ) must be replaced by a function of the detector response ( see , e.g. , @xcite ) . that would be equivalent to ensemble averaging over the final state , which is left beyond the scope of this paper ( see footnote [ exp - before - aver ] ) . at practice , though , the energy resolution of calorimeters is usually high enough for the @xmath21-function approximation in ( [ generic - sum - int ] ) to work satisfactorily . ] @xmath22 the latter distribution ) are known as compound ( or generalized ) poisson distributions @xcite . they are usually defined in terms of self - convolutions of the initial probability density @xmath23 it is easy to see that in eq . ( [ generic - sum - int ] ) , @xmath24 is an @xmath2-fold self - convolution of the single - photon spectrum . ] is already properly normalized : @xmath25 as illustrated in fig . [ fig : integr - domain ] , in the limit of low radiation intensity , distributions ( [ class ] ) and ( [ generic - sum - int ] ) match , since terms @xmath26 in their series are identical ( fig . [ fig : integr - domain]a ) . but the rest of the terms , with @xmath27 , do differ ( cf . fig . [ fig : integr - domain]b ) . distribution ( [ generic - sum - int ] ) is often called multiphoton spectrum , and in contrast to that , ( [ primary ] ) is called single - photon spectrum . integration domains for partial probabilities in eqs . ( [ class ] ) and ( [ generic - sum - int ] ) . ( a ) component @xmath26 . only the point @xmath28 is selected by perfect detector of either kind . ( b ) component @xmath29 . dashed lines mark the region selected by a perfect photon counter and spectrometer . solid line , the region selected by a perfect calorimeter . ] another way of representing the results is via a multiphoton energy spectrum @xmath30 in this paper , though , it will be advantageous for us to work with the probability spectral density . so , in what follows , the term ` spectrum ' will mostly be used in relation to quantities such as @xmath31 and @xmath32 . before proceeding with the resummation of series ( [ generic - sum - int ] ) , it is expedient to inspect its structure more closely . first of all , in ( [ generic - sum - int ] ) , upper integration limits for individual photon energies may actually be lowered down to @xmath20 , granted that if energy of any of the emitted photons exceeds @xmath20 , the energy - conserving @xmath21-function will definitely yield zero . at @xmath26 , one yet needs to ensure that the @xmath21-function falls into the integration interval completely , so the integration upper limit should be written more precisely as @xmath33 . with the forthcoming resummation procedure in mind , it is convenient to let the upper integration limit be the same everywhere , writing : @xmath34 next , even though in eq . ( [ generic - sum - int - upper - omega ] ) upper integration limits for partial probabilities are rendered finite , in integral ( [ w1-def ] ) entering @xmath13 through eq . ( [ w0-def ] ) , the integration extends over an infinite range of @xmath35 . that fact needs care , since at @xmath36 , quantum corrections due to photon recoils invalidate the statistical independence of photon emission acts . for pure coherent radiation , whose spectrum is strongly suppressed beyond a sufficiently low energy @xmath3 , such a problem may be absent . however , there often exists an incoherent bremsstrahlung contribution to @xmath31 , scaling as @xmath37 . therewith , integral @xmath11 logarithmically diverges on the upper limit , whereby factor @xmath38 would nullify the multiphoton radiation spectrum for any finite @xmath20 , if there was no physical end of the bremsstrahlung spectrum , situated at @xmath39 . within the leading logarithmic ( ll ) accuracy , the upper integration limit could be merely replaced by @xmath4 : @xmath40 however , the usage of the leading - log accuracy in eq . ( [ poisson ] ) , where @xmath11 is exponentiated , may be unsatisfactory , as long as it introduces an indeterminate overall factor . thence , it is worth promoting it to the next - to - leading logarithmic ( nll ) accuracy . that is equivalent to effectively replacing eq . ( [ w1-regee ] ) by @xmath41 where parameter @xmath42 must be ab initio calculable in nll . a study delivered in appendix [ app : kappa ] ( though still under the neglect of pair production ) suggests that @xmath43 . greater accuracy may be unnecessary , since @xmath44 only enters in terms a power @xmath45 with @xmath46 ( see sec . [ subsec : incoh - bremsstr ] ) . the incoherent bremsstrahlung contribution also dominates in the opposite extreme @xmath47 , where its behavior @xmath48 makes all the integrals present in eq . ( [ generic - sum - int - upper - omega ] ) diverge on the lower limit . fortunately , though , the infrared ( ir ) divergence cancels for energy - resummed distributions by virtue of the bloch - nordsieck theorem @xcite , as will be demonstrated in the next subsection . resummation of series of type ( [ generic - sum - int ] ) proceeds by applying a laplace transform , which reduces the multiple convolution to a simple product of equal single integrals , and the resulting power series resums to an exponential : @xmath49 [ prior to the resummation , we implemented a proper cutoff parameter @xmath50 , in the manner of eq . ( [ w1-nll ] ) ] . an immediate observation from eq . ( [ generic - exponentiation ] ) is that by virtue of factor @xmath51 tending to zero as @xmath47 , the infrared singularity brought by @xmath31 is suppressed , providing the integral convergence on the lower limit if @xmath52 . if @xmath53 is put to zero , eq . ( [ generic - exponentiation ] ) reproduces normalization condition ( [ total - probab-1 ] ) . the laplace transform is further inverted by computing a mellin integral in the complex @xmath53-plane : @xmath54 here constant @xmath55 is required to be greater than real parts of all singularities of the integrand , otherwise being arbitrary . combining eqs . ( [ generic - mellin - int ] ) and ( [ generic - exponentiation ] ) yields the generic integral representation for the multiphoton radiation spectrum : @xmath56 the appearance in eq . ( [ generic - eq ] ) of a negative term proportional to dirac @xmath21-function does not imply negativity of the resummed spectrum at @xmath57 . it just cancels the corresponding positive singularity in the contour integral , arising because its integrand tends to @xmath58 as @xmath59 . at @xmath60 , the @xmath21-term does not contribute , anyway , but it proves important at derivation of various integral relations involving @xmath32 [ in particular , maintaining the correct normalization ( [ total - probab-1 ] ) ] . the integral on the r.h.s . of ( [ generic - eq ] ) , @xmath61 has an independent physical meaning , representing the distribution function for the radiating electrons , which is singular at @xmath57 , but is normalized to unity : ) , the integral over @xmath20 extends to infinity , because even if the single - photon spectrum terminates at @xmath62 , the spectrum resummed without account of electron energy degradation after photon emission will actually extend beyond the limit @xmath39 ( see discussion in sec . [ subsec : incoh - bremsstr ] ) . ] @xmath63 representation ( [ generic - eq ] ) without the @xmath21-term , i.e. holding for function @xmath64 [ or pertaining to case @xmath65 , @xmath66 , was derived by landau @xcite on the basis of a kinetic equation . next , let us note that contour integral representation ( [ generic - eq ] ) needs not be the only possible one . the example of representation ( [ generic - sum - int - upper - omega ] ) shows that in the exponentiated laplace transform of @xmath31 , the upper integration limit may be arbitrary , provided it exceeds @xmath20 . for instance , lowering it right down to @xmath20 leads to the integral representation @xmath67 indeed , expanding here @xmath68 to taylor series and integrating over @xmath53 termwise leads to decomposition ( [ generic - sum - int - upper - omega ] ) . relation ( [ generic - eq - upperlimitomega ] ) was first noticed in @xcite , wherein it was regarded as only an _ approximate _ consequence of representation ( [ generic - eq ] ) ( in the sense that in the integral over @xmath53 , the contributing @xmath53 are @xmath69 , and hence in @xmath70 the contributing @xmath35 are @xmath71 ) . but here , by deriving both representations ( [ generic - eq ] ) and ( [ generic - eq - upperlimitomega ] ) from the same series , we reveal their exact equivalence . although the derivation of representation ( [ generic - eq - upperlimitomega ] ) given above is illuminating , it may be desirable sometimes to avoid the use of power series , which may be beset by ir divergences . in that case , one may utilize the following generic statement : [ lemma1 ] let @xmath72 be real parameters , and @xmath73 be a positive function of real variable , integrable on interval @xmath74 . also let @xmath75 be an analytic function of complex variable @xmath76 in domain @xmath77 , and such that @xmath78 grows with @xmath79 slower than exponentially . then , contour integral @xmath80 where @xmath81 is greater than real parts of all singularities of the integrand in the @xmath53-plane , is independent of @xmath82 on the semiaxis @xmath83 . differentiation of eq . ( [ lemma ] ) by parameter @xmath82 gives : @xmath84 the condition that function @xmath85 grows with @xmath86 slower than exponentially guarantees the exponential decrease of the integrand of ( [ lemma - diff ] ) when @xmath83 . thus , if the integration contour , defined to pass on the right of all singularities of the integrand , is withdrawn rightwards to infinity , the whole integral exponentially vanishes . therewith , derivative ( [ lemma - diff ] ) is identically zero , which entails independence of integral ( [ lemma ] ) on @xmath82 under the conditions assumed . in our case ( [ generic - eq ] ) , ( [ generic - eq - upperlimitomega ] ) , function @xmath75 has to be identified with an exponential @xmath87 ( where @xmath88 furnishes an ir cutoff , if necessary ) . function ( [ fff ] ) satisfies the conditions of our lemma liberally : since @xmath89 increases at @xmath90 at most logarithmically , @xmath91\|$ ] in this limit can only increase as some power of @xmath53 . as a whole , @xmath75 is actually ir - finite , so after unifying @xmath35-integrals in the exponent of @xmath92 , it is safe to put @xmath93 . by choosing in expression ( [ lemma ] ) @xmath82 arbitrarily small or large , one can deduce eq . ( [ generic - eq - upperlimitomega ] ) from eq . ( [ generic - eq ] ) , or vice versa . we will also employ lemma [ lemma1 ] for simplification of the exponentiated integrals with specific profiles of @xmath94 in the next section . ( [ generic - eq - upperlimitomega ] ) has an appealing property that apart from the factor @xmath95 the multiphoton spectrum involves only contributions from the single - photon spectrum with @xmath96 . that looks natural in view of the positivity of contributing photon energies . in this sense , it seems natural to term eq . ( [ generic - eq - upperlimitomega ] ) ` energetically ordered ' form [ to distinguish it from ` non - ordered ' form ( [ generic - eq ] ) ] . still , one should mind the existence of factor ( [ w0omega - def ] ) , which depends on contributions from @xmath97 , and is nothing but the probability of non - emission of any photon with energy greater than @xmath20 @xcite ( check : @xmath98 with @xmath99 standing for the heaviside unit step function ) . so , the effect of contributions from @xmath97 is ` non - dynamical ' , resulting in a rather trivial explicit factor , but nonetheless , due to this factor , the energetic ordering does not hold strictly for multiphoton probability spectrum . remarkably , in sec . [ sec : reconstr ] we will encounter representations which may be regarded even as energetically anti - ordered . the manifestation of multiphoton effects may be studied in terms of non - linearity of the radiation spectrum dependence on the radiator length ( or the crystalline target thickness ) , @xmath100 ( cf . @xcite ) . in the simplest approximation ( reasonable for long radiators ) , one may assume that @xmath101 . then , @xmath100-dependence of @xmath32 may be studied based on expansion ( [ generic - sum - int ] ) , where factor @xmath13 is yet a non - linear function of @xmath11 . expanding the latter factor to series in @xmath11 leads to a double series for @xmath32 . in fact , a single power series in @xmath100 can be obtained using the following trick . issuing from contour integral representation ( [ generic - eq ] ) , integrate in the exponent by parts : @xmath102 next , expand the @xmath31-dependent part of the exponential to power series : @xmath103 with @xmath104 now , it is straightforward to integrate over @xmath53 termwise with the use of the identity @xmath105 which yields @xmath106 the latter multiple integral converges even if the mean photon number @xmath11 diverges logarithmically ; its multiple derivatives always exist , as well . in contrast to eq . ( [ generic - sum - int ] ) , however , terms of series ( [ dwn - pert - ser ] ) need not be everywhere positive . that is chained to the fact that multiple photon emission leads to a redistribution of the spectrum , not only to pileup of events . from the standpoint of qed , negative terms in the perturbative expansion typically arise from interference of loop ( photon reabsorption ) diagrams with lower - order ones @xcite . ( [ dwn - pert - ser ] ) indicates that for an observable such as the calorimetrically measured spectrum , the structure of those corrections assumes a simple generic form . correspondence of ( [ dwn - pert - ser ] ) with expansion ( [ generic - sum - int ] ) may be established by differentiating in ( [ dwn - pert - ser ] ) the delta - function under the integral sign by formula @xmath107 and subsequently integrating by parts over all @xmath108 . it should be minded thereat that the endpoint terms are non - vanishing , producing the powers of the total single - photon emission probability @xmath11 present in eq . ( [ generic - sum - int ] ) in the exponent . however , if @xmath11 diverges , the integration by parts is no longer possible , wherewith series ( [ generic - sum - int ] ) does not exist , but eq . ( [ dwn - pert - ser ] ) remains valid , anyway . characterization of compact probability distributions is often conducted in terms of their moments . spectrum @xmath32 , however , is not normalized to unity , so it can not be used straightforwardly as a weighting distribution . in capacity of a normalized probability distribution , one should either use electron distribution function ( [ = pi ] ) , or the rescaled radiation spectrum @xmath109 . we will adopt the first option , which leads to simpler final results . the mean emitted photon energy for our resummed spectrum is computed , e.g. , by termwise integration in eq . ( [ generic - sum - int ] ) : @xmath110\equiv\int_0^\infty d\omega \omega\frac{dw}{d\omega}\nonumber\\ & = & e^{-w_1}\sum_{n=1}^\infty\frac{n}{n!}w_1^{n-1}\overline{\omega_1\!}=\overline{\omega_1\!}.\end{aligned}\ ] ] thus , it coincides with that for the single - photon spectrum , @xmath111 that is quite natural , inasmuch as once photon energies in all the events are summed over , it no longer matters whether they were measured by photon counters or calorimeters . [ in particular , that justifies the use of classical formulas for the rate of radiative energy losses ( see , e.g. , @xcite ) under conditions of multiple photon emission . ] to further characterize the spectrum width , asymmetry , and so on , one can refer to higher moments about the mean value ( central moments ) , presuming their existence . those are conveniently calculated based on representation ( [ generic - eq ] ) : @xmath112\qquad\qquad\nonumber\\ & = & \left(-\frac{\partial}{\partial s}-\overline{\omega_1\!}\right)^n e^{\int_0^{e } d\omega_1 \frac{dw_1}{d\omega_1}\left(e^{-s\omega_1}-1\right ) } \bigg\|_{s=0}\nonumber\\ & \equiv&(-1)^n\frac{\partial^n}{\partial s^n}e^{\int_0^{e } d\omega_1 \frac{dw_1}{d\omega_1}\left(e^{-s\omega_1}-1+s\omega_1\right ) } \bigg\|_{s=0}.\,\,\end{aligned}\ ] ] the above representation implies that exponential @xmath113 serves as a generating function @xcite for central moments , while @xmath114 offers a generating function for ordinary moments . applying eq . ( [ higher - moments ] ) for lowest @xmath2 , one finds @xmath115 @xmath116 @xmath117 etc . relations ( [ mean - omega ] ) , ( [ mean - omega2-alt ] ) ( historically known as campbell s theorem @xcite ) were discussed in @xcite with the objective of extracting information about the single - photon distribution from the resummed one . ( [ mean - omega3 ] ) , ( [ mean - omega4 ] ) may serve for the same purpose . in addition , in sec . [ sec : reconstr ] we will formulate a reconstruction procedure for the complete spectrum , not resorting to the notion of moments . a predicament with the use of the above quoted moments is that with the account of an incoherent bremsstrahlung component ( sec . [ sec : coh+incoh ] ) , they all diverge in the ultraviolet . therefore , they may seem to be only useful in the context of pure coherent radiation . nonetheless , later on we will show that treatments of coherent and incoherent radiation components can always be separated , permitting the usage of spectral moments for pure coherent component regardless of the presence of an incoherent one . as was mentioned in the introduction , it is rather difficult in a multiphoton event to practically pin down individual energies of all the photons , i.e. , to measure joint probability distribution @xmath18 . but there exist relatively compact experimental methods allowing to receive partial information about the photon number content . for instance , placement of a thin converter and an ionizing particle counter upstream the calorimeter allows measuring the mean number of photons ( the photon multiplicity ) as a function of the total energy @xmath20 deposited in the calorimeter see fig . [ fig : multty - measur - scheme ] and , e.g. , @xcite . in this subsection , we will extend our resummation procedure to description of the photon multiplicity spectrum . in the spirit of eqs . ( [ mean - n - def ] ) and ( [ generic - sum - int ] ) , to construct the photon multiplicity at a given @xmath20 , one must incorporate in ( [ generic - sum - int ] ) a weighting factor @xmath2 equal to the number of emitted photons , whereupon divide the resulting sum by the unweighted expression : @xmath118 since all the weights @xmath2 in this sum are greater than unity , so must be their mean number : @xmath119 in particular , when @xmath120 , integration phase space volumes for @xmath27 terms shrink to zero , wherefore only term @xmath26 survives . this term cancels with the correspondent term in the denominator , yielding @xmath121 the sum entering eq . ( [ mean - n - omega - def ] ) can again be evaluated by the laplace transform method , as follows : [ mean - n - omega - contour ] @xmath122 [ the latter equality may be inferred from lemma [ lemma1 ] with @xmath123 . this may also be expressed as a convolution of the single - photon spectrum with the resummed one : @xmath124 here the last term stems from the @xmath21-function term of eq . ( [ generic - eq - upperlimitomega ] ) . unlike @xmath32 , however , the multiplicity spectrum is not an ir - safe quantity . but fortunately , the ir divergence of @xmath11 affects only the constant , @xmath20-independent part of @xmath125 . indeed , the contribution from vicinity of the lower integration limit in the convolution term in ( [ mean - n - omega - convol ] ) is @xmath126 , and so is the l.h.s . hence , at @xmath127 , @xmath20-dependent factors in both sides of eq . ( [ mean - n - omega - convol ] ) cancel . to isolate the ir divergence of @xmath128 explicitly , one can rewrite eq . ( [ mean - n - omega - convol ] ) , e.g. , as @xmath129 here the r.h.s . is ir - safe ( thence we put there @xmath93 ) , while in the l.h.s . , the dependence of @xmath128 on the ir cutoff @xmath88 must be additive , to cancel with the additive dependence on ir cutoff of the integral term . on the other hand , the dependence of on @xmath50 for @xmath128 must disappear , as will be demonstrated in the next section . generic resummation formulas established in the previous section are valid for radiation characterized by an arbitrary single - photon profile @xmath94 . in what follows , we will mostly concentrate on a category of coherent radiation spectra involving an admixture of incoherent radiation . in that case , the salient features shared by the single - photon spectra are : * spectral discontinuities at finite @xmath20 in the coherent radiation component ( coherent emission edges ) ; * a ` tail ' towards large @xmath20 brought by the incoherent bremsstrahlung component ; * an infrared divergence , originating from the incoherent bremsstrahlung , or from non - zero net deflection by coherent fields ( see , e.g. , @xcite ) . for simplicity , in this paper we will be only considering the first case . our first task is to investigate how do the above mentioned features modify under multiphoton emission conditions . since multiphoton effects interplay nonlinearly , it is expedient first to analyze them for pure coherent and incoherent components separately . combining them , we can thereupon distinguish effects of their mutual influence . let us first scrutinize the case of purely incoherent bremsstrahlung . its single - photon spectrum must be proportional to bethe - heitler s spectrum of radiation at a single collisionthis assumption admittedly neglects dense - medium effects such as transition radiation due to the change of the dielectric density , and lpm radiation suppression by multiple scattering @xcite . their neglect is justified under typical conditions @xmath130 mm , @xmath131 gev , @xmath132 mev , although some of those conditions can be relaxed at the expense of the others . the influence of forward transition radiation ( noticeable when @xmath133 mm ) , and of lpm suppression on resummed radiation spectra was studied in @xcite . ] @xcite . moreover , in the soft radiation limit we are restricted to , the single - photon spectrum reduces to the well - known semiclassical form @xcite @xmath134 where @xmath135 is proportional to the target thickness @xmath100 . in an amorphous matter , constant @xmath135 equals @xmath136 where the radiation length @xmath137 depends on the atomic density @xmath138 and ( net ) nucleus charge @xmath139 @xcite . a similar estimate applies also for the incoherent component of coherent bremsstrahlung in a crystal . for the case of positron channeling , if their close collisions with atomic nuclei are sufficiently rare , factor @xmath140 should be replaced by the net interplanar electron density @xmath141 . finally , in an undulator , @xmath135 is proportional to the density of the residual air , though in high vacuum that can be negligible . as was argued in the end of sec . [ subsec : stat - indep - emission ] , the integral of ( [ dw1inc ] ) , which is logarithmically divergent in the uv , must be cut off at some photon energy commensurable with the electron energy : @xmath142 , with @xmath143 . we implement this cutoff directly for approximation ( [ dw1inc ] ) , adopting a simplified model : @xmath144 where @xmath145 is the heaviside function . integration contours for resummed spectra . line @xmath146 , used in eq . ( [ generic - eq ] ) , ( [ generic - eq - upperlimitomega ] ) . deformed contour @xmath147 , used in eqs . ( [ eulergamma - def ] ) , ( [ dw - first - interv ] ) . closed loop contour @xmath148 , applicable for pure coherent radiation ( see sec . [ subsubsec : coh - in - fund - interv ] ) . ] a convenient starting point for evaluation of the resummed spectrum corresponding to single - photon spectrum ( [ dw1incoh - cutoff ] ) is the energetically ordered contour integral representation , eq . ( [ generic - eq - upperlimitomega ] ) . therein , the @xmath21-term is suppressed , since @xmath149 , and for the present case it will be neglected . once ( [ dw1incoh - cutoff ] ) is inserted to eq . ( [ generic - eq - upperlimitomega ] ) , ) . that depends on how large are relative fluctuations of parameter @xmath135 per electron passage . if the radiation formation length @xmath150 is @xmath151 , which is plausible under conditions of footnote [ foot : no - medium - effects ] , many electron scatterings in the target contribute to the radiation independently , so fluctuations of @xmath135 are relatively small , justifying the use of resummation for the averaged spectrum . averaging procedure for multiphoton radiation in a thin target will be considered elsewhere . ] all @xmath20- and @xmath50-dependencies factor out after a simple rescaling of the integration variables , giving @xmath152 with @xmath153 to evaluate @xmath154 , one may regularize @xmath155 on the lower limit , then split it in two integrals , and by lemma [ lemma1 ] , extend the upper limit of the integral involving @xmath156 to infinity . thereafter , the ir regulator may again be sent to zero , and the sum of @xmath157-integrals evaluates : @xmath158 the integrand of the latter contour integral is vanishing as @xmath159 , so at complex infinity the contour can be turned to the left from the imaginary axis ( contour @xmath147 in fig . [ fig : contours ] ) . using representation @xmath160 for the euler gamma function , and definition @xmath161 for euler s constant , we get @xmath162 function ( [ fa ] ) is manifestly positive and monotonously decreasing ( see fig . [ fig : phi , nu ] ) , its derivative in the origin being zero : @xmath163 other models for the single - photon spectrum suppression at large @xmath35 yield equivalent results ) one employs a slow exponential factor , @xmath164 with @xmath165 , the insertion of this form to representation ( [ generic - eq ] ) , with the upper limit of @xmath35-integral replaced by infinity , gives @xmath166 apart from factor @xmath167 ( which is close to unity when @xmath168 ) , this structure coincides with eq . ( [ 13 ] ) , once one identifies @xmath169 . there , the single - photon spectrum is expressed by a gamma - distribution @xcite of index 0 , while the resummed spectrum is again a gamma - distribution , but of index @xmath135 . that is a well - known example of functional stability more general than lvy stability ( see , e.g. , @xcite ) . one can as well arrive at result ( [ 13 ] ) from the side of lvy - stable densities @xcite , for which the single - photon distribution is strictly scale - invariant : @xmath170 ; however , lvy distributions at finite @xmath21 , in contrast to ( [ 13 ] ) or ( [ gamma - distr - a ] ) , do not reduce to a product of functions of a single variable . ] . regarding the shape of the resulting distribution , two remarks are in order . first , spectrum ( [ 13 ] ) is suppressed compared to single - photon spectrum ( [ dw1inc ] ) everywhere in the region @xmath171 . at @xmath172 , there formally develops an enhancement of the spectrum , but it is not accurately described by eq . ( [ 13 ] ) , insofar as at @xmath173 quantum effects enhance , while at @xmath39 , the spectrum must strictly terminate at all . hence , there ansatz ( [ dw1incoh - cutoff ] ) for the single - photon spectrum breaks down , so our model description , though self - consistent formally , is inadequate in domain @xmath173 . with this reservation , the uniform suppression of the resummed spectrum at @xmath171 does not contradict , e.g. , the conservation of the mean photon energy at resummation [ eq . ( [ mean - omega ] ) ] . graphs of functions @xmath154 [ eq . ( [ fa ] ) ] and @xmath174 [ eq . ( [ nu - through - digamma ] ) ] . our physical assumptions require that @xmath175 ( see text ) . ] secondly , the dependence of the resummed spectrum on @xmath44 enters eq . ( [ 13 ] ) through parameter @xmath142 , resulting in an overall factor @xmath176 . thus , in order for result ( [ 13 ] ) not to be strongly sensitive to the actual value of @xmath44 , one needs fulfillment of condition @xmath177 ( otherwise , full kinetic equations for the electromagnetic cascade will need to be solved ) . since the accuracy of eq . ( [ 13 ] ) is @xmath178^a=1+\mathcal{o}(a)$ ] , function @xmath179 may be safely approximated by @xmath180 . then , one readily checks that @xmath181 in accord with eq . ( [ total - probab-1 ] ) ( where , again , @xmath127 , owing to ir divergence of @xmath182 ) . property ( [ incoh - norm ] ) implies that despite the smallness of @xmath135 , function ( [ 13 ] ) is not uniformly small , but rather is similar to a @xmath21-function , peaking at @xmath120 where function ( [ 13 ] ) blows up : @xmath183 to prove eq . ( [ delta - lim - ato0 ] ) , and estimate the narrowness of @xmath184 as a function of parameter @xmath135 , it may be expedient to evaluate a median energy @xmath185 at which there accumulates half probability of the resummed incoherent radiation , i.e. @xmath186 that yields @xmath187^{-1/a}e\lll e_e,\ ] ] which at small @xmath135 is exponentially small , thereby validating relation ( [ delta - lim - ato0 ] ) . finally , the ( regularized ) photon multiplicity spectrum can be evaluated for the present case via eq . ( [ mean - n - omega - convol - regul ] ) . inserting ( [ dw1inc ] ) and ( [ 13 ] ) to ( [ mean - n - omega - convol - regul ] ) , and evaluating the integral , we get : @xmath188 with @xmath189\\ & = & 1+a\left[\psi(1)-\psi(1+a)\right]\label{nu - through - digamma}\\ & = & 1-\frac{\pi^2}6 a^2+\mathcal{o}(a^3),\end{aligned}\ ] ] involving digamma - function @xmath190 . hence , photon multiplicity spectrum grows with @xmath20 strictly logarithmically , and proves to be independent of the ultraviolet ( uv ) cutoff @xmath50 . the graph of function @xmath174 is shown in fig . [ fig : phi , nu ] . at low @xmath135 , it is close to unity , which is natural , because no more than one photon is typically emitted per passing electron . next we consider the case of pure coherent radiation . it is principally different from the incoherent radiation case considered above , for as it involves an integrable ( ir- and uv - safe ) , although discontinuous single - photon spectrum . physically , that may correspond to ( gamma - ray ) undulator radiation @xcite , or channeling radiation @xcite . as for coherent bremsstrahlung , usually containing an admixture of incoherent radiation component , too , it is more pertinent to the combined radiation case discussed in the next subsection . in what concerns application to channeling radiation , though , it should be stressed that the spectrum of the latter strongly depends on impact parameters of the charged particles , even when non - channeled passage events are rejected . hence , at passing to multiphoton spectra , at first we have to perform the resummation for a definite impact parameter value , and only at the final step average the result over impact parameters . we can not indulge into such specialized procedures here , leaving them for future studies . so , in application to channeling radiation our results in this paper will only be preliminary . what we wish to reflect in our present study is the transverse oscillatory motion of the radiating particle with respect to the direction of high longitudinal momentum , taking place in all the abovementioned cases . thereat , the radiation spectrum shape depends on the particle oscillation harmonicity . in the simplest case of purely harmonic and small - amplitude oscillatory motion , the electromagnetic radiation spectrum has the structure ( see appendix [ app : coh - rad - sources ] ) @xmath191 where coefficient @xmath192 in a straight crystal / undulator is proportional to the radiator length @xmath100 , and the profile function reads @xcite [ f - def ] @xmath193 [ the latter form demonstrates the function positivity and symmetry with respect to midpoint @xmath194 . in case if aperture collimation is imposed on the photon beam , in view of the unambiguous relation between the photon energy @xmath35 and its emission angle [ see eq . ( [ omega - theta - corresp ] ) ] , function @xmath195 in eq . ( [ dw1coh - real ] ) will change , but still remain finite everywhere . if the transverse oscillatory motion of the electron / positron happens to be highly anharmonic , or is already relativistic , higher harmonics in the single - photon spectrum develop , which can generally be incorporated as @xmath196 in particular , with the increase of the positron energy beyond @xmath197 gev , higher harmonics of channeling radiation in a crystal proliferate , and ultimately render the spectrum a synchrotron - like appearance . this case may also be regarded as universal ; it was dealt with in papers @xcite . henceforth , we will restrict ourselves to the simplest case of a single - harmonic dipole radiation described by eq . ( [ dw1coh - real ] ) . while in our equations the shape of @xmath198 will be handled as generic ( to allow for the possibility of aperture collimation ) , in graphical illustrations specific form ( [ f - def ] ) will be used throughout . ( a ) multiphoton probability spectra computed for dipole harmonic coherent radiation defined through eqs . ( [ dw1coh - real][f - def ] ) ( solid curves ) . red curve , corresponding to intensity parameter @xmath199 , @xmath200 [ small deviations from the single - photon spectrum ( [ f - def ] ) ] . green curve , the same for @xmath201 ( the highest summit reached by the fundamental maximum ) . blue curves , @xmath202 ( onset of high - intensity regime ) . dashed blue curve , gaussian approximation ( [ gauss ] ) . dot - dashed blue curve , corrected gaussian approximation ( [ chebyshev ] ) . ( b ) multiphoton energy spectrum @xmath203 for the same single - photon spectrum as in ( a ) , and intensity parameter values @xmath204 , 2.5 , 6 . the development of secondary maxima for the energy spectrum is more spectacular than for the probability spectrum @xmath205 . ( c ) photon multiplicity spectrum described by eqs . ( [ mean - n - omega - contour ] ) , ( [ dw1coh - real][f - def ] ) . solid curves correspond to same parameters as in ( a ) . dashed blue line is high - intensity approximation ( [ mean - n - at - gauss ] ) . ] inserting entries ( [ dw1coh - real][f - def ] ) to mellin integrals ( [ generic - eq ] ) and ( [ mean - n - omega - contour - non - casual ] ) , and computing them numerically , one can trace the evolution of the resummed spectrum and the multiplicity spectrum with the increase of the intensity . this is illustrated in fig . [ fig : pure - coh - spectrum ] for a few values of @xmath206 . the trends observed in those figures are : smoothening of the spectral shape with the increase of the intensity , the existence of an upper bound for the probability distribution ( fig . [ fig : pure - coh - spectrum]a ) , the appearance of a second maximum in the energy spectrum at moderate intensity ( fig . [ fig : pure - coh - spectrum]b , green curve ) , and monotonic increase of the photon multiplicity spectrum with @xmath20 ( fig . [ fig : pure - coh - spectrum]c ) . explaining the origin of those features will occupy us next . when the single - photon spectrum is described by eq . ( [ dw1coh - real ] ) , at sufficiently low radiation intensity , the multiphoton spectrum will as well be concentrated within the interval @xmath207 ( termed fundamental ) see fig . [ fig : pure - coh - spectrum]a , red and green curves . so , it seems natural to begin with evaluation of the multiphoton spectrum in the fundamental interval . the calculation is straightforward based on representation ( [ generic - eq ] ) , written for finite @xmath208 as @xmath209 here we dropped the @xmath21-term for @xmath60 , and by virtue of lemma [ lemma1 ] , the upper limit in the integral over @xmath35 was replaced by infinity , without the necessity to employ the heaviside step function , which therefore was omitted . if @xmath195 is a polynomial , like ( [ f - def ] ) , the integral in the exponent of ( [ wcoh-1stinterv ] ) evaluates as a finite - order polynomial in @xmath210 : @xmath211 where @xmath212 stands for the @xmath213-th derivative of @xmath198 w.r.t . its argument . inserting ( [ polynom - s ] ) to ( [ wcoh-1stinterv ] ) , and rescaling the integration variable to @xmath214 , yields @xmath215 benefiting from the exponential decrease of the integrand at @xmath159 , along with the absence of cuts in the complex @xmath216-plane , we replaced the infinite integration contour by a closed loop encircling the origin , where the only singularity of the integrand resides ( contour @xmath148 in fig . [ fig : contours ] ) . [ [ infrared - limit . ] ] infrared limit . + + + + + + + + + + + + + + + despite the closed integration contour , and representation of the exponent in the integrand by only a few power terms , integral ( [ dwc - oint ] ) generally does not permit classification in terms of basic special functions . , i.e. , function @xmath198 was constant . such a model may be suitable for quick estimates , but generally is not acceptable numerically . ] but it is instructive at least to obtain its expansion about endpoint @xmath217 . there , one of the exponentials tends to unity , @xmath218 , and if the second exponential is expanded into laurent series @xmath219 the only term surviving after the loop integration is @xmath220 , leaving the result @xmath221 this equation shows that in the limit @xmath120 , the single - photon probability dominates , albeit in conjunction with photon non - emission probability @xmath222 . for the photon multiplicity spectrum , eq . ( [ coh - omegato0 ] ) , when inserted to eq . ( [ mean - n - omega - convol ] ) leads to eq . ( [ mean - n - to1 ] ) . to derive an @xmath223 correction to eq . ( [ coh - omegato0 ] ) , one needs to retain also a linear term in the expansion of @xmath224 : @xmath225\!\!+\!\mathcal{o}(\omega^2)\!\right\}\!.\ ] ] for our specific example ( [ f - def ] ) , @xmath226 . but for a sufficiently high intensity , specifically @xmath227 the coefficient at the linear term in ( [ coh - omegato0-corr ] ) turns positive , wherewith the resummed spectrum in the whole fundamental interval becomes monotonously increasing . that may be associated with the inception of high - intensity regime , which will be examined in the next section . from eqs . ( [ coh - omegato0-corr ] ) and ( [ mean - n - omega - convol ] ) , one similarly infers an expression for the photon multiplicity spectrum @xmath228 this shows that the slope in the origin does not depend on @xmath229 , and is always positive , in accord with eq . ( [ mean - n - gtr1 ] ) . the increase of the slope in the origin with the increase of parameter @xmath192 agrees with fig . [ fig : pure - coh - spectrum]b . [ [ par : fund - max ] ] fundamental maximum . + + + + + + + + + + + + + + + + + + + + the opposite endpoint of the fundamental interval , @xmath230 , is of prime practical significance , as long as it represents the global spectral maximum at moderate radiation intensity . substituting @xmath231 in eq . ( [ dwc - oint ] ) leaves @xmath232 effectively , this is a function of a single dimensionless parameter @xmath233 , inasmuch as all @xmath234 . in fig . [ fig : intensity - in - max ] , the dependence of ( [ max - int - vs - w1c ] ) on @xmath206 [ with @xmath198 defined by eq . ( [ f - def ] ) ] is displayed by the solid gray curve . naturally , with the increase of @xmath192 , at first it rises proportionally , but eventually the rise halts and ends up with a decrease , formally owing to factor @xmath222 . the ultimate suppression of the resumed spectrum at any fixed energy @xmath20 is not surprising , given the saturation of total probability ( [ total - probab-1 ] ) on one hand , and the spread of the spectrum to higher @xmath20 on the other hand . to estimate the maximal summit of function ( [ max - int - vs - w1c ] ) , one can adopt the following approach . to capture the first bend of the nonlinear @xmath192-dependence , expand the contour integral in ( [ max - int - vs - w1c ] ) to second order in @xmath192 ( which is equivalent to keeping in eq . ( [ generic - sum - int ] ) the first two terms ) : @xmath235\nonumber\\ % \qquad\equiv e^{-\omega_0b\int_0 ^ 1dzp(z)}\left[bp(1)+\omega_0b^2\frac12\int_0 ^ 1dzp(z)p(1-z)\right]\\ & = & p(1)e^{-b\omega_0\xi}\left(b\omega_0+\frac{b^2\omega_0 ^ 2}2\eta\right),\,\,\label{40}\end{aligned}\ ] ] where @xmath236 @xmath237 the behavior of approximation ( [ 40 ] ) is illustrated in fig . [ fig : intensity - in - max ] by the dashed gray curve . the maximum of ( [ 40 ] ) can be explicitly found by differentiation by @xmath206 : it is situated at @xmath238 for @xmath198 described by eq . ( [ f - def ] ) ( wherewith @xmath239 , @xmath240 ) , eq . ( [ loc - max ] ) yields @xmath241 , at which @xmath242 . for comparison , the maximum of exact expression ( [ max - int - vs - w1c ] ) is achieved at @xmath243 , and amounts 0.84 ( see fig . [ fig : intensity - in - max ] , solid gray curve ) . hence , at any radiation intensity , the probability spectrum @xmath32 of multiphoton coherent radiation at practice never exceeds unity is such that @xmath157 is exceptionally small , while @xmath244 is sizable . those are typical conditions for strong aperture collimation . but that case is rather trivial , because multiphoton effects are then suppressed as a whole . ] in marked contrast with the unlimited growth of the single - photon spectrum . [ at that , the energy spectrum @xmath245 can indefinitely grow , @xmath246 see sec . [ subsubsec : centr ] , eq . ( [ gauss ] ) ] . the spectral intensity at the fundamental maximum ( @xmath230 ) vs. the radiator length ( in units of @xmath206 ) , for a single - photon spectrum described by eq . ( [ f - def ] ) . solid gray curve , pure coherent radiation [ eq . ( [ max - int - vs - w1c ] ) ] . dashed gray curve , approximation ( [ 40 ] ) . solid black curve , the combination of coherent and incoherent radiation [ eq . ( [ max - dw ] ) ] in proportion @xmath247 , for @xmath248 ( @xmath249 ) . dashed black curve , approximation ( [ corr-1fund - max ] ) . ] besides peaking at @xmath231 , the coherent radiation spectrum encounters at this point a discontinuity , which is actually easier to evaluate . for our simplified case ( [ f - def ] ) , the discontinuity of the single - photon spectrum ( defined to be a positive quantity ) just equals the height of its maximum : @xmath250 with the account of emission of several photons , the spectrum beyond @xmath3 deviates from zero , so higher terms of series ( [ generic - sum - int ] ) demand consideration . but all terms with @xmath27 prove to be continuous - th term of ( [ generic - sum - int ] ) the integrand function @xmath251 is discontinuous on every face of an @xmath2-cubic domain , but the entering @xmath21-function restricts the integration domain to a slicing plane @xmath252 , which is oblique and nowhere parallel to any face of the cube for @xmath27 ( see fig . [ fig : integr - domain]b ) . thereby , for @xmath253 no discontinuities in the @xmath20-dependence can arise . ] . hence , the discontinuity of the total spectrum just equals that of the first term . this term yet involves the total photon non - emission probability factor @xmath254 with integer @xmath27 , the resummed spectrum remains continuous , but its @xmath255-th derivative encounters a discontinuity ( as can be noticed already from figs . [ fig : pure - coh - spectrum ] ) . that also follows from the analysis of the phase space available for integration for terms of series ( [ generic - sum - int]a , b ) . furthermore , even in case if function @xmath198 is vanishing at @xmath256 , in representation ( [ generic - sum - int ] ) the function discontinuity at @xmath231 cancels , but discontinuities of its derivatives persist . we leave the proof of those statements to the reader . as regards the discontinuities of @xmath257 , fig . [ fig : pure - coh - spectrum]c shows that they are similar to those of @xmath205 , but have opposite sign [ because @xmath257 contains @xmath258 in the denominator ] . at low radiation intensity , the multiplicity spectrum exhibits a step - like behavior , approximately amounting to the smallest integer greater than @xmath259 . that is traced to the fact that in an interval @xmath260 @xmath261 , the @xmath2-photon component dominates . as the intensity increases , function @xmath257 smoothens out , given that a @xmath2-photon component becomes competitive with fewer - photon components even in intervals where those components are not yet extinct . having explored separately the shapes of resummed coherent and incoherent radiation spectra , we are now in a position to examine their nonlinear interplay . the physical example when those components are both significant is coherent bremsstrahlung , occurring when an electron or positron crosses a family of crystalline planes at an above - critical ( although small ) angle . thereat , the continuous interplanar potential acts on the electron periodically , evoking coherent radiation @xcite . at the same time , the incoherent scattering on atomic nuclei in the planes causes incoherent bremsstrahlung , similar to that in amorphous matter . in a satisfactory approximation , the single - photon radiation spectrum may be expressed as a sum of two non - interfering , because it is ir- and uv - safe , but that would complicate the structure of @xmath198 . on the other hand , if there is a coherent contribution to the electron r.m.s . deflection angle , it ought to be included to @xmath262 . ] parts : @xmath263 where @xmath262 is given by eq . ( [ dw1inc ] ) . in graphical illustrations , we will continue using eqs . ( [ dw1coh - real][f - def ] ) for @xmath264 , i.e. , neglect higher harmonics . in some cases , those can be small indeed even for coherent bremsstrahlung ( ` one - point ' spectra , see @xcite ) . at the same time , we saw in the previous section that the second maximum in the multiphoton spectrum can be generated even in the absence of secondary harmonics in the single - photon spectrum , so it will be expedient to check whether their effect is anyhow affected by incoherent radiation . for what concerns the ratio of magnitudes of coherent and incoherent radiation components , for coherent bremsstrahlung it obeys a relation @xmath265 with @xmath266 the misalignment angle between the electron momentum and the co - oriented family of atomic planes , and @xmath267 [ cf . ( [ a - amorph ] ) and ( [ dw - cb - fmax ] ) ] . at typical @xmath268 , ratio ( [ a / bomega0 ] ) is pretty small . nevertheless , effects of incoherent radiation in the spectrum may be quite noticeable . that is illustrated by fig . [ fig : high - intensity ] , where in spite of rather large initial ratio @xmath269 , the incoherent bremsstrahlung plateau in the multiphoton energy spectrum @xmath245 is only a few times lower than the spectrum height in the maximum ( fig . [ fig : high - intensity]b ) . handling the effects of incoherent radiation proves to be most convenient with the aid of a convolution representation derived below . once decomposition ( [ coh+inc ] ) is inserted to eq . ( [ generic - eq ] ) , the exponential in the integrand splits into a product of two factors , one depending on @xmath270 , and the other on @xmath271 . expressing then each of those factors via eq . ( [ generic - exponentiation ] ) through laplace transforms of the corresponding resummed radiation spectra , and doing the contour integral , we arrive at a convolution relation . ] : @xmath272 the term in the second line vanishes if @xmath182 divergesat practice , when the incoherent component receives a physical ir cutoff , factor @xmath273 may be not really small , so the last term in ( [ wcoh - times - winc ] ) may remain sizable . to be more quantitative , @xmath274 with @xmath88 the ir cutoff value . but for cutoffs of different nature , and practical energies @xmath4 , ratio @xmath275 can hardly exceed @xmath276 , wherewith @xmath277 . hence , @xmath273 can really be vanishing only provided @xmath278 . we will restrict our consideration to the idealized case when this condition is met . the extension to finite @xmath182 is straightforward , and basically resembles the situation described in the previous subsection . ] , whereupon expression ( [ wcoh - times - winc ] ) becomes linear in @xmath279 . still , the resulting convolution is non - vanishing at @xmath280 , because , as we saw in sec . [ subsec : incoh - bremsstr ] , the multiphoton incoherent radiation spectrum in this limit tends to a @xmath21-function , not to zero [ see eq . ( [ delta - lim - ato0 ] ) ] . more precisely , if median @xmath185 defined by eq . ( [ omegaincoh12 ] ) proves to be much smaller than @xmath3 , convolution ( [ wcoh - times - winc ] ) will be close to @xmath205 . observing further that in eq . ( [ wcoh - times - winc ] ) @xmath281 is a smooth function , whereas the coherent radiation spectrum @xmath282 is not , it may be beneficial to integrate in ( [ wcoh - times - winc ] ) by parts : @xmath283 invoking here eq . ( [ coh - omegato0 ] ) along with identity @xmath284 , the convolution relation recasts @xmath285\qquad\qquad\quad\nonumber\\ -\phi(a)\int_0^{\omega}d\omega ' \left(\frac{\omega'}{e}\right)^a\frac{d}{d\omega'}\left(\frac{dw_\text{c}}{d\omega''}\bigg\|_{\omega''=\omega-\omega'}\right).\end{aligned}\ ] ] the merit of the latter representation is that it picks up discrete contributions from points at which @xmath282 encounters discontinuities , transforming to @xmath21-function - like terms for its derivative . ( a ) multiphoton radiation spectrum involving coherent component with magnitude @xmath201 , and incoherent component with @xmath286 . the uv cutoff parameter is chosen at @xmath248 . solid green curve , the exact result obtained by numerical evaluation of the contour integral . long - dashed black curve , corrected low-@xmath20 approximation ( [ dw - small - omega - nlo ] ) . dotted black curve , approximation next to the coherent emission edge , eq . ( [ smearing - of - disc ] ) . dot - dashed black curve , large-@xmath20 asymptotics ( [ simple - large - omega ] ) . ( b ) the energy spectrum for same conditions as in ( a ) . short - dashed black curve , low-@xmath20 approximation ( [ dw - small - omega ] ) . dot - dashed black curve , large-@xmath20 asymptotics ( [ simple - large - omega ] ) . ( c ) the photon multiplicity spectrum for same parameters as in ( a , b ) , and ir cutoff @xmath287 . solid green curve , exact result . dot - dashed black curve , large-@xmath20 approximation ( [ n - omega - asympt - large ] ) . ] in the limit @xmath120 , the integral term in eq . ( [ convol - byparts ] ) , vanishes due to the integration interval shrinkage , leaving @xmath288.\ ] ] clearly , here the leading term @xmath289 represents the contribution due to incoherent bremsstrahlung , but it receives an extra suppression by factor @xmath222 depending on the coherent radiation . hence , the incoherent component can not be asserted to dominate in this limit alone , in contrast to the case of large @xmath20 ( see sec . [ subsubsec : large - omega - asympt ] below ) . the behavior of asymptotic approximation ( [ dw - small - omega ] ) is illustrated in fig . [ fig : high - intensity]b by the short - dashed black curve . in a similar fashion , one can derive a convolution relation for the multiplicity spectrum , which reads @xmath290\\ + e^{-w_{1\text{c}}}\bar n_\text{i}(\omega)\frac{dw_\text{i}}{d\omega } + \underset{\hookrightarrow0}{e^{-w_{1\text{i}}}}\bar n_\text{c}(\omega)\frac{dw_\text{c}}{d\omega},\nonumber\end{aligned}\ ] ] with @xmath291 and @xmath292 the corresponding multiplicity spectra for pure coherent and incoherent components . in particular , from eqs . ( [ convol - for - n - omega ] ) and ( [ wcoh - times - winc ] ) one observes that both @xmath32 , and @xmath293 involve @xmath50-dependent factors @xmath294 , which cancel between l.h.s . and r.h.s . of eq . ( [ convol - for - n - omega ] ) . thus , @xmath128 is independent of @xmath50 , as anticipated . for an efficient coherent radiator , the coherent spectral component must prevail in the fundamental interval @xmath295 , anyway . it is thus instructive to repeat the analysis of sec . [ subsubsec : coh - in - fund - interv ] for a mixture of coherent and incoherent radiation components in the fundamental interval . we will barely sketch it here , emphasizing the distinctions from the pure coherent radiation case . once expressions ( [ coh+inc ] ) , ( [ dw1inc ] ) , ( [ dw1coh - real ] ) are inserted to eq . ( [ generic - eq - upperlimitomega ] ) , manipulations similar to those used at derivation of eq . ( [ dwc - oint ] ) yield @xmath296 here exponential decrease of the integrand at @xmath159 justifies deformation of the integration contour to shape @xmath147 ( see fig . [ fig : contours ] ) , but its complete enclosure is impossible because of the presence of branching factor @xmath297 . [ [ infrared - limit.-1 ] ] infrared limit . + + + + + + + + + + + + + + + when @xmath298 , contour integral ( [ dw - first - interv ] ) is dominated by large @xmath216 . then , the leading - order term in asymptotic expansion ( [ exp - to - laurent ] ) reproduces result ( [ dw - small - omega ] ) . retaining next - to - leading order terms yields an @xmath223 correction : @xmath299+\mathcal{o}(\omega^2)\bigg\}.\end{aligned}\ ] ] compared to ( [ coh - omegato0-corr ] ) , apart from the overall factor @xmath300 , there emerges a term @xmath301 , whereas other terms remain essentially the same . the term in the second line of eq . ( [ dw - small - omega - nlo ] ) changes its sign under the same condition ( [ monot - after4 ] ) , so inception of high - intensity regime is essentially independent of @xmath135 . the behavior of asymptotic approximation ( [ dw - small - omega - nlo ] ) is illustrated in fig . [ fig : high - intensity]a by the gray long - dashed curve . [ [ reduction - of - the - fundamental - maximum - by - hard - incoherent - bremsstrahlung . ] ] reduction of the fundamental maximum by hard incoherent bremsstrahlung . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in the point of the spectral fundamental maximum , @xmath302 , integral ( [ dw - first - interv ] ) reduces to @xmath303 since @xmath135 and @xmath192 are both proportional to the target thickness , the increase of the latter at fixed @xmath266 [ cf . ( [ a / bomega0 ] ) ] corresponds to a simultaneous increase of @xmath192 and @xmath135 , at ratio @xmath304 held fixed . for an exemplary ratio @xmath305 , the spectral intensity in the fundamental maximum is illustrated in fig . [ fig : intensity - in - max ] by solid gray curve . compared to the pure coherent radiation spectrum , its maximum is achieved at a slightly lower value of @xmath192 , and its height is somewhat lower , too . this owes primarily to factor @xmath306 , where @xmath307 is a small number , but there can also be other appreciable effects of @xmath135 . to assess changes in the height and location of the fundamental maximum due to incoherent radiation , one can , as in sec . [ subsubsec : coh - in - fund - interv ] , expand the contour integral in ( [ max - dw ] ) to second order in @xmath192 and @xmath135 : @xmath308.\end{aligned}\ ] ] it may suffice to account for corrections in @xmath135 within the leading logarithmic accuracy , which gives @xmath309\,\,\,\nonumber\\ + \frac{b^2\omega_0}2\int_0 ^ 1dzp(z)p(1-z ) \bigg\}\quad\end{aligned}\ ] ] ( see fig . [ fig : intensity - in - max ] , dashed black curve ) . here term @xmath310 can be sizable due to the large logarithm , despite the condition @xmath175 . owing to its negative sign , that correction reduces the height of the fundamental coherent radiation maximum , in addition to factor @xmath222 . in the spirit of sec . [ subsubsec : coh - in - fund - interv ] , for approximation ( [ corr-1fund - max ] ) one can determine the target thickness at which the spectral radiation intensity in the fundamental maximum reaches its highest summit . it is still given by formula ( [ loc - max ] ) , with parameter @xmath157 unaltered , but parameter @xmath311 modifies to @xmath312 corrections in the numerator and denominator of ( [ eta - coh+incoh ] ) both result in lowering of @xmath311 , and therethrough in an earlier turnover of the fundamental maximum , as confirmed by fig . [ fig : intensity - in - max ] . the largest effect , of course , stems from the logarithmic term in ( [ eta - coh+incoh ] ) . a principally interesting effect of incoherent radiation concerns discontinuities of the coherent radiation spectrum . as we saw in sec . [ subsubsec : disc - dwc ] , the discontinuity at @xmath231 is damped by factor @xmath313 . with the addition of the incoherent bremsstrahlung component , this factor actually turns to zero , wherefore the discontinuity must be nullified . ] . our analysis of the spectrum behavior then needs extension to a non - vanishing vicinity of point @xmath231 . in vicinity of a singularity of the coherent part of the spectrum , it is convenient to use representation ( [ convol - byparts ] ) . when @xmath20 slightly exceeds @xmath3 , the integrand of ( [ convol - byparts ] ) receives an extra contribution proportional to a @xmath21-function @xmath314 where @xmath315 is given by eq . ( [ disc ] ) . integration of the @xmath21-singularity yields a sizable contribution even at an infinitesimal extension of @xmath20 beyond @xmath3 : @xmath316 eq . ( [ smearing - of - disc ] ) shows that for @xmath317 , factor @xmath318 vanishes as @xmath319 , indeed nullifying the discontinuity brought by factor @xmath320 , and making the whole spectrum everywhere continuous . but the derivative of factor @xmath318 in point @xmath231 diverges , wherewith the resulting spectrum features a sharp spike . the behavior of approximation ( [ smearing - of - disc ] ) is shown in fig . [ fig : high - intensity]a by the dotted black curve . it should be realized , though , that the described subtle effect should be obscured in experiments with limited energy resolution and finite angular divergence of the initial beam , which smear the fundamental peak . besides that , the filling of the dip adjacent to the coherent radiation maximum is also partially provided by the pure coherent multiphoton spectrum , whose derivative beyond @xmath3 is negative ( see fig . [ fig : pure - coh - spectrum]a ) . thus , verification of threshold behavior ( [ smearing - of - disc ] ) may be feasible only with a rather perfect beam , and in a sufficiently thick target , when @xmath135 becomes sizable . to accomplish our analysis in different regions , let us consider the large-@xmath20 asymptotics @xmath321 ( while still obeying condition @xmath322 ) . in this limit , the incoherent bremsstrahlung spectrum decreases by a power law , whereas the multiphoton coherent radiation spectrum , in general , decreases exponentially ( see appendix [ app : periph ] ) . therefore , if eq . ( [ wcoh - times - winc ] ) is rewritten as @xmath323 the main contribution to the integral over @xmath324 comes from vicinity of the lower integration limit . thus , we can expand @xmath325 to taylor series about point @xmath326 , and replace the upper integration limit by infinity : @xmath327 the integral in the second line vanishes due to eq . ( [ mean - omega ] ) , leaving @xmath328.\ ] ] furthermore , in the limit @xmath329 , eq . ( [ large - omega - corr ] ) boils down to @xmath330,\ ] ] which is shown in figs . [ fig : high - intensity]a , b by dot - dashed black curves . ( [ simple - large - omega ] ) implies that the combined spectrum ultimately decreases by the same power law as its ( resummed ) incoherent bremsstrahlung component . that property is in contrast with the infrared limit ( [ dw - small - omega ] ) , which yet involves factor @xmath222 of coherent photon non - emission probability [ note the difference between green and black long - dashed curves in fig . [ fig : high - intensity]a at small @xmath20 ] . finally , it is straightforward to derive large-@xmath20 asymptotics for the photon multiplicity spectrum , with the aid of eq . ( [ convol - for - n - omega ] ) . due to factor @xmath332 in the integrand of ( [ convol - for - n - omega ] ) , again , the dominant contribution comes from the lower integration limit , where it is acceptable to replace @xmath333 $ ] , @xmath334 $ ] , and replace the upper integration limit by infinity . therewith , eq . ( [ convol - for - n - omega ] ) yields @xmath335,\end{aligned}\ ] ] where @xmath336 is given by eq . ( [ mean - n - omega - incoh ] ) . ( terms involving @xmath222 canceled mutually . ) the integral entering ( [ 83 ] ) equals @xmath337 , as can be checked with the aid of eq . ( [ mean - n - omega - contour - non - casual ] ) . canceling in ( [ 83 ] ) the overall factors according to eq . ( [ simple - large - omega ] ) , we are left with @xmath338\left[1+\mathcal{o}\left(\frac{\overline{\omega_{1\!}}_{\text{c}}}{\omega}\right)\right].\ ] ] the behavior of the photon multiplicity spectrum is shown in fig . [ fig : high - intensity]c by the solid green curve . at large @xmath20 , it flattens out , and ultimately enters the regime of logarithmic growth as predicted by eq . ( [ n - omega - asympt - large ] ) ( dot - dashed black curve ) . yet before entering the asymptotic regime , as fig . [ fig : high - intensity]b shows , the multiplicity spectrum first achieves a maximum , then passes through a shallow minimum . the theory for such a behavior will be provided in sec . [ subsec : conv - inc ] . the preceding analysis revealed that multiphoton emission effects redistribute the coherent radiation spectrum , elevating it beyond the fundamental energy @xmath3 , at the expense of reduction below @xmath3 . at practice , though , one should yet be aware of existence of other physical processes which can produce superficially similar effects . in conclusion of this section , we will briefly discuss such effects , too . first of all , next to the coherent emission edge @xmath3 , the spectrum may receive an enhancement not only from multiphoton effects , but also from higher harmonics ( [ dw1coh - higher - harm ] ) . at that , higher harmonics produce also 3rd , 4th maxima , etc . , whereas multiphoton effects can only give rise to a relatively narrow second maximum between @xmath3 and @xmath339 . so , if no maxima are observed beyond @xmath339 , the simplest way to judge about the origin of the secondary peak would be to refer to the photon multiplicity spectrum : the greater its discontinuity at @xmath340 , the greater the significance of multiphoton effects . if measurement of @xmath128 is unfeasible , an indirect criterion may be used : the relative contribution of higher harmonics does not vary with the extent of the radiator , whereas that of multiphoton effects does . therefore , comparing the spectrum shapes at two different radiator lengths , and assessing the length - dependence nonlinearity ( cf . [ subsec : ln ] ) , one can judge about the origin of the spectrum enhancement beyond @xmath3 . another effect is the difference between asymptotics of the resummed energy spectrum @xmath245 at @xmath120 and at @xmath341 [ see fig . [ fig : high - intensity]b , and eqs . ( [ simple - large - omega ] ) and ( [ dw - small - omega ] ) ] . this may either owe to multiphoton effects , or to incomplete lpm - suppression of radiation in a finite - thickness target [ existing as well in an amorphous matter ( tsf - effect ) , see @xcite and refs . therein ] . again , unambiguous discrimination between those effects may rely on the photon multiplicity spectrum behavior : if within the suppression region @xmath128 rises significantly above unity ( cf . , e.g. , fig . [ fig : high - intensity]c ) , the suppression origin ought to be attributed to multiphoton effects , otherwise , it is more likely to be due to lpm - like effects . discrimination criteria based on the target thickness dependence appear to be nonlinear both for multiphoton and incomplete lpm suppressions , and therefore are to be used cautiously . the measure of significance of multiphoton effects in a resummed radiation spectrum is given by the mean photon multiplicity , defined by eqs . ( [ mean - n - def ] ) , ( [ w1-def ] ) . for the dominant coherent radiation component ( [ dw1coh - real][f - def ] ) , it estimates as @xmath342 from the practical viewpoint , it is important further to estimate how large this parameter can be for 3 basic coherent gamma - radiation source types : coherent bremsstrahlung , channeling radiation , and undulator radiation . generic expressions for product @xmath206 for mentioned cases are quoted in appendix [ app : coh - rad - sources ] , so it is now left to assess the entering parameters for conditions of former and future experiments . for coherent bremsstrahlung , parameter @xmath206 is described by eq . ( [ dw - cb - fmax ] ) . early coherent bremsstrahlung spectrum measurements operated with relatively thin crystals ( @xmath343 mm ) and relatively large misalignment angles @xmath344 rad , at which @xmath208 was comfortably small . but with the advent of more practical radiation sources having @xmath345 cm and @xmath346 rad , formidable values @xmath347 were reached ( corresponding to an extensive electromagnetic shower ) . for channeling radiation , the reference equation is ( [ dw - chann - fin ] ) . it tells that at multi - gev positron energies , and @xmath348 , the photon multiplicity must achieve values @xmath349 . multiphoton effects in channeling radiation experiments were found appreciable already when dealing with moderately high energies @xmath350 gev , and moderate - thickness targets , @xmath351 , corresponding to @xmath352 @xcite . in more recent cern channeling experiments @xcite , with @xmath353 gev and @xmath354 , photon multiplicities reached the order of 5 , and the measured spectra were apparently loosing the features characteristic of coherent radiation ( see also @xcite ) . it should be realized , though , that at @xmath197 gev , channeling radiation becomes non - dipole , wherewith simple form ( [ dw1coh - real][f - def ] ) for the single - photon spectrum does not apply [ although contour integral representations ( [ generic - eq ] ) , ( [ generic - eq - upperlimitomega ] ) hold generally ] . for undulator radiation , we must refer to eq . ( [ dw - undul ] ) . for the newest and forthcoming undulators characterized by @xmath355 and @xmath356 @xcite , parameter @xmath206 must be @xmath357 . the only question is whether @xmath3 belongs to the gamma - range , for which the calorimetric method of spectrum measurement is pertinent . in the slac experiment e-166 @xcite , @xmath3 reached @xmath358 mev , though only under moderate photon multiplicity @xmath359 . for the tesla design ( @xmath360 gev , @xmath361 m , @xmath362 , @xmath363 ) , @xmath3 must rise to 25 mev , and correspond to @xmath364 @xcite . such photon energies are admittedly well into the gamma domain . the present estimates show that virtually for all practical coherent radiation sources it appears both feasible and beneficial to enter the regime @xmath349 . in some cases , it may be optimal to confine to values @xmath365 , when the spectrum is not strongly multiphoton yet ( see fig . [ fig : intensity - in - max ] ) . in other cases , such as positron sources @xcite , @xmath208 is demanded to be large indeed . in either case , exploring the asymptotic limit of high @xmath208 is of fundamental interest . its study will occupy us in the remainder of this section . to get feel of the trends for the resummed radiation spectrum behavior at high photon multiplicity , let us first take a look at spectral moments introduced in sec . [ subsec : moments ] . relations ( [ mean - omega ] ) , ( [ mean - omega2-alt ] ) suggest that with the increase of the radiator length , multiphoton spectrum moments grow proportionally : @xmath366 @xmath367 ( and normally @xmath368 ) . therewith , the ratio of the width to mean value decreases : @xmath369 that decrease implies that the multiphoton spectrum becomes more sharply peaked . furthermore , from eq . ( [ mean - omega4 ] ) one infers that the so - called skewness @xcite @xmath370 decreases as well , while the kurtosis @xmath371\ ] ] tends to a constant value 3 ( characteristic of a gaussian ) . ) , ratio @xmath372 appears to be always greater than 3 , and diverges at small @xmath192 , even though the underlying single photon spectrum may well be leptokurtic [ like that defined by eqs . ( [ dw1coh - real][f - def ] ) ] . that owes to our definition of the moments including term @xmath373 in the weighting distribution , which makes any distribution platikurtic . in the high - intensity limit , where @xmath149 exponentially , this definition suits us , anyway . ] the above observations look natural from the point that the process of statistically independent photon emission is a kind of a random walk @xcite in the energy space . for our case , the walk is one - sided , with continuously distributed step size , yet the total probability of each step is less than unity . to substantiate the analogy further , @xmath94 in eq . ( [ generic - eq ] ) may be thought of as depending on the target thickness @xmath100 . then , differentiating both sides of the equation by @xmath100 yields a kinetic equation [ kinetic - eq ] @xmath374 where kernel @xmath375 stands for the radiative cross - section . if we take into account that in the first term of ( [ kinetic - eq1 ] ) @xmath376 equals zero for @xmath97 , we may also rewrite the equation as @xmath377 the initial condition for eq . ( [ kinetic - eq1 ] ) or ( [ kinetic - eq2 ] ) is @xmath378 it is noteworthy that the last term in eqs . ( [ kinetic - eq ] ) makes them inhomogeneous with respect to @xmath32 . that term is necessary to describe the change of the normalization with @xmath100 according to eq . ( [ total - probab-1 ] ) . at small @xmath100 , when @xmath379 , the inhomogeneous term dominates , but at large @xmath100 , it vanishes exponentially . for the electron distribution function @xmath380 ( see sec . [ subsec : cont - int - repr ] ) , the kinetic equation will be strictly homogeneous : [ kinetic - eq - for - pie ] @xmath381 \\ \equiv\int_0^\omega d\omega_1 \pi(e_e-\omega+\omega_1)\frac{\partial}{\partial l}\frac{dw_1}{d\omega_1}\quad \nonumber\\ -\pi(e_e-\omega)\int_0^e d\omega_1 \frac{\partial}{\partial l}\frac{dw_1}{d\omega_1}.\qquad\label{kinetic - eq - for - pie2}\end{aligned}\ ] ] with initial condition @xmath382 in the literature , the kinetic equation is usually quoted in the form of electromagnetic cascade equation ( [ kinetic - eq - for - pie]b ) , but the function entering thereto is sometimes also called the multiphoton spectrum ( which may be misleading ) . in the presence of significant incoherent radiation , though , the inhomogeneous term should vanish , anyway , insofar as @xmath127 . note , incidentally , that integro - differential equation ( [ kinetic - eq - for - pie ] ) is well - suited for numerical solution by monte - carlo method , allowing one to simulate a multiphoton radiation spectrum without actually computing separate @xmath2-photon components . alternative forms ( [ kinetic - eq]a , b ) are suitable for that purpose , too , but granted that their initial condition ( [ kinetic - eq - init - cond-0 ] ) is non - singular , they must be solvable as well by non - random finite - difference methods . in fact , knowledge of @xmath383 may be sufficient to extract also the photon multiplicity spectrum : assuming that the single - photon spectrum is proportional to @xmath100 , i.e. @xmath384 , eq . ( [ kinetic - eq2 ] ) rewrites @xmath385 then , noticing that the first two terms on the r.h.s . of ( [ rewrit ] ) coincide with the r.h.s . of eq . ( [ mean - n - omega - convol ] ) , one arrives at a relation @xmath386 ( where for ir regularization , @xmath387 may need to be replaced by @xmath388 ) . the linear homogeneous 1st - order integro - differential form of equation ( [ kinetic - eq - for - pie ] ) provides the correspondence between our multiphoton emission process and continuous random walks ( a rather general category of markov processes ) in principle may depend on @xmath100 , i.e. , on the electron history , in an arbitrary way , but its further promotion to multiphoton spectrum @xmath389 is sought as solution of a markovian differential equation with respect to increase of @xmath390 , or simply @xmath100 . ] . ( [ kinetic - eq ] ) actually describe a _ driven _ random walk , although it is equivalent to a free one , and belongs to the markovian process category , anyway . now , let us turn to the case of large @xmath100 . then , @xmath13 becomes small , and the radiation process definitely enters the free random walk regime . intuitively , for long random walks , detail of the single - step distribution should fade away after a large number of steps . the central limit theorem @xcite asserts that for a sufficiently long random walk , the particle probability distribution tends to a gaussian , involving only two parameters : the mean value and the variance of the single - step distribution ( single - photon spectrum ) . it must be realized , however , that under the presence of incoherent bremsstrahlung component , the first moments defining the gaussian diverge . that is a familiar situation when the gaussianity breaks down , and the diffusion becomes anomalous @xcite . for such a case , there exists a generalized central limit theorem @xcite , but it is not really relevant for our conditions , as long as the incoherent component intensity , quantified by parameter @xmath135 , is not supposed to become large ( @xmath391 ) . a reasonable way out of the present situation may be to treat the incoherent radiation component as a single - jump process , rather than a multistep random walk . to this end , one may first cope with the pure coherent radiation in the high multiplicity regime independently , where the problem of divergent moments is not encountered ( that may also be of intrinsic interest in application to undulator radiation , where @xmath135 is often negligible ) . at the final step of the calculation , the contribution from the incoherent radiation component can be restored via convolution relation ( [ wcoh - times - winc ] ) . following this approach , in the next subsection we deal with the pure coherent radiation case . since typical @xmath20 in the coherent radiation spectrum at high intensity are @xmath392 , energetically ordered representations offer in this case no advantage . it is thus simpler to issue from contour integral representation ( [ generic - eq ] ) , @xmath393 where we explicitly introduced the coherent emission edge @xmath3 as the finite upper limit for @xmath35-integration . obviously , as @xmath264 grows large , the exponential integrand of ( [ cont - int - for - wc ] ) steepens as a function of @xmath53 , peaking somewhere between the integration contour endpoints . such a situation suggests application of the steepest descent method , which is a common tool in the random walk theory . for our case , though , the corresponding procedure yet involves certain subtleties which will be highlighted below . the steepest descent method ( see , e.g. , @xcite ) determines the asymptotics of a contour integral by deforming its integration contour in the complex plane so that it passes through a point , to both sides from which the integrand decreases by its absolute value , without significant oscillations . for an analytic integrand , this must be a regular extremum ( saddle point in the complex plane ) , where the derivative vanishes . hence the equation for saddle point @xmath394 of the integrand of ( [ cont - int - for - wc ] ) emerges as ) involves no preexponential , the equation for the saddle point is basically unequivocal . ] @xmath395\bigg\|_{s = s_0}\nonumber\\ = \omega-\int_0^{\omega_0 } d\omega_1 \omega_1\frac{dw_{1\text{c}}}{d\omega_1}e^{-s_0\omega_1}=0.\end{aligned}\ ] ] once solution @xmath396 to eq . ( [ saddle - point - eq ] ) is found , the exponent in eq . ( [ cont - int - for - wc ] ) is further expanded around this point to second order : @xmath397+\frac12 b[s_0(\omega)](s - s_0)^2%\nonumber\\ + \mathcal{o}\left[(s - s_0)^3\right]\!.\quad\end{aligned}\ ] ] to express coefficients @xmath398 and @xmath399 , we write in the exponent of ( [ cont - int - for - wc ] ) @xmath400,\end{aligned}\ ] ] and notice that term @xmath401 , when convolved with @xmath264 , cancels with term @xmath402 by virtue of eq . ( [ saddle - point - eq ] ) . that produces the expected structure ( [ hd ] ) , wherefrom we read off @xmath403=\int_0^{\omega_0 } d\omega_1 \frac{dw_{1\text{c}}}{d\omega_1}\left[e^{-s_0\omega_1}(1+s_0\omega_1)-1\right],\ ] ] and @xmath404=\int_0^{\omega_0 } d\omega_1 \omega_1 ^ 2\frac{dw_{1\text{c}}}{d\omega_1}e^{-s_0\omega_1}=-\frac1{s_0}\frac{\partial a}{\partial s_0}.\ ] ] the neglect in the exponent of ( [ cont - int - for - wc ] ) of taylor terms higher than quadratic reduces the integrand to a gaussian , and the contour integral evaluates @xmath405 one should yet be aware that the saddle - point equation in the complex plane may have several solutions . for our edgy single - photon spectrum ( [ dw1coh - real ] ) , the number of solutions is actually infinite . to demonstrate that , the distribution of saddle points in the complex @xmath53-plane , and their motion with the variation of @xmath20 , is portrayed in fig . [ fig : re - plot3d ] for specific single - photon spectrum profile ( [ dw1coh - real][f - def ] ) . there , black solid curves correspond to the @xmath20-independent imaginary part of eq . ( [ saddle - point - eq ] ) , giving saddle points trajectories . dashed color curves correspond to the real part of eq . ( [ saddle - point - eq ] ) ; their intersections with the solid curves give saddle point locations , which depend on @xmath20 . the figure suggests that for any @xmath20 , the number of saddle points must be infinite , so instead of eq . ( [ dwdomega - saddle ] ) one ougth to write @xmath406 where the summation runs over all the saddle points , and @xmath407 , @xmath408 . ( a ) the distribution of saddle points in the complex plane of @xmath53 . solid black curves are solutions of equation @xmath409 ( saddle - point trajectories ) . dashed curves are solutions of equation @xmath410 ( magenta , @xmath411 ; green , @xmath412 ; cyan , @xmath413 ) . intersections of solid curves with dashed ones give saddle points at a given @xmath414 . the density plot , function @xmath415 , for @xmath416 given by eq . ( [ h ] ) ( brighter regions correspond to greater values of @xmath417 ) . ( b ) values of function @xmath415 in the saddle points . [ for @xmath413 ( cyan ) , the plotted function is multiplied by 0.02 . ] ] nevertheless , in the limit @xmath418 , series ( [ dwdomega - saddle - sum ] ) must be dominated by a term with the greatest @xmath417 . the diagram in fig . [ fig : re - plot3d]b shows the values of @xmath419 in saddle points . for any @xmath20 , the saddle point at @xmath420 appears to be the highest , and therefore asymptotically dominant . thus , in the limit @xmath421 , eq . ( [ dwdomega - saddle ] ) is recovered , with @xmath422 being the unique _ real _ solution of eq . ( [ saddle - point - eq ] ) . another issue is the robustness of the saddle point approximation itself . that approximation breaks down in vicinity of the point where coefficient @xmath399 vanishes , i.e. at small @xmath20 corresponding to large @xmath394 . to determine the condition of applicability of the saddle - point approximation , one needs to assess higher terms in @xmath423 in eq . ( [ hd ] ) . that leads to the inequality @xmath424 the l.h.s . thereof admits interpretation as an order of magnitude of the number of emitted photons with energies below @xmath20 . thus , ineq . ( [ saddle - point - cond ] ) implies that for the diffusion mode to develop , the same ` destination ' ( total energy ) @xmath20 must be reachable by many radiative scenarios with commensurable and sizable probability . if all the premises listed above are fulfilled , eq . ( [ dwdomega - saddle ] ) offers an explicit solution for the resummed spectrum , provided the real solution to equation ( [ saddle - point - eq ] ) is found and substituted to eqs . ( [ h][d ] ) . an impediment is that even with simplest shapes of @xmath264 , like ( [ f - def ] ) or a uniform distribution , equation ( [ saddle - point - eq ] ) is transcendental . on the other hand , in the crudest approximation , it may suffice to know its behavior around single point corresponding to the maximum of expression ( [ dwdomega - saddle ] ) . the latter region , called central , will be scrutinized in the next paragraph . the problem of finding the maximum of distribution ( [ dwdomega - saddle ] ) is essentially equivalent to tracing that for exponent @xmath398 , as long as the exponential should be an asymptotically steeper function than its preexponential factor . inspection of integral ( [ h ] ) reveals that for any positive kernel @xmath264 , the integral achieves a maximum in point @xmath425 , since there the @xmath394-dependent factor @xmath426 $ ] turns to zero , while being negative elsewhere . from eq . ( [ saddle - point - eq ] ) , one finds that value @xmath425 corresponds to @xmath427 . expansion the of r.h.s . of eq . ( [ h ] ) by deviations of @xmath394 from zero ( beginning from a quadratic term ) yields @xmath428= -\frac12 \overline{\omega^2_{1\!}}_{\text{c } } s_0 ^ 2+\mathcal{o}\left(\overline{\omega^3_{1\!}}_{\text{c}}s_0 ^ 3\right).\ ] ] if eq . ( [ saddle - point - eq ] ) in turn is linearized in @xmath394 , @xmath429 it solves as @xmath430 combining eqs . ( [ hcoh - quadr ] ) and ( [ saddle - point - eq - lin - soln ] ) , we obtain the explicit asymptotic expression for the probability spectrum in vicinity of the maximum ( central region ) : @xmath431 with @xmath432 as anticipated , this limiting distribution is merely a gaussian , with the mean value and variance complying with eqs . ( [ mean - omega][mean - omega2-alt ] ) , for an arbitrary shape of @xmath271 ( see , e.g. , @xcite ) . since dependencies @xmath433 for different shapes and magnitudes of @xmath264 and intensities are related by a linear change of the variables ( i.e. , the multiphoton spectrum shape is invariant under changes of @xmath264 ) , the central region is also called scaling one @xcite . so far , the approach to gaussian asymptotics remains largely unexplored experimentally , although spectra resembling gaussians were obtained , e.g. , in . approximation ( [ gauss ] ) holds as long as in eq . ( [ saddle - point - eq - lin ] ) nonlinear terms in @xmath394 keep relatively small , i.e. @xmath434 , the gaussian approximation must work in a sufficiently broad interval about its maximum . the corresponding photon multiplicity spectrum @xmath435 is further evaluated by referring to representation ( [ mean - n - omega - convol ] ) . its second term vanishes due to practically vanishing @xmath13 , or due to vanishing @xmath264 for @xmath436 , while in the first term , the integration is effectively limited to interval @xmath437 , where factor @xmath264 differs from zero . noting that within such an interval , the second factor @xmath438 varies slowly , one can taylor - expand it to first order , say , about the origin : @xmath439 wherewith @xmath440 inserting here expression ( [ gauss ] ) , we obtain an utterly simple result : @xmath441 the linear function in the r.h.s . of ( [ mean - n - at - gauss ] ) has positive intercept @xmath442 , and slope , but at small @xmath20 , its slope is @xmath443 [ see eq . ( [ n0=1 + ] ) ] . ] @xmath444 ; the latter implies that @xmath257 increases by more than unity per each interval @xmath445 . it is also evident that in vicinity of the gaussian maximum , @xmath446 which , according to ( [ mean - n - def ] ) , amounts the mean photon multiplicity for the whole coherent radiation process . an approximately linear increase of @xmath257 in the range of typical @xmath20 was observed in experiment @xcite . there was also observed a turnover of @xmath257 , one of the possible reasons for which will be explained later . in fig . [ fig : pure - coh - spectrum ] , the exact resummed spectrum shape for radiation intensity parameter @xmath202 ( blue solid curve ) is compared with the behavior of gaussian approximation ( [ gauss ] ) , ( [ mean - n - at - gauss ] ) ( blue dashed curve ) . at the chosen value of @xmath206 , a perfect convergence for the resummed spectrum ( fig . [ fig : pure - coh - spectrum]a ) is not reached yet . at the same time , for @xmath257 ( fig . [ fig : pure - coh - spectrum]c ) , the agreement is already excellent . remarkably , the linearity of the latter function holds well even far away from the central region , where the spectrum is no longer gaussian [ @xmath422 dependence gets highly nonlinear , and variation of the preexponent @xmath399 becomes significant , as well ] . to improve the accuracy of the diffusive approximation , one may be prompted to derive a correction to eq . ( [ gauss ] ) . to this end , note that eq . ( [ gauss ] ) corresponds to expanding the integral in the exponent in eq . ( [ cont - int - for - wc ] ) to second order in @xmath53 . a natural improvement thus results if in the expansion of the exponentiated integral a cubic term in @xmath53 is retained : @xmath447 the latter contour integral is reducible to an airy function , but exponentiation of a small correction may be an excess of precision . if we simply linearize the exponential , @xmath448 and evaluate the contour integral , it leads to an additive correction dating back to chebyshev ( see , e.g. , @xcite ) : @xmath449,\ ] ] with @xmath450 defined by eq . ( [ skewness ] ) , and @xmath451 the hermite polynomial of order 3 @xcite . the behavior of approximation ( [ chebyshev ] ) is illustrated in fig . [ fig : pure - coh - spectrum ] and fig . [ fig : logplot ] ( for a greater @xmath206 ) by dot - dashed curves , and is appreciably closer to the exact distribution than a simple gaussian ( dashed curves ) . the evaluated skewness correction mildly breaks the scaling , and serves to describe the residual asymmetry of the spectrum . one of its consequences is the redshift of the spectral maximum location to the value @xmath452 where the correction term is independent of the radiation intensity [ for the spectral shape ( [ f - def ] ) , @xmath453 . however , the correction to the height of the maximum in this approximation is zero : @xmath454.\ ] ] fig . [ fig : logplot]a also indicates that even though the gaussian approximation , especially with skewness correction ( [ chebyshev ] ) , works fairly well , away from the central region the falloff law differs from the gaussian , and the scaling property gets strongly violated . in principle , this deviation can still be calculated based on the steepest descent method of sec . [ subsubsec : steepest - descent ] , provided one copes with the saddle point equation at @xmath455 . approximations derived along these lines for sub- and trans - central regions are described in appendix [ app : periph ] . let us finally inspect how does the multiphoton spectrum shape modify upon incorporation of an incoherent component . as in sec . [ subsec : coh+incoh ] , this is accomplished by convolving the evaluated pure coherent radiation contribution with pure incoherent component ( [ 13 ] ) via eq . ( [ wcoh - times - winc ] ) . given the concentration of the coherent radiation probability in the central region , it may be reasonable to employ in ( [ wcoh - times - winc ] ) the simplest gaussian form ( [ gauss ] ) . that leaves us with the expression @xmath456 ( term @xmath457 has been neglected as exponentially small at large @xmath208 . ) ( a ) log plot of the resummed pure coherent radiation spectrum , at intensity parameter @xmath458 . solid purple curve , exact distribution evaluated by eqs . ( [ cont - int - for - wc ] ) , ( [ dw1coh - real][f - def ] ) . dashed black curve , gaussian approximation ( [ gauss ] ) . dot - dashed black curve , gaussian approximation with chebyshev correction ( [ chebyshev ] ) . ( b ) photon multiplicity spectrum under same conditions . dashed black line , approximation ( [ mean - n - at - gauss ] ) . ] the encountered integral still depends on rather many parameters , but under the high - intensity condition @xmath459 , their number can be effectively reduced . the upper limit of the integral is always on the right of the maximum of the gaussian entering the integrand , hence at the upper integration limit , the gaussian is decreasing . granted that this decrease is rapid , one can replace this integration limit by infinity , unless the integration interval is too short . more precisely , the gaussian varies across the integration interval significantly if @xmath460 i.e. , @xmath461 condition ( [ small - omega - range ] ) is not very restrictive as long as its r.h.s . is independent of the radiation intensity , while its l.h.s . is typically @xmath462 . thus , in the high photon multiplicity limit , it appears legitimate virtually at all @xmath20 to replace the upper integration limit in eq . ( [ 1st - conv ] ) by infinity . therewith , utilizing the value of the integral @xmath463 defining parabolic cylinder function @xmath464 @xcite , we arrive at the closed - form representation @xmath465 since function ( [ parab - cyl ] ) , similarly to ( [ gauss ] ) , depends on @xmath20 only through a linearly related variable @xmath466 , this may be regarded as extension of the scaling property beyond the vicinity of the maximum . but it should be remembered that the spectrum shape also depends on parameter @xmath135 , so here we have a 2-parameter scaling . the corresponding functions , though , differ from lvy distributions , rather being intermediate between lvy distributions and gaussian distributions . the accuracy of approximation ( [ parab - cyl ] ) may be assessed from fig . [ fig : loglogplot - a]a . there one observes that the parabolic cylinder approximation ( dashed black curve ) holds fairly well almost everywhere , except the small-@xmath20 limit . based on the obtained compact formula ( [ parab - cyl ] ) , we may now assess the deviation from the gaussian behavior in the central region . in our equations , parameter @xmath135 must always remain small , and it is straightforward to check from definition ( [ d - def ] ) that @xmath467 , wherefore , at @xmath280 distribution ( [ parab - cyl ] ) returns to gaussian form ( [ gauss ] ) . in the next approximation , the maximum of function ( [ d - def ] ) satisfies the equation @xmath468 at small @xmath135 , the solution of eq . ( [ eq - for - zmax ] ) is small , too , and in the linear approximation it equals @xmath469 in terms of @xmath20 , it casts as @xmath470 one may as well add here the skewness correction ( [ omega - max - skew ] ) ( provided both corrections are small ) , getting : @xmath471 note that those corrections have opposite signs , i.e. , the incoherent radiation component gives a blueshift , while skewness a redshift , and they partially compensate each other . evaluating the second derivative in the maximum , one can also show that the effect of @xmath135 is to broaden the spectrum . ( a ) resummed radiation spectrum for coherent component described by eqs . ( [ dw1coh - real][f - def ] ) with @xmath458 , and incoherent radiation component with @xmath472 , @xmath473 . solid purple curve , exact distribution ; dashed black curve , parabolic cylinder approximation ( [ parab - cyl ] ) . ( b ) the photon multiplicity spectrum for same conditions as ( a ) , and @xmath474 . dashed black curve , parabolic cylinder approximation ( [ convol - for - n - omega2 ] ) . dot - dashed black curve , high-@xmath20 asymptotics ( [ n - omega - asympt - large ] ) . ] at large @xmath20 , the spectrum deviates from a gaussian strongly , developing a power - law ` tail ' . the parabolic cylinder function at large values of its argument has asymptotics @xmath475+\mathcal{o}\!\left(\!e^{-\rho^2\!/2}\right)\!,\ ] ] entailing for the spectrum @xmath476 that agrees with eq . ( [ large - omega - corr ] ) modulo exponentially small terms . the calculation of @xmath128 proceeds along the same lines when based on convolution relation ( [ convol - for - n - omega ] ) . employing there gaussian approximations ( [ gauss ] ) , ( [ mean - n - at - gauss ] ) along with power - law and logarithmic expressions ( [ 13 ] ) , ( [ mean - n - omega - incoh ] ) for the pure incoherent component , we are led to the integral representation @xmath477.\quad\end{aligned}\ ] ] by the same reasoning as for eq . ( [ 1st - conv ] ) , the upper integration limit here may be extended to infinity . evaluation of the integral gives the result @xmath478 where @xmath479 and @xmath480 ( a ) plots of function @xmath481 , for parameter @xmath482 , 0.03 , 0.1 , 0.3 ( top to bottom ) . ( b ) plots of @xmath483 for same values of @xmath135 . ] shapes of functions ( [ r - def ] ) , ( [ s - def ] ) for several values of parameter @xmath135 are illustrated in figs . [ fig : ra , sa]a , b . function @xmath484 appears to be monotonous , though step - like , whereas @xmath485 features a hump . the maximum of @xmath485 rises with the decrease of parameter @xmath135 , though only logarithmically . note that function @xmath486 enters eq . ( [ convol - for - n - omega2 ] ) with a large coefficient , thence the shape of the maximum of @xmath128 is primarily described by @xmath485 . an additive correction comes from @xmath484 , whose unit jump in the turnover region of @xmath485 partially compensates the drop of the photon multiplicity spectrum beyond its maximum . the analysis of behavior of functions ( [ r - def ] ) , ( [ s - def ] ) is alleviated by the knowledge of their values in the origin , @xmath487 @xmath488,\ ] ] and asymptotics at large @xmath489 : @xmath490 @xmath491 @xmath492 @xmath493 with @xmath174 defined by eq . ( [ nu - def ] ) . at @xmath494 , i.e. asymptotic trans - central @xmath20 , function @xmath484 dominates , and along with terms @xmath495 reproduces result ( [ n - omega - asympt - large ] ) . finally , we recall that the humpy structure of @xmath128 manifested itself already at moderate radiation intensity ( cf . fig . [ fig : high - intensity]c ) . that may be non - accidental , since according to ineq . ( [ saddle - point - cond ] ) , the accuracy of the saddle - point approximation improves at large @xmath20 , where the maximum of @xmath128 is achieved . although measurements of radiation spectra by the calorimetric method are ubiquitous , for further applications of the extracted photon beam it is often preferable to describe it in terms of photon number spectrum ( [ class ] ) , anyway . needless to say , the rate of any reaction ( such as @xmath0 creation or nuclear reactions ) caused by a photon beam in a target is proportional to the photon number per energy interval , irrespective of temporal correlations between photons . besides that , if the radiation spectrum is used for diagnostics of the radiator , the single - photon spectrum is more directly related to the classical radiation spectrum and the charged particle trajectory , and hence to the radiator parameters , than the multiphoton one . thus , there arises a need for reconstructing single - photon spectra from calorimetrically measured spectra . such a problem had been addressed some time ago in @xcite , where a procedure for reconstructing first moments ( the mean and the variance ) of the single - photon spectra from those of a multiphoton spectra was proposed . here we will develop a reconstruction procedure for the complete spectrum . the prerequisite , however , is fair independence of the spectrum on initial state parameters to be averaged over , wherewith the outcoming multiphoton spectrum is representable purely by contour integral ( [ generic - eq ] ) . as was argued in sec . [ subsec : pure - coh ] , such conditions can be met for coherent bremsstrahlung or undulator radiation , but are unlikely for channeling radiation . let us begin with the observation that integral relation ( [ generic - exponentiation ] ) , with the aid of eq . ( [ total - probab-1 ] ) can be cast in form where the l.h.s . depends only on @xmath31 , while the r.h.s . only on @xmath32 : @xmath496 ( as long as @xmath32 is supposed to be inferred from experimental data , the upper limit of integration over @xmath20 is set to @xmath4 , which is the physical end of the spectrum . ) to further obtain a linear integral equation for the single - photon spectrum , it suffices to take the logarithm of both sides of eq . ( [ 90 ] ) : @xmath497.\end{aligned}\ ] ] after integration in the l.h.s . by parts , this can also be written @xmath498.\,\,\end{aligned}\ ] ] if @xmath11 is ir - finite , one can deal with eq . ( [ simpler ] ) arranged with the aid of relation ( [ total - probab-1 ] ) in a simple form @xmath499 since here the l.h.s . is merely a laplace transform of @xmath31 , it inverts in the standard way : [ 110 ] @xmath500 @xmath501 now , for @xmath502 ( when the @xmath21-term does not contribute ) , ir divergences do not appear in eq . ( [ reconstr ] ) , so the initial assumption of ir finiteness of @xmath11 may be relaxed . ( [ 110 ] ) might as well be derived from eq . ( [ finite ] ) , which is a bit more lengthy ( solving a 1st - order differential equation ) , but does not need presuming ir finiteness of @xmath11 . if @xmath13 is finite , integral ( [ log(1 + ... ) ] ) converges better , since the logarithm in the integrand vanishes at infinite integration limits . on the contrary , if @xmath503 , representation ( [ log(1 + ... ) ] ) is untenable , and ( [ reconstr ] ) ( without the @xmath21-term ) is the representation of choice . from our experience with the resummation procedure in sec . [ sec:2 ] , one may expect ambiguity in the contour integral representation for the reconstructed spectrum , as well . indeed , applying lemma [ lemma1 ] with @xmath504 , or @xmath505 , , the argument of the logarithm , @xmath506 , tends to zero as @xmath507 . thereat , @xmath91\|=\frac1{\|z(s)\|}$ ] increases , but only by a power law , @xmath508 , still satisfying the conditions of lemma [ lemma1 ] . ] one can see that the upper integration limit for the laplace transform of @xmath32 in ( [ 110 ] ) may be an arbitrary number greater than @xmath35 , e.g. equal @xmath509 : @xmath510 relation @xmath97 in integral ( [ reconstr - caus ] ) may seem to be physically counterintuitive , because it is energetically _ anti_-ordered . but as we emphasized in sec . [ subsec : cont - int - repr ] , the energetic ordering property does not hold exactly for our distributions , anyway . another way to arrive at eq . ( [ reconstr - caus ] ) is to expand the logarithm in eq . ( [ log(1 + ... ) ] ) to power series , and evaluate the contour integral termwise : [ reconstr - power - series - together ] @xmath511 here , as in passing from eq . ( [ generic - sum - int ] ) to ( [ generic - sum - int - upper - omega ] ) , one may replace upper limits @xmath4 by @xmath512 : @xmath513 in contrast to eq . ( [ generic - sum - int ] ) , however , series ( [ reconstr - power - series - together ] ) involves no factorial in the denominator , whence it must have a finite convergence radius , such that @xmath514 . the latter property is natural , as long as resummed spectra are always bounded from above ( see sec . [ par : fund - max ] , [ subsubsec : comb - in - fund - int ] ) . series ( [ reconstr - power - series - together ] ) at sufficiently small intensities of @xmath11 may prove to be actually even more convenient than contour integral representations ( [ 110 ] ) . in particular , in the limit @xmath47 , only the first term of ( [ reconstr - power - series - caus ] ) survives , giving @xmath515 which is equivalent to eq . ( [ coh - omegato0 ] ) . keeping @xmath516 terms , one can derive an analog of eq . ( [ coh - omegato0-corr ] ) , etc . for application of contour integral representations ( [ 110 ] ) , it is also highly desirable to map singularities . ] , as well as sectors of growth and decrease of the integrand , to which the integration contour can be adjusted . as long as the radiation spectrum obeys eq . ( [ 108 ] ) , the integrand singularities in eqs . ( [ 110 ] ) must coincide with those of laplace transform of the single - photon spectrum . but the latter transform may not have singularities in the right half - plane , since @xmath517 converges at any positive @xmath86 , so it is an analytic function of @xmath53 at @xmath518 . in case if with the increase of @xmath35 , @xmath31 decreases faster than exponentially [ e.g. , strictly vanishes beyond a coherent emission edge @xmath3 , like in eq . ( [ dw1coh - real ] ) ] , the l.h.s . of ( [ 108 ] ) exists at any @xmath53 , so no singularities in the @xmath53 plane can emerge at all . in case if @xmath31 decreases with @xmath35 exponentially , its laplace transform possesses singularities at finite negative @xmath86 . finally , if @xmath31 decreases by a power law , the l.h.s . of ( [ 108 ] ) diverges at @xmath519 , so a singularity is situated at point @xmath520 . as concerns the convergence sectors , the integrand of ( [ log(1 + ... ) ] ) obviously diverges at @xmath521 because of factor @xmath522 . at @xmath523 , the integrand may decrease only provided the argument of the logarithm , after analytic continuation to the negative @xmath86 half - plane , does not increase faster than exponentially . in the latter case , the integration contour may be deformed to shape @xmath147 ( see fig . [ fig : contours ] ) . but that appears to be impossible in the important case when @xmath31 is anticipated to decrease at @xmath524 faster than exponentially ( e.g. , pure single - harmonic coherent radiation ) . then , as we noted above , the integrand of ( [ 110 ] ) may not have singularities in the whole complex plane , so if the integration contour might be deformed to @xmath147 , the result of the integration would identically equal zero , which is a clear contradiction . in this case , the only option for the integration contour in eqs . ( [ 110 ] ) is a line parallel to the imaginary axis ( @xmath146 ) . at practice , when @xmath32 is measured as a number of events per bin of deposited energy , in eq . ( [ reconstr ] ) the integral involving @xmath32 must be replaced by the corresponding finite sum . that is harmless for evaluation of the contour integrals , since the resulting function of @xmath53 remains analytic in the @xmath53 plane . but it is mandatory that the spectrum measurements are absolute , providing correct normalization of the spectrum . at high intensity , when any resummed coherent spectrum tends to gaussian form , our reconstruction method will ultimately become unsustainable . to appreciate the encountered difficulties , consider a toy model for the single - photon spectrum : @xmath525 ( with @xmath526 ) . the corresponding resummed spectrum evaluates in terms of a modified bessel function : @xmath527 laplace transform thereof equals @xmath528 now suppose that a calorimetric spectrum measurement gave instead of @xmath32 some slightly deviating distribution @xmath529 . inserting this to eq . ( [ log(1 + ... ) ] ) leads to the reconstructed underlying spectrum @xmath530 clearly , as @xmath192 increases , signal term @xmath531 becomes exponentially small , and ultimately inferior to the error term ; hence , for sufficiently large @xmath532 the method will loose efficiency . the best one can do in this situation , probably , is to extract first spectral moments as in ref . @xcite , which may appear insufficient for complete characterization of a highly structured single - photon spectrum . in the latter straitened circumstances , it may also be worth engaging information contained in the photon multiplicity spectrum @xmath128 . examination of structure of integral ( [ mean - n - omega - contour - non - casual ] ) shows that it is more complicated than ( [ generic - eq ] ) , and does not allow expressing @xmath94 explicitly . but in the high - multiplicity limit , when the reconstruction method based on @xmath258 is obstructed , one can at least extract from @xmath128 an additional parameter @xmath208 , say , through eq . ( [ n - to - wi1c ] ) . its knowledge , along with the knowledge of moments @xmath533 and @xmath534 ( and possibly @xmath535 ) extracted from @xmath32 , may constrain the shape of @xmath94 more tightly . the techniques developed in the present paper can be further employed to calculate spectra of specific coherent radiation sources , including channeling radiation . for undulator radiation , and partially for coherent bremsstrahlung , the results presented in this paper may already be applicable directly . let us summarize the main physical lessons learned at this stage : * the fundamental maximum of the multiphoton spectrum does not rise indefinitely with the increase of the radiator length @xmath100 , but eventually saturates , and subsequently decreases . instead , there is an elevation of the spectrum beyond the fundamental maximum , and its general spread ( sec . [ sec : coh+incoh ] ) . the high-@xmath20 spectral tail due to incoherent bremsstrahlung is the least affected by multiphoton effects , and its ratio w.r.t . the fundamental maximum thereby increases . * multiphoton effects alone can give rise to a second maximum in the coherent radiation spectrum , at @xmath536 , even when secondary harmonics in the single - photon spectrum are absent . * the low-@xmath20 part of the multiphoton spectrum , as well as discontinuities of the spectrum for coherent radiation , are suppressed by the factor of photon non - emission probability ( sec . [ sec : coh+incoh ] ) . therein , manifestations of the incoherent radiation component in the spectrum are particularly pronounced . * at high radiation intensity ( in a long radiator ) , the resummed spectrum of pure coheernt radiation tends to a gaussian form . but under presence of an incoherent component , it features a power - law tail towards large @xmath20 ( sec . [ sec : high - intensity ] ) . such a diffusion regime may be regarded as _ weakly _ anomalous ( characterized by @xmath537 ) . the corresponding limiting multiphoton spectrum is described by a parabolic cylinder function , being an intermediate asymptotic case between gaussian and lvy distributions . * the photon multiplicity spectrum @xmath128 is an informative observable complementary to @xmath32 , and is measurable with the modern state of the technology . the growth of @xmath128 with @xmath20 at low radiation intensity is stepwise , whereas at high intensity it smoothens out and tends to a linear law . in the region where the incoherent contribution overtakes the coherent one , @xmath128 saturates , and slightly subsides , whereupon continues rising , but only logarithmically ( secs . [ sec : coh+incoh ] , [ sec : high - intensity ] ) . promising are also the formulas for reconstruction of the single - photon spectrum from the calorimetrically measured one , derived in sec . [ sec : reconstr ] . the applications thereof will be considered elsewhere . it should also be remembered that although the present framework is well suited for fast evaluation of multiphoton emission effects , it is limited to conditions of soft radiation , which do not yet cover all the variety of interesting physical problems ( e.g. , hard coherent electromagnetic cascades , with deposited energies close to @xmath4 , investigated in experiments @xcite ) . nonetheless , for practical coherent radiation sources , as discussed in appendix [ app : coh - rad - sources ] and the beginning of sec . [ sec : high - intensity ] , conditions of soft radiation are quite typical . at the same time , albeit many experiments in the soft radiation domain found multiphoton effects to be appreciable , none of those experiments addressed multiphoton effects systematically . the need for such dedicated experimental studies thus remains acute . in sec . [ subsec : incoh - bremsstr ] , we introduced parameter @xmath142 as an effective uv cutoff for the single - photon scale - invariant incoherent bremsstrahlung spectrum . the value of this parameter can be inferred from a theory accurately describing acts of large radiative energy loss . for our purposes , we need only the nll soft asymptotics of this theory . as long as we deal solely with spectrum @xmath32 ( and are not concerned with the photon multiplicity spectrum ) , the coherent radiation component at large @xmath20 can be entirely neglected . thereby , we return to the well - known electromagnetic cascade theory in an amorphous medium , but with the due account of the electron spin . for the single - photon bremsstrahlung spectrum , instead of ( [ dw1incoh - cutoff ] ) we take the ultra - relativistic and completely screened approximation of the bethe - heitler formula : @xmath538 where @xcite @xmath539 for simplicity , the pair production process will be neglected , as before . then , the kinetic equation for electron energy distribution @xmath540 remains one - component : @xmath541 where dimensionless parameter @xmath135 is proportional to the target thickness [ see eq . ( [ a - amorph ] ) ] . integro - differential equation ( [ shower - eq ] ) is endowed with the initial condition ( [ kinetic - eq - init - cond - for - pie ] ) : @xmath542 once the solution of eq . ( [ shower - eq][shower - init - cond ] ) is attained , the multiphoton radiation spectrum , equal to the radiative energy loss spectrum , is determined by the correspondence rule ( [ def - pie ] ) : @xmath543,\ ] ] and for this function we are interested in the limit @xmath544 . note that by a change of integration variables , eq . ( [ shower - eq ] ) may can be recast @xmath545 which turns to recoilless kinetic equation ( [ kinetic - eq2 ] ) in the limit @xmath546 , when @xmath547 . however , form ( [ shower - eq ] ) is more convenient for solution . solution of eq . ( [ shower - eq ] ) is obtained by applying mellin transform @xmath548 in terms of which the integro - differential equation becomes ordinary differential : @xmath549 with @xmath135-independent coefficient @xmath550 and initial condition @xmath551 integrating eq . ( [ shower - eq - mell ] ) , and applying the inverse mellin transform @xmath552 , leads to the solution @xmath553 by correspondence rule ( [ rel - with - multiphot - spectr ] ) , the multiphoton bremsstrahlung spectrum ensues @xmath554 in the last line , we linearized @xmath555 for small @xmath556 we are concerned with , and changed the integration variable to @xmath557 . function @xmath558 at large arguments needed for us increases logarithmically : @xmath559\qquad\qquad\nonumber\\ = \ln s+\gamma_{\text{e}}+\!\int_0 ^ 1\!\frac{dz'}{z'}[w_{\text{bh}}(z')\!-\!w_{\text{bh}}(0)]+\mathcal{o}\!\left(s^{-1}\right)\!.\nonumber\\\end{aligned}\ ] ] inserting this approximation to eq . ( [ pi - solution ] ) , and doing the elementary contour integral , from comparison with eq . ( [ 13 ] , [ fa ] ) we ultimately infer the value for @xmath44 : @xmath560}=e^{-5/8}\approx0.5.\ ] ] according to eq . ( [ kappa58 ] ) , the effective uv cutoff must be set at about half the initial electron energy . with this value for @xmath44 , our eqs . ( [ 13 ] , [ fa ] ) agree with eqs . ( 2.11 ) , ( 2.19 ) of @xcite . in this appendix , we will collect equations for spectral parameters of main coherent radiation sources , needed for numerical estimates of significance of multiphoton effects in sec . [ sec : high - intensity ] . any coherent radiation source we consider may be viewed as an ultrarelativistic electron ( or positron ) radiating under the influence of periodic force @xmath561 ( having period @xmath562 ) , directed predominantly transverse to the electron s large mean momentum ( which is taken parallel to @xmath563 ) . if the radiation is of dipole type , i.e. produced by a weakly accelerated electron remaining non - relativistic in its average rest frame ( e.r.f . ) , the emitted electromagnetic radiation in this frame will be monochromatic . in e.r.f . , period @xmath562 is lorentz - contracted to @xmath564 ( @xmath1 stands for the electron lorentz factor ) , corresponding to frequency , where frequency and mass have dimension of energy , while the electron charge is dimensionless . ] the radiation in the e.r.f . is not strictly isotropic , as long as the driving force is known to be transverse w.r.t . @xmath563 . for our purposes , we need only the angular distribution averaged over the azimuthal angles , which will thus obey a @xmath566 law , where @xmath567 is the angle between @xmath563 and the photon emission direction in e.r.f . this information suffices to write the angle - differential radiation intensity : @xmath568 where @xmath569 and @xmath570 are the electron mass and charge . [ to check the numerical coefficient , note that integration of ( [ i - dipole ] ) over photon emission angles with @xmath571 , and identification of @xmath572 with the electron acceleration , leads to the famous larmor formula . ] ( [ i - dipole ] ) gives an average ( over the oscillation period ) energy emitted by the electron in its rest frame per unit time . dividing this by the photon energy @xmath573 gives the probability density , and the total probability is obtained by multiplying by the period value and the number of periods @xmath574 : @xmath575 the latter quantity must be lorentz invariant , which is opportune for expressing the desired radiation spectrum in the lab . the advantage of referring to e.r.f . variables is that @xmath576 is related to the photon momentum projection on the light front in @xmath76-direction , and therethrough to the photon energy in the lab frame : @xmath577 where @xmath3 is the maximal photon energy in the lab , corresponding to @xmath578 . expressing @xmath576 from ( [ omega - theta - corresp ] ) , inserting it to ( [ i - dipole ] ) , and differentiating , we convert the angular distribution in e.r.f . to the spectral distribution in the lab : @xmath579 with @xmath198 being function ( [ f - def ] ) , thus explaining its origin . product @xmath580 entering the time integral is lorentz invariant , too , so it can be written as @xmath581 , where @xmath582 is the force , and @xmath583 the time in the lab . comparing ( [ 152 ] ) with eq . ( [ dw1coh - real ] ) , we infer @xmath584 where @xmath562 is the motion period in the lab , and obviously , @xmath585 . although product @xmath206 is all we need for estimates of the photon multiplicity , it is also expedient to evaluate the photon energy scale @xmath3 independently . the latter is obtained from e.r.f . value @xmath586 by boosting along @xmath563 with the lorentz factor @xmath1 . that results in multiplication by @xmath587 , so the maximal photon energy in the lab equals @xmath588 exceeding by factor of @xmath589 the driving force frequency in this frame . the analysis of photon multiplicity for specific radiation sources further reduces to estimating in eq . ( [ dw - lab ] ) the driving force magnitude and period . note that generally , the square of the integral in ( [ dw - lab ] ) is @xmath590 , while @xmath591 . thus , in total , the radiation source brightness is proportional to its period . on the other hand , according to eq . ( [ 2gamma2 ] ) , the photon energy scales as @xmath592 , though that is generally compensable by an appropriate increase in @xmath1 . for an electron crossing a family of atomic planes by a near - straight trajectory at a small misalignment angle @xmath266 , the time in eq . ( [ dw - lab ] ) expresses as @xmath593 , where @xmath594 is the coordinate transverse to the planes , and the oscillation period in the lab equals @xmath595 @xmath596 being the plane spacing . therewith , eq . ( [ dw - lab ] ) becomes @xmath597 where @xmath582 as a function of time was replaced by the corresponding function of @xmath594 . if @xmath598 within a period describes roughly as @xmath599 evaluation of the integral in eq . ( [ dw - cb ] ) gives @xmath600}{\chi},\ ] ] where we used numerical estimates @xmath601 , and @xmath602 . parameters @xmath100 and @xmath266 are not completely independent : they must obey inequality ( cf . @xcite ) @xmath603 in order for the multiple scattering spread @xmath604 not to blur the coherent peak . at the same time , @xmath266 correlates with @xmath4 , since under assumption @xmath605 , eqs . ( [ 2gamma2 ] , [ t - cb ] ) yield @xmath606 for practical misalignment angles @xmath607 , this constrains the electron energy to @xmath608 gev . ( [ cb - ineq ] ) and ( [ echi - ll20 ] ) are only compatible provided @xmath609 , which is essentially the condition ( [ all1 ] ) adopted in our paper . for the case of channeling , the oscillation period depends on the particle energy : @xmath610 for a harmonic potential [ corresponding to force ( [ fx ] ) ] , the period will be constant , but with the account of anharmonicity , it may acquire also some dependence on the particle transverse energy in the channel . for our estimates , it will suffice to adopt the harmonic approximation , letting @xmath611 where @xmath612 depends on the particle impact parameters , in the spirit of eq . ( [ fx ] ) . as an order of magnitude estimate , we may use @xmath613 . substituting ( [ f - chann ] ) to eq . ( [ dw - lab ] ) , we get @xmath614 at practice , the length of a crystal for channeling is limited by dechanneling effects . for positrons , the dechanneling length may be estimated crudely as @xcite @xmath615 in terms of which eq . ( [ dw - chann ] ) assumes the form @xmath616\right)^{3/2}\frac{l}{l_{\text{d}}}.\ ] ] in an undulator , the particle is moving in a fixed channel , similarly to channeling , but the motion period , determined by the magnet spacing , is energy - independent , like in coherent bremsstrahlung . since @xmath562 is macroscopically large , according to eqs . ( [ dw - lab][2gamma2 ] ) , the source must be bright , but the photon energy be relatively low . nonetheless , gamma - range undulators were created recently , with periods @xmath617 cm , and electron energies @xmath618 gev @xcite . for description of the undulator field strength , it is customary to introduce parameter @xmath619t[\text{cm}],\ ] ] where @xmath620 is the field amplitude , @xmath621 . in terms of @xmath622 , eq . ( [ dw - lab ] ) reads @xmath623 at practice , parameter @xmath622 can be @xmath624 ( somewhat violating the dipole radiation condition @xmath625 ) , whilst the number of periods can amount several hundreds or thousands . the latter number is particularly large for undulator - based positron sources , where @xmath626 @xcite . as had been pointed out in sec . [ subsubsec : centr ] , beyond the central region specified by condition ( [ central - region ] ) , the gaussian approximation breaks down , and the multiphoton spectrum becomes dependent on all the detail of the single - photon spectrum . even though in those regions the radiation spectrum is relatively faint ( and at high @xmath20 it may be overwhelmed by the incoherent bremsstrahlung ` tail ' ) , it may still prove measurable under high absolute radiation intensity . therefore , it would be useful to find some approximations for peripheral regions , as well . suitable approximations for coherent radiation in peripheral regions can be built based on the same saddle - point approximation described in sec . [ subsubsec : steepest - descent ] , provided transcendental saddle - point equation ( [ saddle - point - eq ] ) is solved at least approximately . that requires development of a special approach on each side from @xmath533 . let us consider the region @xmath627 first . from the viewpoint of eq . ( [ saddle - point - eq ] ) , this corresponds to large and negative @xmath394 ( see fig . [ fig : s0c - vs - omega ] ) . it is therefore beneficial to rewrite eq . ( [ saddle - point - eq ] ) as @xmath628,\ ] ] where the r.h.s . is smaller than any of the terms on the l.h.s . ( granted that the exponential within the integration domain is @xmath629 , achieving unity only at the endpoint @xmath630 ) . thence , the r.h.s . can be treated as a perturbation . graphical real solution of saddle - point equation ( [ saddle - point - eq ] ) , for @xmath264 given by eqs . ( [ dw1coh - real][f - def ] ) . solid purple curve , exact solution . dotted black line , central region approximation ( [ saddle - point - eq - lin - soln ] ) . purple point , the central solution . long - dashed black curve , large-@xmath20 approximation ( [ s0c - log-1 ] ) . short - dashed black curve , improved large-@xmath20 approximation ( [ s0c - log-2 ] ) . dot - dashed black curve , low-@xmath20 approximation ( [ quadr - eq ] ) . ] to solve eq . ( [ saddle - point - eq - large - omega ] ) iteratively , first , we express @xmath631 from equating the l.h.s . to zero ( leading log approximation ) : @xmath632 substituting this to the r.h.s . of ( [ saddle - point - eq - large - omega ] ) gives a nll approximation for @xmath631 : @xmath633+\mathcal{o}\left(\ln\ln\ln\frac{\omega}{b\omega_0 ^ 2}\right).\ ] ] its behavior is illustrated in fig . [ fig : s0c - vs - omega ] by the long - dashed black curve . it does not seem to be accurate enough , which compels us to proceed to nnll : substitute @xmath631 in the r.h.s . of eq . ( [ saddle - point - eq - large - omega ] ) by that expressed from the l.h.s . , whereupon replace the remaining @xmath631 in the r.h.s . by ll expression ( [ s0c - log ] ) : @xmath634^{1-z}\bigg\}.\end{aligned}\ ] ] the latter solution is to be inserted to eqs . ( [ h][d ] ) , which are then inserted to ( [ dwdomega - saddle ] ) . the resulting expression is rather bulky , and will be omitted . its accuracy can be visualized from fig . [ fig : s0c - vs - omega ] , and may be regarded as satisfactory . the qualitative corollary from eq . ( [ s0c - log-2 ] ) is that the multiphoton pure coherent spectrum decreases by the law close to linear exponential [ since in ( [ dwdomega - saddle ] ) @xmath635 . the opposite peripheral region @xmath636 corresponds to positive @xmath394 ( see fig . [ fig : s0c - vs - omega ] ) . there , the main contribution to the integral entering eq . ( [ saddle - point - eq ] ) stems from the lower integration limit . asymptotically , at large @xmath394 , @xmath637 but the latter asymptotic solution becomes accurate at @xmath638 only , whereas we need to cover the whole region @xmath639 . therefore , eq . ( [ saddle - point - eq ] ) needs to be solved accurately on the whole interval @xmath640 , which is challenging insofar as this equation is transcendental . in the present situation , one of the simplest approaches may be to interpolate the solution between asymptotes ( [ 1/s2 ] ) and ( [ saddle - point - eq - lin ] ) , e.g. , writing it in form @xmath641 the behavior of the latter interpolation is illustrated in fig . [ fig : s0c - vs - omega ] by the dot - dashed black curve , which is rather close to the exact solution on the interval of interest . expressing @xmath394 ( or @xmath642 ) from quadratic equation ( [ quadr - eq ] ) and inserting to eq . ( [ h][d ] ) , we get an approximation holding up to the central region ) . qualitatively , at large @xmath394 , @xmath643 , which may be interpreted in the sense that the function rises faster than a gaussian . h. motz , p. luchini . _ undulators and free - electron lasers_. oxford : university press , 1990 ; p. rullhusen , x. artru , p. dhez . _ novel radiation sources using relativistic electrons_. world scientific , 1998 ;
the electrons that carry charge current in electronic circuits also have spins . in non - ferromagnetic samples , the spins are usually randomly oriented and do not play a role in the behavior of the device . however , when ferromagnetic components are incorporated into a device , the flowing electrons can become partially spin polarized and these spins can play an important role in device function . due to spin - based interactions between the ferromagnets and electrons , the orientations of the magnetization for ferromagnetic elements can determine the amount of current flow . by means of these same interactions , the electron spins can also influence the orientations of the magnetizations . this last effect , the so - called spin transfer torque , is the topic of the following series of articles . the goal of these articles is to provide an introduction to the topical scientific issues concerning the theories , experiments , and commercial applications related to spin transfer torques . these articles are written by some of the leaders in the study of spin transfer torques and we hope that they will be useful to beginning students starting their study as well as of interest to experts in the field . the authors of the succeeding articles were asked to consider the field from their particular point of view , and to provide a `` preview '' including discussion of interesting unsolved problems rather than merely a review of completed work . hopefully , when taken together these articles provide a broad and representative view of where the field is and where it may be going . the goal of the present article is to provide basic background and context for the succeeding articles . we start with a very brief history of the field and provide references to background material . then , in section [ sec : ferro ] , we introduce aspects of ferromagnetism that are important to the discussion , particularly for transition - metal ferromagnets , and discuss how spin polarized currents arise . section [ sec : stt ] describes how spin transfer torques can be understood as resulting from changes in spin currents . section [ sec : multi ] explains some of the issues associated with how spin transfer torques affect magnetic - multilayer devices and tunnel junctions . section [ sec : wires ] describes how they affect domain walls in magnetic nanowires . finally , section [ sec : outlook ] introduces the succeeding articles in the context provided by the rest of this article . the first work to consider the existence of spin transfer torques occurred in the late 1970 s and 1980 s , with berger s prediction that spin transfer torques should be able to move magnetic domain walls @xcite , followed by his group s experimental observations of domain - wall motion in thin ferromagnetic films under the influence of large current pulses @xcite . the phenomenon did not attract a great deal of attention at the time , largely because very large currents ( up to 45 a ! ) were required , given that the samples were quite wide on the scale of mm . however , with advances in nanofabrication techniques , magnetic wires with 100 nm widths can now be made readily , and these exhibit domain wall motion at currents of a few ma and below . research on spin - transfer - induced domain wall motion has been pursued vigorously by many groups since the early days of the 21st century @xcite . the present state of studies of spin - torque - driven domain wall motion is described in the articles by beach , tsoi , and erskine , by tserkovnyak , brataas , and bauer , and by ohno and dietl . widespread interest in magnetic nanostructures began in 1986 with the discovery of interlayer exchange coupling @xcite . interlayer exchange coupling @xcite is the interaction between the magnetizations of two ferromagnetic layer separated by an ultrathin , non - ferromagnetic spacer layer . of particular interest was the discovery of antiferromagnetic coupling in the fe / cr / fe system by grnberg et al . because it led shortly thereafter to the discovery of the giant magnetoresistance ( gmr ) effect by grnberg s group and fert s group @xcite . grnberg and fert shared the 2007 nobel prize in physics for this discovery . gmr is the change in resistance that occurs when the relative orientation of the magnetizations in two ferromagnetic layers changes . for example , when the magnetizations of two fe layers separated by cr are antiparallel to each other as they are when antiferromagnetically coupled , the sample has a relatively large resistance . the magnetizations of the fe layers can be brought into parallel alignment by an external magnetic field , and this decreases the resistance . at intermediate angles , the resistance is also intermediate between these maximum and minimum values . see fig . [ fig : gmr ] . the overall size of the change in resistance is typically up to a few 10 s of percent , which is `` giant '' compared to the @xmath0 1 % magnetoresistance changes of pure magnetic metals by themselves ( due to anisotropic magnetoresistance @xcite ) . most of the early work on gmr focused on the sample geometry in which the current flows in the plane ( cip ) of the multilayer sample . another geometry , first explored in 1991 @xcite has the current flow perpendicular to the planes ( cpp ) of the multilayer and gives larger fractional resistance changes @xcite . the cpp geometry is of particular interest in the present context because spin transfer effects are more important than they are in the cip geometry . in studying the gmr effect , et al_. discovered in 1990 that the interlayer exchange coupling oscillates as a function of the thickness of the spacer layer @xcite . the oscillations could be quite dramatic ; up to sixty changes in sign of the coupling were seen in single wedge - shaped samples allowing the simultaneous study of a range of thicknesses @xcite . comparison of the oscillations with calculations @xcite confirmed that this coupling was an exchange interaction mediated by the electrons in the spacer layer and that the oscillation periods were determined by the geometry of the spacer layer s fermi surface . in 1989 , slonczewski @xcite calculated the interlayer exchange coupling for the case in which the spacer layer is an insulating tunnel barrier . while there were no measurements of exchange coupling across insulators at the time , his article has two features of particular interest in regard to spin - torque physics . first , slonczewski calculated the exchange coupling by determining the spin current flowing through the tunnel barrier . a spin current flows even with zero applied bias across the tunnel junction whenever the magnetizations of the two electrodes are non - collinear , and the source of the exchange coupling can be understood to be the transfer of angular momentum from this spin current to each magnet . the second feature of interest is that he considered the additional coupling that results when a voltage is applied across the junction . this was the first calculation of a spin transfer torque in a multilayer geometry with current flowing perpendicular to the plane . however , there was little immediate experimental follow - up , because the technology for making magnetic tunnel junctions at that time was still rather primitive , and provided only tunnel barriers that were too thick to permit the large current densities needed to excite spin - torque - driven magnetic dynamics . the papers most influential in launching the study of spin transfer torques came in 1996 , when slonczewski @xcite and berger @xcite independently predicted that current flowing perpendicular to the plane in a metallic multilayer can generate a spin transfer torque strong enough to reorient the magnetization in one of the layers . since the metallic magnetic multilayers used for gmr studies have low resistances ( compared to tunnel barriers ) , they could easily sustain the current densities required for spin transfer torques to be important . slonczewski predicted that the spin transfer torque from a direct current could excite two qualitatively different types of magnetic behaviors depending on the device design and the magnitude of an applied magnetic field : either simple switching from one static magnetic orientation to another or a dynamical state in which the magnetization undergoes steady - state precession . his subsequent 1997 patent @xcite was remarkably far - seeing , providing detailed predictions for many of the applications that are currently being pursued . measurements of current - induced resistance changes in magnetic multilayer devices were first identified with spin - torque - driven excitations in 1998 by tsoi _ et al_. , for devices consisting of a mechanical point contact to a metallic multilayer @xcite , and in 1999 by sun in manganite devices @xcite . observation of magnetization reversal caused by spin torques in lithographically defined samples occurred shortly thereafter @xcite . [ fig : switching ] shows comparisons between spin - torque - driven magnetic switching and magnetic - field driven switching for a metallic multilayer and a magnetic tunnel junction . phase - locking between spin - torque - driven magnetic precession and an alternating magnetic field was detected in 2000 @xcite and direct measurements of steady - state high - frequency magnetic precession caused by spin torque from a direct current were made beginning in 2003 @xcite . figure [ fig : oscil ] shows an example of voltage oscillations due to spin - torque - driven magnetic precession . the article by berkov and miltat describes some of these results and shows to what extent it is currently possible to make a detailed comparison between theory and experiment . the article by silva and rippard discusses some of the open questions pertaining to spin - torque - driven precession in the point - contact sample geometry . at the same time that interest was beginning to grow regarding spin transfer torques in metallic multilayers , so was the interest in magnetic tunnel junctions , starting with the observation in 1995 of substantial tunnel magnetoresistance ( tmr , the difference in resistance between parallel and antiparallel orientation for the electrode magnetizations of a magnetic tunnel junction ) at room temperature @xcite . most of the early studies used aluminum oxide as a tunnel barrier . since aluminum oxide barriers are heavily disordered it is difficult to study them theoretically . in 2001 , et al_. and mathon and umerski calculated the tunneling properties of the fe / mgo / fe system @xcite , which can be lattice - matched and potentially well - ordered . they found that the symmetry of the system and the relevant electronic states lead to the possibility of an extremely large tmr . in 2004 , values of tmr greater than those observed for aluminum oxide barriers were published @xcite and the values of tmr demonstrated experimentally have continued to increase rapidly until this day . techniques have now been developed to make both aluminum oxide and mgo tunnel barriers sufficiently thin to support the current densities needed to produce magnetic switching with spin transfer torques @xcite . work is underway to investigate spin - torque - driven precession in tunnel barriers , as well . one reason for the interest in spin - torque effects in tunnel junctions is that tunnel junctions are better - suited than metallic magnetic multilayers for many types of applications . tunnel junctions have higher resistances that can often be better impedance - matched to silicon - based electronics , and tmr values can now be made larger than the gmr values in metallic devices . the science and technology of spin transfer torques in tunnel junctions are discussed in the articles by katine and fullerton and by sun and ralph . the possibility of commercial application has been a strong driving force in this field from the beginning . grnberg @xcite filed a german patent for applications of gmr in 1988 even before the effect was published in the scientific literature , and the phrase `` giant magnetoresistance '' now appears in over 1500 us patents . devices based on the gmr and tmr effects have already found very widespread application as the magnetic - field sensors in the read heads of magnetic hard disk drives , and a non - volatile random access memory based on magnetic tunnel junctions has recently been introduced . slonczewski s 1997 patent @xcite for devices based on spin transfer torques has been referenced by 64 subsequent patents . applications of spin transfer torques are envisioned using both types of magnetic dynamics that spin torques can excite . magnetic switching driven by the spin transfer effect can be much more efficient than switching driven by current - induced magnetic fields ( the mechanism used in the existing magnetic random access memory ) . this may enable the production of magnetic memory devices with much lower switching currents and hence greater energy efficiency and also greater device density than field - switched devices . the steady - state magnetic precession mode that can be excited by spin transfer is under investigation for a number of high - frequency applications , for example nanometer - scale frequency - tunable microwave sources , detectors , mixers , and phase shifters . one potential area of use is for short - range chip - to - chip or even within - chip communications . spin - torque - driven domain wall motion is also under investigation for memory applications . parkin has proposed a `` racetrack memory '' @xcite which envisions storing bits of information using many domains arranged sequentially in a magnetic nanowire and retrieving the information by using spin transfer torques to move the domains through a read - out sensor . the article by katine and fullerton discusses in detail the opportunities and challenges for potential applications . there are several books that can provide more background information for this set of articles on spin transfer torques . one resource is the four volume series _ ultrathin magnetic structures _ edited by heinrich and bland @xcite . these volumes contain articles on almost all aspects of magnetic thin films and devices made out of them . another useful book , written at a more pedagogical level , is _ nanomagnetism : ultrathin films , multilayers and nanostructures _ , edited by mills and bland @xcite . the three volume series _ spin dynamics in confined magnetic structures _ edited by hillebrands , ounadjela , and thiaville covers dynamical aspects of magnetic nanostructures @xcite . the five volume set , handbook of magnetism and advanced magnetic materials , edited by kronmller and parkin , covers the entire field of magnetism including the topics of interest here @xcite . _ concepts in spin electronics _ , edited by maekawa , provides another recent overview @xcite . finally , the journal of magnetism and magnetic materials published a collection of review articles including several on topics related to magnetic multilayers in volume 200 @xcite . specific chapters in these books and other review articles on spin transfer torques will be mentioned throughout this article . _ the origin of ferromagnetism . _ ferromagnetism occurs when an electron system becomes spontaneously spin polarized . in transition metals , ferromagnetism results from a balance between atomic - like exchange interactions , which tend to align spins , and inter - atomic hybridization , which tends to reduce spin polarization . an accurate accounting of both effects is quite difficult @xcite . however , a qualitative understanding is straightforward . in isolated atoms , hund s rules describe how to put electrons into nearly degenerate atomic levels to minimize the energy . hund s first rule says to maximize the spin , that is , to put in as many electrons with spins in one direction into a partially filled atomic orbital before you start adding spins in the other direction . the energy gain that motivates hund s rule is that pauli exclusion keeps electrons with the same spin further apart on average , thereby lowering the coulomb repulsion between them . this energy is called the atomic exchange energy . in accordance with hund s first rule , essentially all isolated atoms with partially filled orbital levels have non - zero spin moments . non - zero values of orbital angular momentum can also contribute to the magnetic moment of isolated atoms . in solids , on the other hand , electron states on neighboring atoms hybridize and form bands . band formation acts to suppress the formation of magnetic moments in two ways . first , hybridization breaks spherical symmetry for the environment of each atom , which tends to quench any orbital component of the magnetic moment . second , band formation also inhibits spin polarization . if one starts with a system of unpolarized electrons and imagines flipping spins to create alignment , then there is a kinetic - energy cost associated with moving electrons from lower - energy filled band states to higher - energy unoccupied band states . as a result , most solids are not ferromagnetic . there are , of course , exceptions . for example in materials with tightly bound 4f - orbitals , the hybridization is so weak that those levels do become spin polarized much as they do in the atomic state . the transition metal ferromagnets iron , cobalt , nickel , and their alloys , having partially filled d - orbitals , are the exceptions of particular interest in these articles . the transition metal ferromagnets have both strong exchange splitting and strong hybridization . the exchange splitting can stabilize a spin - polarized ferromagnetic state , even in the presence of band formation , by generating a self - consistent shift of the majority - electron - spin band states to lower energy than the minority - electron - spin states , so as to more than compensate for the kinetic - energy cost associated with the formation of the polarization . _ models of ferromagnetic materials . _ the local spin density approximation ( lsda ) @xcite accurately describes much of the important physics in these systems . it treats the atomic - like exchange and correlation effects in mean field theory and treats the hybridization exactly . without any fitting parameters it accurately predicts @xcite many of the properties of transition metal ferromagnets like the magnetic moment . in this approach , the electron density and spin density are the fundamental degrees of freedom and the wave functions are formal constructs that allow calculation of the density . as such , there is no formal justification for using the lsda wave functions as the physical wave functions . however , the wave functions are a solution to an accurate mean field theory ( lsda ) and in practice they can serve as a good approximation to the real wave functions in many cases . many calculations of spin transfer torques are based on using the wave functions found from the lsda ; see in particular the article from haney , duine , nez , and macdonald for examples . there are two simplified models of ferromagnetism that are sometimes used to give descriptions of the physics of spin transfer torques and for calculations . the first is the free - electron stoner model . this assumes that the electron bands for spin - up and spin - down electrons have a relative shift in energy due to an exchange interaction but otherwise they both have a free - electron dispersion , @xmath1 . here @xmath2 is the pauli spin matrix and @xmath3 is the exchange splitting . the second simplified model is the s - d model . it was originally introduced @xcite to describe local moment impurities in a non - magnetic host . the `` s '' electrons describe the delocalized conduction band states of the host and the `` d '' electrons describe the localized magnetic states , which are weakly coupled to the s electrons . frequently , each d electron shell is treated as a local moment @xmath4 , which interacts with the conduction - electron spin density @xmath5 through a weak local interaction @xmath6 . neither the stoner model nor the s - d model are well - justified approximations for describing the transition - metal ferromagnets . see fig . [ fig : bs ] for a comparison of the band structures of these two models with a more realistic band structure for co computed using the lsda . the simplified models can be very useful for illustrating physical concepts , and sometimes for estimates , but they are far from realistic . the band structure in transition - metal ferromagnets is considerably more complicated than that of a single free - electron band as assumed in the stoner model . as for comparisons to the s - d model , in real materials the hybridization within the d bands and of the d bands with the s bands are quite strong , and so the d electrons can not be considered localized . on the other hand , the s - d model is one of several models that have been used to describe ferromagnetic semiconductors , like those discussed in the article by ohno and dietl . in these systems , the mn substitutions are believed to act very much like local moments . the origins of spin - polarized currents in magnetic devices and the giant magnetoresistance ( gmr ) effect can be understood as consequences of the difference in band structures for the majority - spin and minority - spin states in magnets , as predicted by lsda calculations and illustrated in a over - simplified way by the exchange splitting in the free - electron stoner model . spin - polarized currents can come about because the spin - dependent electron properties in ferromagnets allow magnetic thin films to act like spin filters . consider the example of cr / fe multilayers in which gmr was first discovered . for electrons incident from cr into a fe layer , minority spins have a greater probability to be transmitted through the fe film than majority spins , due to differences scattering caused by the band - structure mismatch at the interface and also due to band - structure - induced differences in the strength of scattering from defects and impurities in the fe layer @xcite . therefore the current transmitted through the fe layer in a cr / fe / cr device is partially spin - polarized in the minority direction , while the current reflected from the layer is partially polarized in the majority direction . many other factors can also affect the ultimate polarization of the currents , including the layer thickness and the spin flip scattering rates . once in a non - ferromagnetic metal like cr or cu , a spin - polarized current persists on the scale of the spin diffusion length , typically on the order of 100 nm or more in cu and about 5 nm in cr . when two fe layers in a cr / fe multilayer are spaced more closely than this and have parallel magnetic moments , minority electrons have a high probability of being transmitted through both layers , resulting in a relatively low overall device resistance even though the majority electrons are scattered strongly . in effect , the minority electrons `` short out '' the structure , giving a low resistance . when nearby magnetic layers have antiparallel moments , both majority and minority electrons scatter strongly in either one layer or the other , and the resulting resistance is higher . this is the origin of gmr . other combinations of normal metals and magnetic layers can act similarly , although for combinations like cu / co or cu / ni it is the majority electrons that are more easily transmitted through the magnetic layer , rather than the minority electrons @xcite . a number of different theoretical approaches have been used to calculate the transport properties of magnetic multilayers and their gmr @xcite . these techniques will be discussed in more detail in section [ sec : multi ] . _ micromagnetics . _ in order to describe the equilibrium configuration of the magnetization in a ferromagnet , or the dynamical response to an applied magnetic field or a spin - transfer torque , it can be important to take into account that the magnetization distribution may become spatially non - uniform . micromagnetics @xcite is a phenomenological description of magnetism on a mesoscopic length scale designed to model such non - uniformities in an efficient way . it does not attempt to describe the behavior of the moment associated with each atom , but rather adopts a continuum description much like elasticity theory . its utility arises because the length scales of interest in magnetic studies are frequently much longer than atomic lengths . atomic scale calculations become impractical in this case . in equilibrium , the magnetization direction aligns itself with an effective field , which can vary as a function of position . there are generally four main contributions to this effective field : the externally applied magnetic field , magnetocrystalline anisotropy , micromagnetic exchange , and the magnetostatic field . each of the fields is most easily described in terms of an associated contribution to the free energy . the total effective field is then the functional derivative of the free energy with respect to the magnetization @xmath7 . the magnetocrystalline anisotropy arises from the spin - orbit interactions and tends to align the magnetization with particular lattice directions . generally speaking the anisotropy field is a local function of the magnetization direction and has a different functional form for different lattices and materials . the micromagnetic exchange @xcite is the interaction that tends to keep the magnetization aligned in a common direction , adding an energy cost when the magnetization rotates as a function of position . the magnetostatic interaction is a highly non - local interaction between the magnetization at different points mediated by the magnetic field produced by the magnetization . together , the four free energies can be written @xmath8 where @xmath9 , @xmath10 , @xmath11 is the saturation magnetization , @xmath12 is the exchange constant , and @xmath13 is the anisotropy constant . here we have taken the specific example of a uniaxial anisotropy with an easy axis along @xmath14 . the total effective field derived from eq . ( [ eq : mumage ] ) is @xmath15 the atomic - like exchange , which drives the formation of the magnetization and which is not explicit in these expressions , places a strong energetic penalty on deviations of the magnitude of @xmath16 away from @xmath11 . this interaction is generally taken into account by treating @xmath16 as having the fixed length @xmath11 . _ magnetic domains . _ the interactions within eqs . ( [ eq : mumage ] ) and ( [ eq : mumagh ] ) can compete with one another in determining the orientation of @xmath17 as a function of position . different interactions can dominate on different spatial scales , with the consequence that the magnetic ground state is often spatially non - uniform , containing non - trivial magnetization patterns even in equilibrium . the micromagnetic exchange and magnetocrystalline anisotropy both represent relatively short - ranged or local interactions . the micromagnetic exchange tends to keep the magnetization spatially uniform and the magnetocrystalline anisotropy can tend to keep it directed in particular lattice directions . for the energy functional above , with uniaxial anisotropy , those directions are @xmath18 . for the materials of interest for spin transfer applications the magnetocrystalline anisotropy is frequently weak and does not play an important role ( although there are exceptions @xcite ) . anisotropies in spin transfer devices are more commonly the result of the sample shape . magnetostatic interactions favor magnetization orientations aligned in the plane of thin - film samples and along the long axis of samples with non - circular cross - sections . because of the dipole pattern of long - ranged magnetic fields , the magnetostatic interaction can also favor antiparallel alignment of the magnetization in distant parts of a sample , and this can cause the magnetization pattern to become non - uniform . on short length scales the magnetostatic interaction is relatively weak in comparison to micromagnetic exchange . however the magnetostatic interaction is long - ranged so that it can eventually dominate in large enough samples . this causes the magnetization pattern to depend on the sample size . for magnetic thin - film samples smaller than about 100 nm to 200 nm in diameter , the ground state is approximately ( but not exactly ) uniform , with the micromagnetic exchange dominating the magnetostatic interactions @xcite . for samples slightly bigger than this , magnetostatic interactions become more important and the ground state can be a vortex state . for still larger samples , the magnetization pattern may break up into regions which have different magnetic orientations but within which the magnetization is roughly uniform . these regions are called domains @xcite . the border region between domains is referred to as a domain wall . here the magnetization rotates over a relatively short distance from one domain s orientation to the other s . the more gradual is this rotation , the less the cost in exchange energy . however , wider domain walls deviate from the low - energy orientation for magnetocrystalline anisotropy over a larger volume , and therefore cost more anisotropy energy . the domain wall width is determined by a compromise which minimizes the total energy of exchange + anisotropy , and can be characterized roughly by the scale @xmath19 . this length is strongly material dependent , ranging from @xmath0 1 nm for hard magnetic materials to more than 100 nm for soft magnetic materials . in cases with weak anisotropy , domain wall widths are determined by a competition between exchange and magnetostatic interactions . the domain wall width can also depend on sample geometry , and in a narrow contact between electrodes the wall width can be narrowed in proportion to the contact diameter @xcite . in thin - film wires , typical of those used to study current - induced domain wall motion , domain walls typically take one of two structures , `` transverse '' walls or `` vortex '' walls . on either side of the wall , the magnetization lies in - plane and points along the length of the wire to minimize the magnetostatic energy , and there is a net 180@xmath20 rotation of the magnetization at the wall . in a transverse wall , the magnetization simply rotates in the plane of the sample from one domain to the other . however , the competition between the micromagnetic exchange energy and the magnetostatic energy causes the wall width to be narrow ( on the scale of the exchange length , @xmath21 4 nm to 8 nm , where @xmath12 is the prefactor of the micromagnetic exchange in eq . ( [ eq : mumage ] ) and @xmath11 is the saturation magnetization ) on one edge of the wire and wide ( on the scale of the width of the wire ) on the other edge . a vortex wall is even more complicated . here , the magnetization wraps around a central singularity @xcite , the vortex core , giving a circulating pattern to the magnetization . the competition between these two wall structures is studied in ref . beach , tsoi , and erskine describe how the detailed structure of domain walls plays a crucial role in their motion when they are driven by either an applied magnetic field or a spin transfer torque . _ magnetic dynamics in the absence of spin transfer torques . _ when a magnetic configuration is away from equilibrium , the magnetization precesses around the instantaneous local effective field . in the absence of dissipation , the magnetization distribution stays on a constant energy surface . in order to account for energy loss , landau and lifshitz @xcite introduced a phenomenological damping term into the equation of motion and gilbert @xcite introduced a slightly different form several decades later . both forms of the damping move the local magnetization vector toward the local effective field direction : @xmath22 where @xmath23 is the gyromagnetic ratio , @xmath24 is the landau - lifshitz damping parameter , and @xmath25 is the gilbert damping parameter . these two forms are known to be equivalent with the substitutions , @xmath26 and @xmath27 . in spite of this equivalence , there has been an ongoing debate about which is more correct . this debate has been rekindled with the interest in current - driven domain wall motion ( one of the present authors is guilty of contributing to the debate ) and is mentioned in the articles by tserkovnyak , brataas , and bauer and by berkov and miltat . part of the fervor of the debate arises from the fact that it is not experimentally testable . appropriate equations of motion can be formulated with either form of damping at the expense of slight modifications to other terms in the equation of motion . this point is discussed further in sec . [ sec : wires ] . note that both the precession and damping terms rotate the magnetization , but do not change its length . this is consistent with treating the magnetization as having a fixed length . there have been many attempts to compute the damping parameters from models for various physics processes @xcite . some mechanisms are intrinsic to the material , such as those due to magnetoelastic scattering @xcite , and others are considered extrinsic like two - magnon scattering from inhomogeneities @xcite . it appears that a model due to kambersky @xcite for electron - hole pair generation describes the dominant source of intrinsic damping in a variety of metallic systems including the ferromagnetic semiconductors @xcite and transition metals @xcite primarily of interest for spin transfer torque applications . for a magnetic element in metallic contact with other materials , there can also be a contribution to the damping from `` spin - pumping '' the emission of spin - angular momentum from the precessing magnet via the conduction electrons @xcite . magnetization dynamics is most easily investigated using the macrospin approximation . the macrospin approximation assumes that the magnetization of a sample stays spatially uniform throughout its motion and can be treated as a single macroscopic spin . since the spatial variation of the magnetization is frozen out , exploring the dynamics of magnetic systems is much more tractable using the macrospin approximation than it is using full micromagnetic simulations . the macrospin model makes it easy to explore the phase space of different torque models , and it has been a very useful tool for gaining a zeroth - order understanding of spin - torque physics . however , for many systems of interest , even ones with very small magnetic elements , the macrospin approximation breaks down . for a full understanding of magnetic dynamics , a micromagnetic approach is therefore necessary . the article by berkov and miltat discusses some of the circumstances in which the macrospin approach is and is not a reasonable approximation to the true dynamics . for analyses of magnetic domain wall dynamics , there is a different simplified description that can sometimes be useful for qualitative understanding , as an alternative to full micromagnetic calculations . the walker ansatz @xcite restricts the variation in the domain wall to uniform translations , uniform rotations out of plane , and in some versions a uniform scaling of the wall width . for the simplest version , the dynamics of the domain wall can then be described by two degrees of freedom . the article by beach , tsoi and erskine discusses when this approximation is valid and when it is not for current - induced domain wall motion . in some situations it is useful to consider the dynamics of a magnetic sample in terms of the normal modes of the system , known as spin waves or magnons , instead of directly integrating the equations of motion at every point on a closely spaced grid designed to model the sample , as is usually done in micromagnetic calculations . a spin wave is a small amplitude oscillation of the magnetization around its average direction . in macroscopic magnetic systems , the spectrum of spin waves is essentially continuous as a function of frequency , but when thin - film magnetic elements are shrunk to the scale of 100 nm in diameter , the spin wave spectrum becomes measurably discrete @xcite . in fact , the recent development of spin - transfer - driven ferromagnetic resonance has made it possible to measure the frequencies of these normal modes within individual nanostructures @xcite . interestingly , calculations show that uniform precession , of the type assumed in the macrospin approximation , is generally not a true normal mode because the magnetostatic field is generally not uniform across the sample . an analysis in terms of the normal modes can provide a strategy for simulating magnetic dynamics that is more efficient than standard micromagnetic simulations and somewhat more accurate than the macrospin approximation . the dynamics of a magnetic excitation can be approximated by expanding the excitation using a finite set of normal modes as a basis set , and determining the time dependence based on the dynamics of the individual modes and their nonlinear couplings , rather than by integrating the full equation of motion directly . in cases when the lowest - frequency most - spatially uniform normal mode dominates the magnetic dynamics , the results are typically very similar to the predictions of the simplest macrospin descriptions . the relevance of the spin wave modes is discussed in the articles by berkov and miltat and by sun and ralph . thus far we have discussed the dynamics of magnets in the absence of the spin transfer torque . spin transfer torques arise whenever the flow of spin - angular momentum through a sample is not constant , but has sources or sinks . this happens , for example , whenever a spin current ( created by spin filtering from one magnetic thin film ) is filtered again by another magnetic thin film whose moment is not collinear with the first . in the process of filtering , the second magnet necessarily absorbs a portion of the spin angular momentum that is carried by the electron spins . changes in the flow of spin angular momentum also occur when spin - polarized electrons pass through a magnetic domain wall or any other spatially non - uniform magnetization distribution . in this process , the spins of the charge carriers rotate to follow the local magnetization , so the spin vector of the angular momentum flow changes as a function of position . in either of these cases , the magnetization of the ferromagnet changes the flow of spin angular momentum by exerting a torque on the flowing spins to reorient them , and therefore the flowing electrons must exert an equal and opposite torque on the ferromagnet . this torque that is applied by non - equilibrium conduction electrons onto a ferromagnet is what we will call the spin transfer torque . its strength can be calculated either by considering directly the mutual precession of the electron spin and magnetic moment during their interaction ( an approach discussed in the article by haney , duine , nez , and macdonald ) or by considering the net change in the spin current before and after the interaction ( the approach we will emphasize ) . our discussion in this section will consist of two parts . first we will consider how it is that a spin - polarized current can apply a torque to a ferromagnet . this will be straightforward since a torque is simply a time rate of change of angular momentum , considerations of angular momentum conservation can be used to relate the spin transfer torque directly to the angular momentum lost or gained by spin currents . we will use two simple toy models to illustrate some of the physics involved in this process . the second part of our discussion will describe how to incorporate the spin transfer torque into the equation of motion for the magnetization dynamics . this step of the argument will involve some more - subtle points , related to the connection between the magnetization of a ferromagnet and its total angular momentum . to explore these points fully , we will consider how one might derive the equation of motion for the magnetization , @xmath28 , within a rigorous quantum mechanical theory . _ definition of the spin current density . _ the primary quantity on which we will focus our interest will be the spin current density @xmath29 . this has both a direction in spin space and a direction of flow in real space , so it is a tensor quantity . for a single electron , the spin current is given classically by the outer product of the average electron velocity and spin density @xmath30 . for a single - electron wavefunction @xmath31 , the spin current density may be written @xmath32 where @xmath33 is the electron mass , and @xmath34 represents the pauli matrices @xmath35 , @xmath36 , and @xmath2 . the form of the spin current density is similar to the more - familiar probability current density @xmath37 . for a spinor plane - wave wavefunction of the form @xmath38 where @xmath39 is a normalization volume , the spatial part of the spin current points in the @xmath40 direction , and the three spin components take the simple forms @xmath41 by conservation of angular momentum , one can say that the spin transfer torque acting on some volume of material can be computed simply by determining the net flux of non - equilibrium spin current through the surfaces of that volume , or equivalently by integrating the divergence of the spin current density within an imaginary pillbox surrounding the volume in question : @xmath42 where @xmath43 is the in - plane position and @xmath44 is the interface normal for each surface of the pillbox . ( note that since @xmath29 is a tensor , its dot product with a vector in real space leaves a vector in spin space . ) if one prefers to think in terms of the differential form of eq . ( [ eq : pillbox ] ) , it states that the spin torque density is the divergence of the spin current density . _ toy model # 1 . _ our first simple model is meant to illustrate that when a spin polarized current interacts with a thin ferromagnetic layer and undergoes spin filtering the result , in general , is that a spin transfer torque is applied to the magnetic layer . consider the problem of a single - electron state with wave vector @xmath45 in the @xmath46 direction and spin oriented in the @xmath40-@xmath47 plane at an angle @xmath48 with respect to the @xmath47 direction , which is incident onto a thin magnetic layer whose magnetization is pointed in the @xmath47 direction ( see fig . [ fig : toy](a ) ) . we will initially not be concerned about what goes on inside the magnetic layer , but we will account for its spin filtering properties simply by assuming that it can be described by overall transmission and reflection amplitudes for spin - up electrons ( @xmath49 , @xmath50 ) that are different from the transmission and reflection amplitudes for spin - down electrons ( @xmath51 , @xmath52 ) , and that no spin - flipping processes occur . under these assumptions , the incident part of the wavefunction is @xmath53 this can be derived , for example , by starting with the @xmath54 state and applying the appropriate rotation matrix for a spin-1/2 system @xcite . the transmitted and reflected parts of the scattering wavefunction are @xmath55 the components of the spin current density can be determined using the expressions given in eq . ( [ eq : qs ] ) . the flows of spin density in the @xmath40 spatial direction for the incident , transmitted , and reflected parts of the wavefunction take the forms @xmath56 \nonumber \\ { \bf q}_{\rm trans } & = & { \hbar^2 k \over 2m{\omega } } \sin(\theta){\rm re}(t_{\uparrow}t_{\downarrow}^ * ) { \bf \hat{x } } \nonumber\\ & & + { \hbar^2 k \over 2m{\omega } } \sin(\theta ) { \rm im}(t_{\uparrow}t_{\downarrow}^ * ) { \bf \hat{y } } \nonumber\\ & & + { \hbar^2 k \over 2m{\omega } } \big [ \|t_{\uparrow}\|^2 \cos^2(\theta/2)- \|t_{\downarrow}\|^2 \sin^2(\theta/2)\big ] { \bf \hat{z } } \\ { \bf q}_{\rm refl } & = & - { \hbar^2 k \over 2m{\omega } } \sin(\theta){\rm re}(r_{\uparrow}r_{\downarrow}^ * ) { \bf \hat{x } } \nonumber\\ & & - { \hbar^2 k \over 2m{\omega } } \sin(\theta ) { \rm im}(r_{\uparrow}r_{\downarrow}^ * ) { \bf \hat{y } } \nonumber\\ & & - { \hbar^2 k \over 2m{\omega } } \big [ \|r_{\uparrow}\|^2 \cos^2(\theta/2)- \|r_{\downarrow}\|^2 \sin^2(\theta/2)\big ] { \bf\hat{z}}. \nonumber\end{aligned}\ ] ] it is then clear that the total spin current is not conserved during the filtering process : the spin current density flowing on the left of the magnet @xmath57 is not equal to the spin current density on the right @xmath58 . by eq . ( [ eq : pillbox ] ) , we can say that the spin transfer torque @xmath59 on an area @xmath60 of the ferromagnet is equal to the net spin current transferred from the electron to the ferromagnet , and is given in this toy model by @xmath61 { \bf \hat{x } } \nonumber\\ & & - { a \over \omega}{\hbar^2 k \over 2 m } \sin(\theta ) { \rm im}(t_{\uparrow}t_{\downarrow}^ * + r_{\uparrow}r_{\downarrow}^ * ) { \bf \hat{y } } . \label{eq : model1}\end{aligned}\ ] ] we have used the fact that @xmath62 and @xmath63 . there is no component of spin torque in the @xmath47 direction . we find the general result that the spin transfer torque is zero when @xmath64 and @xmath65 ( in which case the `` magnetic '' layer would provide no spin filtering ) or when the incoming spin orientation is collinear with the magnetization of the layer , @xmath66 or @xmath67 . however , for any non - collinear spin orientation , when the magnet does provide spin filtering , it is a direct consequence of the spin filtering that the spin transfer torque acting on the ferromagnetic layer is non - zero . this torque is perpendicular to the magnetization of the layer ( no @xmath68 component ) , and for an individual incident electron the torque may have components in both the @xmath46 and @xmath69 directions , depending on the values of the transmission and reflection coefficients . it is possible to extend this type of 1-d toy model to calculate the torque applied to a magnetic thin film in a realistic 3-dimensional sample . the calculation proceeds by summing the torque contributed by electron waves incident onto the magnetic thin film from throughout the fermi surface of the non - magnetic metal , corresponding to electrons incident from many directions in real space @xcite . this requires summing the contributions to the @xmath46 and @xmath70 components of the torque in eq . ( [ eq : model1 ] ) . in terms of the terminology commonly used in this field @xcite , the sum over the @xmath46-component contributions is proportional to the real part of the `` mixing conductance '' and gives the `` in - plane '' torque ( the plane defined by the moments on the ferromagnet and the incoming spin ) , and the sum over the @xmath69-component contributions is proportional to the imaginary part of the mixing conductance and gives a perpendicular torque . _ toy model # 2 . _ next we consider a second simple 1-dimensional toy model @xcite , to illustrate some of the processes that occur near a normal metal / ferromagnetic interface and influence the spin torque . here we again first assume a single incoming spin - polarized electron wavefunction of the form given by eq . ( [ eq : wavefunction ] ) incident onto a magnetic layer whose magnetization is in the @xmath47 direction . however , in this case we will use a stoner - model approach to describe the magnetic layer . that is , we will assume that the electrons inside the ferromagnetic layer experience an exchange splitting @xmath3 which shifts the states in the minority - spin band ( down electrons ) higher in energy than the majority - spin band ( up electrons ) , but that both bands have a free - electron dispersion . the physics near the interface can then be modeled as a simple scattering problem in which the electron scatters from a rectangular potential - energy step ( at position @xmath71 ) that has different heights for spin - up and spin - down electrons ( see fig . [ fig : toy](b ) ) . for simplicity , we will assume that the height of the potential - energy step is 0 for up spins and @xmath3 for down spins , and we will consider an electron energy @xmath72 which is greater than @xmath3 . by matching wavefunctions and their derivatives at the interfaces , it is an elementary problem to calculate the transmitted and reflected parts of the scattering - state wavefunction with energy eigenvalue @xmath73 : @xmath74 where @xmath75 and @xmath76^{1/2}/\hbar < k$ ] . the incident , transmitted , and reflected spin currents are @xmath77 { \bf \hat{x } } \nonumber\\ & & - { \hbar^2 \over 2m{\omega } } \sin(\theta ) k \sin[(k_{\uparrow } - k_{\downarrow})x ] { \bf \hat{y } } \\ & & + { \hbar^2 \over 2m{\omega } } \left [ k \cos^2(\theta/2)- k_{\downarrow } \left({2k \over k + k_{\downarrow}}\right)^2 \sin^2(\theta/2)\right ] { \bf \hat{z } } \nonumber \\ { \bf q}_{\rm refl } & = & { \hbar^2 \over 2m{\omega } } k \left ( { k- k_{\downarrow } \over k + k_{\downarrow } } \right)^2 \sin^2(\theta/2 ) { \bf \hat{z}}. \nonumber\end{aligned}\ ] ] there are two points of physics that we wish to illustrate with this example . first , the transverse ( perpendicular to @xmath47 ) spin component of the reflected spin current density is equal to zero . since the total spin current density is continuous at the interface , this means that all of the of the transverse component of the incident spin current density is transmitted through the interface ; none is reflected . in this toy model , this result follows from our assumption that for spin up electrons the height of the potential - energy step at the interface is zero , so that the reflection amplitude for spin up electrons is zero and the reflected part of the wavefunction is purely spin down . for models in which both components of spin experience a non - zero potential - energy step , some of the incident @xmath46 component of the spin current density will be reflected . however , for many of the materials combinations used commonly in metallic gmr devices , like cu / co , cu / ni , or cr / fe , one of the spin components actually does have a reflection amplitude close to zero over a large part of the fermi surface @xcite , so it is a reasonable approximation in these cases that almost all of the transverse ( @xmath46 ) component of the spin current density will be transmitted into the ferromagnet . the second important point of physics illustrated by the model concerns what happens to the transverse component of the spin current density after it enters the ferromagnet . the oscillatory @xmath40 and @xmath69 terms in the transmitted spin current density represent precession of the spin about the @xmath47 axis as a function of position as it penetrates through the magnet . in any model in which there is a difference in exchange energy between majority and minority spin states , the two spin components of a wavefunction for a given eigenvalue @xmath73 must have different kinetic energies , so that @xmath78 and the spin state inside the magnet will precess . the same phenomenon is therefore present in more rigorous models . one can view this effect as simply the precession of the spin in the exchange field of the magnet . the period of the precession , @xmath79 is very short for a typical transition metal ferromagnet , on the scale of a few atomic lattice spacings . this is important because in real 3-dimensional samples many electrons are incident on the magnetic layer from a variety of directions , corresponding to states from all parts of the fermi surface , and therefore different electrons take different paths through the magnetic layer . even if all of the electrons begin with perfectly aligned spins at the normal - metal / ferromagnet interface , electrons reaching a given depth inside the magnet will have traveled different path lengths to get there . the result is classical dephasing . electron spins that have traveled different path lengths will have precessed by different angles around the @xmath47 direction , and therefore their @xmath40 and @xmath70 components will not add constructively . for locations more than a few atomic lattice constants into a magnetic layer , when one sums over electrons from all relevant parts of the fermi surface in calculations that include first - principles computations of the transmission amplitudes @xcite , the transverse ( to @xmath47 ) components of the spins average to zero . as a consequence of this classical dephasing , there is no net transmission of transverse spin angular momentum through the ferromagnet . the transverse angular momentum that enters into the ferromagnet is effectively absorbed within a few atomic layers from the interface . for a full calculation of the spin torque at an interface in a real 3-dimensional sample , it is important to take into account not just propagating wavefunctions , but also evanescent scattering states at the interface . if our toy model # 2 is generalized to three dimensions , evanescent scattering wavefunctions are required in cases where the incident electron approaches the interface from a glancing angle , so that the part of the kinetic energy associated with the perpendicular wavevector is less than the step height @xmath3 . in calculations with more realistic band structures , both evanescent and propagating scattering states can couple to incident bloch states more generally , and it is necessary to take the evanescent states into account to guarantee the continuity of the wavefunction and its first derivative on the atomic scale . although evanescent scattering states do not carry charge current , they do carry spin current , so that they can contribute a significant spin torque even when the net charge flow through the interface is zero . this point can be illustrated by our toy model if we consider a case in which the spin - dependent step heights are sufficiently high that both of the spin components are completely reflected . the transmitted and reflected parts of the scattering wavefunction are then @xmath80 where @xmath81 and @xmath82 are decay constants for the evanescent states in the ferromagnet . because the transmission and reflection amplitudes are now complex , with ( in general ) different complex phases for spin up and spin down electrons , the transmitted and reflected spin current densities will contain @xmath83 as well as @xmath46 components . the transmitted spin current density also decays exponentially to zero as a function of the penetration distance into the ferromagnet . what this means is that , in effect , the electron penetrates into the ferromagnet a distance on the order of @xmath84 and precesses around the exchange field as it does so , so that when it emerges from the magnet it is rotated away from its original orientation . consequently , during the process of reflection an electron can apply a torque to the magnet in both the in - plane and perpendicular directions . in calculations with realistic band structures for normal - metal / ferromagnet interfaces , when one sums over the fermi surface to determine the total value of the transverse part of the reflected spin current there is significant ( but not perfect ) classical dephasing , so that the overall net flow of reflected transverse angular momentum is close to zero @xcite . this means that the incident transverse angular momentum that couples into the evanescent states can not end up flowing away from the interface through the reflected states . instead , this transverse spin angular momentum is deposited in the interfacial region of the ferromagnet via the torque from the evanescent states . this contribution to the torque , which is not directly mediated by the propagating states , can be described as due to spin filtering . if one accounts for the evanescent states when constructing the scattering states but then ignores their contribution to the spin current , the spin - filtering contribution corresponds to the resulting interfacial discontinuity in the part of the spin current density carried by just the propagating states . the net result of the classical dephasing that occurs for both transmitted and reflected electron waves at a normal - metal / ferromagnet interface is that the total transmitted and reflected spin currents , summed over all relevant states on the fermi surface , are approximately collinear with the ferromagnetic layer s magnetization ( in the @xmath47 direction , in our toy model ) . since , to a good approximation , no transverse angular momentum flows away from the magnet , this collinearity means that approximately the _ entire _ incident transverse spin current is absorbed near the normal - metal / ferromagnet interface , and the spin transfer torque ( eq . ( [ eq : pillbox ] ) ) becomes @xmath85 in terms of the parameters used in our first toy model above , it is correct to say that when summing or averaging over all contributions from around the fermi surface that to a good approximation for a typical metallic interface the dephasing leads to @xmath86 , and to a somewhat less - accurate approximation @xmath87 , so that on average for our one electron @xmath88 and the spin torque acting on the magnet per unit area is equal to the full component of incident spin current that is transverse to the ferromagnet s moment . the result in eq . ( [ eq : intabsorb ] ) is a good approximation for metallic interfaces , like cr / fe or cu / co , but the processes that lead to the simple form @xmath89 may be different for magnetic semiconductors or for tunnel junctions like fe / mgo / fe . most electrons that scatter from tunnel barriers reflect , whether they are majority or minority . in addition , tunneling is dominated by electrons that are largely from particular parts of the fermi surface , so the classical dephasing processes that are important for metallic junctions may be weaker in tunnel junctions . in fact , there is good evidence that @xmath90 in tunnel junctions , so that for large applied biases there can be a significant spin torque component in the @xmath70 direction ( perpendicular to the plane defined by the incoming electron spin and the ferromagnet s moment ) @xcite . this is discussed in more detail in the article by sun and ralph . before moving on to consider how the spin torque will affect the ferromagnet s magnetization orientation , we wish to re - emphasize one last important point . spin currents can flow within parts of devices even where there is no net charge current . consequently , a spin transfer torque can also be applied to magnetic elements that do not carry any charge current @xcite . we have already noted two examples of this effect , in slonczewski s original calculation of interlayer exchange coupling in a magnetic tunnel junction @xcite and in our toy model # 2 for the case when both spin - up and spin - down components of the wavefunction are completely reflected . another important example occurs in multiterminal normal - metal / ferromagnet devices , which are designed so that a charge current flows only between two selected terminals , but diffusive spin currents may also flow throughout the rest of the device . the groups of johnson , van wees , and others have demonstrated non - local spin accumulation in nonmagnetic wires using multiterminal devices @xcite . kimura et al . have used a similar lateral device design to demonstrate spin - torque - driven switching of a thin - film magnetic element which carries no charge current @xcite . one can view this effect as due simply to a flow of spin - polarized electrons penetrating by diffusive motion into a magnetic element and transferring their transverse component of spin angular momentum , while an equal number of electrons exit the magnet with an average spin component collinear with the magnet so that they give no spin torque . in this way there can be a non - zero net spin current and therefore a non - zero spin torque on the magnet , even when there is no net charge current . thus far the switching currents required in the devices of kimura _ et al_. are larger than those needed to switch comparable magnetic elements in standard magnetic - multilayer pillar devices , because the magnitude of the spin current densities achieved in the lateral devices ( per unit injected charge current ) is smaller than in the multilayers . _ spin transfer torque and the landau - lifshitz - gilbert equation . _ to calculate the effects of the spin transfer torque on magnetic dynamics , in practice a term @xmath91 is generally simply inserted as an additional contribution on the right side of the landau - lifshitz - gilbert equation ( eq . ( [ eq : llg ] ) ) . however , this step deserves some careful consideration , as it involves a few subtle points of physics . first , this insertion assumes that the all of the angular momentum transferred from the transverse spin current density acts entirely to reorient the orientation of the ferromagnet rather than , for example , being absorbed in the excitation of short - wavelength magnon modes or being transferred directly to the atomic lattice . this seems to be a reasonable approximation for describing the experiments performed to date ; however , as we note below , the existing spin - torque measurements in metallic multilayer samples are not particularly quantitative . a second simple matter to keep straight is the sign of the torque . an electron s magnetic moment is opposite to its spin angular momentum @xmath92 , where @xmath4 is the total spin and @xmath93 , and likewise in transition metal ferromagnets the magnetization is generally opposite to the spin density @xmath94 , where @xmath5 is here the spin density and @xmath95 is typically in the range -2.1 to -2.2 @xcite . we have defined @xmath96 as a time rate of change of angular momentum . since the landau - lifshitz - gilbert equation is stated in terms of magnetization , the contribution of the spin transfer torque should enter this equation with a sign opposite to the change in angular momentum . in the end , the sign of the spin transfer torque is such as to rotate the spin angular momentum density of the ferromagnet toward the direction of the spin of the incoming electrons , or equivalently to rotate the magnetization of the ferromagnet toward the direction of the moment of the incoming electrons . a third , potentially much more consequential , subtlety involves the orbital contribution to the magnetization . if we assume that all of the angular momentum in the ferromagnet is due to its spin density , then conservation of angular momentum implies that the effect of spin transfer torque on the magnetization can be described simply by inserting @xmath97 into the right side of the landau - lifshitz - gilbert equation ( eq . ( [ eq : llg ] ) ) . here @xmath98 is the volume of the ferromagnet ( free layer ) over which the spin torque @xmath96 is applied . ignoring any orbital contribution is a reasonable first - order approximation , because the orbital moments in transition metal ferromagnets are largely quenched by the strong hybridization of the d electrons . however , spin - orbit coupling does give rise to a weak orbital moment , typically less than a tenth of the spin moment as indicated by the deviation of @xmath95 from -2 . in a more precise treatment that takes the orbital contribution into account , the total angular momentum density would be @xmath99 , and the magnetization would be @xmath100 , so that there might not be any simple proportionality between the total angular momentum density and @xmath101 . this would necessitate a significantly more complicated picture , as described in more detail immediately below . however , in most analyses of spin transfer torques , the potential effects of orbital moments are ignored , and it is assumed that the spin torque is simply described by eq . ( [ eq : mdot ] ) . the appropriateness of neglecting orbital angular momentum is discussed briefly in the article by haney , duine , nez , and macdonald . _ a more rigorous approach to the equation of motion for magnetization . _ in this section we will discuss how one might take a more systematic approach to deriving the equation of motion for a ferromagnet under the influence of a spin transfer torque . this exercise will give additional insights into what might be required to account more accurately for effects like orbital angular momentum and spin - orbit coupling . this approach also provides a more natural framework for considering spin transfer torques at domain walls and in other spatially non - uniform magnetization distributions . our starting point is that the equation of motion for any variable in quantum mechanics can be determined by taking the commutator of the operator corresponding to that variable with the hamiltonian and then evaluating the expectation value of the result . to explore how this process works , we will consider first the charge density , then the spin density , and finally ( briefly ) the magnetization . in second quantized notation @xcite the charge density and spin density operators are @xmath102 in terms of the creation @xmath103 and destruction @xmath104 operators for an electron at point @xmath105 and spin @xmath106 . these fermion operators obey the anti - commutation relations @xmath107 . for the present purposes we consider the hamiltonian for non - interacting electrons as in the mean field lsda approach @xmath108 \nonumber\\ & & + \int d^3 r \mu_0 { g\mu_{\rm b } \over \hbar } ( { \bf h}_{\rm ext } + { \bf h}_{\rm dip } ) \cdot \hat{{\bf s } } .\end{aligned}\ ] ] the first term in the above equation is the kinetic energy , the second term is the potential , including the local exchange field , and the third term is the coupling with the applied field @xmath109 and the dipolar field @xmath110 due to the rest of the spins . more generally , the potential , the local exchange field , and the dipolar field are many - body terms , but for the present example they are treated as effective single particle interactions . for the moment we are also ignoring spin - orbit coupling in this hamiltonian . let us first derive the equation of motion for the electron charge density . the only term in the hamiltonian that does not commute with the charge density operator is the kinetic energy and it gives rise to a term that is the divergence of the charge current density @xmath111 \nonumber\\ & = & { -e\hbar \over 2im } \sum_\sigma \int d^3 r \bigg\ { \hat\psi^\dagger_{\sigma ' } ( { \bf r } ' ) [ \bm\nabla_{{\bf r}'}\delta(({\bf r}-{\bf r}'))]\delta_{\sigma,\sigma ' } \bm\nabla\hat\psi_\sigma ( { \bf r } ) \nonumber\\ & & - \bm\nabla\hat\psi^\dagger_\sigma ( { \bf r } ) [ \bm\nabla\delta(({\bf r}-{\bf r}'))]\delta_{\sigma,\sigma ' } \hat\psi_{\sigma ' } ( { \bf r } ' ) \bigg\ } \nonumber\\ & = & - \bm\nabla \cdot \hat{\bf j}.\end{aligned}\ ] ] the charge current density operator is @xmath112 .\end{aligned}\ ] ] taking the expectation value of eq . ( [ eq : continuity ] ) gives simply the continuity equation for the charge density , as is required by charge conservation . the time rate of change of the charge density in some volume is given by the net flux of electrons into that volume . finding the time evolution of the spin density can be done using the same method , but this exercise is somewhat more complicated because the spin density is a vector and there are additional terms in the hamiltonian beside the kinetic energy with which it does not commute . if we ignore spin - orbit coupling we get @xmath113 where @xmath114 is the tensor spin current density operator @xmath115 , \end{aligned}\ ] ] and the dot product in eq . ( [ eq : qdef ] ) connects to the spatial index of the spin current . similar to the contribution in the continuity equation , there is a contribution to the time rate of change in the spin density due to the net flux of spins in and out of a volume ( the term involving @xmath114 ) . in addition , the spin density precesses in the local fields @xmath116 . there is no contribution from the local exchange field because ( at least in the lsda mean field theory ) it is exactly aligned with the expectation value of the local spin density . the cross product between these two quantities is then identically zero . comparing eq . ( [ eq : qdef ] ) to eqs . ( [ eq : mumagh ] ) and ( [ eq : llg ] ) , the astute reader will notice some differences . ( [ eq : qdef ] ) does not include a contribution from the magnetocrystalline anisotropy , but this is just because for the present we are ignoring the spin - orbit coupling . more importantly , eq . ( [ eq : qdef ] ) appears not to include any term to account for micromagnetic exchange . the explanation for this is that there are actually two contributions to the spin current density @xmath29 and its divergence within a ferromagnet having a spatially non - uniform magnetization . the first is the non - equilibrium spin current that flows with an applied bias and which is the contribution of interest in this series of articles . a second contribution is present in the absence of any applied bias whenever the magnetization is non - collinear . this contribution can be viewed as the mediator of the micromagnetic exchange interaction in analogy to slonczewski s calculation @xcite of the exchange coupling across a tunnel barrier . in general in the discussion of spin transfer torques , and in particular in the rest of this article and the accompanying articles , this contribution is taken into account by including explicitly in the equation of motion a micromagnetic exchange contribution in the form @xmath117 { \bf s } \times \nabla^2 { \bf m}$ ] , so that the spin current contribution then describes only the non - equilibrium component . we expect that the equations of motion become even more interesting if one were to include spin - orbit coupling in the hamiltonian . first , there are additional terms in the equation of motion of the spin density , eq . ( [ eq : qdef ] ) , because the contribution of spin - orbit coupling to the hamiltonian will not commute with the spin density operator . one new term generated by the spin - orbit coupling is straightforward : the magnetocrystalline anisotropy gives a contribution @xmath118 . the damping term in eq . ( [ eq : llg ] ) emerges as well , when a coupling to a source of energy and angular momentum is also included . the article by tserkovnyak , brataas , and bauer describes the derivation of such terms . however , when the orbital angular momentum in a ferromagnet is appreciable , one should recognize that the quantity of primary interest in the landau - lifshitz - gilbert equation is the magnetization rather than the spin density , and these quantities need no longer be simply proportional to each other . the magnetization operator is @xmath119 , where @xmath120 is the orbital angular momentum density operator . when taking the commutator of @xmath121 with the hamiltonian , the spin part of the magnetization will generate the same divergence of the spin current written in eq . ( [ eq : qdef ] ) but there will be a large number of additional terms due to spin - orbit coupling and @xmath122 . these complications due to spin - orbit coupling are likely to play an important role in the dynamics of ferromagnetic semiconductors discussed in the article by ohno and dietl , because spin - orbit coupling is much more significant in these materials than in transition metal ferromagnets . as the spin - orbit coupling is comparable to the exchange splitting in the ferromagnetic semiconductors , the band structure does not divide cleanly into majority and minority bands . this leads to additional complications in calculating transport properties @xcite , even beyond the complications discussed above . in these materials , it is not even clear that the spin current is the most appropriate current to consider . this question is related to the issues of interest in the study of the spin hall effect , see @xcite for a review . complications from spin - orbit coupling are also likely to be amplified in ferrimagnetic samples as discussed in the articles by haney , duine , nez , and macdonald and by sun and ralph . _ device geometries . _ for understanding the behavior of spin - torque devices , the simplest geometry to consider consists of two magnetic layers separated by a thin non - magnetic spacer layer . one magnetic layer serves to spin - polarize a current flowing perpendicular to the layer interface ( this spin filtering can occur either in transmission or reflection ) , and then this spin - polarized current can transfer angular momentum to the other magnetic layer to excite magnetic dynamics . the spacer layer can either be a non - magnetic metal or a tunnel junction . in order that magnetic dynamics are excited in one magnetic layer but not both , typically devices are designed to hold the magnetization in one magnetic layer ( the `` fixed '' or `` pinned '' layer ) approximately stationary at least for low currents . this is done either by making this layer much thicker than the other , so that it is more difficult to excite by spin torque , or by fabricating it in contact with an antiferromagnetic layer , which produces an effective field ( `` exchange bias '' ) and increases the damping , which both act to keep the ferromagnetic layer pinned in place . the strength of the spin torque acting on the thin `` free layer '' can be increased by sandwiching it between two different pinned magnetic layers ( with moments oriented antiparallel ) , so that a spin - polarized current is incident onto the free layer from both sides , @xcite . spin transfer devices must be fabricated with relatively small lateral cross sections , less than about 250 nm in diameter for typical materials , in order that the spin torque effect dominates over the oersted field produced by the flowing current @xcite . ( the oersted field is often not negligible even in samples for which the spin torque effect dominates it can be included in micromagnetic simulations of the magnetic dynamics as an additional contribution to @xmath123 in the landau - lifshitz - gilbert equation as described in the article by berkov and miltat . ) small device sizes are also convenient for understanding spin - torque - generated magnetic dynamics because the macrospin approximation can become a reasonable approximation , and of course small devices are desired for many applications . two general experimental approaches are used to direct current flow through an area @xmath124 nm wide in a magnetic multilayer . one is to make electrical contact to an extended multilayer substrate in a `` point contact '' geometry , made either mechanically with a very sharp tip @xcite or using lithography techniques @xcite . the other approach is to make `` nanopillar '' devices in which at least the free magnetic layer and sometimes both the free and the fixed magnetic layers are patterned to a desired cross section . see fig . [ fig : geom ] for a schematic illustration of both geometries . nanopillars can be made by electron - beam lithography and ion milling @xcite , stencil techniques @xcite , or electrodeposition of layers inside cylindrical nanopores @xcite . samples in which the free layer is patterned but the fixed layer is left as an extended film are convenient for some experiments because this geometry minimizes the magnetostatic coupling between the magnetic layers . in general , point contact devices require much more current density to excite magnetic dynamics with spin torque , because the excitations must reorient a small region in an extended magnetic film , working against strong micromagnetic exchange . typical critical current densities for excitations in point contact devices are @xmath125 a/@xmath126 to @xmath127 a/@xmath126 , while nanopillar devices have achieved current densities of @xmath128 a/@xmath126 . as far as we are aware , the record low intrinsic ( extrapolated to @xmath129 ) critical current in a nanopillar device with a magnetic free layer that is thermally stable at room temperature is 1.1 @xmath130 10@xmath131 a/@xmath126 @xcite . two of the nice features of nanopillars are that they can be made to have two different stable magnetic configurations at zero applied magnetic field for memory applications , and recently they have enabled direct x ray imaging of current - driven magnetic dynamics @xcite . point contacts have an advantage in that they give narrower linewidths ( @xmath132 ) as a function of frequency @xmath133 when used to make spin - transfer nano - oscillators , as described below . both point contacts @xcite and nanopillars @xcite have also been used to study samples with a single ferromagnetic layer . the article by katine and fullerton describes some of the strategies which can be employed to decrease the critical current density for magnetic switching in nanopillars . the article by silva and rippard discusses research on point contact devices . _ methods for calculating spin torques . _ from a theoretical point of view , a consensus has developed @xcite for the general approach to be used for calculating spin torques in metal multilayers . we note however , that there is not universal agreement that this model is correct @xcite . to solve for the spin transfer torque acting on a static magnetic configuration , the consensus view is that one should determine the spin current density @xmath29 through an analysis of the spin - dependent electron transport in the device structure , and then identify the torque from the divergences of @xmath29 near magnetic interfaces or in regions of non - uniform magnetization . ( as we noted above , this assumes that no angular momentum is lost to the excitation of short - wavelength spin wave modes or to other processes . ) to calculate the spin torque on a moving magnetic configuration , a somewhat more sophisticated procedure is in principle necessary , for a rigorous treatment , in order to take into account the `` spin - pumping '' effect @xcite . spin pumping refers to the fact that a precessing magnetization can produce a non - zero spin current density in a magnetic device , in addition to the spin current density generated directly by any applied bias . the total spin torque on the moving magnetization should be determined from the divergence of the total spin current arising from both sources . however , typically the primary effect of spin pumping is merely to increase the effective damping of the ferromagnet in a way that is independent of bias ( although it may depend strongly on the precession angle @xcite ) , so that spin pumping can be taken into account approximately just by renormalizing the damping constant . then the spin transfer torque even on a moving magnetization configuration can be determined , approximately , from the divergence of just the bias - dependent part of the spin current density . once the spin torque is calculated , the response of the magnetization is generally determined by inserting this torque as an additional contribution to the classical landau - lifshitz gilbert equation of motion ( eq . ( [ eq : llg ] ) ) . the article by haney , duine , nez , and macdonald describes this consensus approach as the `` bookkeeping theory . '' spin pumping can also be incorporated into fully self - consistent solutions without too much extra difficulty , because spin pumping is local in time , depending only on @xmath101 and @xmath134 , and it can be calculated within the same type of theoretical framework needed for calculating the direct bias - generated spin torque @xcite . depending on the amount of disorder in the device and other details , there are a number of strategies for computing the spin current density , some of which are compared in ref . @xcite . if scattering within a layer is weak enough that most electrons do not scatter except at interfaces , the transport is called ballistic . the opposite limit , in which most electrons scatter several times while traversing a layer , is called diffusive . calculations for particular devices can be complicated by the fact that some layers may be in the diffusive limit while others are in the ballistic limit and some layers are in between . the ballistic regime can be treated by constructing the scattering states of the system @xcite , by the keldysh formalism @xcite or by non - equilibrium green s functions @xcite . references @xcite and @xcite have considered some of the potential effects of quantum coherence in the ballistic limit , but these effects are generally not expected to be important in the types of metallic multilayer samples typically studied experimentally , because their interfaces are not sufficiently abrupt and perfect . in principle , scattering - state formalisms , the keldysh method , and non - equilibrium green s functions techniques can all be extended into the diffusive regime , but the strong scattering in the diffusive regime is generally easier to treat using the semiclassical approaches discussed next . the boltzmann equation @xcite neglects coherent effects , but is accurate for both ballistic and diffusive samples , and can interpolate between these limits . in the boltzmann - equation approach , the transport is described in terms of a semiclassical distribution function so that the behavior of electrons on different parts of the fermi surface are tracked separately . other calculation strategies are based on approximations that sum over this distribution function and compute the transport in terms of just its moments . these include the drift - diffusion approximation @xcite and circuit theory @xcite . for the non - collinear magnetic configurations considered here , these approximations are extensions of valet - fert theory @xcite , which is widely used to describe gmr for collinear magnetizations . in all of these approaches , the transport across the interfaces is described in terms of spin - dependent transmission and reflection amplitudes . for electron spins that are collinear with the magnetization , these processes are incorporated into the transport calculations as boundary conditions between the solutions of the transport equations in each layer , and the results are spin - dependent interface resistances @xcite ( or conductances ) . for spins that are not collinear , transmission and reflection give rise both to boundary conditions on the transport calculations and also give the spin transfer torque . in the drift - diffusion approach , the non - collinear boundary conditions are that the transverse spin current is proportional to the transverse spin accumulation @xmath135 where @xmath14 is the interface normal , @xmath136 is a characteristic velocity , @xmath137 is the transverse spin accumulation , and the @xmath138 superscript indicates that the transverse spin density and spin current are evaluated in the non - magnet as they are both zero in the ferromagnet . this boundary condition has a straightforward interpretation . since the interface acts as an absorber of any transverse spin component that scatters from it , there is no out - going spin current to cancel the incoming spin current , so there must be a net accumulation of transverse spins in the non - magnet near the interface . since the incident transverse spin current is equal to the spin transfer torque it is also the case that the transverse spin accumulation is proportional to the spin transfer torque . the relation leads some authors to discuss separate spin current and spin accumulation mechanisms for the spin transfer torque . however , from this discussion , we see that the two are intimately related , and the spin transfer torque can be fully accounted for in terms of the spin current ; there is not an extra separate contribution from spin accumulation . for the case of a symmetric two - magnetic - layer device with a metal spacer , slonczewski @xcite calculated the spin transfer torque using a simplified boltzmann equation grafted with circuit theory . he found that the torque on the free layer magnetization @xmath101 due to the misalignment with fixed layer magnetization @xmath139 can be described by adding to the landau - lifshitz - gilbert equation ( eq . ( [ eq : llg ] ) ) a term of the form @xmath140 where @xmath141 is the current , @xmath98 is the free - layer volume on which the spin torque acts , @xmath142 , @xmath121 and @xmath143 are unit vectors in the directions of @xmath101 and @xmath139 ( not operators ) , and @xmath144 . all of the details of the layer structure are buried in the constants @xmath145 , @xmath60 , and @xmath146 . very similar results have also been found by a variety of other theoretical approaches . below , we will sometimes refer loosely to @xmath147 as a `` torque '' , even though strictly speaking its units are _ ( magnetic moment)/(volume _ @xmath148 _ time ) _ rather than _ ( angular momentum)/time_. we note that the direction of @xmath147 indicated by eq . ( [ eq : nst ] ) is exactly what is expected from the simple picture of eq . ( [ eq : intabsorb ] ) , based on the approximately complete absorption of the transverse spin current by the magnetic free layer . when the current has the sign that electrons flow from the fixed layer to the free layer in a multilayer like co / cu / co , the electron spin moment incident on the free layer is in the same direction as @xmath139 and the double cross product in eq . ( [ eq : nst ] ) represents just the transverse component . when the current is reversed , it is the electrons reflected from the fixed layer that apply a torque to the free layer ; their moments are on average oriented antiparallel to @xmath139 on account of the reflection , and therefore the transverse component of spin current incident on the free layer changes sign . subsequent calculations @xcite have generalized eq . ( [ eq : nst ] ) for asymmetric structures . ref . @xcite compares these simple forms to full calculations using the boltzmann equation and shows that they agree for typical layer thicknesses but break down in some limits . that paper also shows where a drift - diffusion approximation fails to reproduce the results of the boltzmann equation calculations . _ spin - transfer - driven magnetic dynamics . _ the qualitative types of magnetic dynamics that can be excited by @xmath147 with the form given by eq . ( [ eq : nst ] ) can be understood using the diagrams shown in fig . [ fig : precess ] and fig . [ fig : traject ] . we first consider the simplest possible geometry , in which the free layer magnetization @xmath101 is assumed to move as one macrospin and the magnetic - field direction and the fixed layer moment @xmath139 both point along @xmath68 . we also assume , initially , that there is no magnetic anisotropy . we will consider the problem in terms of a linear stability analysis , to see , when the free layer moment is initially perturbed slightly from the field direction , whether it returns to rest or whether a current can destabilize it to generate large - angle dynamics . this analysis can be achieved using the landau - lifshitz - gilbert equation of motion ( eq . ( [ eq : llg ] ) ) . in the absence of any spin transfer torque or damping , if the free layer moment @xmath101 is instantaneously tilted away from @xmath47 then it will precess in a circle , due to the torque from the applied magnetic field . ( for real thin - film devices , magnetic anisotropies generally cause the precessional motion to trace out an ellipse . ) if there is damping in addition to an applied magnetic field ( but still no applied current ) , the torque due to damping will push @xmath101 back toward the low - energy configuration along @xmath47 . consequently , if @xmath17 is perturbed away from @xmath47 , then at @xmath149 it will precess with gradually decreasing precession angle back toward @xmath47 along a spiral path . this type of magnetic trajectory is depicted in fig . [ fig : traject](b ) . when a current is applied , the direction of the spin transfer torque predicted by eq . ( [ eq : nst ] ) is either parallel to the damping torque or antiparallel to it , depending on the sign of the current ( see fig . [ fig : precess ] ) . ( for the more realistic case of elliptical precession in the presence of magnetic anisotropies , the instantaneous orientations of the spin torque and the damping are not always collinear , but on average over each cycle the spin torque can still be understood as either reinforcing or acting opposite to the damping . ) for the sign of the current that produces a spin - torque contribution @xmath147 in the same direction as the damping , there are no current - induced instabilities in the free - layer orientation . the current increases the value of the effective damping , and @xmath101 simply spirals more rapidly back to the @xmath47 direction after any perturbation . for small currents of the opposite sign , such that @xmath147 is opposite to the damping but weaker in magnitude , the spin torque just decreases the effective damping , and again nothing exciting happens , at least at zero temperature . ( effects of thermal fluctuations are discussed below . ) when @xmath147 is opposite to the damping torque and of greater magnitude , then following any small perturbation @xmath101 will spiral away from the low - energy configuration along @xmath47 to increasing angles the current destabilizes the orientation with @xmath101 parallel to @xmath139 and may excite large - angle dynamics . in effect , a sufficiently large current drives the damping to be negative , which leads to the amplification of any deviations of @xmath101 from equilibrium . past this point of instability , the large - angle dynamics excited by spin transfer can fall in two broad classes within the macrospin approximation , depending on the angular dependencies of the spin transfer torque , the damping torque , and the magnetic anisotropy . one possibility , which may occur if the damping torque increases with precession angle faster than the spin torque , is that the initial increase in precession angle may eventually be limited , so that @xmath101 may achieve a state of dynamical equilibrium , precessing continuously at some fixed average angle in response to the direct current ( see fig . [ fig : traject](c ) ) . in this state , the energy gained from the spin torque during each cycle of precession is balanced by the energy lost to damping . the second possible class of spin - torque - driven magnetic dynamics is that the precession angle may be excited to ever - increasing values until eventually it reaches 180@xmath20 , meaning that @xmath101 is reversed ( see fig . [ fig : traject](d ) ) . in real samples , the magnetic anisotropy can generally not be ignored , as we have done so far in the discussion above . the typical sample formed as part of a thin - film magnetic multilayer will have biaxial magnetic anisotropy , consisting of a strong in - plane anisotropy and a weaker uniaxial anisotropy along one of the in - plane axes . in fig . [ fig : traject](e)-(g ) we consider some of the magnetic trajectories that can be excited by spin transfer for this type of sample , where the easy magnetic axes are assumed to lie along the @xmath150 directions , and @xmath139 and the applied magnetic field are also assumed to point along @xmath151 instead of the @xmath152 direction as in fig . [ fig : traject](a)-(d ) . within the macrospin approximation , the dynamics of @xmath101 excited by the spin transfer torque can still be calculated by integrating the landau - lifshitz - gilbert equation ( eq . ( [ eq : llg ] ) ) with the spin - transfer - torque term eq . ( [ eq : nst ] ) included . different behavior occurs in different regimes . consider first in - plane magnetic fields less than the coercive field needed to produce magnetic - field - induced switching . at low temperature , as the current increases , there is generally first a critical current which leads to states of dynamical equilibrium in which @xmath101 undergoes steady - state precession along an approximately elliptical trajectory ( see fig . [ fig : traject](g ) ) @xcite . for slightly larger currents , the precessing state becomes unstable , and the precession angle for @xmath101 increases until it reaches 180@xmath20 , thereby achieving switching to the state with @xmath17 antiparallel to @xmath153 ( fig . [ fig : traject](h ) ) . for applied magnetic fields larger than the coercive field , usually only steady - state precessional dynamics are observed ( fig . [ fig : traject](g ) ) . additional static and dynamic magnetic states may occur at even larger values of current ( see below ) . if one starts with the free - layer moment @xmath101 oriented antiparallel to the fixed layer , rather than parallel ( _ i.e. _ , with @xmath101 in the @xmath154 direction rather than @xmath155 ) , a reversed sign of current is required to produce a negative effective damping , and this can excite steady - state dynamics or switch @xmath101 back to the parallel orientation . for thin - film samples with the magnetizations of both the fixed and free layers oriented in plane and with their easy axes aligned with the applied magnetic field , it is possible to estimate the threshold current for small - angle excitations in a macrospin picture . the estimation proceeds by integrating over the elliptic orbit for a small precession angle to determine both the energy lost due to damping and the energy gained from the spin torque during each cycle , and determining at what value of current the spin torque overcomes the damping . this leads , for example , to the critical current expression for excitations from an initially parallel magnetic orientation ( @xmath66 ) @xmath156 referred to by the article of katine and fullerton . here @xmath25 is the gilbert damping , @xmath157 is the applied magnetic field and @xmath158 the strength of the within - plane magnetic anisotropy @xcite . the saturation magnetization @xmath11 appears in the last factor because of the large thin - film demagnetization effect which favors an in - plane orientation for @xmath17 ; @xmath159 is typically much larger than @xmath157 or @xmath158 . it is important to note that this threshold describes only the first instability of the free layer to small angle precession . there is a separate threshold for magnetic switching at slightly larger currents , which has a different dependence on magnetic field . approximate analytical expressions been derived to describe the switching threshold and other boundaries between the different dynamical states that can be driven by spin torques @xcite . these phase boundaries are illustrated in fig . [ fig : pd ] for the case of a magnetic field applied in - plane along the easy axis of the magnetic free layer . one common result of all semiclassical calculations of transport and spin transfer torque for metallic multilayers @xcite is that they predict a fairly substantial asymmetry between the differential torque near parallel and antiparallel alignment . that is , @xmath160 in other words , @xmath146 is comparable to @xmath60 in @xmath161 . this asymmetry arises from the different amounts of spin accumulation for alignments close to parallel as compared to those close to antiparallel . the results suggest that the critical currents for switching out of the antiparallel state in metallic multilayers should be significantly less than those for switching out of the parallel state . however , typically the asymmetry found in the measured critical currents is much less than expected @xcite . the degree of asymmetry seems to depend on the sample structure ( _ e.g. _ , on the amount of taper in nanopillar sidewalls ) , so the discrepancy could result in part from deviations from a completely uniform magnetization state , which might affect either the mechanism of reversal or the degree of spin accumulation . on the other hand , measurements of noise instabilities for magnetizations close to perpendicular @xcite suggest an asymmetry consistent with the theoretical expectations . if @xmath139 and @xmath101 are misaligned within the sample plane by an applied magnetic field or an exchange bias , the critical currents will depend on the angle @xmath162 between the equilibrium orientations of @xmath101 and @xmath153 . this can be understood in a simple way by taylor - expanding the instantaneous value of the spin torque in the sample plane : @xmath163 ( \theta - \theta_0)$ ] . for current pulses much larger than the threshold needed to produce excitations , the first term can dominate the fast magnetic dynamics that are produced . however , for currents near the critical current this term will generally just cause the equilibrium angle to shift by a small amount as a function of @xmath141 , because it can be opposed by strong anisotropy forces . near the excitation threshold , it is actually the second term of the taylor expansion which governs the stability of the static state , because it determines whether the spin torque has the sign to increase or decrease the magnitude of deviations from the equilibrium angle . dynamical excitations will occur at zero temperature when @xmath164 has the correct sign and becomes sufficiently large to overcome the intrinsic damping , so as to cause the free - layer moment to spiral away from its equilibrium orientation with increasing precession angle , as discussed above . in a macrospin picture @xmath165 \approx 1/\cos(\theta_0)$ ] @xcite . an additional subtlety that can complicate understanding spin - torque - driven magnetic dynamics is that the magnetization in ferromagnetic devices is never completely spatially uniform , but even in thin - film devices with very small cross section there is some spatial dependence to the magnetization within the layer . this has the consequence that even in devices with electric current flowing strictly perpendicular to the plane the spin current density will have a non - zero spatial component within the plane . these lateral spin currents can produce additional instabilities that lead to the excitation of spatially non - uniform magnetization dynamics within a magnetic layer @xcite , and in fact these excitations have been observed in devices containing only a single magnetic layer @xcite and standard two - magnetic - layer devices @xcite . magnetic excitations that are non - uniform through the thickness of magnetic layers may also be possible @xcite . initial attempts to incorporate lateral spin transport self - consistently with spatially dependent magnetic dynamics in micromagnetic calculations of multilayer devices are under way . _ measurements of magnetic dynamics . _ the main experimental probe of the spin - torque - driven magnetization dynamics is measuring the resistance . the resistance reflects the magnetic configuration through the gmr effect . one signature that a current - driven change in magnetic configuration is due to the spin transfer effect is that it is asymmetric in the direction of the current , for the reasons explained above . in the first identification of a spin transfer effect @xcite , tsoi _ et al_. measured a peak in the differential resistance for one direction of the current and not the other . in switching devices @xcite , for large currents flowing from the fixed layer to the free layer ( electrons flowing from free to fixed ) , the free layer magnetization is driven to be antiparallel to the fixed layer , resulting in the high resistance state , while the opposite current drives the sample to the parallel , low - resistance state . the sign of the torque observed experimentally agrees with the sign predicted by theory . a sample excited into a state of dc - driven magnetic precession naturally emits a substantial microwave - frequency signal at its electrical contacts . the resistance is changing at microwave frequencies and a dc current is applied , so by ohm s law a microwave voltage is generated . this can be detected either in the frequency regime @xcite or directly in time - domain measurements @xcite . this mode of steady - state precession is of interest for applications which might benefit from a nanoscale oscillator or microwave source that can have a narrow linewidth and is tunable in frequency , see fig . [ fig : spectra ] . the phase diagram of the spin - transfer - driven dynamics as a function of current , magnetic field , and the direction of magnetic field contains several distinct types of precessional modes and static magnetic states @xcite . when one crosses from one mode to another by varying current or field , the frequency of precession can jump , the microwave power output can increase or decrease dramatically , the linewidth can change , and the resistance measured by standard low - frequency techniques generally changes by a small amount , too . when a device is biased near the boundary between two different modes , it often exhibits time - dependent switching behavior between them @xcite . the dynamical phase diagram has been mapped in some detail for both in - plane and perpendicular magnetic fields in nanopillar devices @xcite , and it shows surprisingly good agreement even with the simplest macrospin models , with some exceptions at large currents and for devices in which the magnetic configuration begins in a vortex state @xcite . in addition to studies in which the equilibrium orientation of the moments is in the plane , devices have recently been produced in which the magnetic anisotropy of one or more of the magnetic layers is manipulated so that the magnetization points out of plane @xcite . such devices may enable ways to decrease the critical current for switching in memory devices , to increase the switching speed @xcite , or to produce improved nano - oscillators . frequency- and time - domain measurements of the different modes of spin - transfer - driven dynamics provide very fertile ground for comparisons with sophisticated micromagnetic simulations of the dynamics , as reported by berkov and miltat in their article . an alternate measurement approach that has recently been exploited is to apply a high frequency input current to drive resonant magnetic precession and look at the static output voltage generated by mixing , a technique referred to as spin - transfer ferromagnetic resonance ( st - fmr ) @xcite . these measurements , as applied to magnetic tunnel junctions , are discussed in the article by sun and ralph . _ effects of non - zero temperature . _ spin - torque devices have sufficiently small sizes that temperature - induced fluctuations can have important effects on the magnetic dynamics . at room temperature , the reduced effective damping that can be produced by spin transfer torque can enhance the amplitude of thermal fluctuations even for currents well below the @xmath129 critical current . these enhanced thermal magnetic oscillations can be measured accurately in tunnel junctions @xcite . in both tunnel junctions and metallic devices , room temperature thermal fluctuations can also cause magnetic switching to occur at currents that are well below the intrinsic zero - temperature threshold current for spin - transfer - driven instabilities ( eq . ( [ eq : criti ] ) ) . in order to determine the intrinsic critical current for switching , it is necessary to make direct measurements of switching on the typical time scale of ferromagnetic precession near 1 ns , or measure the switching current as a function of temperature and extrapolate to zero temperature , or measure the statistics of switching using current pulses of various lengths and extrapolate to the ns scale . the linewidths of the microwave signals generated by spin - torque - driven magnetic precession appear to be governed largely by thermal fluctuations that produce deviations from perfectly periodic motion @xcite . point contact devices , in which magnetic precession is excited in a local area of an extended magnetic film can produce narrower linewidths than typical nanopillar devices @xcite , perhaps because the micromagnetic exchange coupling to the extended film makes them less susceptible to thermal fluctuations . if the effects of thermal fluctuations can be reduced or eliminated , it is likely that the ultimate limit on the linewidths of spin - torque nano - oscillators will be chaotic dynamics of the magnetization @xcite . in terms of theory , the role of temperature is an area that is still under development . transport calculations and determinations of the torque are typically done assuming zero temperature . this approximation is expected to be reasonable because not much about the transport is expected to change with temperature except for scattering rates . there have been a number of theoretical studies of the temperature dependence based on the macrospin approximation @xcite . however , as the temperature increases , the macrospin approximation becomes worse . it is possible to include thermal effects into full micromagnetic simulations , but the calculations become quite time consuming and it is very difficult to capture meaningful statistics . _ on the perpendicular component of the spin torque vector . _ in discussing the result of eq . ( [ eq : intabsorb ] ) , _ i.e_. that one expects essentially all of the transverse spin angular momentum incident onto a normal - metal / ferromagnet interface to be absorbed in producing the spin torque , we made the point that this was only approximately correct ; it is not exact . the most important caveat is that classical averaging over the fermi surface need not necessarily eliminate all transport of transverse spin density away from the interface , particularly for the reflected part of the scattering wavefunction ( or for very thin magnetic layers in transmission ) . one consequence is that the amplitude of the `` in - plane '' component of the torque can differ somewhat from eq . ( [ eq : intabsorb ] ) . in addition , there is the possibility of an additional contribution due to the spin torque in the form @xmath166 oriented perpendicular to the plane defined by @xmath101 and @xmath153 , rather than within this plane as in eq . ( [ eq : nst ] ) . note that here the symbol @xmath167 refers to the direction perpendicular to the plane of the magnetizations as opposed to the usage in eq . ( [ eq : intabsorb ] ) where it refers to both components of the spin current transverse to the free layer magnetization . in fig . [ fig : precess ] , this new component of torque would point in or out of the page . the out - of - plane torque component is sometimes referred to as an `` effective field '' contribution because its form is similar to the torque that would result from a field aligned with the fixed layer magnetization . it can be viewed as a consequence of a small amount of average precession into the @xmath69 direction for reflected electrons in our toy models from section [ sec : stt ] . calculations incorporating transmission and reflection coefficients computed by _ ab initio _ techniques find that the perpendicular component of the torque is small for the materials generally used in metallic multilayer devices . for electrons of a given energy , the out - of - plane component of spin transfer is predicted to be less than 10 % , and typically 1 % to 3 % of the in - plane component @xcite . the final magnitude of the bias - dependent part of the out - of - plane torque is expected to be smaller still , due to a cancellation that arises when computing the bias dependence . this cancellation is maximal in the case of a symmetric n / f / n / f / n junction , where it can be understood from a simple symmetry argument . consider a junction whose layer structure is perfectly symmetric about a plane at the midpoint of the device , see fig . [ fig : geomperp ] . assume that the magnetic moments of the two ferromagnetic layers are oriented in the plane of the sample layers , with an arbitrary angle @xmath48 between them . when a bias is applied to the junction , the resulting spin current density in each lead ( each of the outer n layers ) will be aligned with the magnetization of the neighboring ferromagnetic layer provided that the electrons that transmit into the lead from the rest of the devices are completely classically dephased . in this case , the spin currents in the outer n layers are oriented in the plane of the two magnetizations , and an out - of - plane spin component of the spin current density can exist only in the middle n layer . if there are no sources or sinks of angular momentum except for the spin torques applied to the magnetic layers ( _ e.g. _ , we assume that there is no angular momentum lost to the excitation of short - wavelength spin wave modes ) , then angular momentum conservation allows the perpendicular component of spin torques on the ferromagnetic layers to be determined solely from the out - of - plane spin current traveling between the magnetic layers . as a consequence , the out - of - plane spin torques on the two magnetic layers must be equal and opposite . ( to be clear , our argument applies only to the perpendicular component of the spin torque and not the in - plane component , because the in - plane spin component of the spin - current density is not zero in the outer n layers , and hence the in - plane spin torques on the ferromagnetic layers are not generally equal and opposite . ) for an applied bias such that electrons flow from left to right , suppose that the perpendicular spin torque on the right magnetic layer has the value @xmath168 . now imagine that the bias is reversed , so that electrons flow from right to left . because of the symmetry of the device , the spin current density in this case can be determined from the first case simply by an appropriate rotation ; the answer is that both the direction of the out - of - plane spin component that flows between the two layers and the direction of its flow are reversed , so that ( multiplying the two negative signs ) the out - of - plane spin current density is actually _ the same _ as in the first case . consequently , the perpendicular spin torque acting on the right magnetic layer due to electrons flowing right to left will again be @xmath168 , the same magnitude and the same sign as for the perpendicular torque on the same ( right ) magnetic layer due electron flow from left to right . it follows that the perpendicular spin torque on a magnetic layer in a perfectly - symmetric junction is an even function of the bias @xmath169 , as long as conservation of angular momentum can be applied to the spin - transfer process and the classical dephasing processes are complete . consequently , there is no contribution that is linear in @xmath169 at zero bias in a symmetric junction , in spite of the fact that _ ab initio _ calculations show that electrons at a given energy incident onto a magnetic interface can give non - zero contributions to the perpendicular torque . in a more microscopic picture , one can see how this result comes about by considering a calculation in which the total perpendicular spin torque is determined by integrating the contributions from all electrons incident onto the symmetric junction from both sides . following the logic of the symmetry argument above , the perpendicular torque due to incoming electrons in the energy range @xmath170 incident from the left should be the same , @xmath171 , as for electrons of the same energy incident from the right , as long as we assume that the effect of the applied bias on the scattering potential is sufficiently weak that the electron transmission and reflection amplitudes do not depend explicitly on @xmath169 . for a symmetric junction , any applied voltage bias will offset the electron chemical potential on the two sides of the device by @xmath172 . therefore , if we consider only elastic scattering processes , the bias - dependent part of the perpendicular torque can be calculated by subtracting the contributions of electrons in the energy range @xmath173 to @xmath174 that are no longer incident on the junction from one side , and adding the contribution of the extra electrons in the range @xmath174 to @xmath175 that are incident from the other side : @xmath176 the contributions from the states in the energy ranges @xmath173 to @xmath174 and @xmath174 to @xmath175 cancel to first order in @xmath177 , so that the lowest - order contribution to the perpendicular spin torque should go only as @xmath178 , symmetric in @xmath169 with no linear term . even at non - zero biases the perpendicular component of the spin torque is probably very small in most metallic multilayers . predictions for non - zero values of the perpendicular torque typically arise only under the assumption of coherent transport between ideal interfaces , but in real devices this coherence is generally not present due to interface disorder @xcite . for tunnel junctions , the bias - dependent part of the out - of - plane spin torque is predicted to be larger than for metallic multilayers @xcite . it remains an open question as to how well our symmetry argument applies to tunnel junctions , because the current may be dominated by a small part of the fermi surface . in such systems , classical dephasing may not be as complete and therefore in principle a small linear - in-@xmath169 out - of - plane torque may remain even for a symmetric junction . in addition , for tunnel junctions at high bias , angular momentum loss from hot electrons to the excitation of short - wavelength spin waves might become significant , which would invalidate our arguments because they are based on angular momentum conservation between the conduction electrons and the approximately uniform magnetization mode . nevertheless , a tight - binding calculation designed to model a symmetric magnetic tunnel junction ( in the absence of any short wavelength spin - wave excitations ) did predict a bias dependence @xmath179 @xcite , consistent with the symmetry argument . for a tunnel junction device with a non - symmetric layer structure none of the symmetry arguments apply , and in this case one should expect that there may be a perpendicular spin torque with a linear dependence on @xmath169 near zero bias . the potential absence of complete classical dephasing in magnetic tunnel junctions , the consequences of spin - wave excitation by hot electrons , and the effects of non - symmetric layer structures are all interesting questions for future theoretical work , as is the effect of disorder on the out - of - plane torque . the experimental literature regarding the perpendicular component of spin torque contains some contradictory results . recent spin - transfer ferromagnetic resonance measurements ( st - fmr ) on metallic nife / cu / nife devices find no sign of a perpendicular torque at the level of 1 % of the in - plane torque @xcite , in agreement with the theoretical expectations . however , zimmler _ et al_. interpreted a measurement on co / cu / co devices @xcite of a non - zero slope _ vs_. current for the critical magnetic field for switching @xmath180 at 4.2 k as demonstrating a perpendicular torque linear in @xmath169 with a magnitude about 20 % of the in - plane torque . we are skeptical of this interpretation , and suspect that the observation might be an artifact of heating . the measured slope of @xmath180 was not actually constant as a function of @xmath141 as would be expected from a perpendicular torque , but rather the slope of @xmath180 was approximately zero except for combinations of current polarity and magnetic orientation for which the spin torque decreases the effective magnetic damping . this asymmetric rounding of the critical field line can be viewed as a signature of thermally induced fluctuations @xcite , in that a decreased effective damping can increase the amplitude of thermal fluctuations and therefore decrease the measured switching field at non - zero temperature . et al_. attempted to correct for this temperature effect in their analysis , but did not consider that the temperature might be varying as a function of current due to heating at 4.2 k. spin - transfer fmr measurements on symmetric magnetic tunnel junctions differ from experiments on magnetic multilayers in that the perpendicular torques found in the tunnel junctions are significant @xcite . in agreement with our symmetry argument and with more detailed calculations @xcite , these experiments find that to good accuracy @xmath179 at low bias for symmetric tunnel junctions , with no contribution to the perpendicular torque that is linear in @xmath169 . an earlier report by tulapurkar _ et al_. @xcite of a perpendicular torque linear in bias near @xmath181 is now believed to be incorrect ; the same group has said more recently that this observation may have been an artifact due to a spatially non - uniform magnetic state @xcite . the article by sun and ralph describes in more detail the similarities and differences between spin torques in metal multilayers and tunnel junctions . _ comparison between theory and experiments . _ while there is a general consensus about the correct approach for calculating spin transfer torques in magnetic multilayers , there is as of yet no fully quantitative comparison between theory and experiment in metallic multilayers . the primary difficulty is that the torques themselves are generally not measured directly , but only the resulting dynamics as inferred from the time dependent resistance . spin - transfer - driven ferromagnetic resonance may eventually enable more direct quantitative measurements , but so far this has not been achieved for the in - plane component of the spin - torque in metal multilayers . as described in the article by berkov and miltat , attempting to infer the dynamics of a magnetic sample from its time dependent resistance is highly non - trivial and may hide the details of the torque . the problem is particularly complicated because it is difficult to determine simply from resistance measurements the degree to which spin - torque induced excitations are spatially non - uniform . the most sophisticated calculations can make very good qualitative predictions concerning , for example , the different types of dynamical modes ( static or precessional ) that are observed experimentally . however , they currently do not match experiment well in more quantitative comparisons , for instance in comparing to the current - dependent amplitude of the microwave signals emitted in the steady - state precession regime ( see the article by berkov and miltat ) . another complication is that there is still some variation experimentally between the behaviors of nominally identical samples . this degree of variability has improved since spin transfer effects were first measured , so that quantities like switching currents and precession frequencies now show good reproducibility . however , other quantities , _ e.g. _ microwave linewidths and the positions of transitions between different precession modes , are much more variable . the variability that still exists highlights the importance of achieving improved control over materials and lithography . in this section we provide a very brief introduction to current - induced domain wall motion . this section is brief compared to the previous section because the articles by beach , tsoi , and erskine , by tserkovnyak , brataas , and bauer , and by ohno and dietl together provide a rather comprehensive introduction , review , and summary of open questions for this topic . experimentally , current - induced domain wall motion is typically studied in lithographically defined , narrow magnetic wires . the wires are frequently curved or the ends are designed to make it possible to controllably introduce a domain wall into the wire using an applied field . in early experiments , the positions of the domain walls were imaged using techniques such as magnetic force microscopy ( mfm ) @xcite or scanning electron microscopy with polarization analysis ( sempa ) @xcite . measuring the positions before and after a current pulses allows estimates of the wall velocities . alternatively , the locations of the wall could be determined in real time using the magneto - optic kerr effect ( moke ) @xcite or electrically using gmr sandwich structures @xcite or through the extra resistance due to anisotropic magnetoresistance ( amr ) in a domain wall @xcite . using these various techniques , experimentalists can determine the wall velocity as a function of current and applied magnetic field and compare with theoretical predictions . the level of the comparison is discussed in detail in the article by beach , tsoi , and erskine . to a first approximation , understanding current - induced domain wall motion is quite simple @xcite . assuming that the electron spins adiabatically follow the magnetization direction , the divergence of the spin current in eq . ( [ eq : qdef ] ) gives a torque ( actually a torque density , but for the rest of the article we refer to it simply as a torque ) on the magnetization of @xmath182 . here , @xmath183 is a vector in the current direction with magnitude @xmath184 , where @xmath185 is the polarization of the current . if the current is uniform , this torque density simply translates the domain wall in the direction of electron flow with a speed @xmath186 . there are several factors that complicate this simple description . these factors are described in the three articles by beach , tsoi , and erskine , by tserkovnyak , brataas , and bauer , and by ohno and dietl . the degree to which the spins adiabatically follow the magnetization has been computed in several models @xcite . the deviations are small except for rather narrow domain walls . when they are non - negligible , there is an additional torque in the @xmath187 direction . this torque is referred to as a non - adiabatic torque because it derives from the inability of the electron spins to adiabatically follow the magnetization direction . much of the debate on the theoretical description of current - induced domain wall motion is associated with how to describe the damping and whether there is an additional torque in the direction of the non - adiabatic torque @xcite that arises from the same processes that lead to damping . this torque , while not related to a true non - adiabatic torque is still sometimes called a non - adiabatic torque . alternatively , sometimes it is called the beta term , because @xmath188 is the dimensionless parameter that is frequently used to characterize its strength . this contribution is extensively discussed by tserkovnyak , brataas , and bauer . one technical aspect , whether the adiabatic spin transfer torque can be derived from an energy functional , is addressed in the article by haney , duine , nez , and macdonald . one of the present authors believes that their argument is incorrect , but disagreement is one of the things that makes an issue an open question . just as the dynamics of the free magnetic layers in nanopillars are complicated by the possibility of spatially non - uniform magnetization patterns , so are the dynamics of domain wall motion complicated by non - trivial wall structures . when the domain patterns distort , the motion is no longer simple . domain walls distort in response to non - adiabatic torques , damping , and the presence of non - uniformities in the sample . the non - uniformities are frequently described as pinning centers . they have been studied extensively in the context of field driven domain wall motion as discussed by both beach , tsoi , and erskine and ohno and dietl . much experimental effort is spent on trying to minimize and understand the effects of these pinning centers . in this article , we have tried to highlight some of the open questions in the study of spin transfer torques and provide background material for the succeeding articles . as katine and fullerton describe , spin transfer torques should start having commercial impact in the very near future , but there are still important scientific issues to work through . it seems likely that tunnel junctions will be an important system for applications of spin transfer torques . sun and ralph discuss what is known about spin transfer torques in magnetic tunnel junctions as well as what we still need to learn . the dynamics that result from spin transfer torques is the topic of the articles by berkov and miltat and by silva and rippard . the former focuses primarily on the nanopillar geometry with a free layer that is patterned to have a finite extent . the latter focuses on the fabricated nanocontact geometry in which the free layer is part of an extended thin film . both show that the agreement between theory and experiment is still incomplete . the developments in current - induced domain wall motion are more recent than those in magnetic multilayers and tunnel junctions , so applications are still more uncertain . as beach , tsoi , and erskine describe , the experiments are rapidly becoming more sophisticated and more meaningful results are appearing . tserkovnyak , brataas , and bauer describe the rapid progress in the theory of spin transfer torques in these systems . it is likely that experimental progress will challenge these theories and more work will be required . most research on spin transfer torques has focused on transition metal ferromagnets , but there are other potentially interesting materials systems . ferromagnetic semiconductors are potentially exciting because they have magnetizations much smaller than transition metal magnets , so that their moments might be manipulated using spin transfer from much smaller currents than transition metal magnets . while the feasibility of a room - temperature dilute ferromagnetic semiconductor remains to be established , there are fascinating scientific questions to be understood about these materials and their interactions with current . ohno and dietl describe the measurements of current - induced domain wall motion in a dilute magnetic semiconductor and a theory that can be used to characterize the electrical and magnetic properties of these systems . haney , duine , nez , and macdonald describe a different way of computing spin transfer torques and apply it to novel systems . in particular , they raise the possibility of large spin transfer torques in antiferromagnets . we thank piet brouwer , bob buhrman , paul haney and christian heiliger for valuable conversations about this manuscript , and many other colleagues ( too many to list ) who have helped to educate us about spin transfer torques . we also thank julie borchers , arne brataas , yongtao cui , zhipan li , jabez mcclelland , jacques miltat , josh parks , vlad pribiag , kiran thadani , hsin - wei tseng , and yaroslav tserkovnyak for reading the manuscript and offering suggestions for improvement , and sufei shi for help with figures . the spin - torque work of dcr is supported by the office of naval research , the nsf ( dmr-0605742 and eec-0646547 ) , and darpa . this version of the tutorial is modified from the originally - published version . ( _ i _ ) we have corrected algebraic errors in eq . ( 13 ) and modified the discussion in the following paragraphs to correct a mistaken claim that the spin current density in the toy model can be discontinuous at the interface . we thank claire baraduc for pointing out the error . ( _ ii _ ) we have corrected an error in the units for the y axes in figures 10a and 10b . m. d. stiles , j. magn . magn . mater . * 200 * , 322 ( 1999 ) ; p. bruno , j. phys . : condens . matter * 11 * , 9403 ( 1999 ) ; d. e. brgler , p. grnberg , s. o. demokritov , and m. t. johnson , in _ handbook of magnetic materials , vol . 13 _ , ed . k. h. j. buschow ( 2001 ) ; m. d. stiles in _ ultrathin magnetic structures iii _ , j. a. c. bland and b. heinrich ( eds . ) ( springer verlag berlin 2005 ) p 99 ; m. d. stiles in , _ nanomagnetism : ultrathin films , multilayers and nanostructures _ , d. l. mills and j. a. c. bland ( eds . ) ( elsevier amsterdam 2006 ) , p. 51 . t. mcguire and r. potter , ieee trans . * 11 * , 1018 ( 1975 ) ; i. a. campbell and a. fert , transport properties of ferromagnets , in _ ferromagnetic materials , vol . 3 _ , e. p. wohlfarth ( ed . ) , north holland , amsterdam , 1982 . , j. a. c. bland and b. heinrich ( eds . ) ( springer verlag berlin 1994 ) ; _ ultrathin magnetic structures ii _ , b. heinrich and j. a. c. bland ( eds . ) ( springer verlag berlin 1994 ) ; _ ultrathin magnetic structures iii _ , j. a. c. bland and b. heinrich ( eds . ) ( springer verlag berlin 2004 ) ; _ ultrathin magnetic structures iv _ , b. heinrich and j. a. c. bland ( eds . ) ( springer verlag berlin 2005 ) . , b. hillebrands and k. ounadjela ( eds , ) ( springer verlag berlin heidelberg 2001 ) ; _ spin dynamics in confirnd magnetic structures ii _ , b. hillebrands and k. ounadjela ( eds , ) ( springer verlag berlin heidelberg 2003 ) ; _ spin dynamics in confirnd magnetic structures iii _ , b. hillebrands and a. thiaville ( eds , ) ( springer verlag berlin heidelberg 2006 ) . in discussions of ferromagnetism , the word `` exchange '' is used to mean several things , all fundamentally related but nevertheless distinct . this can be confusing . exchange can be used to describe the electron interactions that give rise to the formation of a moment . it can be used to describe the interaction between a moment and a spin , as is done in the s - d model . it can also be used to describe the interaction between moments on neighboring sites ( @xmath189 and @xmath190 ) that tends to keep the moments aligned , @xmath191 . when we speak of exchange in the context of micromagnetism , it is this last meaning to which we refer . d. houssameddine , u. ebels , b. delat , b. rodmacq , i. firastrau , f. ponthenier , m. brunet , c. thirion , j .- p . michel , l. prejbeanu - buda , m .- c . cyrille , o. redon , and b. dieny , nature mater . * 6 * , 447 ( 2007 ) . a vortex core is truly only a singularity when the magnetization is required to stay within the plane . in real vortex cores , the magnetization rotates out of the plane and the magnetization pattern is continuous everywhere . y. tserkovnyak , a. brataas , g. e. w. bauer , phys . lett . * 88 * , 117601 ( 2002);y . tserkovnyak , a. brataas , g. e. w. bauer , phys . rev . b * 66 * , 224403 ( 2002 ) ; y. tserkovnyak , a . brataas , g. e. w. bauer , b. i. halperin , rev . phys . * 77 * , 1375 ( 2005 ) . engel , e. i. rashab , and b. i. halperin , in _ handbook of magnetism and advanced magnetic materials , _ h. kronmller and s. s. p. parkin eds . ( john wiley & sons ltd . , chichester , uk , 2007 ) , p. 2858 ( arxiv : cond - mat/0603306 ) . for a thin - film spin transfer device with a circular cross section , a fixed layer structure , and current flowing perpendicularly to the plane , the threshold for spin - torque - driven excitations is given by a particular current density , so that the critical current scales @xmath192 , where @xmath193 is the device radius . reorientations produced by the current - induced oersted field will generally occur at a particular field magnitude @xmath194 , so that the threshold current for field - generated reorientations scales @xmath195 , approximately . therefore a sufficiently small device radius is needed if one desires that the threshold for spin - torque excitations should occur at a lower current than thresholds for magnetic - field - driven reorientations . j. zhang , p. m. levy , s. zhang , v. antropov , phys . lett . * 93 * , 256602 ( 2004 ) ; p. m. levy , j. zhang , phys . rev . b 70 , 132406 ( 2004 ) ; j. zhang , p. m. levy , phys . b * 71 * , 184417 ( 2005 ) ; j. zhang , p. m. levy , phys . rev . b * 71 * , 184426 ( 2005 ) . j. grollier , v. cros , a. hamzic , j. m. george , h. jaffres , a. fert , g. faini , j. ben youssef , and h. legall , appl . 78 * , 3663 ( 2001 ) ; j. grollier , v. cros , h. jaffres , a. hamzic , j. m. george , g. faini , j. ben youssef , h. le gall , and a. fert , phys . b * 67 * , 174402 ( 2003 ) ; a. fert , v. cros , j. m. george , j. grollier , h. jaffres , a. hamzic , a. vaures , g. faini , j. ben youssef , and h. le gall , j. magn . magn . mater . * 272 - 276 * , 1706 ( 2004 ) . k. m. schep , j. b. a. n. van hoof , p. j. kelly , g. e. w. bauer , and j. e. inglesfield , j. magn . magn . mater . * 177 * , 1166 ( 1998 ) ; m.d . stiles , and d.r . penn , phys . b 61 3200 ( 2000 ) ; k. xia , p. j. kelly , g. e. w. bauer , i. turek , j. kudrnovsky , v. drchal , phys . rev . b * 63 * , 064407 ( 2001 ) ; g. e. w. bauer , k. m. schep , k. xia , and p. j. kelly , j. phys . d , * 35 * , 2410 ( 2002 ) .
magnetic reconnection is a process in which magnetic energy is converted into kinetic energy as the topology of the magnetic field changes . this process can produce fast energy transfer and accelerate particles to high energies . in the relativistic reconnection regime ( see @xcite for a review ) where the ratio of the magnetic energy to the total enthalpy of the particles ( the magnetization @xmath0 ) is much larger than 1 , emission from these high - energy particles has been hypothesized to produce observed high - energy emission in various astrophysical contexts . systems whose emission may be explained by relativistic reconnection include the prompt phase of gamma - ray bursts ( grbs ) @xcite pulsar wind nebulae @xcite , and active galactic nucleus ( agn ) jets @xcite . one of the features of magnetic reconnection is that it may be able to evade a constraint on particle acceleration called the synchrotron burnoff limit @xcite . this constraint can be derived by equating the synchrotron energy loss from a particle ( which increases with the lorentz factor of a particle as @xmath1 ) , with the energy gain from electrical acceleration ( which is independent of @xmath2 ) . this produces a maximum energy reachable by synchrotron radiation from electromagnetically accelerated particles of @xmath3 where @xmath4 is the electron mass , @xmath5 is the fine - structure constant , @xmath6 is the speed of light , @xmath7 is the typical electric field that accelerates particles , and @xmath8 is the typical magnetic field should be @xmath9 because only the magnetic field perpendicular to the particle s motion is responsible for synchrotron radiation . but unless there are special geometric constraints the perpendicular magnetic field is similar to the background magnetic field . in the case of harris - like magnetic reconnection @xcite that is investigated in this paper , the motion of accelerating particles is primarily perpendicular to the reversing field . unless a very strong guide field is present , @xmath8 is therefore approximately equal to both the reversing field and the background field . ] that is responsible for synchrotron radiation . in ideal plasmas like those involved in the simplest models of shock acceleration @xmath10 , so synchrotron emission can not be produced above @xmath11 mev , the maximum synchrotron burnoff limit . yet there are many systems that appear to go above this limit . observations of gamma - ray flares in the crab nebula @xcite present a clear case of emission beyond the synchrotron burnoff limit , with peak energies of 375 mev with significant radiation up to 1 gev @xcite and there is no evidence of ultrarelativistic motion there @xcite to boost the photon energy . there are also other cases of high - energy emission which may involve synchrotron radiation . the extended emission at energies of up to 300 gev in grbs @xcite seems to require rest - frame photon energies above 160 mev , although the emission may be inverse compton radiation ( e.g. , * ? ? ? tev flares observed in agn ( e.g. , * ? ? ? * ) seem to require relativistic motion within the jet to escape scattering in the emission region @xcite , but they may also involve rest - frame synchrotron emission above the burnoff limit . in order to produce emission beyond 160 mev , it is necessary to decouple particle acceleration and cooling so that the constraint @xmath10 no longer applies . in the context of relativistic shocks , particles may accelerate in the background field but radiate in spatially varying or microturbulent fields near the shock in the synchrotron @xcite or jitter @xcite regimes . but numerical simulations indicate that particle acceleration in relativistic shocks is unlikely to be efficient enough to produce synchrotron radiation beyond 160 mev @xcite , and particles in these shocks are unlikely to emit in the jitter regime @xcite . in contrast , magnetic reconnection naturally produces emission above the maximum synchrotron burnoff limit . in the diffusion region of magnetic reconnection , where much of the particle acceleration takes place , @xmath12 and @xmath13 . significant cooling only occurs after the particle has left the acceleration region . thus , particle acceleration and cooling are naturally decoupled in magnetic reconnection . analytical studies of magnetic reconnection @xcite and a test particle study in a fixed reconnection geometry @xcite indicate that particle acceleration through this mechanism is not limited by cooling . these studies suggest that particle acceleration in magnetic reconnection can produce synchrotron radiation beyond the maximum burnoff limit . particle - in - cell ( pic ) simulations are the most common method for probing the acceleration of particles in fully dynamic magnetic reconnection . these simulations , which focus on the pair - plasma case which is easier to simulate and is directly applicable to the crab nebula , have shown that relativistic reconnection can produce hard non - thermal tails of highly accelerated particles that are typically interpreted as power laws @xcite . in general , the physics of reconnection is similar in two and three dimensions @xcite unless the initial reconnection structure is inherently three - dimensional , justifying the use of two - dimensional simulations which reduce computational cost and allow larger - scale simulations to be carried out . studies of the radiation produced by accelerated particles in reconnection that do not include the effect of cooling on particles @xcite indicate that radiation above the synchrotron burnoff limit should be highly beamed , and therefore highly variable , which is consistent with observations of fast variability in the crab flares @xcite . several researchers have used pic codes with a form of the radiation reaction force included @xcite to investigate the effects of cooling on particle acceleration in relativistic reconnection @xcite . they find that relativistic reconnection produces beamed particles and highly variable radiation above the synchrotron burnoff limit in both two @xcite and three @xcite dimensions . relativistic reconnection also produces high - energy radiation above the synchrotron burnoff limit in the presence of a guide field @xcite as long as it is weaker than the reversing field , and in reconnection produced by the collapse of a two - dimensional flux rope system @xcite . these simulations demonstrate that the production of radiation beyond the synchrotron burnoff limit is robust in relativistic magnetic reconnection . however , they are limited in dynamic range to @xmath14 decades in lorentz factor @xmath2 . the peaks of the corresponding radiation spectra are located at or below the burnoff limit , with the high - energy tail reaching above the limit by less than a factor of 10 in frequency . therefore , these simulations do not predict how far above the synchrotron burnoff limit radiation can be produced in relativistic reconnection . we overcome this problem in our simulations using the new test particle comparison ( tpc ) method . in this paper we use two - dimensional pic simulations including the radiation reaction force to reexamine the extent to which particles accelerated in relativistic reconnection can produce radiation above the synchrotron burnoff limit . in the absence of a large dynamic range in lorentz factor , which is difficult to produce from an initial thermal spectrum even in the largest simulations @xcite , we instead choose to trace particle trajectories to separate the effects of cooling during acceleration ( which limit the maximum radiation energy ) from the effects of cooling after acceleration is complete ( which do not limit it ) . but tracing particles throughout the whole simulation imposes significant computational costs , and it is nontrivial to correct for the different sizes and characteristics of dynamically changing acceleration regions throughout the system , which are expected to influence the effects of cooling on particle trajectories and acceleration . therefore , we carry out two simulations of relativistic reconnection in which the ( uncooled ) structural particles and fields that determine the evolution of the reconnection are identical , and compare the trajectories of test particles incident on a single typical reconnection region with and without cooling . this tpc method allows us to isolate the effects of cooling on accelerating particles in a fully dynamic reconnection region , because its properties are the same in both simulations . we calculate a prescription for the effects of cooling based on a particle s uncooled trajectory in the reconnection region and the acceleration it experiences there . we apply this prescription to analytically predict the effects of cooling on an uncooled power law energy spectrum produced in reconnection over a large dynamic range . we qualitatively confirm our results over a limited dynamic range by comparing the particle spectra in these uncooled simulations with the spectrum in a simulation in which structural particles are cooled . our paper is organised as follows . in section [ sec : methods ] , we discuss our simulation methods , including our treatment of the radiation reaction force and a detailed discussion of the tpc method . in section [ sec : results ] , we discuss the results , and in section [ sec : conclusions ] , we discuss the conclusions of our research . we use the particle - in - cell method to simulate reconnection . this method includes kinetic effects by discretizing the equations of electrodynamics ( the mean - field maxwell s equations and the lorentz force law ) , replacing groups of physical particles with larger macroparticles and tracking fields only on the vertices of a grid . we implement the pic simulations using the tristan - mp code @xcite , which uses current filtering to greatly reduce noise even at relatively small macroparticle densities . we initialize all of our simulations with a simple harris current sheet configuration that is susceptible to reconnection , described in section [ sec : init ] . our first two simulations have identical evolution for the structural particles ( which do not experience cooling ) and fields . after nonlinear reconnection has developed , we inject test particles that duplicate structural particles flowing into a single reconnection region in the current sheet with ( simulation tc ) and without ( simulation tu ) applying cooling to those test particles . using this tpc method allows us to isolate the effects of cooling and predict its effects in large - scale simulations . to do this , we first find a prescription for the energy loss due to cooling for an accelerating particle as a function of the particle s trajectory . this is parameterized by @xmath15 here the brackets indicate an average ( described in section [ sec : tpc ] ) over the particle trajectory in the acceleration region , @xmath16 and @xmath17 are the local electric and magnetic fields at each point in the trajectory , and @xmath7 and @xmath8 are the characteristic electric field of the reconnection region and the background magnetic field . we call @xmath18 the burrowing parameter , because it represents the extent to which the particle has burrowed into the current sheet and experienced a lower magnetic field than the background field . by finding the burrowing parameter as a function of the ( uncooled ) final energy of the test particles in the acceleration region , we can analytically predict the cooled particle energy spectrum from the uncooled spectrum , as explained in section [ sec : tpc ] . we also carry out a third simulation ( simulation ac ) in which cooling affects structural particles to evaluate the effects of cooling on the overall reconnection configuration during nonlinear reconnection and verify our analytical results . table [ tab : overview ] summarizes the properties of the three simulations . ccccc & y&562.5&n&n + tc&y&562.5&n & y + ac&n&-&y&n the spatial domain is rectangular with @xmath19 , @xmath20 , and the boundary conditions are periodic in all directions . the initial configuration contains two relativistic harris current sheet equilibria @xcite without guide field at @xmath21 and @xmath22 with equal and opposite currents . each current sheet has the magnetic field profile @xmath23 where @xmath24 is the half - thickness of each current sheet and @xmath25 is its center . the density profile of the structural particles consists of a specially varying , drifting current population with maximum density @xmath26 centered at each current sheet , plus a background population of stationary particles of density @xmath27 : @xmath28 densities are defined including both species . the resulting drift velocity of the positively and negatively charged current sheet particles is given by @xmath29 . , where @xmath30 is the charge of the macroparticle . for the magnetic field and drift velocity profiles , the sign in @xmath31 is positive for the vicinity of the current sheet at @xmath21 and negative near the current sheet at @xmath32 . the drifting and stationary populations of both species begin in relativistic maxwellians at temperature @xmath33 . the two most important parameters that determine the physics of reconnection in the absence of cooling are the magnetization in the background plasma @xmath34 @xmath35 and the ratio @xmath36 of the current sheet width to the plasma skin depth of particles in the center of the current sheet , @xmath37 here @xmath38 is the average enthalpy of particles , @xmath39 is the mean particle lorentz factor in the current sheet , and @xmath40 is the particle pressure . in order for fast relativistic reconnection to occur , one must have @xmath41 and @xmath42 . we choose parameters @xmath43 and @xmath44 . we choose a low value for @xmath34 to probe the @xmath45 regime for which the current sheet structure does not depend significantly on the initialization the current sheet becomes significantly wider and inflows become relativistic if the pressure gradient in the magnetic field is balanced by density gradients , but not if it is balanced by temperature gradients . ] of the current sheet @xcite . @xmath33 corresponds to @xmath46 and enthalpy @xmath47 , so @xmath43 gives @xmath48 . the size of the simulations is @xmath49 , large enough for nonlinear reconnection to occur while the boundary conditions do not affect the evolution . test particles are injected in simulations tu and tc at @xmath50 , and all simulations are run at least until @xmath51 , a time at which nonlinear reconnection begins to be affected by the boundary conditions . this is enough time to observe the effects of acceleration and cooling on the injected particles and evaluate the effects of cooling on the overall structure of the x - point regions . to ensure sufficient resolution to capture the physics of reconnection , we use a density of 8 macroparticles / cell / species throughout the plasma and a grid size @xmath52 . tests of the code @xcite show that the physics of reconnection and the evolution of the current sheet is similar for macroparticle densities up to 50 macroparticles / cell / species and grid sizes as small as @xmath53 . we simulate the cooling of particles using the reduced landau - lifshitz radiation reaction force @xcite . we express this force in dimensionless units with electric and magnetic fields normalized to a fiducial magnetic field @xmath8 and times normalized to @xmath54 ( setting @xmath55 ) , where @xmath56 is the nonrelativistic gyrofrequency of particles in the field @xmath8 . in terms of the normalized momentum @xmath57 the radiation reaction force is given by @xmath58\gamma \mathbf{u } \label{eq : reactionforce}\\ & + \mathbf{e}\times \mathbf{b}+\frac{1}{\gamma } \mathbf{b } \times ( \mathbf{b}\times \mathbf{u})+\frac{1}{\gamma}\mathbf{e}(\mathbf{u } \cdot \mathbf{e})\bigg\},\nonumber\end{aligned}\ ] ] where @xmath59 . the terms on the first line represent the @xcite radiation reaction force used by many other authors @xcite . these terms typically dominate the radiation reaction force , but we include the additional reduced landau - lifshitz terms on the second line to increase the accuracy of the simulation without significant computational cost . we have tested the cooling code extensively in the synchrotron , bremstrahlung , and curvature radiation regimes , finding good agreement with analytical expectations . we do not include the effects of inverse compton cooling , because in a realistic case in which electrons emit close to the burnoff limit , the radiation is deep in the klein - nishina regime . because maxwell s equations and the lorentz force law are linear in the fields @xmath60 , @xmath61 , charge density @xmath62 , and current density @xmath63 in the absence of cooling , @xmath8 and @xmath64 can be scaled out of the equations of motion using a similar procedure to that in @xcite . we first normalize the electric and magnetic fields to @xmath8 . we then scale times to @xmath54 ( setting @xmath55 ) and distances to @xmath65 . we normalize the phase space density to the background number density @xmath66 and then express it in terms of dimensionless positions , momenta , and times as @xmath67 where @xmath68 . our choice of normalization for the phase space density determines the normalization of its moments : the number density is normalized to @xmath64 , the charge density to @xmath69 and the current density to @xmath70 . we find that the resulting dimensionless vlasov - maxwell equations for the normalized fields in the absence of the radiation reaction force are @xmath71 @xmath72 where the differential operators are defined as @xmath73 and @xmath74 . the only parameter that appears in the equations is @xmath75 . the fiducial magnetic field @xmath8 and background number density @xmath27 thus appear only in combination , and either may be varied without changing the form of the simulation as long as a corresponding adjustment is made to the other . the setup of the initial conditions also depends on the normalized current sheet width @xmath76 and the initial normalized temperature @xmath77 . combined with the size of the simulation , these parameters completely characterise our simulations in the absence of radiative feedback . with the radiation reaction force added , @xmath78 is a new dimensionless parameter which sets the limiting lorentz factor that corresponds to synchrotron emission at the burnoff limit . the value of @xmath78 also constitutes the choice of a physical value of the fiducial magnetic field @xmath8 for a given species of physical particle , because the choice of a particle sets @xmath79 . we derive the limiting lorentz factor as follows : a relativistic particle accelerated in the typical electric field @xmath7 has an energy gain per unit time of approximately . @xmath80 while the same particle radiating in the typical background magnetic field @xmath8 loses energy at a rate @xmath81 the critical lorentz factor for which the acceleration and cooling are equal is then given by @xmath82 the synchrotron peak of particles with this lorentz factor is at the burnoff limit @xmath83 , so we call @xmath84 the fiducial burnoff limit for particle energy . we can express @xmath78 in terms of the limiting lorentz factor as @xmath85 this allows us to choose a value for @xmath78 ( and thus , @xmath8 ) based on the ratio of the reconnection electric field to the background field @xmath86 and the desired value for @xmath84 . in simulations tc and ac , we have chosen @xmath87 . we find similar qualitative results in both simulations for a larger value of @xmath88 . we now check the validity of our expression ( [ eq : reactionforce ] ) for the radiation reaction force . in order for its derivation to be valid , it is necessary that particles are not strongly perturbed by the magnetic field on the light - crossing time of their classical electromagnetic radii . this condition , stated mathematically , is @xmath89 where @xmath90 is the magnetic field measured in the rest frame of the particle . using equation ( [ eq : lorentzbo ] ) , we find the radiation reaction force equation ( 7 ) is valid if throughout the simulation all particles have @xmath91 where @xmath92 is the critical magnetic field at which equation ( [ eq : reactionforce ] ) becomes invalid . for @xmath93 , we find that @xmath94 . we now verify that this condition is met in our simulations . for all of our traced particles in simulation tc , we use lorentz transformations to calculate @xmath90 at the particle location from the local value of @xmath17 in the simulation frame . over the complete particle trajectories , we find that in all cases @xmath95 , which is factor of @xmath96 below @xmath97 . because our test particle analysis is based only on the particle trajectories during acceleration , this is actually an overestimate of the relevant values of @xmath90 . for all particles during the acceleration phase , we find that at all times @xmath98 , which is a factor of @xmath99 below @xmath97 . thus , our treatment of the radiation reaction force is accurate in our test particle analysis . in realistic physical situations where @xmath100 , @xmath90 will be even smaller compared to @xmath97 for particles near the burnoff limit because @xmath101 . while quantum effects are expected to set in below @xmath97 ( at @xmath102 for electrons ) , they are unimportant in this realistic case . therefore , including quantum effects in our simulations would be incorrect and we do not do so . we also calculate @xmath90 for all particles in simulation ac at time @xmath50 . due to the presence of regions of very strong magnetic field of up to @xmath103 at the edges of islands , a significant number of particles do experience fields approaching @xmath97 . quantitatively , approximately @xmath104 of the accelerated particles with @xmath105 have @xmath106 . however , particles in x - point regions in this simulation always have @xmath107 , so the physics of particle acceleration and the high - energy part of the particle energy spectrum are unaffected by inaccuracies arising from the use of the classical radiation reaction force . we then compare of simulation ac with a simulation that is identical except for larger @xmath108 to verify that cooling does not greatly change the current sheet structure . we find that the small qualitative effects resulting from cooling found in simulation ac are verified in this additional simulation , and they are quantitatively even smaller as expected due to the weaker cooling . the reason that the inaccuracies have little effect is that they merely affect the details of how particles lose energy in islands after the end of particle acceleration . because the purpose of simulation ac is merely to test that particles can be accelerated beyond the burnoff limit and verify that cooling does not strongly affect the current sheet structure , we reserve more rigorous fully cooled simulations for future work . background particles in simulation ac are mildly relativistic and can cool slowly over the time of the simulation . this can distort the reconnection configuration in a way that would not occur in a realistic case in which the burnoff limit corresponds to a much higher lorentz factor @xmath109 . therefore , we turn off cooling in simulation ac for particles that are below @xmath110 to prevent background cooling from affecting the current sheet structure . the tpc method compares initially identical test particles entering a reconnection region with ( simulation tc ) and without ( simulation tu ) cooling to calculate the effect of cooling on a chosen uncooled spectrum over a large dynamic range . it assumes that the trajectories of these test particles , and the values of @xmath18 ( defined below ) that parameterize those trajectories , do not depend significantly on whether cooling is present . using this assumption , we can predict the maximum lorentz factor reached by a particle in simulation particle as a function of its trajectory and final lorentz factor in the uncooled simulation . the local limit on particle acceleration resulting from cooling at a given point is given by equation ( [ eq : lorentzbo ] ) , with the local values of the fields replacing @xmath7 and @xmath8 . the resulting local burnoff limit is given by @xmath111 , where @xmath112 . we hypothesize that the global limiting lorentz factor for the particle is @xmath113 , where @xmath18 , called the burrowing parameter , is an average of @xmath114 over the particle s trajectory . because we are interested in the values of @xmath114 at locations where the particle radiates , we weight the average by the synchrotron power of the particle . in a simple approximation , synchrotron power is proportional to @xmath115 , so we define @xmath116 we then predict that the final lorentz factor @xmath117 reached by a particle in the cooled simulation is given by @xmath118 the f subscript is present because we are referring to the final lorentz factor measured for the particle at the end of acceleration in each simulations . because both @xmath18 and @xmath119 on the right side of the equation are calculated in the uncooled simulation tu , this equation can be used to predict the effects of cooling in a simulation with a given value of @xmath84 without actually carrying out that simulation . the final step of the tpc method is the calculation of the distribution function for @xmath18 at any given value of @xmath120 using the measured trajectories . for any given uncooled particle energy spectrum , we can use this distribution function and the prescription ( [ eq : prescription ] ) to derive the cooled particle energy spectrum . in section 3.5 , we describe in detail how to derive the cooled spectrum corresponding to an uncooled power law energy spectrum and the distribution for @xmath18 at each @xmath121 that we find in our simulations . we begin in section [ sec : structure ] with a description of the overall structure of the current sheet in simulation tu as well as the characteristics of the x - point region we focus on in our analysis . then we describe our analysis of cooling during particle acceleration using the tpc method . this consists of a study of the effects of cooling on typical particle trajectories ( section [ sec : trajectories ] ) , a model of the effects of cooling on particle acceleration ( section [ sec : model ] ) , and a calculation of the relationship between particle trajectories and particle acceleration in the absence of cooling ( section [ sec : relationship ] ) . in section [ sec : powerlaw ] , we use this analysis to analytically compute the effect of cooling on a power law particle energy distribution . finally , in section [ sec : comparison ] , we compare the current sheet structure and particle energy spectra in simulations tu without cooling and ac with cooling to test the effects of cooling in a full simulation . the left part of figure [ fig : overall ] shows the structure of simulation tu ( which is the same as that of simulation tc ) during the nonlinear phase of magnetic reconnection at time @xmath50 when particle tracing begins . as in other simulations of collisionless reconnection , the current sheet has broken up into many x - point regions ( where reconnection takes place ) and high - density magnetic islands ( where outflows from the x - point regions meet ) . the right part of the figure shows an inset of the asymmetric x - point region into which test particles are injected in both simulations . the size of the x - point region is small compared to the overall box size , ensuring that its dynamics and our test particle results are unaffected by the boundary conditions . the structure of the x - point region is best indicated by comparing the plot of @xmath122 ( top left ) with that of @xmath123 ( top right ) . the narrow reconnection region at , @xmath124 in the central part of the x - point region is identical in both plots , indicating that the magnetic field is far below the background as expected in reconnection regions . in contrast , in the outflow regions of the figure with @xmath125 the magnetic field is similar to that in the background but the density is enhanced , producing a lower @xmath126 that is consistent with expectations for outflow regions where particles are being ejected into the magnetic islands . the lower left panel shows the momentum structure of the reconnection region . this clearly indicates that the center of the x - point is indeed at @xmath127 . the maximum inflow momentum we find for this x - point region is @xmath128 , while typical outflow bulk momenta are @xmath129 . this is similar to the values found in @xcite for @xmath130 . in order to calibrate our analysis of the test particle trajectories , we must find the value of @xmath86 in the x - point region . the lower right panel of the figure shows that the electric field is approximately uniform with @xmath131 throughout the central reconnection region ; therefore , we set @xmath132 for this region . since we chose @xmath87 for our simulations , the fiducial burnoff limit for particles in this x - point region is @xmath133 . @xmath86 is also a measurement of the reconnection rate , because in a steady state reconnection equilibrium @xmath134 . we also consider the results of two other methods of measuring the reconnection rate . one involves calculating the ratio of the inflow velocity to the alfvn velocity in the inflow region ( which is close to @xmath6 for relativistic plasmas ) ; with this definition , the typical reconnection rate shown in the lower left panel of figure 1 is @xmath135 . however , this is a local measure that overestimates the overall rate of energy conversion . another , global , method involves comparing the total kinetic energy gain to the maximum possible gain @xcite , with the equation @xmath136 where @xmath137 is the total kinetic energy and @xmath138 is the total initial magnetic energy . with this definition , we find @xmath139 , which is consistent with ( but somewhat smaller than ) the reconnection rate calculated from @xmath86 . this indicates that the energy conversion rate in our chosen x - point is representative of that in the overall simulation . we select a sample of @xmath140 particles from the test particles the were initiated in the shaded regions in figure [ fig : overall ] and trace their trajectories within our chosen reconnection region in the cooled simulation tc and uncooled simulation tu the particles selected are approximately uniformly distributed in @xmath141 at @xmath142 well after all of the particles have left the x - point region . this enables us to calculate the properties of particles that have undergone differing amounts of acceleration with reasonable number statistics . we choose to calculate the properties of the electrons in the rest of this work , but the properties of the positrons are similar . we calculate the final lorentz factors for particles in the uncooled ( @xmath119 ) and cooled ( @xmath117 ) simulations as follows . to ensure that we calculate the final lorentz factor for the main acceleration episode for each particle and properly estimate the values of the fields during this episode , we restrict our analysis of trajectories to particles in the central x - point region at @xmath124 , where they accelerate in a region of low magnetic field . this restriction does not have a strong effect on our results , because particles do not spend much time or undergo a great deal of acceleration in the region @xmath143 . we then estimate the lorentz factor at the end of the acceleration episode by calculating the location where it reaches @xmath144 of its maximum value . this allows us to avoid including locations where the particle lorentz factor plateaus , resulting in an underestimation of the average electric field acceleration . because particles in the uncooled simulation may undergo multiple acceleration episodes , we add an additional stipulation : if the rate of energy gain by a particle falls below a fraction @xmath145 of the fiducial gain given by @xmath146 and the particle has reached @xmath147 of its maximum value in the x - point , we estimate that acceleration has ended at that location . figure [ fig : trajectories ] compares the trajectories within the chosen reconnection region of three particles cooled simulation tc and uncooled simulation tu ; all of the particles chosen were significantly accelerated in the current sheet with @xmath148 and significantly cooled before leaving the current sheet in the cooled simulations . the fact that a significant number of particles can reach such high energies is an indication that radiation does not directly prevent acceleration beyond the fiducial burnoff limit in magnetic reconnection . the figure shows that all of the particles undergo speiser orbits : they are accelerated by the reconnection electric field while oscillating across the current sheet in the @xmath149 direction . the oscillation amplitude decreases with time as the particles are accelerated in accordance with momentum conservation constraints @xcite . the first and second panels show that the trajectories in the cooled and uncooled simulations are typically similar at the beginning of acceleration and while the particles are at the center of the x - point region , but sometimes differ after most acceleration is complete . particles cooled and uncooled trajectories typically differ most in the phase of their oscillation across the current sheet , but cooled particles also tend to leave the x - point slightly earlier than the uncooled particles , probably because cooling has reduced their inertia against deflection out of the x - point region . the third panel shows the relatively rare case in which the cooled and uncooled particles exit in different directions . such particles are only around @xmath150 of all particles , but they are a somewhat larger proportion ( @xmath151 ) of the highly accelerated particles with @xmath148 . because few particles have trajectories that are strongly modified by cooling , it should be possible to use the properties of the particles in the uncooled simulation to predict the acceleration and cooling of the particles in the cooled simulation . the tpc method can be used to predict the final energy reached by cooled particles from the corresponding uncooled particle energies and trajectories . it relies on the assumption that @xmath18 is the same in both the cooled and uncooled simulations . we verify that @xmath18 is indeed nearly the same in the cooled simulation tc and the uncooled simulation tu . this result is in agreement with the similarity of the particle trajectories in the two simulations , discussed in section [ sec : trajectories ] . figure [ fig : predcool ] shows a two - dimensional histogram of the relationship between the predicted and actual lorentz final lorentz factors in the cooled simulation . it shows that the model indeed predicts the cooled particle lorentz factor of most particles very well , although it tends to slightly underestimate ( by up to @xmath152 ) the amount of cooling for high - energy particles . even restricting the calculation to particles that have undergone significant cooling so that @xmath153 produces a similar relationship . the great accuracy of the model indicates that our prescription ( [ eq : prescription ] ) captures the physics of radiative cooling of particles within an x - point region . figure [ fig : rhogamma ] shows the relationship between @xmath18 and the uncooled final lorentz factor @xmath154 . it shows that the minimum value of @xmath18 for all values of @xmath119 is @xmath155 , and the probability distribution for @xmath156 is uniform up to a maximum value of @xmath18 that increases with @xmath119 . a small proportion ( @xmath157 ) of the particles with @xmath158 have @xmath159 ; they have undergone multiple acceleration episodes and therefore spent significant time in regions of low @xmath18 while still undergoing acceleration . we calculate @xmath160 by finding the value of @xmath18 where the distribution falls to @xmath145 times the peak at each interval of 0.1 in @xmath161 . a power law fit for @xmath162 yields @xmath163 . this distribution can be compared with analytical models of speiser orbits , which fall into two classes . in the first , the particle s oscillation amplitude is always larger than the current sheet width @xmath24 , so that @xmath164 . in the other case , in which the amplitude of oscillation is comparable to or smaller than the current sheet width and @xmath165 , where @xmath166 is the amplitude of oscillation for the particle . we can calculate the value of @xmath18 by finding @xmath167 during the particle trajectory analytically and using the relation @xmath168 which for @xmath169 gives @xmath170 . we calculate @xmath167 using modified version of the analytical arguments of @xcite . we assume that the angle @xmath171 at which the particle exits the current sheet center during each oscillation is small and the particle s velocity in the @xmath172 direction is close to @xmath6 , so that @xmath173 and @xmath174 . defining @xmath175 and @xmath176 , the equations of motion of the particle are then @xmath177 @xmath178 using an analagous argument to that in @xcite , we find as they do that @xmath179 . but we find a different expression for @xmath166 : @xmath180 combining these expressions with equation ( [ eq : speisbsq ] ) , we obtain @xmath181 and @xmath182 . in figure [ fig : rhogamma ] , we see that there are always particles with @xmath183 for all @xmath119 ; these particles typically never burrow into the current sheet and correspond to our first regime . the particles that do burrow into the current sheet have @xmath162 with @xmath184 , which is slightly larger than the analytical value of @xmath182 in the second regime . we note that in the test particle simulations of @xcite , the oscillation amplitude declined somewhat faster than analytical expectations in the regime where radiation is negligible . that result is consistent with what we find here . we now extrapolate the results of our simulation using analytical techniques to calculate the effects of cooling on particle energy spectra over many decades in dynamic range . because we are calculating a steady - state distribution , we omit the subscript f for the cooled and uncooled lorentz factors . we assume that the uncooled distribution of the particles is a pure power law of the form @xmath185 , where @xmath2 here is the uncooled lorentz factor of a particle normalised to the minimum lorentz factor @xmath186 of the power law . we assume that @xmath18 has a characteristic value @xmath187 that increases with @xmath2 as a power law , so @xmath188 , as discussed in the previous section . the measured distribution of @xmath156 in figure [ fig : rhogamma ] indicates that the conditional distribution of @xmath18 at a given value of @xmath2 is approximately uniform over @xmath189 , and 0 elsewhere , so that @xmath190 . the actual value of @xmath191 in the figure , but we assume @xmath192 in our analysis for simplicity . this does not significantly affect , the locations of the breaks in the particle energy spectrum in equation ( [ eq : fulldist ] ) are scaled by a factor of @xmath193 and @xmath194 , but the functional forms do not change . ] our results . the resulting joint probability distribution @xmath195 for @xmath196 is @xmath197 we also set @xmath198 where @xmath199 here refers to the delta function , so that @xmath200 for @xmath201 . there are three ranges of the distribution for @xmath202 , with two breaks present at @xmath203 and @xmath204 . for @xmath205 , the second break does not exist , so most of our following discussion assumes that @xmath206 . for @xmath207 , @xmath208 in all cases , so @xmath209 . for @xmath210 , @xmath211 , so @xmath212 . @xmath213 is given by @xmath214 where @xmath215 is the hypergeometric function . the last expression comes from an expansion of that function that applies for @xmath216 , which is the relevant limit for @xmath217 . finally , for @xmath218 , @xmath219 where the bars indicate substitution and we have made use of an identity for the distribution of @xmath220 . note that @xmath221 is the _ marginal _ distribution of @xmath18 , and is not the same as the uniform conditional distribution for @xmath18 mentioned earlier . using the same expansion for the hypergeometric function , we can calculate the full distribution for @xmath206 : @xmath222where @xmath223 is a parameter that always falls in the range @xmath224 . one important thing to note in the analytical distribution is the slight discontinuity at @xmath203 , where @xmath225 . this discontinuity results from the presence of particles with @xmath226 at all values of @xmath2 ; such particles all cool down to @xmath203 , but no further . for @xmath227 , @xmath228 . therefore the discontinuity at @xmath203 is only important locally . because the electric field varies slightly with location in a realistic x - point region , there will be variation in @xmath84 that smears out the distribution and eliminates this discontinuity . as a result , we do not expect cooling to have strong effects on the particle distribution for @xmath229 . the typical power law index @xmath230 of particles in the uncooled simulation is approximately @xmath231 ( see below ) , and our previous analysis shows that @xmath232 . therefore , we expect that @xmath233 , and that the power law index above this break is about @xmath234 . therefore , cooling effects on particle acceleration in magnetic reconnection occur only very far above the burnoff limit , and they produce only a moderate break in the particle energy spectrum . we now discuss the validity of our analytical spectrum for real physical systems with @xmath109 . the analytical particle spectrum we have derived in this section is based on our test particle simulation results using equation ( 7 ) for the radiation reaction force . therefore , its results will only be valid for @xmath235 . in the physical case for which our extrapolation is relevant , quantum effects may become important before the classical perturbation result breaks down . the more restrictive schwinger limit for electrons corresponds to @xmath236 and @xmath237 . because we do not know the value of @xmath90 for particles in a real physical system , we conservatively estimate that the accelerating particles experience a field @xmath8 in the plasma rest frame . because @xmath238 even in the diffusion region where @xmath239 , the lorentz transformation into the frame of a particle with cooled lorentz factor @xmath240 yields @xmath241 . combining this result with equation ( [ eq : bpcrit ] ) and the definition of @xmath242 then yields the condition for the validity of our classical analytical spectrum @xmath243 our main conclusion that cooling has no effect below @xmath244 is valid if @xmath245 . this condition may be expressed as @xmath246 taking @xmath247 and @xmath184 , this equation gives @xmath248 , which corresponds via equation ( 10 ) to @xmath249 g. reconnection models that produce observed emission through the synchrotron mechanism in grb prompt emission @xcite , agn flares @xcite and the crab flares @xcite invoke magnetic field strengths far below this value ( 1 g-@xmath250 g , @xmath2511 g , and @xmath2511 mg , respectively ) . we also note that the presence of quantum effects for magnetic fields approaching @xmath252 is likely to lessen the strength of the radiation reaction force rather than increasing it , because the classical force includes photons with energy larger than the electron rest energy whose production will be suppressed by quantum effects . therefore , our derived spectrum below @xmath244 should accurately estimate the effects of cooling on particles accelerated in magnetic reconnection in these systems . we now compare the overall structure of simulation ac with cooling and simulation tu without cooling . we find that in simulation ac the evolution of the current sheet is somewhat slower , but the energy conversion rate is somewhat faster than in simulation tu . the global reconnection rate is approximately @xmath253 , which is @xmath254 higher than in the uncooled simulation . the typical reconnection electric field is @xmath255 , which is @xmath256 higher than in the uncooled simulation . figure [ fig : overallcomp ] compares the densities and the average kinetic energies in the uncooled simulation tu and the cooled simulation ac at time @xmath50 . the density structures of the current sheets in the two runs are similar , except for the slower evolution , as are most other properties of the reconnection . the only significant difference is that the particles in the islands and outside of the center of the current sheet have been cooled to low energies @xmath257 ; as a result , the particle density at the very center of magnetic islands in simulation ac is extremely high , up to @xmath258 times the background density at the center point of islands . this cooling is in line with what we expect given that @xmath133 and particles are able to cool for many plasma times if they are not being accelerated in reconnection regions . we conclude that the structure of reconnection regions will not vary greatly due to cooling , and our test particle results may be approximately applied to uncooled simulations to estimate the effects of cooling . figure [ fig : energyspectra ] compares the particle energy spectra in the two simulations in the ( half ) box , in the x - point regions , and in the islands . the figure clearly shows that there is no cutoff at @xmath133 in any of these locations . the cutoff of the power laws is instead located at @xmath259 in both simulations and is a result of the saturation of acceleration due to low @xmath0 @xcite . for a magnetic field equal to @xmath8 this corresponds to a cutoff synchrotron energy of over 16 times the burnoff limit in the cooled simulation . the cutoff synchrotron energy may be even higher than this because the edges of islands often have a magnetic field significantly larger than @xmath8 . thus , acceleration well beyond the burnoff limit is present even in these limited simulations . the power law index in the uncooled simulation is @xmath260 , and that in the cooled simulation is @xmath261 , with no obvious break in either spectrum . in contrast , our analytical power law model with @xmath184 predicts that because the beginning of the power law in simulation tu is located at @xmath262 , there will be a break in the cooled spectrum at @xmath263 with a power law index above the break of @xmath264 . the disagreement between the analytical model and the actual results is not surprising given the limited dynamical range of the simulations , but it is encouraging that the slope of the power law is indeed moderately steeper in the cooled simulation . the population of particles in both the x - point regions and the islands is larger in the cooled simulation ac than in the uncooled simulation tu , reflecting the difference in reconnection rate . in the uncooled simulation tu , the islands have far more particles than the x - points at almost all energies , with the x - points becoming important only at the highest energies of @xmath265 . this is because most particles eventually escape from the x - points after accelerating and enter islands without losing significant energy in the uncooled simulation , and can even undergo significant additional acceleration there @xcite . only at the highest energies are the x - points important , because particles that remain there are capable of accelerating for a longer time period than other particles . in the cooled simulation ac , in contrast , the x - point population makes a significant contribution to the overall energy spectrum for @xmath266 and becomes dominant for @xmath267 . this is because the strong cooling experienced by particles in the islands greatly reduces their contribution to the high - energy particle spectrum . nevertheless , the island particle spectrum still does not have a cutoff at @xmath133 , because particles are always flowing into the islands from the x - point regions as long as reconnection is ongoing , so that high - energy particles are continuously replenished there . overall , the burnoff limit does not limit the particle energy anywhere in the system in our fully cooled simulation ac , indicating that it is not an important limit at any stage of particle acceleration in magnetic reconnection . we do note that this conclusion is tentative in magnetic islands because particles there may have been affected by the presence of fields with @xmath268 , but the islands do not strongly affect the overall high - energy spectrum because the island and x - point populations are comparable there . this confirms that our analytical results obtained using the tpc method are correct , and indicates that particles can be accelerated above the burnoff limit by orders of magnitude . in this paper , we investigated the effect of cooling on particle acceleration in relativistic magnetic reconnection by using the tpc ( test particle comparison ) method , in which we compare test particle orbits in the same reconnection region with ( simulation tu ) and without ( simulation tc ) cooling . we also carried out an additional simulation ( simulation ac ) to find the effect of cooling on reconnection overall and confirm the utility of the tpc method and the parameterization of particle trajectories using the burrowing parameter @xmath18 . our conclusions are as follows : * the trajectories of the vast majority of accelerating particles are weakly affected by cooling . the most prominent effect of cooling is that the phase of oscillation across the current sheet is different for cooled particles , and cooled particles leave the x - point regions slightly earlier due to reduced inertia as a result of cooling . in rare cases , cooling causes a particle to exit the x - point region in the opposite direction from its uncooled path . * the acceleration of a particle in a cooled simulation can be derived from its acceleration in the uncooled simulation and its burrowing parameter @xmath18 using the prescription @xmath269 . here , the burrowing parameter @xmath270 indicates the extent to which the particle spends time in the current sheet and experiences a lower magnetic field ( and therefore a reduced amount of cooling ) . * the burrowing parameter @xmath18 of particles varies between approximately unity and @xmath271 , with @xmath184 . this is roughly consistent with analytical speiser orbit calculations , which predict that @xmath182 . * if the uncooled particle energy spectrum is a power law of the form @xmath272 , we predict analytically that when cooling is present the distribution @xmath273 will be the same up to @xmath274 . this is far above the burnoff limit of @xmath84 . above this break , the power law index becomes @xmath275 , which is only a moderate break in the spectrum for @xmath276 . the unimportance of @xmath84 for the particle energy spectrum is confirmed in fully cooled simulation ac . * in the cooled simulation ac , the effect of overall cooling is to speed up the rate of energy transfer relative to the uncooled simulation tu while slightly slowing the current sheet evolution . however , there are no significant qualitative differences in the reconnection structure . * the particle energy spectrum in the fully cooled simulation ac is similar to that found in the uncooled simulation tu , and does not show any cutoff at @xmath133 . only in the islands is the form of the spectrum significantly different in the two simulations , because that is where accelerated particles experience significant cooling . * the analytical results from the tpc method are confirmed by the fully cooled simulation ac . we conclude that particles can be accelerated above the burnoff limit by many orders of magnitude . our analytical calculations of the effect of cooling on power law energy spectra , together with the measured particle energy spectra in the fully cooled pic simulation , confirm the conclusion of @xcite that particle acceleration is not limited by cooling . therefore , emission beyond the synchrotron burnoff limit is indeed well explained by a magnetic reconnection model . we now turn to the applications of our results in astrophysical contexts . we show here that reconnection is that the particle energy spectrum and the resulting radiation spectrum have no direct dependence on the burnoff limit . the tev emission in agn is far beyond the synchrotron burnoff limit ( even accounting for relativistic boosting of the bulk of the jet ) . this is consistent with being produced by reconnection . on the other hand , if the crab flares are produced by reconnection it is necessary to explain why their peak energies are only a factor of @xmath277 above the maximum synchrotron burnoff limit of 160 mev . this difficulty can be resolved if the typical sizes of the flaring regions are small enough that electrons can not reach extreme energies except in the largest flares . alternatively , if typical flaring regions have a strong guide field perpendicular to the plane of reconnection , most regions will not produce significant acceleration beyond 160 mev @xcite . if future studies of the morphology of the crab nebula during flaring activity are consistent one of these scenarios , they will provide further evidence that high - energy flares above 160 mev in the crab nebula are produced by particles accelerated in relativistic magnetic reconnection . this research was supported by the i - core center for excellence in research in astrophysics and by an isa grant . tp was partially supported by a cnsf - isf grant . dk and en were partially supported by an erc starting grant ( grb / sn ) and an isf grant ( 1277/13 ) .
the radial instability of the electron - degenerate , chandrasekhar - mass core marks the beginning of the final episode in the life of a massive star with zero - age main - sequence ( zams ) mass in the range @xmath3 . collapse ensues and , once fully dynamical , separates the core into a subsonically homologously contracting inner core and a supersonically collapsing outer core . at nuclear density , the repulsive component of the nuclear force leads to a stiffening of the equation of state ( eos ) . this stabilizes the inner core , which overshoots its new equilibrium , then rebounds into the outer core , launching a strong hydrodynamic shock wave from its edge . this instant in time is referred to as _ core bounce_. the inner core has a mass of @xmath4 at bounce and this material becomes the unshocked core of the protoneutron star . the shock formed at core bounce propagates into the outer core , but the dissociation of nuclei in the accreted material into neutrons and protons and the electron capture on free protons in the region behind the shock ( the _ postshock _ region ) soon sap its might , driving it into submission to the ram pressure of accretion . the shock stalls and turns into a standing accretion shock that must be revived to unbind the stellar envelope and drive a core - collapse supernova explosion . neutrinos play a pivotal and dominant role in stellar collapse and core - collapse supernovae . neutrinos and antineutrinos of all flavors carry away the @xmath5@xmath6 ( @xmath7 ) of gravitational binding energy of the remnant neutron star over tens of seconds after core bounce . aided by multi - dimensional fluid instabilities , they probably deposit , within a few hundred milliseconds after bounce , sufficient energy in the region behind the stalled shock to revive the shock and make typical @xmath5@xmath8 core - collapse supernova explosion ( @xcite and references therein ) . only hyper - energetic ( i.e. , @xmath9 ) explosions may require a different mechanism @xcite , e.g. , rapid rotation combined with strong magnetic fields , which may lead to energetic jet - driven explosions ( e.g. , @xcite ) . for a galactic or near - extragalactic core - collapse supernova , neutrinos offer the unique possibility of directly observing the dynamics and thermodynamic conditions prevalent in the supernova core . together with gravitational waves ( see , e.g. , @xcite ) they will herald the next nearby supernova possibly hours before any telescope sensitive to electromagnetic waves will notice the event . in the probable case that the next galactic supernova occurs in a dust - enshrouded region and/or close or behind the galactic center , the supernova may be impossible to observe in broad bands of the electromagnetic spectrum , making neutrino and gravitational - wave observations even more important . the observation of neutrinos from sn 1987a in the large magellanic cloud @xcite confirmed the basic picture of core collapse and early protoneutron star evolution ( e.g. , @xcite and references therein ) , but the small number and poor timing of the observed interactions did not allow far - reaching and robust conclusions on core - collapse supernova dynamics and the involved neutrino physics , nuclear physics and astrophysics . the situation will be completely different when the neutrino burst from a galactic core - collapse supernova reaches current and near - future neutrino detectors on earth . super - kamiokande @xcite , icecube @xcite , lvd @xcite , borexino @xcite , kamland @xcite , sno+ @xcite , no@xmath10a @xcite and others will together see many thousands of neutrinos from a core collapse event at @xmath11 @xcite . distance estimates based on the observed neutrino flux @xcite , sky localization @xcite , and triggering of gravitational - wave searches by the reconstruction of the time of core bounce @xcite will likely all be possible . well - timed high - statistics coincident neutrino observations will allow to probe in detail a broad range of supernova astrophysics , nuclear physics , and neutrino physics ( see @xcite for overviews ) . fast characteristic temporal variations in the preexplosion neutrino fluxes would be tell - tale signs of multi - dimensional fluid instabilities in the postshock region @xcite and/or early postbounce ring - down oscillations of a rapidly spinning protoneutron star @xcite . while artificially - driven , spherically - symmetric core collapse simulations suggest that a sudden deep drop of the accretion - driven component of the neutrino luminosity ( primarily in @xmath12 and @xmath13 ) within a few hundred milliseconds would indicate the onset of explosion ( e.g. , @xcite ) , the situation is likely different in nature as multidimensional simulations show a much shallower drop due to simultaneous accretion and explosion of material . this leads to a persistent accretion luminosity even after the explosion has begun @xcite . the spectral characteristics and long - term spectral evolution of the neutrino flux could provide important constraints on the nuclear eos @xcite and/or the spin of the progenitor core @xcite . the ability of some detectors to distinguish interactions of different neutrino flavors would lead to constraints on the neutron - to - proton ratio in the neutrino - driven wind phase , allowing an observational test of core - collapse supernovae as potential sites for @xmath14-process nucleosynthesis @xcite . the neutrino signature of core collapse , of the subsequent core - collapse supernova evolution and of the protoneutron star cooling phase , is invariably intertwined with neutrino oscillation physics . the robustness of all of the above mentioned observational conclusions will depend on our understanding of the impact of neutrino flavor oscillations . neutrinos propagating from their emission site to detectors on earth may experience ( _ i _ ) so - called vacuum oscillations driven by neutrino mass differences @xcite , ( _ ii _ ) oscillations mediated by a resonance in @xmath10@xmath15 scattering ( the mikheyev - smirnov - wolfenstein [ msw ] effect ; @xcite ) , and ( _ iii _ ) oscillations due to @xmath10@xmath10 scattering ( @xcite ; see @xcite for a review ) . vacuum and msw oscillations are well understood and their outcomes depend essentially only on neutrino mixing parameters , in particular the neutrino mass hierarchy and the mixing angles . the @xmath10@xmath10-scattering driven oscillations , on the other hand , have a non - linear hamiltonian that may lead to so - called collective oscillations with very complex spatial and temporal outcome that remains to be fully understood ( see , e.g. , @xcite and references therein ) . however , a number of recent studies suggest that collective oscillations may be completely or at least partially suppressed in the preexplosion accretion phase of ordinary core - collapse supernovae @xcite , but see @xcite and @xcite for discrepant results . provided that collective oscillations can be ignored in the preexplosion phase and that the @xmath16 mixing angle indeed has the large value suggested by recent measurements @xcite , the neutrino mass hierarchy may be inferred from the qualitative shape of the early postbounce neutrino signal @xcite . a preexplosion accretion phase with suppressed collective oscillations would also offer the opportunity to probe the structure of the progenitor star on the basis of the observed neutrino signal . the details of the preexplosion neutrino emission have been discussed carefully , e.g. , by @xcite and @xcite and we shall not repeat them here . it is , however , necessary to outline its most salient features . for simplicity , we neglect neutrino oscillations in the following . in core collapse and in the subsequent postbounce evolution , emission of @xmath12 and @xmath13 occurs via charged and neutral currents , while heavy - lepton neutrinos @xmath17 are created exclusively via thermal neutral - current pair processes . before core bounce , only @xmath12 are emitted from electron capture in the collapsing core . milliseconds after core bounce , the shock breaks out of the @xmath12 neutrinosphere ( where the optical depth is @xmath18 ) and a strong burst of @xmath12 is emitted for @xmath5@xmath19 from rapid electron capture on the freshly abundant free protons behind the shock . @xmath20 are copiously created in the hot interior of the protoneutron star after bounce and begin to diffuse out , leading to a steep rise , quick leveling and subsequent slow decay of the @xmath20 luminosity ( @xmath21 ) . @xmath13 production via charged - current positron capture is initially suppressed due to the high degeneracy of the electrons . the latter is partially lifted after bounce at the moderate - density , hot edge of the protoneutron star and @xmath22 rises , reaching or surpassing the value at which @xmath23 levels off after the neutronization burst decays . the subsequent preexplosion luminosity can roughly be split into a diffusive component from the core and accretion luminosity ( @xmath24 \dot{m } / r_\nu$ ] , where @xmath25 is an approximate neutrinosphere radius ) from or from above the neutrinosphere @xcite . @xmath21 is primarily diffusive , while @xmath23 and @xmath22 are dominated by accretion . in general , @xmath26 , but @xmath27 . @xmath20 have the lowest opacity , since they interact only via neutral currents . they decouple from matter at the smallest radii and highest temperatures and thus have the highest average energies @xmath28 . @xmath13 have a slightly lower opacity than @xmath12 , leading to the well established neutrino energy hierarchy in the preexplosion phase @xmath29 . the mean energy of all species grows with increasing postbounce time , reflecting the recession of the neutrinospheres due to the contraction of the protoneutron star . considering that the accretion luminosity will scale with the postbounce accretion rate @xmath30 , one would naturally expect an increase of the detected neutrino interactions with increasing mass of the stellar core . since higher accretion rates correspond to more material compressing and settling more rapidly on the protoneutron star , the latter s outer regions will be hotter . thus the thermal neutral - current emission will be enhanced , leading to higher luminosities and higher mean neutrino energies . the variation of the preexplosion neutrino signal with progenitor star zams mass was first discussed by @xcite based on the pioneering simulations of @xcite and @xcite . these authors provided total emission characteristics and spectra that show a systematic increase of total energy emitted in neutrinos and mean @xmath13 energy with zams mass in the range from @xmath31 . @xcite , before sn 1987a , carried out simulations of a range of progenitor stars with zams mass in the range @xmath32 . they found that the @xmath12 neutronization burst shows little dependence on the progenitor , due to the rather universal homologous collapse and bounce dynamics . furthermore , they mentioned , though did not discuss in detail , that the luminosities and mean neutrino energies increase as a function of iron core mass ( and not zams mass ) . a more detailed and clear physical discussion was provided by @xcite , who contrasted the predicted neutrino signal from the early postbounce phase in two different progenitor core models with neutrino observations of sn 1987a . he noted that there are significant uncertainties in connecting a given zams mass to precollapse structure . instead of a progenitor with an associated zams mass , he considered a massive ( and high - entropy ) @xmath33-@xmath34 iron core model and a lower - mass ( and lower - entropy ) @xmath35-@xmath34 iron core model in his spherically - symmetric ( 1d ) neutrino radiation - hydrodynamic simulations . he showed that the more massive core leads to a consistently higher @xmath13 luminosity in both the accretion and diffusion sectors . the water cherenkov detectors that observed neutrinos from sn 1987a are most sensitive to the inverse beta decay ( ibd ) reaction @xmath36 . @xcite predicted a factor of two difference in the integrated number of early ibd interactions between the massive and the low - mass core in these detectors . he concluded that the neutrino signal observed by these detectors from sn 1987a was most consistent with the low - mass core . @xcite , who carried out a parameter study of quasi - hydrostatic protoneutron star cooling , considering various initial masses , ad - hoc accretion rates , and different nuclear eos , found a similar trend . he showed that more massive cores , higher accretion rates , and softer eos lead to stronger , higher - energy neutrino emission . some of his strongest emitters were cases in which eventually a black hole was formed . liebendrfer and collaborators carried out a sequence of studies of the progenitor dependence of the neutrino signal using modern general relativistic 1d radiation - hydrodynamics simulations @xcite . they showed that the @xmath12 neutronization burst is indeed almost independent of progenitor structure ( as first suggested by @xcite ) . they also qualitatively and quantitatively connected the evolution of the postbounce preexplosion luminosity to the postbounce accretion rate , but did not discuss observational implications . their results were corroborated by similarly sophisticated subsequent studies of @xcite . of these , @xcite presented the most comprehensive analysis and also compared between 1d and axisymmetric ( 2d ) results . they found that in 2d , convection in the protoneutron star alters the structure of the latter , affecting the neutrino emission starting @xmath5@xmath37 after bounce , but preserving the overall systematics with accretion rate . @xcite also were the only authors to suggest that the accretion - rate dependence of luminosity and total emitted energy in the preexplosion phase could be used to infer the structure of the progenitor . the other studies , being focused on aspects such as neutrino oscillations , black hole formation , or the late - time post - explosion evolution , did not consider observational consequences . @xcite , using a limited set of three progenitor models ( \{@xmath38 , @xmath39 , @xmath40}@xmath41 at zams ) , found similar systematics as the aforementioned studies , but also carried out an analysis of the expected signal in various neutrino detectors in the first @xmath42 after bounce . they computed ibd interaction rates for their @xmath40-@xmath34 and @xmath38-@xmath34 and found a factor of two more ibd interactions for the former , which would allow a high - confidence distinction between these progenitors for a galactic core collapse event . however , their @xmath39-@xmath34 model yielded a postbounce neutrino signal very similar to that of their @xmath38-@xmath34 model and would be indistinguishable by neutrino observations alone . this suggests that zams mass is not a good parameter to describe presupernova stellar structure ( cf . , @xcite ) . in this article , we present a fresh look at the progenitor dependence of the neutrino signature in the preexplosion accretion phase of core - collapse supernovae . we perform 1d general relativistic radiation - hydrodynamics core collapse simulations of 32 progenitor models from the single - star solar - metallicity presupernova model suite of @xcite and follow the postbounce preexplosion evolution for @xmath43 . in zams mass , these models range from @xmath0 to @xmath1 , but guided by the previous results discussed in the above , we choose not to parameterize our simulations by zams mass . instead we employ the compactness parameter @xmath44 ( for a relevant mass scale @xmath45 , measured at the time of bounce ) . as shown in @xcite , and further explored in @xcite , @xmath46 is a quantitative stellar structure parameter that describes the postbounce accretion evolution to a remnant mass scale @xmath45 . we demonstrate that the preexplosion neutrino emission is very well parameterized by the compactness . the preexplosion luminosities and mean energies of all neutrino species increase essentially monotonically with increasing @xmath46 . we compute predicted integrated ibd interactions for a galactic core - collapse supernova in the super - kamiokande detector and show that the clear systematics governed by @xmath46 carries over to observation , even when standard msw neutrino oscillations are taken into account . our results thus indicate that in the absence of complicated collective neutrino oscillations a high - statistics detection of neutrinos from the preexplosion phase will allow , in principle , a tight constraint of the compactness of the progenitor star s core . this , however , will require knowledge of the nuclear eos and of the rotation rate of the collapsed core , since , as we show , both can dilute the otherwise clear compactness - dependent neutrino emission systematics . this paper is structured as follows . in [ sec : methods ] , we discuss our general relativistic hydrodynamics code gr1d and introduce its extension to neutrino radiation - hydrodynamics in the two - moment approximation , nugr1d . the initial models and the employed eos are discussed in [ sec : models ] . in [ sec : boltzmanncompare ] , we present results from a benchmark collapse and postbounce simulation that allows us to compare with the previously published code comparison of @xcite to assess nugr1d s ability to reproduce results of full boltzmann neutrino transport . we present the results of our simulations in [ sec : results ] , analyze the dependence of the neutrino signal on progenitor compactness , discuss predicted ibd signals from a galactic core collapse event in the super - kamiokande detector , and explore potential degeneracies introduced by eos and rotation . finally , in [ sec : discussion ] , we critically summarize our work and conclude by contrasting our results with the early neutrino signal observed from sn 1987a . we make use of the open - source 1d general relativistic hydrodynamics code ` gr1d ` ( @xcite ; available at http://www.stellarcollapse.org ) outfitted with an energy - dependent multi - species m1 neutrino transport scheme in which the zeroth and first moments of the neutrino distribution function are evolved . we refer the reader to @xcite for details on ` gr1d ` and describe in the following our current implementation of the transport scheme . for this first application , we neglect the computationally - expensive energy - coupling neutrino interactions and transport terms these terms are undoubtedly important for making highly accurate predictions of the neutrino signature ( see , e.g. , @xcite ) , but are unlikely to affect the general trends we observe . we will address them in future work , but provide a discussion on the consequences of neglecting these terms via a comparison to full boltzmann neutrino transport simulations in . our m1 scheme closely follows @xcite , who formulate the m1 evolution equations in a closed covariant form . the scheme is simplified greatly by neglecting the energy - coupling terms . this further requires that the velocity dependent terms are also ignored . in this limit , and using the schwarzschild - like metric and radial - polar slicing of ` gr1d ` and setting @xmath47 , the coordinate frame evolution equations for the neutrino energy density , @xmath48 , and the neutrino flux vector , @xmath49 , simplify from equations 3.37 and 3.38 of @xcite to @xmath50 and @xmath51 where @xmath52 is the neutrino interaction source term ( see below ) , @xmath53 is the lapse function and @xmath54 . @xmath55 is the neutrino pressure tensor and is taken to be an interpolation between the two limiting cases of free streaming and diffusion . we follow @xcite , who express @xmath56 as @xmath57 where @xmath58 is the eddington factor , taken here to be the maximum entropy closure in a closed , analytic form @xcite , @xmath59 in the no - velocity limit for ` gr1d ` , @xmath60 . the free streaming and diffusion limits of the neutrino pressure tensor are @xmath61 and @xmath62 , respectively . we set out to solve the system of equations via standard hyperbolic methods borrowed from conservative hydrodynamic schemes @xcite . transport variables live at cell centers and we employ piece - wise linear reconstruction to cell interfaces with van leer s limiter @xcite and the hlle approximate riemann solver @xcite for calculating the intercell fluxes . a complication arises when solving the neutrino moment equations in the high - opacity limit . in this case , the standard fluxes returned from the riemann solver are dominated by a numerical diffusion term . we follow @xcite and modify the fluxes to correct for this . essentially , the modified fluxes return the diffusion - limit flux in the high opacity limit . the modified form of the neutrino energy density flux through the @xmath63 interface is given by @xmath64 the corresponding modified interface flux for the neutrino flux evolution equation is , @xmath65 with @xmath66 in these equations , @xmath67 controls the modification to the fluxes to account for the high opacity . following @xcite , we take @xmath68 where @xmath69 is the sum of the scattering and absorptive opacities . these opacities are strong functions of energy and are also species dependent . we note that when @xmath67 is 1 , the intercell fluxes reduce to the standard hlle approximation . the characteristic speeds needed for the hlle scheme are calculated in the same spirit as the neutrino pressure tensor @xcite , @xmath70 where in the zero velocity limit , @xmath71 and @xmath72 . @xmath73 and @xmath74 are the maximum and minimum values , respectively , of these characteristic speeds evaluated from both the right and left reconstructed variables . finally , the source terms in and [ eq : fluxevolution ] are taken from @xcite . in the zero velocity limit , @xmath75 where @xmath76 , @xmath77 , and @xmath78 are the neutrino emissivity , neutrino absorption opacity , and the neutrino scattering opacity , respectively we precompute the neutrino interaction terms for each neutrinos species ( we treat @xmath12 , @xmath13 and @xmath79 ) and neutrino energy group in dense tabular form as a function of density @xmath80 , temperature @xmath81 , and electron fraction @xmath82 . we then use linear interpolation for efficient on - the - fly interpolation . we include all standard iso - energetic scattering processes , charged - current absorption and emission , and thermal pair - production processes @xcite in the calculation of the neutrino interaction terms . since the neutrino matter interactions for heavy - lepton neutrinos and antineutrinos are slightly different , ` nulib ` averages the two values of the emissivities and opacities . our library of neutrino interaction routines , which we call ` nulib ` , is open source and available as a github repository at http://www.nulib.org . ` nulib ` requires an eos for the evaluation of the emissivities and opacities . our treatment of thermal pair - processes in gr1d warrants some comments . since we do not currently consider energy ( or species ) coupling for thermal emission processes such as electron positron annihilation to a neutrino antineutrino pair , we compute an emissivity based on the thermal content of the matter ignoring any final state neutrino blocking . to limit the neutrino energy density to the equilibrium value ( where neutrino - antineutrino annihilation rates are in equilibrium with the thermal pair production rates ) , we use kirchhoff s law to derive an _ effective _ absorption opacity for neutrino antineutrino annihilation from the thermal emissivity , @xmath83 where @xmath84 is the thermal energy density of neutrinos with energy @xmath48 and @xmath85}+1)$ ] is the equilibrium neutrino distribution function with chemical potential @xmath86 . as we shall see , this method performs well at predicting the thermal neutrino flux of the heavy - lepton neutrinos during the preexplosion phase . in ` nugr1d ` , we first update the hydrodynamic variables to the @xmath87-th timestep . we then compute the neutrino opacities and emissivities associated with the updated hydrodynamic variables . we update the radiation field operator - split . the flux term is solved explicitly , using the radiation moments of the @xmath88-th timestep . we calculate the neutrino matter interaction terms using the @xmath87 radiation moments via a local implicit update . with the @xmath87 radiation energy density source term , we then update the energy density and electron fraction of the matter . we use 24 energy groups , with lowest - energy group centers at 0.5mev and 1.5mev , and then spaced logarithmically up to 200mev for @xmath12 , @xmath13 and @xmath20 . we note that for the highest energy bins it occasionally occurs that the evolved neutrino flux vector exceeds the evolved neutrino energy density . this tends to occur in the most dynamic phases of our simulations and where the opacities vary significantly from one zone to the next . when this is the case we limit the neutrino flux to the neutrino energy density . we extract the radiation quantities in the coordinate frame at a radius of 500 km . we employ the most recent non - rotating solar - metallicity single - star model set from the stellar evolution code ` kepler ` @xcite . this model set contains the presupernova configuration of 32 stars ranging in zams mass from 12@xmath34 to 120@xmath34 . we denote individual models by @xmath89wh07 , where @xmath90 corresponds to the integer zams mass of the model , e.g. , @xmath91wh07 is the @xmath92-@xmath34 model of this model set . in @xcite , we investigated this and other model sets in the context of black hole formation . under the assumption of a failed core - collapse supernova , we found a strong empirical relation between the properties of the presupernova structure and the evolution of the failing supernova , e.g. , the time to black hole formation . this led to a clear prediction : if we observe black hole formation in a failed core - collapse supernova via neutrinos , the lifetime of the protoneutron star ( and thus of the neutrino signal ) relays direct information about the presupernova structure . however , such a prediction , ( _ i _ ) requires a failed supernova , which may not be the norm , and ( _ ii _ ) has a strong dependence on the nuclear equation of state . the empirical parameter introduced in @xcite is the compactness of the progenitor , measured at the time of core bounce . it is an inverse measure of the radial extent of a given mass coordinate at the time of bounce , @xmath93 where @xmath94 is the radial coordinate that encloses a baryonic mass of @xmath45 at the time of core bounce . in @xcite , we chose @xmath95 , since this is the relevant mass scale for black hole formation , i.e. , a typical maximum baryonic mass at which a range of eos can no longer support a neutron star against gravity . in this study , we primarily use @xmath96 . the motivation for this is that during the postbounce preexplosion phase , the relevant mass scale , especially for models with relatively small compactness , is much less than 2.5@xmath34 . in this study , we choose 1.75@xmath34 because this is close to the average baryonic mass inside the shock at 200300ms after bounce for all models : in the two extreme models that span the space in compactness parameter ( model s12wh07 , [ @xmath97 and @xmath98 , on the lower end ; model s40wh07 [ @xmath99 and @xmath100 on the upper end ) , the baryonic mass accreted through the shock at 250 ms after bounce is 1.45 m and 2.05 m , respectively . we further justify our motivation of using @xmath96 over @xmath101 in . in , we plot both @xmath96 and @xmath101 versus zams mass for all 32 considered models . @xmath96 is provided in for all models . for , one notes that while @xmath96 and @xmath101 differ quantitatively , there is no significant qualitative difference between them . the overall trends transcending individual models remain , including the two regions of high compactness near 2225@xmath34 and 3545@xmath34 . @xmath96 simply provides a more fine - grained parameterization at the lower mass scale relevant in the first few hundred milliseconds after bounce . note , however , that there are a few models that have similar @xmath101 , but rather different density structure at small enclosed masses and radii and , hence , a different @xmath96 . models @xmath10214wh07 and @xmath10216wh07 are examples . and @xmath101 . the mapping between zams mass and precollapse structure is highly non - monotonic , making the former an ill - suited parameter for describing progenitor structure in core collapse simulations . ] in this study we perform core collapse simulations with each progenitor and two eos . we use the eos of @xcite with a nuclear incompressibility of 220mev . the ls220 eos is based on a compressible liquid - drop model of the nucleus . of the publicly available nuclear eos , the ls220 eos best matches the constraints from nuclear theory and astrophysical observations ( see fig . 1 of @xcite and @xcite ) . we also employ the relativistic mean field eos of @xcite that is based on the tm1 parameter set . it is very different from the ls220 eos . the maximum neutrino - less @xmath103-equilibrium cold neutron star gravitational masses are 2.04@xmath34 and 2.24@xmath34 for the ls220 and hshen eos , respectively . the radius of a neutrino - less @xmath103-equilibrium cold neutron star with a gravitational mass of 1.4@xmath34 using the ls220 eos is 12.7 km . for the hshen eos , the corresponding radius is 14.6 km . for details on our particular implementation and the treatment of the low - density eos , we refer the reader to @xcite . the eos tables , reader and interpolation routines are available from http://www.stellarcollapse.org . since our implementation of neutrino transport is new and approximate , a comparison with published results of full boltzmann neutrino transport is warranted . this will allow us to test the ability of our code to reproduce the neutrino luminosities and spectral properties in the preexplosion phase . we compare ` nugr1d ` with the results of @xcite , a comparison study between two boltzmann neutrino transport codes ] . the two codes , @xcite , and ` vertex ` @xcite , approach the neutrino transport problem in very different ways . their results compare very well in the newtonian limit , but show significant quantitative differences in the general relativistic case . subsequent modifications to the approximate general relativistic potential used in ` vertex ` @xcite have since removed many of the quantitative differences between the codes . the general relativistic test case of @xcite was the collapse and early postbounce evolution of a @xmath39-@xmath34 ( at zams ) solar - metallicity progenitor of @xcite , referred to as model @xmath10215ww95 in the following . they employed the ls180 eos @xcite and a baseline set of neutrino - matter interactions , including coupling of energy groups via inelastic scattering processes @xcite . we repeat their test here , using the same initial conditions and eos , with our current approximations and compare the neutrino observables . we emphasize again that the current version of ` nugr1d ` lacks inelastic neutrino electron scattering and velocity dependent transport terms . both are included in the simulations of @xcite . our transport scheme evolves only the zeroth and first moment of the neutrino distribution function , using an analytic closure to truncate the series of moment equations , whereas @xcite solve the full boltzmann equation for neutrino transport . 15ww95 progenitor . these luminosities ( top panels ) and energies ( bottom panels ) correspond to the comparison study of @xcite . the left panels contain results for @xmath12 , the right panels show @xmath13 ( thick lines ) and @xmath20 ( thin lines ) results . the inset plot in the upper left panel shows the @xmath12 luminosity around core bounce . shown in solid black lines are luminosities and root mean squared energies obtained with ` nugr1d ` . the blue dashed lines and red dashed - dotted lines are the results from @xcite using the ` vertex ` code @xcite and code @xcite , respectively . the thin dashed black lines ( labeled as ` nugr1d ` * ) are curves from a test simulation using profile data from the ` vertex ` simulation 5ms before bounce as starting data . a detailed discussion of the differences is provided in the text . ] in , we show the luminosities ( top panels ) and root mean squared energies ( bottom panels ) of three neutrino species : @xmath12 ( left panels ) , and @xmath13 and @xmath20 ( right panels ) . both the luminosity and root mean squared energies are defined in @xcite , section 4 . the black solid lines are the results obtained with ` nugr1d ` , the red dashed - dotted lines are the predictions of , and the blue dashed lines are the ` vertex ` results . overall , the agreement is good , however there are are several systematic differences : ( _ i _ ) the magnitudes of the @xmath12 and @xmath13 luminosities in the early postbounce phase predicted by ` nugr1d ` agree well with the results but they are systematically lower than the ` vertex ` results . this discrepancy , which also exists between ` vertex ` and the , has been further investigated in @xcite and has since been resolved . the updated ` vertex ` code employs an improved general relativistic potential and gives comparable amplitudes to , and hence ` nugr1d ` . ( _ ii _ ) the time when the silicon oxygen interface ( located at a baryonic mass coordinate of 1.43@xmath34 in model @xmath10215ww95 ) accretes through the shock , which is marked by a sharp drop in the @xmath12 and @xmath13 luminosities , is earlier ( at @xmath104ms ) in our simulations than in the simulations of @xcite ( @xmath105ms in the ` vertex ` simulations and @xmath13 luminosities begins around the same time as in the ` vertex ` simulations . ] ) . we have carried out a number of tests to try to understand this difference . we have varied the zero point of the internal energy , which affects the relativistic enthalpy entering the momentum and energy equations and alters the collapse time . we have attempted different mappings of the presupernova stellar structure to our eulerian grid , we have tested variations in precollapse electron capture by parameterizing @xmath82 as a function of density in the collapse phase @xcite , and we have replaced the low density compositions with a pure silicon gas ( which replaces the nse equivalent of mainly iron and therefore better represents the low density matter ) . none of these tests led to a change of the postbounce time to silicon oxygen interface accretion by more than @xmath510ms . without currently having the modeling technology to test it , we suspect that the apparent difference may be due to the only other obvious difference between ` nugr1d ` and ` vertex ` : our lack of detailed neutrino physics during the collapse phase . two facts lead us to this conclusion . the collapse time discrepancy between the ` vertex ` ( @xmath106ms ) and ( @xmath107ms ) and ` nugr1d ` ( @xmath108ms ) can be explained by the lack of inelastic scattering . @xcite gives a @xmath109ms in when neutrino - electron inelastic scattering is neglected . this collapse time difference , @xmath540ms , is the difference in the silicon - oxygen interface accretion times . also , when we map in the ` vertex ` profiles from @xcite at 5ms before bounce to ` nugr1d ` and continue the evolution we reproduce the postbounce time that the silicon - oxygen interface reaches the shock . we show the result of this simulation with a thin dashed curve , marked as nugr1d * , in . it is worth noting that there may be additional differences due to the simplistic treatment of the eos at low densities . below a density of @xmath110g@xmath111 , ` vertex ` replaces the nuclear statistical equilibrium ( nse ) eos with an eos that specifically depends on the composition of the matter . assumes all matter below this density is silicon . both have some treatment of nuclear burning . in ` gr1d ` , as discussed in @xcite , we assume nse compositions from the nuclear eos . at densities below the validity regime of the nuclear eos , we take the compositions at the lowest density point from the nuclear eos and use the timmes eos @xcite . for reference , the initial density of the silicon oxygen interface is 0.41.0@xmath112g@xmath111 , where the range represents the extent in density space . ( _ iii _ ) the @xmath12 and @xmath13 root mean squared energies , predicted by ` nugr1d ` agree very well with the boltzmann transport results during the postbounce phase . the difference seen in the @xmath20 root mean squared energy is similar to that observed by @xcite and @xcite when investigating the effects of inelastic neutrino electron scattering . in the postbounce evolution , this interaction is expected to predominately affect the @xmath20 neutrino . we currently ignore this process in ` nugr1d ` and note that the @xmath20 luminosity predicted by ` nugr1d ` still agrees well with the full boltzmann results . ( _ iv _ ) another difference between the evolution in ` nugr1d ` and the full boltzmann transport results arises in the collapse phase . the lack of velocity terms and inelastic @xmath113 scattering significantly effects the composition of the inner core . in simulations with inelastic @xmath113 scattering , neutrinos from electron capture on free protons down - scatter off of electrons to lower energies . since the optical depth is lower , these neutrinos can then escape , deleptonizing the core . in our simulations , these high - energy neutrinos can not down - scatter and therefore can not escape . deleptonization is suppressed until later phases , when central density and temperature are higher . the lack of velocity - dependent terms delays full trapping by neutrino advection that would normally begin to occur at @xmath114g@xmath111 until nuclear densities , allowing for further deleptonization . at bounce the central value of @xmath82 are @xmath50.22 compared to @xmath50.29 in ` vertex ` and . the root mean squared @xmath12 energies predicted by ` nugr1d ` are higher during the prebounce phase , because the neutrinos do not experience the down - scattering via inelastic neutrino electron scattering . overall , we find that differences in the inner core structure and composition at bounce do not strongly present themselves in the neutrino signal after the collapse phase . therefore , for this study , we find the current version of ` nugr1d ` to be acceptable , since our primary focus is the neutrino signal of the preexplosion accretion phase . we do note , however , that the lack of energy - coupling terms in our transport can cause qualitative differences near black hole formation . when energy - coupling terms , such as gravitational redshift , are included , then the @xmath12 and @xmath13 luminosities drop off at times very close to black hole formation , as seen in @xcite . this effect can be captured with ` gr1d ` s leakage scheme in which it is trivial to include redshift terms @xcite . capturing redshift in an energy - dependent transport scheme requires energy - group coupling . we have also compared our m1 scheme to the results of @xcite for the 40@xmath34 model from @xcite using the ls180 eos . in this model a black hole forms within 500ms of bounce . we find differences in the @xmath12 and @xmath13 luminosities of @xmath51020% in the last @xmath550ms , but good agreement in the early postbounce phase . rcccccccccccc @xmath10212 & 0.235 & 19.24 / 14.19 / 7.73 & 1.02 / 0.92 & 2.13 / 1.78 + @xmath10213 & 0.383 & 22.18 / 16.58 / 8.76 & 1.25 / 1.09 & 2.53 / 2.07 + @xmath10214 & 0.537 & 25.19 / 19.35 / 9.24 & 1.36 / 1.20 & 3.06 / 2.49 + @xmath10215 & 0.580 & 25.51 / 19.59 / 9.17 & 1.30 / 1.16 & 3.13 / 2.59 + @xmath10216 & 0.338 & 18.91 / 13.72 / 8.20 & 1.11 / 0.95 & 2.00 / 1.68 + @xmath10217 & 0.383 & 19.93 / 14.54 / 8.57 & 1.20 / 1.02 & 2.13 / 1.78 + @xmath10218 & 0.738 & 28.66 / 22.26 / 10.30 & 1.55 / 1.36 & 3.62 / 2.92 + @xmath10219 & 0.544 & 23.36 / 17.49 / 9.37 & 1.43 / 1.23 & 2.67 / 2.18 + @xmath10220 & 0.944 & 29.39 / 22.70 / 11.08 & 1.79 / 1.55 & 3.64 / 2.91 + @xmath10221 & 0.325 & 18.48 / 13.41 / 8.09 & 1.04 / 0.91 & 1.95 / 1.65 + @xmath10222 & 0.972 & 30.06 / 23.29 / 11.26 & 1.82 / 1.58 & 3.76 / 3.00 + @xmath10223 & 1.256 & 42.00 / 33.88 / 14.68 & 2.29 / 2.00 & 5.99 / 4.67 + @xmath10224 & 1.167 & 39.14 / 31.35 / 13.66 & 2.16 / 1.89 & 5.44 / 4.24 + @xmath10225 & 1.040 & 32.30 / 25.28 / 11.87 & 1.93 / 1.68 & 4.15 / 3.28 + @xmath10226 & 0.727 & 24.97 / 18.84 / 10.10 & 1.59 / 1.38 & 2.88 / 2.33 + @xmath10227 & 0.783 & 25.67 / 19.45 / 10.21 & 1.61 / 1.40 & 3.01 / 2.42 + @xmath10228 & 0.640 & 25.86 / 19.83 / 9.54 & 1.36 / 1.21 & 3.16 / 2.62 + @xmath10229 & 0.556 & 23.24 / 17.38 / 9.43 & 1.45 / 1.22 & 2.64 / 2.17 + @xmath10230 & 0.760 & 26.61 / 20.30 / 10.19 & 1.60 / 1.40 & 3.20 / 2.58 + @xmath10231 & 0.687 & 25.23 / 19.09 / 9.95 & 1.56 / 1.35 & 2.96 / 2.40 + @xmath10232 & 0.883 & 28.10 / 21.58 / 10.70 & 1.71 / 1.49 & 3.43 / 2.75 + @xmath10233 & 0.965 & 30.36 / 23.57 / 11.28 & 1.82 / 1.58 & 3.82 / 3.04 + @xmath10235 & 1.129 & 35.93 / 28.50 / 12.77 & 2.06 / 1.79 & 4.83 / 3.79 + @xmath10240 & 1.328 & 48.32 / 39.45 / 17.51 & 2.49 / 2.21 & 7.23 / 5.75 + @xmath10245 & 1.300 & 46.42 / 37.81 / 16.56 & 2.42 / 2.14 & 6.86 / 5.44 + @xmath10250 & 0.701 & 25.60 / 19.42 / 10.00 & 1.57 / 1.32 & 3.02 / 2.45 + @xmath10255 & 0.577 & 23.56 / 17.65 / 9.52 & 1.48 / 1.25 & 2.69 / 2.20 + @xmath10260 & 0.461 & 21.52 / 15.88 / 9.07 & 1.33 / 1.12 & 2.36 / 1.95 + @xmath10270 & 0.755 & 25.83 / 19.58 / 10.17 & 1.60 / 1.39 & 3.04 / 2.46 + @xmath10280 & 0.591 & 23.75 / 17.80 / 9.63 & 1.48 / 1.26 & 2.71 / 2.22 + @xmath102100 & 0.792 & 29.90 / 23.33 / 10.59 & 1.61 / 1.42 & 3.82 / 3.07 + @xmath102120 & 0.474 & 22.31 / 16.64 / 9.02 & 1.33 / 1.15 & 2.52 / 2.07 + [ tab : results ] we perform core collapse and early postbounce evolutions of all 32 models introduced in using both the ls220 and the hshen eos . in we present the three neutrino luminosities and average energies for each model and eos as a function of postbounce time . we do not expect a clear trend in the neutrino observables with zams mass . however , we do expect trends based on the presupernova structure of the star , which is well encapsulated by the compactness parameter introduced in [ sec : models ] . in the top set of panels we show simulations run using the ls220 eos . in the bottom panels , simulations performed with the hshen eos are shown . to highlight that there is indeed a trend with presupernova structure , we color - code individual models according to their compactness parameter . the mapping between line color and @xmath96 is provided on the right . to more directly highlight the eos dependence , we include the luminosity and average energies of two models run with the ls220 eos in the hshen eos panels with thick dashed lines . these models , @xmath10212wh07 and @xmath10240wh07 , have the lowest and highest compactness parameter in our model set , respectively . we find that there is little variation in the peak luminosity of the @xmath12 neutronization burst signal . for all 32 models simulated using the ls220 ( hshen ) eos , the peak amplitude varies by less than 3% ( 5% ) from the average . this reflects the universal nature of the collapse of the inner core @xcite . after the neutronization burst , the postbounce luminosities of all species increase systematically with increasing compactness parameter . models with higher @xmath96 have higher temperatures throughout the protoneutron star @xcite . this increases the diffusive neutrino luminosity and is best seen in the @xmath20 luminosities . the postbounce accretion rate also increases with the compactness parameter . higher accretion rates , and the deeper gravitational potential due to the higher protoneutron star mass , increase the accretion luminosity , which is most directly reflected in the @xmath12 and @xmath13 signals . after roughly 100ms , the average energy of the emitted neutrinos also shows an increasing trend with the compactness parameter . the matter temperature at the neutrinosphere is higher in models with larger compactness , therefore a higher average neutrino energy is observed at infinity . the neutrino luminosities and average energies from simulations using the hshen eos are systematically lower than the luminosities and average energies from simulations of the same model run with the ls220 eos . this is clearly seen in the bottom set of panels in : models @xmath10212wh07-ls220 and @xmath10240wh07-ls220 have luminosities and average energies that are comparable to or larger than in the corresponding hshen models . for a fixed accretion rate ( or fixed progenitor model ) , the location of the neutrinosphere of each species influences the emitted luminosities and spectra . in models evolved with the stiff hshen eos , the neutrinospheres are located systematically at larger radii and lower matter temperatures than in models run with the softer ls220 eos . for example , in the @xmath10212wh07 simulations , the rosseland - mean @xmath12 neutrinospheres at 200ms after bounce have radii and temperatures of @xmath535.3 km and @xmath54.86mev ; and @xmath539.5 km and @xmath54.46mev for the ls220 and the hshen eos , respectively . the larger neutrinosphere radii are responsible for the lower accretion luminosity since the latter is set essentially by the product of the mass accretion rate and the gravitational potential at the protoneutron star surface @xcite . the latter is located at larger radii in simulations using the hshen eos . the differences in the neutrinosphere radii and temperatures between the ls220 and hshen eos also give an explanation for the systematically lower average neutrino energies seen in the hshen simulations . matter at larger radii has been compressed less and therefore is cooler . this leads to average neutrino energies that can be up to 5mev lower for the hshen eos than for the ls220 eos for the same progenitor model ( see ) . the difference in the neutrino luminosities and average energies between the two eos is largest for models with @xmath115 , in which the high accretion rates lead to the accumulation of @xmath52@xmath34 of material inside the shock within @xmath5200300ms of bounce . in the case of the ls220 eos , this leads to very high temperatures throughout the protoneutron star as it becomes more and more compact and closer to gravitational collapse to a black hole . in our simulations , the most compact model @xmath10240wh07 ( @xmath99 , @xmath116 ) forms a black hole 503ms after bounce . the slightly less compact model @xmath10245wh07 ( @xmath117 , @xmath118 ) forms a black hole 563ms after bounce . the high temperatures present in the ls220 simulations at these times will not be obtained until postbounce times @xmath119 in models using the hshen eos40wh07 and @xmath10245wh07 models are @xmath51.3s and @xmath51.4s , respectively , when using the hshen eos . ] . + most neutrino detectors are most sensitive to the electron antineutrino luminosity through the dominant ibd interaction , @xmath120 . in the left panel of , we consider the cumulative emitted @xmath13 energy for each model using the ls220 eos . we color code the models based on their compactness parameter and include two reference models that use the hshen eos , model @xmath10212wh07 and model @xmath10240wh07 . the graphs shown in this panel are the integral of the graphs shown in the top center panel of . it is obvious that the cumulative amount of @xmath13 energy emitted during the preexplosion phase strongly correlates with the compactness of the progenitor model . for example , the amount of emitted @xmath13 energy from model @xmath10240wh07 ( @xmath99 ) is _ always _ between two and three times of that of model @xmath10212wh07 ( @xmath97 ) . we make this point more quantitative in the center and right panels of . in the center ( right ) panel we plot the cumulative emitted @xmath13 energy at 100 , 200 , 300 , and 400ms after bounce for both eos as a function of @xmath96 ( @xmath101 ) . for reference , we present a subset of these numbers in . we see a very clear correlation that depends only weakly on the chosen eos . note , however , that for models with small compactness parameter ( @xmath121 ) , the correlation between the total emitted @xmath13 energy and the compactness parameter is not as strong after 400ms of postbounce evolution . comparing the center and right panels justifies our choice of @xmath96 over @xmath101 as explained in . the onset of an explosion will break the correlation observed in . once it is launched , the accretion luminosity effectively turns off and only the diffusion luminosity remains . one also expects this diffusion luminosity to show a correlation with the compactness of the progenitor , since the remnant protoneutron star s thermodynamic conditions , such as the central entropy and its mass , are essentially set by the presupernova structure . however , it is currently unclear whether one should obtain a correlation between explosion time and the compactness parameter . clarification will require a more complete understanding of the core - collapse supernova explosion mechanism and may require extensive parameter studies with fully self - consistent three - dimensional simulations . and is provided in . the dashed lines are results for models @xmath10212wh07 and @xmath10240wh07 run with the hshen eos . in the inset we show the cumulative ibd interactions as a function of @xmath96 for each model and eos at four postbounce times : 100 , 200 , 300 , and 400ms . ] we use the publicly available software ` snowglobes ` ] to predict the neutrino signal observed in earth - based neutrino detectors . ` snowglobes ` @xcite , which in turn relies on ` globes ` @xcite , is a set of routines that compute the interaction rates of supernova neutrinos in user - specified detector configurations . variables include detector material ( e.g. , water , scintillator , argon , and lead ) , detector volume , detector response functions , and a host of relevant neutrino interactions . for this investigation , we consider only ibd interactions in a water cherenkov detector . we choose a detector mass of 32kt , the mass of water in super - kamiokande that is sensitive to core - collapse supernova neutrinos @xcite . for reference , we use the ` wc100kt30prct ` smearing rates and efficiencies provided with ` snowglobes ` . we construct ` snowglobes ` initial fluence data from our simulations binned in 5ms intervals . we provide these energy - dependent fluences at http://www.stellarcollapse.org/m1prog for all models , neutrino species , and both eos in 5ms intervals up to 450ms after bounce . we assume a fiducial galactic supernova distance of 10kpc . in , we show the cumulative number of interactions for each model run with the ls220 eos and , for reference , two models run with the hshen eos . the lines are color - coded according to compactness parameter . note that the vertical scale in this figure is in thousands of interactions . we note that our answers agree with the total number of expected detected interactions in super - kamiokande from a galactic core - collapse supernova at 10kpc , which is estimated to be @xmath57000 @xcite . to arrive at this number from our results , consider the lowest zams mass progenitor in our model set , model @xmath10212wh07 . after 450ms of evolution , 15b of @xmath13 energy has been radiated ( ) , which corresponds to 2000 ibd detected interactions ( ) . for 50b of released energy ( @xmath51/6 of 300b , the fiducial energy released in neutrinos over the entire cooling phase ) , one would then expect @xmath57000 detected interactions . however , as is clear from , the number of detected interactions from the next galactic supernova may be higher than this fiducial number . more importantly , the rate of interactions in the preexplosion phase will give us detailed information on the progenitor core structure . in order to more directly quantify the differences between variations in progenitor compactness and variations in the nuclear eos , we plot in the inset of the number of expected ibd detected interactions in a super - kamiokande - like water cherenkov detector at various postbounce times versus @xmath96 . there is a well defined trend : the number of ibd interactions detected in the first 100 , 200 , 300 , and 400ms increases with the compactness parameter of the models . for reference , we include the expected number of interactions at 200 and 400ms for both eos in . we find that the eos dependence of the expected number of interactions is similar to the eos dependence of the total emitted @xmath13 energy : the hshen eos leads to a lower number of interactions ( compare the inset of to the center panel of ) . the dependence on eos is somewhat stronger here , since the the lower average @xmath13 energy predicted from stiffer eos translates into a reduced cross section in earth - based detectors . in addition to the total number of interactions , a water cherenkov detector measures individual energies , and thus , allows for the reconstruction of the cumulative emitted @xmath13 energy over time . this reconstruction will depend on the detector s response function and efficiency . an additional independent path to experimentally probing the inner structure of the progenitor is via the total neutrino energy emitted in all species over the first 10s of seconds after the initial collapse . this method requires a measurement of the total fluence of neutrinos of all species , not just electron antineutrinos . examples of neutrino interactions capable of relaying such information are the mono - energetic de - excitation of a neutral - current neutrino - excitation of @xmath122c @xcite or neutrino - proton elastic scattering interactions @xcite . such measurements would require good energy resolution , a significant source of carbon and/or a low energy threshold , for example , a liquid scintillator neutrino detector . we note that even with a liquid scintillator detector , the dominant neutrino interaction is still ibd @xcite . if such a measurement was made , and there is not a significant amount of rotation ( see the discussion on rotation in ) , one can immediately infer the gravitational binding energy of the remnant , since neutrinos carry away the vast majority ( @xmath1231% of the energy is predominantly shared among the kinetic energy of the explosion , the original binding energy of the unbound stellar mantle , and the binding energy of the iron core at the onset of collapse . ] ) of the gravitational binding energy . for typical nuclear eos like the ones considered here , this results in a one - to - one mapping of the released gravitational binding energy to the baryonic mass of the remnant , and , hence , the gravitational mass of the remnant . this is most easily seen by fitting the gravitational binding energy of a cold ( t=0.1mev ) , neutrino - less @xmath103-equilibrium , non - rotating neutron star to its baryonic mass . from cold neutron star tov solutions using the ls220 eos one can obtain an empirical fit to better than 3% above a baryonic mass of 1.15@xmath34 , @xmath124 a similar fit for the hshen eos gives , @xmath125 and is accurate to 5% above baryonic masses of 1.15@xmath34 . below @xmath126 , the empirical quadratic fit is not as accurate . however , all models considered here reach a baryonic protoneutron star mass of 1.15@xmath34 within @xmath510ms of bounce . hence , we believe that the above fits are acceptable for the iron - core core collapse events considered here . we now make the assumption that an explosion launched at a particular postbounce time will result in a neutron star remnant with a baryonic mass equal to the baryonic mass that has accreted through the shock up until the time of the explosion . this neglects any late - time fallback of material onto the protoneutron star which would lead to additional neutrino emission . @xcite and @xcite predict fallback masses @xmath1275 - 10% of the initial protoneutron star remnant mass . we also neglect any asymmetric mass accretion which may occur in the early explosion phase . in , we convert the baryonic mass enclosed by the shock to the _ total _ emitted neutrino energy using and [ eq : ebind_hshen ] . we plot this for all progenitor models ( run with the ls220 eos ) as a function of the hypothetical time of explosion . as a concrete example , consider the situation where 300b ( shown as the dashed line in ) of total neutrino energy was observational inferred . this could correspond ( @xmath128 ) to a progenitor with a high compactness parameter that exploded at an early time , e.g. , model @xmath10240wh07 at 70ms or ( @xmath129 ) to a low progenitor with low compactness parameter that exploded at late times , e.g. , model @xmath10212wh07 at 400ms . if we have an estimate of the explosion time , e.g. , via characteristic features in the neutrino observables , then we can use the combined measurement to probe the progenitor core structure . this is further quantified in , where we choose three total emitted neutrino energies , 250 , 300 , and 350b , and determine the time at which the explosion must have been launched for a given compactness and the respective total emitted energy . we plot this explosion time versus @xmath96 for all models and both eos . in general , for a fixed total emitted energy , as the compactness parameter of the progenitor increases , the explosion time must decrease . for progenitors with high @xmath96 ( @xmath130 ) , there is a clear mapping between the explosion time and the compactness parameter , given a specific total emitted neutrino energy and an eos . as was the case for the total ibd rates , there is some scatter at low @xmath96 ( @xmath131 ) , but there is still an overall trend . , and the baryonic mass enclosed in the shock at any given time . this defines the explosion time to be , in a lagrangian sense , the time at which the outermost final neutron star mass element accretes through the shock . the color coding corresponds to @xmath96 , the color coding is provided in . ] shows that there is a very strong eos dependence . while the baryonic mass inside the shock as a function of postbounce time does not vary strongly with eos , the gravitational binding energy released does . based on the above empirical fits to the gravitational binding energy , the ls220 eos leads to @xmath132 more energy release than the hshen eos for the same baryonic mass . this method of determining the compactness parameter ( by combining an estimate of the explosion time and the total emitted neutrino energy ) can be used together with the methods described above ( using the postbounce , preexplosion ibd rates ) as a consistency check , or to break degeneracies , which we will discuss next in . 14% energy for a given baryonic mass . ] there are a number of degeneracies and uncertainties that may prevent fully conclusive statements regarding the mapping from detected signal to progenitor core structure . these include , nuclear eos , rotation , viewing angle , distance , and neutrino oscillations ( including collective oscillations ) . we will discuss each one of these consecutively and independently , although all may be relevant in a generic situation . _ nuclear eos _ : we have already briefly explored the dependence of the neutrino observables on the nuclear eos by taking two very different eos and comparing the emitted neutrino signal . the total number of interactions predicted to be detected from a given progenitor model varies with eos ( cf . , ) in such a way that a high compactness model paired with the hshen eos produces the same number of interactions as a model with a slightly lower compactness paired with a softer eos , like the ls220 . hence , there is a clear degeneracy between the progenitor compactness and the equation of state . however , a water cherenkov detector will not only detect a given number of ibd interactions , but also their energy distribution . from this distribution , and the detector response , one can work backwards to reconstruct both the emitted average @xmath13 energy and the luminosity . we show in how knowledge of the emitted energy spectrum can break the degeneracy with cumulative emitted @xmath13 energy and eos , provided the distance to the core collapse event is known ( see the discussion of distance uncertainties later in this section ) . for each model and eos , we plot the average @xmath13 energy as a function of cumulative emitted @xmath13 energy at three select postbounce times : 100 , 200 , and 300ms . the data from each of the 32 progenitors fall on a unique line that is parameterized by the compactness . in principle , provided very large statistics and ultimate predictions of neutrino observables , this allows one to break the eos - progenitor degeneracy between eos . in reality , achieving an instantaneous measurement of the neutrino average energy to within a few % will be difficult , and there will likely be a large suite of nuclear eos that could reproduce the observables within error . energy versus cumulative emitted @xmath13 energy each eos and progenitor . this is shown for three postbounce times , 100ms ( left panel ) ; 200ms ( center panel ) ; and 300ms ( right panel ) . solid ( open ) points correspond to the ls220 ( hshen ) eos . the degeneracy between eos and cumulative @xmath13 energy is broken when the average energy is taken into account . we find that is a robust prediction across all progenitors and postbounce preexplosion times . ] _ rotation and viewing angle _ : the effect of rotation on the neutrino signal may be more difficult to disentangle from the effects of progenitor structure . to explore the effect of rotation , we perform 1.5d simulations of rotating core collapse . these simulations treat rotation in a spherically symmetric way , using ` shellular rotation ' @xcite . this approximation only captures the spherically - averaged centrifugal effect of rotation on the matter . as the amount of rotation is increased , the protoneutron star becomes more and more centrifugally supported . increased rotation leads to lower densities and temperatures throughout the protoneutron star . this in turn effects the neutrino signal . in , we plot the @xmath13 luminosities and average @xmath13 energies determined from ` nugr1d ` for 12 rotating core collapse simulations . we use model @xmath10215wh07 paired with the ls220 eos . the initial rotation rate is assigned via @xcite @xmath133^{-1}\times 10^{16}\ \mathrm{cm}^2\ \mathrm{s}^{-1}\,,\label{eq : rotationlaw}\ ] ] where @xmath134 is the radius that encloses 1@xmath34 , which for the @xmath10215wh07 progenitor is 703 km , and @xmath135 is the specific angular momentum at infinity , in units of @xmath136 . we vary @xmath135 from 0 to 3 in increments of 0.25 . for the @xmath10215wh07 progenitor , the initial central angular velocity is @xmath137rads@xmath138 , giving a range of @xmath139 from 0 to 6.08rads@xmath138 . at bounce , the @xmath140rads@xmath138 model has a rotation period of @xmath52.4ms , which decreases nearly linearly with increasing initial @xmath135 . increased rotation leads to the overall reduction of both the @xmath13 luminosity and average energy . for example , with @xmath139 = 0 , 2.03 , 4.05 , and 6.08rads@xmath138 the @xmath13 luminosity at 100ms is @xmath555 , @xmath552 , @xmath540 , and @xmath525bs@xmath138 , respectively . while we do not show the @xmath12 and @xmath20 luminosities and average energies , they follow the @xmath13 trends . the only exception to this is the @xmath12 neutronization burst where rotation in our 1.5d approach does not significantly alter the @xmath12 luminosity . our 1.5d approach to rotation does not capture the anisotropy of the neutrino emission , rather just the angle - averaged value . @xcite considered a model with an initial rotation rate of @xmath141rads@xmath138 , roughly corresponding to our @xmath142 model . they found that the ratio of polar to equatorial luminosity can be as large as 34 and that average neutrino energies in polar regions can be harder by @xmath14312mev than on the equator . we find that it is not possible with the cumulative neutrino signals alone to break the degeneracy between rotation and the progenitor star eos combinations . the angle - averaged @xmath13 luminosity and average energy from a rotating system mimic the @xmath13 luminosity and average energy from a non - rotating system with a progenitor with lower compactness and/or a stiffer eos . in addition , the rotational energy stored in the neutron star remnant can lead to an underestimate of the total emitted neutrino energy , again mimicking a lower - compactness progenitor and/or a stiffer eos . however , all hope is not lost : stellar evolution and current constraints on pulsar birth spins ( e.g. , @xcite ) suggest that most massive stars will be spinning too slowly for rotation to have a strong effect on dynamics and neutrino signal . @xcite , for example , predict a presupernova rotation rate of 0.2rads@xmath138 for a 15@xmath34 star , which is smaller than all rotation rates considered here . if , on the other hand , the progenitor is rapidly spinning , the degeneracy in the neutrino signal may be broken by coincident observations in gravitational waves that are able to constrain the rotation rate of the collapse core @xcite . 15wh07 for various initial central angular velocities . we assign angular momentum to the @xmath10215wh07 progenitor via with values of @xmath135 ranging from 0 to 3 in 0.25 increments . this gives initial central rotation rates of 0 to @xmath56.08 rad s@xmath138 in increments of @xmath50.51 rad s@xmath138 . in our simulations , increased rotation lowers both the neutrino luminosity and the average energy . ] _ distance _ : if the electromagnetic signal is blocked , by , for example , the galactic center , we may not be able to obtain a reliable distance to the supernova . the @xmath13 flux at earth follows an inverse square law , while the spectral distribution will not change over the distances of relevance . this means that degeneracy with distance is also hard to break with the ibd neutrino signal alone for the following reason : a high - compactness progenitor at a large distance can produce the same energy - integrated flux at earth as a low - compactness progenitor . less neutrinos are emitted in the latter but the flux has not been diluted as much due to the closer distance . if the nuclear eos is not known , the observed neutrino flux and energy spectra could be associated with a range of nuclear eos distance combinations . a progenitor at a large distance could produce the same neutrino flux and energy distribution as a lower - compactness progenitor with a softer eos . here , the softer eos , which will give rise to an increased average energy , compensates for the decrease in the average energy produced by the smaller compactness . future constraints on the nuclear eos could break this degeneracy since one could more reliably associate an observed energy distribution with a particular progenitor compactness ( ) . this may even provide a distance estimate , as would a detection of the neutronization burst signal @xcite . it does not show a strong dependence on the progenitor model , but is difficult to detect since the @xmath12 cross sections in water cherenkov detectors are much lower than the ibd cross section , resulting in few interactions . some detector materials have significantly larger @xmath12 cross sections and would be better suited to detect the neutronization burst , such as liquid argon @xcite and lead @xcite . _ neutrino oscillations : _ first , considering only matter - induced neutrino oscillations , in the normal mass hierarchy , the electron antineutrino signal at earth is a composite spectrum of @xmath14470% of the original @xmath13 neutrinos and @xmath14530% of the original @xmath146 spectrum , where @xmath147 is the mixing angle between the mass eigenstates @xmath148 and @xmath149 . in the inverted neutrino mass hierarchy , the entire @xmath13 signal is replaced with the original @xmath150 signal @xcite . in either case , we still expect the total number of ibd interactions to increase with the compactness parameter of the progenitor since both the @xmath13 and the @xmath20 luminosity increase . if the hierarchy of the neutrino mass eigenstates is not determined by neutrino experiments before the next nearby core - collapse supernova , the early postbounce , preexplosion neutrino luminosities may provide an answer @xcite . this relies on the systematically different rise times of the @xmath20 and @xmath13 signals . we also point out a degeneracy in disentangling the spectral properties of the emitted neutrino spectra from the matter - induced oscillated neutrino spectra observed in earth detectors @xcite . much more troublesome are the collective neutrino oscillations that arise from coherent neutrino - neutrino forward scattering . collective oscillations are very sensitive to the energy spectra ( both the distribution and magnitude ) of all neutrino flavors and the background matter density . since the governing equations are highly non - linear , there are currently no simple analytic expressions predicting the neutrino signal at earth based on the output of core - collapse simulations . recent studies suggest that during the early postbounce , preexplosion phase , collective neutrino oscillations may be suppressed @xcite , however the community has not yet reached consensus , see , e.g. , @xcite and @xcite . the next nearby core - collapse supernova will be extremely well observed in neutrinos . super - kamiokande alone will observe @xmath5@xmath151 electron antineutrinos from a typical core - collapse supernova at a fiducial galactic distance of @xmath11 . future detectors of the scale of the proposed hyper - kamiokande may see in excess of @xmath152 interactions . such high - statistics observations will provide rich information on the neutrino signal . comparison with theoretical model predictions will allow to falsify or constrain a broad range of hypotheses in core - collapse supernova astrophysics and nuclear / neutrino physics . unexpected signal features may lead to the discovery of new physics . in this study , our focus has been on the imprint of the progenitor star s structure on the neutrino signal in the postbounce preexplosion phase of core - collapse supernovae . we have carried out a large set of spherically - symmetric radiation - hydrodynamics simulations of core collapse and the early postbounce phase with the goal of studying trends in the neutrino signal with variations in progenitor structure . our results show , in agreement with previous work ( e.g. , @xcite ) , that the @xmath12 signal from the neutronization burst emerging shortly after bounce has very little progenitor dependence , due to the universal nature of homologous inner core collapse . the neutrino signal in the postbounce preexplosion phase is determined primarily by the accretion luminosity of outer iron core , silicon shell , and oxygen shell material . the postbounce accretion rate depends on the inner structure of the progenitor at the presupernova stage . our results show that the preexplosion neutrino signal has an essentially monotonic dependence on progenitor structure described by a single parameter , the compactness @xmath153 ( where @xmath45 is a typical baryonic mass reaching the center over the timescale of interest ) . the greater a progenitor s @xmath46 , the higher are the emitted luminosities and average energies of all neutrino species . scaling in the same way is the total emitted energy in neutrinos over the entire protoneutron star cooling phase for a given explosion time at which accretion is shut off . these trends are robust and independent of the nuclear eos . they are also rather insensitive to the particular choice of the reference mass @xmath45 in @xmath46 as long as it is in the range of typical neutron star baryonic masses ( @xmath5@xmath154 ) and we find @xmath96 to be a good choice . and @xmath20 spectra . the color coded lines denote different progenitors and are generated with ` snowglobes ` based on our simulations with the ls220 eos . more details are provided in the text . we overlay the cumulative detected interactions observed in kamiokande ii from sn 1987a , assuming the first interaction denotes the time of core - bounce . we caution the reader by noting that the very low number statistics make it hard to draw any conclusions regarding the progenitor of sn1987a . ] the monotonic dependence of the preexplosion neutrino emission on progenitor compactness translates directly to the neutrino signal observed by detectors , provided collective neutrino oscillations do not lead to complicated swaps of flavor spectra that brake the dependence of the observed signal on progenitor structure . neutrino observations of the next nearby core collapse event thus may , in principle , allow quantitative constraints on the inner structure of the progenitor star . as an example with real neutrino data , we consider the early postbounce neutrino signal observed from sn 1987a by the kamiokande ii experiment @xcite . of the eleven interactions that were observed , the first four occurred within 323ms of each other . all interactions observed by kamiokande ii are consistent with being ibd interactions @xcite . we assume that the first interaction occurs at the onset of the postbounce phase , although the actual bounce time is likely somewhat earlier . in , we plot the cumulative number of detected interactions observed from sn 1987a in the first 500ms along with the ` snowglobes ` prediction from our simulations with the ls220 eos using a 2.14kt water cherenkov detector at 51.4kpc . we use the efficiencies quoted in @xcite and the smearing matrices from the ` snowglobes ` detector configuration ` wc100kt30prct ` . to show the full range of the possible effects of msw neutrino oscillations , we show the expected number of interactions assuming no oscillations and assuming a complete switch of the @xmath13 and @xmath20 spectra ( as would be the case in the inverted mass hierarchy ; the normal hierarchy is a combination of these two signals ) . during the accretion phase the expected number of detected interactions from an oscillated @xmath20 spectrum can be significantly smaller than the electron - type neutrino luminosity due to the smaller neutrino luminosity in the accretion phase . while the quantitative results obviously depend on neutrino oscillation details , the qualitative trend with @xmath46 is unbroken . comparing our predictions with the interactions observed from sn 1987a one notes ( but must keep the very small - number statistics in mind ) that either the explosion must have occurred early in the postbounce phase and/or the progenitor must have had a relatively low @xmath96 . stated another way , that data strongly disfavors a late - time explosion in a high compactness model . both of these statements are broadly consistent with the previous work of @xcite and @xcite . it is also not inconsistent with the proposed zams mass of @xmath5@xmath155 for sanduleak @xmath156 @xmath157 ( e.g. , @xcite ) , the blue supergiant progenitor star of sn 1987a . however , the mapping between zams mass and stellar structure ( i.e. , compactness ) at the presupernova stage appears to be highly non - monotonic ( cf . [ fig : xis ] ; @xcite ) . this makes it very difficult to link the observed neutrino signal to zams mass without additional constraints from classical astronomical observations in the electromagnetic spectrum . in the discussion of sn 1987a , we have ignored uncertainties regarding nuclear eos , rotation of the progenitor core , distance , and collective neutrino oscillations . all affect the preexplosion neutrino signature in ways that may be degenerate with variations in the compactness parameter . however , as we have shown , large variations in the stiffness of the nuclear eos in the density and temperature regime relevant in the preexplosion phase can be disentangled from @xmath46 via observations of the emitted neutrino spectrum . rapid rotation , which decreases both ( angle - averaged ) luminosities and average neutrino energies , can be constrained by gravitational wave observations of a galactic event ( e.g. , @xcite ) . distance uncertainties from electromagnetic observations , which translate into uncertainties in the absolute luminosities , can be reduced by exploiting the generic neutronization burst as a standard candle @xcite . collective neutrino oscillations are not yet fully understood and have not been directly incorporated in neutrino radiation - hydrodynamics simulations . they may lead to one or multiple energy - dependent swaps of spectra between flavors and could thus complicate the mapping between neutrino signal and compactness parameter . however , the recent understanding suggests that collective oscillations may not be significant in the preexplosion phase ( @xcite ; but : @xcite and @xcite ) . our goal with this study was to highlight overall trends of the preexplosion neutrino signal with progenitor structure in the limit of spherical symmetry . we expect these overall trends to be robust and to carry over to the multi - dimensional case . we did not aim at making precise and robust quantitative predictions for any individual model . these are not possible with our current spherically - symmetric radiation - hydrodynamics treatment , which neglects inelastic scattering , redshift and velocity dependent terms ( e.g. , @xcite ) . rotation is included only in an angle - averaged manner and other important multi - dimensional dynamics , in particular convection and the standing - accretion shock instability @xcite , can not be captured . future work to remove these limitations will be necessary to produce the reliable signal predictions necessary for drawing quantitative conclusions from neutrino observations of the next nearby core - collapse supernova . however , even with fully realistic and complete simulation codes , large parameter studies will be necessary to account for prevailing uncertainties in the nuclear eos and/or neutrino interaction physics . we acknowledge helpful discussions with and input from john beacom , adam burrows , luc dessart , ken nomoto , ryan patterson , kate scholberg , mark vagins , and stan woosley . cdo thanks the kavli institute for the physics and mathematics of the universe for hospitality while work on a draft of this article was carried out . the computations were performed at caltech s center for advanced computing research on the cluster `` zwicky '' funded through nsf grant no . phy-0960291 and the sherman fairchild foundation . furthermore , computations were performed on louisiana optical network infrastructure computer systems under allocation loni_numrel06 , on the nsf xsede network under allocation tg - phy100033 , and on resources of the national energy research scientific computing center , which is supported by the office of science of the u.s . department of energy under contract no . de - ac02 - 05ch11231 . eoc is supported in part by a post - graduate fellowship from the natural sciences and engineering research council of canada ( nserc ) . this research is supported in part by the national science foundation under grant nos . ast-0855535 and oci-0905046 , by the alfred p. sloan foundation , and by the sherman fairchild foundation . , a. , beroz , f. , carr , r. , duan , h. , friedland , a. , kaiser , n. , kneller , j. , moss , a. , reitzner , d. , scholberg , k. , webber , d. , & wendell , r. 2011 , snowglobes , revision 1373 , http://www.phy.duke.edu/~schol/snowglobes/
the emergence of slow modes of relaxations is a major signature of the glassy behavior . @xcite the first consequence of large equilibration time is that the properties of a glass evolve on experimental timescales , the system ages . despite the huge amount of work that has been devoted to glassy systems , a general understanding of slow relaxation processes and of their relation to macroscopic rheology is still lacking . for example , experimental and numerical data converge to a stretched exponential shape of relaxation phenomena , the so - called kohlrausch - williams - watts ( kww ) relaxation , @xcite while mode coupling theory predicts power law behavior . @xcite when it comes to extend these results to non - linear rheology , the situation is even more difficult from the theoretical point of view , since the system is driven far from equilibrium . at an applied strain rate , power law viscosities or the emergence of a yield stress , are common features of complex fluids . @xcite md simulations of lennard - jones systems @xcite and recent experiments on colloidal glasses @xcite are also consistent with such properties . recent theoretical approaches to non - linear rheology @xcite have called on ideas originating in mode coupling theory , or trap models of spin glasses . @xcite these approaches , however , fail to provide a description of non - linear rheology that is consistent with experimental observations ; @xcite moreover , these models , based only on a phase - space picture of glassiness , do not lead to a clear identification of real - space physical mechanism that are responsible for an observed macroscopic behavior . some other attempts have been directed towards phenomenological approaches , @xcite but those are still unsatisfactory , and do not offer a microscopic picture , even heuristic , of the undergoing physical mechanisms . the current work lies at the convergence point of several recent theoretical ideas in various fields . firstly , it is based on the idea by liu , nagel and coworkers , that some unique mechanism lies behind the profound similarities displayed by structural glassy materials , be they colloidal glasses , granular materials , foams , @xcite drawing on this idea , the fundemental assumption underlying the current work is that macroscopic properties of dense materials results from major _ dynamical _ properties of structural rearrangement of the contact network . another important element of the current work comes from the study of density relaxation in granular materials . @xcite under vertical tapping , granular materials present a time - logarithmic relaxation of the occupied volume ; this is in strong analogy with free - volume relaxation , commonly observed in glasses , and of central importance for aging . @xcite however , in the case of tapped granular materials , density relaxation can not result from thermal activation since there is no thermal bath : the logarithmic density relaxation has been explained by introducing an equation of motion for free - volume , which characterizes the internal state of the granular material . finally , this work uses a theory introduced recently by falk and langer to account for elasto - plastic transition in amorphous solids . @xcite the so - called shear transformation zone ( stz ) theory , originates from ideas by argon , spaepen and others to describe creep in metallic alloys ; @xcite it has been primarily developed in the framework of fracture mechanics , and provides a general scheme for jamming as resulting from a structuration of the material by creation of arrangements oriented along principal directions of an applied stress . stz theory calls to free - volume arguments , but in fact , free - volume remains a parameter and enters constants of the theory . these ideas lead me to propose a general approach to the rheology of dense materials , leading to a simple set of constitutive equations . @xcite the purpose of the this article is to elaborate the theoretical arguments underlying these equations a to detail some of their consequences . two types of internal state variables are used : free - volume , which is an isotropic property of a molecular packing , and populations of arrangements , which are related to the distortion of the contact network between molecules ( hence introduce anisotropy ) . those variables evolves dynamically by free - volume activated processes : two types of rearrangements are associated to compaction and shear motion . free - volume enters activation factors that control both types of elementary rearrangements , and it is also a dynamical quantity . kww relaxation has been shown to result from the existence of a distribution of timescales ; @xcite and it seems that this is often thought to be a necessary requirement . the first contribution of the current work is to show that it is not necessary , and that kww relaxation also result from a more simple mechanism : the interplay between free - volume logarithmic relaxation and free - volume activated shear deformation . the second contribution of this work is to establish a relation between kww relaxation , time - logarithmic density relaxation , and power law viscosity of a strongly sheared glassy material . a single parameter of the theory determines these three types of phenomena . power law viscosity will be shown to result from the existence of shear induced dilatancy . i propose one of the simplest set of equations that can account for those three properties is a single framework . the third contribution of this work is to present a theoretical framework which accounts for two coexisting mechanisms for jamming . in mode - coupling theory , jamming results from an entropy crisis : at low temperature , the system supports any applied stress because no motion of molecules is allowed . resistance to shear is therefore an intrinsic property of a material , determined by the thermodynamic parameters of the system , and results from a phase transition . in stz theory , jamming results from a structuration of the contact network : jamming requires some amount of plastic deformation to induce the creation of anisotropic structures that can support a static stress . those two mechanisms are shown here to be incorporated in a unique theory . the difference between hard and soft glassy materials will be identified with the cases when one or the other process dominates . the interplay between those two mechanisms lead to a novel interpretation of brittleness and ductility , and provide age - dependent dynamical yield criteria . the main assumption made in this work is that the out - of - equilibrium dynamics of a glass can be modeled in the framework of a mean - field approximation . by mean - field , i mean that spatially averaged quantities account for the local state of the material . constitutive equations are proposed that relate the average shear stress , to the average shear strain . the internal state of the material is determined by free - volume and by densities of arrangements ( to be defined further ) . the forces in the material are supposed to be characterized by a stress tensor , which is written @xmath0 where @xmath1 is the pressure , and @xmath2 the shear stress . the deformation tensor is defined accordingly , and composed of shear , and of isotropic deformations , directly related to dilatancy . the average free - volume per molecule is denoted @xmath3 , and is related to the total volume @xmath4 by , @xmath5 where @xmath6 is the volume per molecule of the material in random close packing , and @xmath7 is the number of molecules . under strong shear , soft glassy materials , like clays , foams , or granular materials , can achieve large deformations ; the evolution of the system is more similar to the flow of a liquid than to the deformation of a solid . in this work i present a theory of deformation that is expected to hold for large deformations of amorphous solids or soft glassy materials . since there is a long - standing matter of debate in the mechanical community about linear versus non - linear deformations , @xcite it seems necessary to clarify this issue . i show here that the notion of inherent states @xcite offers a simple physical interpretation of small and large deformations . at low temperature , a supercooled liquid or a glassy material evolves in a complex energy landscape which can be partitioned in domains of influence of local minima . for almost all geometrical configuration of the molecules , a unique inherent state is defined as the local minimum to which the systems relaxes if suddenly quenched to zero temperature , @xcite and in the absence of forcing . when a material is forced , a constant stress or strain pulls the system through the energy landscape . each geometrical configuration reached lies in the domain associated with an instantaneous inherent state . the time series of instantaneous inherent states provides a coarse - grained description of the deformation of the material , and is taken here as the definition of irreversible , plastic , deformation . the instantaneous inherent state is a natural reference state for a given molecular configuration , and deviation from the instantaneous inherent state defines an _ inherent ( shear ) deformation _ starting from a known initial configuration , the total shear deformation , @xmath9 , results both from inherent deformation , and from the _ flow _ of inherent states in the phase space . by definition , the deformation @xmath10 is measured between two inherent states : it is not a state variable , since it is defined only from the knowledge of a configuration designed as `` initial '' . @xmath8 , however , is a state variable , since it is well defined for any instantaneous configuration of the system . moreover , @xmath8 is expected to be small , since it measures the deformation between two nearby configurations in phase space ; elasticity results from a harmonic approximation for every local minimum : it permits to write the stress as given by a hooke law , @xmath11 . the equation that governs @xmath2 can finally be written : @xmath12 the rate of plastic deformation @xmath13 will come out of a statistical analysis of hopping motion in the phase - space . it will be written as a function of the state variables of the system , @xmath2 , @xmath3 , an others , to define the constitutive law of the material . this work relies on the idea that , in dense materials , macroscopic deformation results from free - volume activated rearrangements . this is expected because temperature is so low that thermal activation becomes irrelevant , and activated processes are controlled by entropic fluctuations of the free - volume . a rearrangements occurs if sufficient space exists in the neighborhood of a given point ; the transition probabilities are thus determined by the size distribution of voids . in order to work at a mean - field level , the volume distribution is characterized by the single scalar value @xmath3 ; in order to write activation factors , it is necessary to provide an ansatz for the distribution of voids in the material . in the original works by cohen and turnbull on free - volume theory , it was argued that the voids in a material are given by a poisson distribution . @xcite in their picture , the motion of a molecule in a glassy material is permitted only if it can move into a hole of size , say @xmath14 . the probability to find a hole larger than @xmath14 , is proportional to @xmath15 : this leads to estimate a diffusion constant @xmath16 , which is a basis of free - volume approaches . important differences , however , exist between various uses and interpretations of free - volume theory . falk and langer use free - volume activation factors in the definition of stz theory . @xcite their argument borrowed from original ideas by edwards and coworkers for granular materials : @xcite the only extensive variable is supposed to be the volume @xmath4 ; the number of states available to the system is roughly proportional to @xmath17 , where @xmath18 is an arbitrary constant with dimension of a volume ; the entropy is defined as , @xmath19 and the intensive quantity , analogous to temperature , is @xmath20 : @xmath21 free - volume thus enters activation factors of the form @xmath15 . in stz theory , this discussion turns out to be somewhat formal since free - volume is taken constant : @xmath3 is absorbed in the phenomenological constants of the resulting equations . it also leads to complicated expressions for transition rates as functions of the applied stress ( because @xmath14 is supposed to be a function of @xmath2 ) . however , free - volume should vary when the material dilates or contracts . this idea has been considered recently by several authors in the analysis of density relaxation of granular materials subjected to vertical tapping . @xcite in these works , free - volume is understood as a purely dynamical quantity , and relaxes by the motion of single grains in holes created by volume fluctuations . a poisson distribution of volume fluctuations is assumed at all times : this leads to an equation of motion of the form , @xmath22 \quad . \label{eqn : vf:0}\ ] ] the activation factor @xmath23 $ ] is the probability that a hole of size larger than @xmath14 is present at a given point in the material ; @xmath24 is determined by the tapping frequency . the current work borrows from those different lines of thought . drawing on the works initiated on granular compaction , free - volume is a dynamical quantity ; voids are expected to redistribute fast enough in the material ; fast means , faster than free - volume activated collective rearrangements themselves . this is justified because redistribution of free - volume requires very tiny displacements of the molecules , and voids , considered as particles , are expected to diffuse faster than molecules themselves : around a void , there is by definition an excess of free - volume . there is an important difference , however , from the picture used in @xcite where compaction results from the motion of a single grain in a hole . here , borrowing on the argument of stz theory , compaction results from elementary rearrangements , involving several molecules at a mesoscopic scale . probabilities of transitions are estimated by entropic arguments . due to the redistribution of voids , the number of states available to the system is supposed to be proportional to @xmath17 at all times : the fast redistribution of voids allows the system to realize a quasi - equilibrium thermodynamic ensemble , determined by an intensive variable , @xmath3 . the existence of an intensive variable , associated to the fluctuations of molecular configurations , is supported by the observation of an effective fluctuation - dissipation theorem in sheared fluids . @xcite for reasons of clarity , and to be specific , i reserve the word temperature for the intensive quantity associated with the fluctuations of energy , or for the specific kinetic energy;therefore , i will not comply with to the recent use of `` effective temperature '' for something that is neither thermodynamic nor granular temperature . nevertheless , the observation of an effective fluctuation - dissipation theorem indicates that such an intensive quantity exists , without identifying what this quantity actually is . here , it is identified to @xmath3 and is associated with entropic fluctuations of molecular configurations at a mesoscopic scale . the important novelty introduced in the current work it that @xmath3 enters activation factors and is also a dynamical quantity . this is in weak analogy with granular temperature , which is a temperature , and evolves dynamically . but granular temperature couples with properties of the material only as far as it determines a collision frequency , @xcite while free - volume enters exponential prefactors in transformation rates . another important element of the current theory is that the dynamics of free - volume results both from density relaxation by activated rearrangements which exists in the absence of forcing and from dilatancy induced by the macroscopic shear flow which occurs when the system is driven out - of - equilibrium . the non - linear coupling with the mean - flow will be addressed later ; i will first continue the description of rearrangement processes . the typical elementary compaction process , corresponds to the system going through a saddle point in phase space . the rate of such transformations is controlled by free - volume fluctuations and this saddle is characterized by a typical activation volume @xmath14 , which depends on the shape of the molecules , and on the details of microscopic interactions . the activated hopping of the system through this type of saddle permits density relaxation , and leads to an equation of motion for @xmath3 of the form ( [ eqn : vf:0 ] ) . other types of elementary rearrangements occur in the material , in particular leading to elementary shear motion . before presenting a more elaborate scheme for macroscopic shear deformation , let me use a very minimal argument : shear strain is supposed to be proportional to shear stress : @xmath25 ; where @xmath26 is proportional to a free - volume activation factor . the current approach relies on the remark that the relative motion of molecules that permit compaction and shear are different ; different types of collective rearrangements , should require different activation volumes . in phase space , shear and compaction correspond to two types of saddle points , and those saddle points are entropic barriers of different heights . the activation volume for an elementary shear is denoted @xmath27 , and _ a priori _ differs from @xmath14 . this leads to @xmath28\,\sigma \quad . \label{eqn : dotgamma:0}\ ] ] one major purpose of this work is to study the consequences of introducing complexity , not through a non - trivial distribution of timescales , @xcite but through the existence of two types of entropy barriers of unequal heights . the set of equations ( [ eqn : sigma:0 ] ) , ( [ eqn : vf:0 ] ) , and ( [ eqn : dotgamma:0 ] ) will be shown to account for logarithmic relaxation of @xmath3 and for a kww relaxation of the stress after a strain increment . [ [ shear - transformation - zones . ] ] shear transformation zones . + + + + + + + + + + + + + + + + + + + + + + + + + + + equation ( [ eqn : dotgamma:0 ] ) accounts for a free - volume dependent viscosity : the material described is a liquid ( in fact a non - newtonian liquid ) . it is however , a very rough description of shear motion . the theoretical description of shear deformation that i present is directly inspired by stz theory . @xcite the first assumption on which stz theory is based is that elementary shear deformation does not occur just anywhere in the material , but in zones , where a rearrangement is permitted by the local configuration of molecules . a shear transformation zone ( stz ) is thus defined as a locus within a material , where an elementary shear is possible . another essential assumption is that stz s are two - state systems . this can be understood as follows : an elementary arrangement corresponds to the opening and closing of some contacts between molecules ; once such a rearrangement has occurred somewhere in the material , these molecules can not shear further in the same direction , but they can shear backwards . a symmetry is thus introduced by shearing motion , and ( at least ) two types of arrangements can be identified that are transformed into one another , as pictured here : = 0.0011 ( 100,60)(0,0 ) ( 50,40)(0,0)@xmath29 ( 50,0)(0,0)@xmath30 ( -50,0 ) note that an stz is not supposed to contain only four molecules , but it is a mesoscopic object that accounts for the structure of several neighboring molecules ( say of order 10 ) . to simplify the analysis , only one pair of types of arrangements is considered , aligned along the principal directions of the stress tensor . macroscopic motion results from the statistics of elementary rearrangements , and reads : @xmath31 where @xmath32 denote the number density of arrangements , and where @xmath33 denote the rate of transition @xmath34 . those rates result from activation processes determined by free - volume , but also by force fluctuations . a stz shears if there is sufficient free - volume at its location , and if the local bias of the force network triggers the shear in the appropriate direction . in this work , force and volume fluctuations are supposed to be uncorrelated ; the probabilities associated with those two types of fluctuations factorize : @xmath35 . from the preceding discussion , @xmath36 . in order to estimate the factors @xmath37 associated with the fluctuations of the force network , let me first remark that @xmath38 is a positive , increasing function of @xmath2 . moreover , an elementary shear is expected to be triggered by _ large forces_. the distributions of static forces in glassy systems have been shown to share strong similarities with the force distribution in a granular material , @xcite in which case , large forces are distributed exponentially . @xcite this justifies an exponential dependency of @xmath37 on @xmath2 : @xmath39\,\exp\left[{\pm{\sigma\over\bar\mu}}\right ] \quad,\ ] ] with @xmath40 , the update frequency of microscopic processes . note however , that the exponential form for the activation factors @xmath37 plays no role in the following , since these factors will be linearized . the parameter @xmath41 measures the typical stress that must be overcome to trigger a rearrangement . in hard - sphere systems , it is expected to be proportional to the pressure @xmath1 . in general , @xmath41 is essentially the energy required to break a bond between two molecules , times the density of bonds . here , @xmath41 will be taken constant , either because @xmath1 fixed , or because it depends on some interaction potential . at some point , @xmath41 will be set to unity , which fixes the unit of forces . [ [ dynamics - of - arrangements . ] ] dynamics of arrangements . + + + + + + + + + + + + + + + + + + + + + + + + + following @xcite , the equation of motion for the populations @xmath32 is written , @xmath42 the first two terms of the rhs account for the transitions between both types of stz . the last term accounts for the renewal of molecular configuration by the macroscopic flow : stz constantly appear and disappear during plastic deformation . the renewal of configurations is supposed to be proportional to the work of plastic deformation , @xmath43 , which is thus a common factor of creation and annhilitation terms . [ [ shear - induced - dilatancy . ] ] shear - induced dilatancy . + + + + + + + + + + + + + + + + + + + + + + + + in this work , the dynamics of rearrangements is coupled to free - volume dynamics . part of this coupling results from the factor @xmath44 that controls shear transformation , but the macroscopic flow is also expected to influence the relaxation of free - volume . free - volume dynamics is derived in analogy with the dynamics of the populations @xmath32 : plastic deformations corresponds to the evolution of the system in the phase - space along trajectories that do not necessarily permit a minimization of free - volume . in real space , plastic deformation requires the constant renewal of local arrangements , and new arrangements are local configurations of large entropy ; by definition , their density is not optimized . therefore , the renewal of configurations by the shear flow leads to dilatancy . in experiments on suspensions , for example , the system is fluidized by applying a strong shear which rejuvenates the glassy structure . @xcite in order to quantify the average dilatancy induced by the shear flow , i assume that a fraction of the work of plastic deformation , @xmath45 in used for enthalpy production , @xmath46 . the equation for @xmath3 is : @xmath47 + { { \cal a}_v\over p}\,\sigma\,\dot\gamma \label{eqn : vf:1}\ ] ] this equation is an essential element of the current theory . the non - linear coupling with the macroscopic flow shows up when the system is strongly driven out - of - equilibrium . the competition between self - relaxation and shear induced dilatancy will be shown to lead to a power law viscosity . from ( [ eqn : stz : npm ] ) , the equation of motion for the total density , @xmath48 , of arrangements reads , @xmath49 the creation and destruction terms account for an entropic mixing of molecular configurations . the previous equation shows that this mixing leads to a convergence of @xmath50 to an asymptotic value , @xmath51 ; therefore , @xmath52 defines a state of maximal entropy of molecular configurations . for this reason , i do not expect @xmath50 to be different from @xmath53 unless in very special circumstances , _ e.g. _ if some level of crystallization exists in the material . in particular , an amorphous solid is usually produced by a rapid quench of a liquid at low temperature : the system comes from a state of high entropy , therefore , in the resulting glass , @xmath50 is expected to achieve its asymptotic value . in this work , i assume that @xmath52 at all times . the dynamics of local arrangements is thus determined only by the bias @xmath54 between the populations of stz . equations ( [ eqn : stz:0 ] ) and ( [ eqn : stz : npm ] ) are written in a more suitable form by introducing the variable @xmath55 and parameters , @xmath56 , @xmath57 , and @xmath58 . it is also convenient to introduce @xmath59 , @xmath60 and @xmath61 , and to rescale free - volume as , @xmath62 the complete set of constitutive equations is finally : @xmath63\ , \left(\sinh\left({\sigma\over\bar\mu}\right)-\delta\,\cosh\left({\sigma\over\bar\mu}\right)\right)\\ \label{eqn : stzdil:2 } \dot\delta&=&{\ddpl\over\epsilon_0}\,\left(1- \mu_0\,\sigma\,\delta\right)\\ \label{eqn : stzdil:3 } \dot\chi & = & -e_1 \exp\left[-{\kappa\over\chi}\right ] + \alpha\,\sigma\,\ddpl \quad.\end{aligned}\ ] ] those equations involve two state variables , @xmath64 and @xmath20 which play very different roles . @xmath64 evolves only when the material is sheared , therefore accounts for some sort of structural memory induced by deformation imposed on the material . on the contrary , @xmath20 evolves spontaneously , and relaxes towards 0 . since this relaxation is slow , @xmath20 also encodes some sort of memory , but a very different type of memory than @xmath64 , a memory which constantly evolves as the system ages . note that the asymptotic value for the relaxation of @xmath20 ( which here is 0 ) may , in fact , be temperature dependent ; this question is eluded in the current work : it seems a very reasonable assumption for materials like suspensions if not for glasses below the glass transition , because the logarithmic slow - down of density relaxation around @xmath65 prevents it to reach asymptotic values far below ( @xmath66 ) . the parameters involved in the theory should depend on pressure and temperature , and this dependency results from particularities of a given material ; for example the update frequencies @xmath24 and @xmath40 are expected to be proportional to @xmath67 ( where @xmath68 is the temperature ) in a dense hard - sphere material . @xcite such questions , however , are not the focus of this study ; in this article , the values of the parameters are constant and can be understood as a given thermodynamical state of the system . equations ( [ eqn : stzdil:1]-[eqn : stzdil:3 ] ) present non - linear couplings between shear motion and density relaxation . before studying the complete set of equations in various tests , i start with two simple cases , when one or the other process dominates . i show here that jamming transitions are captured by equations ( [ eqn : stzdil:1]-[eqn : stzdil:2 ] ) , and free - volume relaxation by equation ( [ eqn : stzdil:3 ] ) . in the absence of shear stress , @xmath69 and @xmath70 ; the system ( [ eqn : stzdil:1]-[eqn : stzdil:3 ] ) reduces to the single equation : @xmath71 \quad,\ ] ] which accounts for density relaxation . at short time , this process is dominated by the initial value , @xmath72 : @xmath73 \quad;\ ] ] this defines a time scale for short - time free - volume relaxation : @xmath74 \quad.\ ] ] the long time behavior is estimated from the integral expression : @xmath75 \d \chi \simeq \kappa\,\exp\left[{\kappa\over \chi(t)}\right]\ ] ] at long time , the integral is dominated by the small values of @xmath20 , _ i.e. _ by the ultimate value reached , @xmath76 . therefore , @xmath77 free - volume relaxes logarithmically with time : in the absence of any forcing , the system ages . in order to isolate the stz mechanism for jamming , i now assume that free - volume is constant , at some value @xmath20 ; the evolution of the system is governed by equations ( [ eqn : stzdil:1 ] ) and ( [ eqn : stzdil:2 ] ) only , and a constant stress @xmath2 is applied . these equations admit two stationary solutions : a jammed state for which , @xmath78 , and @xmath79 a steady deformation regime , for which , @xmath80 and , @xmath81\ , \left(\sinh\left({\sigma\over\bar\mu}\right)-{1\over\mu_0 \sigma}\,\cosh\left({\sigma\over\bar\mu}\right)\right ) \quad.\ ] ] the latter solution is unstable at small stresses , in which case , the material jams . for large stresses , the jammed state loose stability , and steady deformation results . the two solutions exchange stability at the yield stress @xmath82 , which is the solution of : @xmath83 in the following , i will consider several limiting cases of the general constitutive equations ( [ eqn : stzdil:1]-[eqn : stzdil:3 ] ) . in particular , the exponential dependency of @xmath84 as a function of @xmath2 will not be studied , although this non - linearity might be important in some cases . @xcite activation factors involving @xmath2 will thus be linearized ; this is valid under the assumption that @xmath85 ; @xmath41 is then incorporated into constants by taking it to unity . it leads to : @xmath86\,\left(\sigma-\delta\right)\\ \label{eqn : stzdil : lin:2 } \dot\delta&=&{\ddpl\over\epsilon_0}\,\left(1- \mu_0\,\sigma\,\delta\right)\\ \label{eqn : stzdil : lin:3 } \dot \chi & = & -e_1 \exp\left[-{\kappa/\chi}\right]+\alpha\,\sigma\ddpl \quad.\end{aligned}\ ] ] i will also consider an _ isotropic ( liquid ) limit _ , where the structural anisotropy , measured by @xmath64 plays no role . this is the case either at timescales when the dynamics of @xmath64 is too slow , @xmath87 ( and @xmath88 ) or because the typical stress @xmath89 vanishes . in this case , the rheology is determined by a single state variable : @xmath90\,\sigma\\ \label{eqn : stzdil : iso:2 } \dot \chi & = & -e_1 \exp\left[-{\kappa/\chi}\right]+\alpha\,\sigma\ddpl \quad.\end{aligned}\ ] ] the sets of equations ( [ eqn : stzdil : lin:1]-[eqn : stzdil : lin:3 ] ) and ( [ eqn : stzdil : iso:1]-[eqn : stzdil : iso:2 ] ) will be helpful to emphasize the roles of @xmath64 versus @xmath20 . many aspects of glassiness will be shown to be captured by equations ( [ eqn : stzdil : iso:1 ] ) and ( [ eqn : stzdil : iso:2 ] ) . jamming , however , is not captured at this level of simplification , and requires the existence of the variable @xmath64 . a word about the preparation of the sample : at time @xmath91 , the system is quenched from a highly mixed state , in which , @xmath88 and @xmath20 is large . the sample then ages during a waiting time , @xmath92 without any forcing . after this relaxation , some test is performed on the system . two main experimental procedures consist in forcing the stress , or forcing the deformation . if a stress , @xmath2 , is applied , the plastic deformation @xmath93 is determined by constitutive equations , and the total deformation is , @xmath94 . if the deformation , @xmath95 , is forced , the evolution of @xmath2 is given by equation ( [ eqn : sigma:0 ] ) , and couples to constitutive equations . if a small perturbation is applied to the sample , the response of the system results from first order terms in the linearized constitutive equations ( [ eqn : stzdil : lin:1]-[eqn : stzdil : lin:3 ] ) , or ( [ eqn : stzdil : iso:1]-[eqn : stzdil : iso:2 ] ) . in relaxation processes , however , the linear approximation is not always valid at long times , @xcite because , the activation factors can become small , possibly smaller than non - linear terms . in the current theory , all non - linear terms involve the work of plastic deformation @xmath96 . therefore , the linear analysis is valid so long as this work is small compared to relaxation processes . stress relaxation is monitored after a small strain step . the step can be understood as a ramp in the deformation , which occurs on a very small time interval around @xmath92 . the strain @xmath97 determines the initial value of the stress , @xmath98 . during the relaxation , @xmath99 ; the subsequent evolution of the stress is governed by @xmath100 , and accounts for the plastic adaptation of the material to the imposed strain . the relaxation modulus , @xmath101 is the correlator associated with rearrangement processes : it measures the correlation between the initial state of the material , and its state at any following time , @xmath102 . assuming that the diffusion of molecules is dominated by collective rearrangements , stress relaxation can then be compared with the decay of self - intermediate scattering functions , which measure the spatial decorrelation resulting from the diffusion of particles in the dense medium . @xcite the long - time behavior of correlation functions is one of the major properties of glassy materials . @xcite ( see also recent reviews by debenedetti and stillinger , @xcite or angell _ et al_. @xcite ) . in the isotropic limit , the constitutive equations are ( [ eqn : stzdil : iso:1 ] ) and ( [ eqn : stzdil : iso:2 ] ) . stress relaxation is governed by the linearized equations : @xmath103\sigma\\ \dot\chi & = & -e_1 \exp\left[-{\kappa\over\chi}\right ] \quad.\end{aligned}\ ] ] to determine the relaxation modulus , @xmath104 , the only relevant initial condition is the initial value of @xmath20 , when the strain is applied ; this value is also determined by the initial free - volume @xmath72 at time @xmath91 and by the waiting time @xmath92 . moreover , either @xmath105 or @xmath106 can be incorporated in the time scale , therefore , only two effective parameters determine all the possible outcomes of those equations : @xmath65 , and @xmath107 . while @xmath65 measures the relative heights of entropy barriers , @xmath107 measures the relative update frequencies of shear and compaction processes . at long time , the relaxation of @xmath20 is given by ( [ eqn : vf : relax ] ) , and stress relaxation verifies : @xmath108\sigma \simeq -\mu\,e_0\ , \left({e_1\,t\over\kappa}\right)^{-1/\kappa}\sigma \quad.\ ] ] the response modulus @xmath109 is , @xmath110\ ] ] with an exponent @xmath111 which is directly related to the ratio @xmath65 between the heights of entropy barriers : @xmath112 the validity of linear approximation is checked by considering the time - dependency of the work of plastic deformations @xmath96 ; this term should be compared with the first term is free - volume relaxation , @xmath113\propto { \kappa\over t } \quad.\ ] ] the stress relaxation , leads to the following evolution for the work of plastic deformations : @xmath114 \quad.\ ] ] if @xmath115 , or @xmath116 , the relaxation of @xmath96 is dominated by the exponential factor ; if @xmath117 , the stress goes to a constant , but @xmath118 decays faster than @xmath119 . in all cases , the non - linear term decays faster than the linear term , which justifies the linear approximation at all times . for @xmath116 , _ i.e. _ @xmath115 , the stress undergoes kww relaxation . this property results directly from the existence of two types of saddle points of unequal heights , when hopping motion is controlled by a logarithmically relaxing free - volume . measured values of exponent @xmath111 in glassy materials , ranging from @xmath120 to @xmath121 @xcite are consistent with , @xmath122 $ ] . the ratio @xmath65 is related to the shape of molecules , and to the details of their interactions . there is no _ a priori _ reason , from a phase space point of view , to assume a particular value for @xmath65 . in this work i will systematically study both cases in order to draw a complete picture of the phenomenology described by the proposed constitutive equations . if @xmath117 , then @xmath123 , and the previous calculation shows that the stress assumes a non - vanishing asymptotic value after relaxation . in this case , @xmath124 , and the stress relaxation towards its ultimate constant value can be further estimated as a power law : @xmath125\\ & \simeq & \exp\left[-\|a\|\,t_w^{-\|\beta\|}\right]\;\left(1+\|a\|\,t^{-\|\beta\|}\right ) \quad.\end{aligned}\ ] ] interpreting @xmath126 as a correlation function , it also means that the material never completely decorrelates from its initial state . this non - ergodicity is an indication of the glass transition as understood in mode - coupling theory @xcite ; it can be characterized by the asymptotic value @xmath127 , the edwards - anderson parameter : @xmath128 \quad,\ ] ] which increases with the age of the sample . such a dependency of the edwards - anderson parameter of the age of the sample has been observed experimentally by bonn and coworkers on a colloidal glass at very low concentrations . @xcite at short time , if free - volume can be assumed large or constant , the exponential term @xmath129 $ ] does not vary very much , and stress undergoes an exponential relaxation : @xmath130\;\left(t - t_w\right)\right ] \quad.\ ] ] this relaxation occurs on a time scale of order , @xmath131 \quad,\ ] ] which should be compared with the time scale of free - volume relaxation , @xmath132 . if @xmath133 , free - volume relaxes faster than stress , and the long time , free - volume controlled , stress relaxation can be observed . if @xmath134 , the stress relaxes exponentially before free - volume achieves sufficiently low values to slow down stress relaxation . the crossover between exponential and kww relaxation is determined by , @xmath135 or , @xmath136 \sim { \chi_0\over e_1}\;\exp\left[{\kappa\over\chi_0}\right ] \quad.\ ] ] for a large initial free - volume , this reduces to , @xmath137 kww relaxation can be observed for a larger value of @xmath105 , or for sufficiently small initial value of the free - volume , @xmath138 , ( that is for sufficiently large waiting time @xmath92 ) . in order to look at the short and long time response of the system , equations ( [ eqn : sigma:0 ] ) , ( [ eqn : stzdil : iso:1 ] ) and ( [ eqn : stzdil : iso:2 ] ) are integrated numerically with a runge - kutta algorithm with adaptive timestep . the system is prepared in a dilute state , @xmath139 , which is allowed to relax during time @xmath92 , in the absence of any forcing ( @xmath140)=0 $ ] ) . at time @xmath92 , a strain @xmath141 is applied and produces an elastic response , @xmath142 . the dynamics of free - volume and stress are then monitored . [ [ crossover - stress . ] ] crossover stress . + + + + + + + + + + + + + + + + + in a first series of tests , @xmath92 is set to 0 , and the relaxation spectra are displayed for different values of the parameter @xmath105 . parameters @xmath143 , @xmath144 and @xmath145 are set to 1 ; @xmath146 and @xmath147 . the results are displayed figure [ fig : relax1 ] . for the first value @xmath148 , only exponential relaxation is observed because all the stress has been released before @xmath20 entered the logarithmic relaxation . the value @xmath149 corresponds to the crossover @xmath150 : kww relaxation is observed at later stages of the relaxation , with a non - vanishing values of @xmath126 . for a given @xmath72 , the stress release during the exponential part of the relaxation is proportional to @xmath106 , hence inversely proportional to @xmath105 . the plateau in the relaxation spectrum can be observed only for sufficiently large values of @xmath151 so that @xmath126 is still observable by the time @xmath20 enters the logarithmic relaxation . these relaxation curves are very similar to those observed in md simulations , @xcite or in experiments ; @xcite the general appearance of those curves depends sensibly on the parameters considered . from a practical point of view , it means that it depends on the time window that is available , numerically or experimentally . = 0.005 ( 100,110)(5,0 ) ( 10,5 ) ( 5,95)(0,0)@xmath152 ( 5,50)(0,0)@xmath126 ( 90,2)(0,0)@xmath153 [ [ aging ] ] aging + + + + + the crossover between exponential relaxation can also be studied by varying the initial free - volume . this amounts to varying the waiting time @xmath92 for a given @xmath154 . in order to observe a plateau in the relaxation spectrum , parameter @xmath105 is set to 20 , and parameters @xmath143 , @xmath144 and @xmath145 are set to 1 ; @xmath146 . in order to obtain an exponential relaxation for short waiting times , and to observe a wide variety of behavior , the initial value of @xmath20 is set to 100 . the results are displayed figure [ fig : relax2 ] . with this choice of parameter , the curves resemble experimental data . @xcite such curves strongly depend on the available experimental window , and on the width of the crossover region which is unobservable in a log - lin plot for smaller values of @xmath151 . to emphasize this point , another set of relaxation spectra is presented figure [ fig : relax ] for values of the parameters : @xmath155 , @xmath156 , and @xmath146 . the initial condition is @xmath157 and waiting times are : @xmath158 . the relaxation spectra are presented in log - lin , log - log , and log - log(log ) plot ; the crossover between exponential and kww relaxation is observable around values of the relaxation modulus of order @xmath159 , which is featureless in a log - lin plot . = 0.005 ( 100,110)(5,0 ) ( 10,5 ) ( 5,95)(0,0)@xmath152 ( 5,50)(0,0)@xmath126 ( 90,2)(0,0)@xmath160 = 0.005 ( 100,110)(5,0 ) ( 10,5 ) ( 5,95)(0,0)@xmath126 ( 5,50)(0,0)@xmath161 ( 90,2)(0,0)@xmath160 [ [ ergodic - to - non - ergodic - transition . ] ] ergodic to non - ergodic transition . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + i now consider the dependency of relaxation spectra on the parameter @xmath65 . this parameter is of primary importance in the current approach and is directly related to the existence of entropy barriers of unequal heights in the phase space . the parameters are taken to be , @xmath155 , @xmath162 , the initial condition is @xmath157 , and the waiting is set to @xmath163 . this choice is primarily motivated to obtain easily observable features on the relaxation spectrum . the results are displayed figure [ fig : relax3 ] . when @xmath65 goes to 1 , the system presents a transition to non - ergodic behavior . for @xmath117 , the asymptotic value of the stress is no longer vanishing . at the point @xmath164 , the long time relaxation of @xmath2 verifies : @xmath108\sigma \simeq -\mu\,{e_0\over e_1\,t}\sigma \quad,\ ] ] whence an asymptotic power law relaxation : @xmath165 = 0.005 ( 100,110)(5,0 ) ( 10,5 ) ( 5,95)(0,0)@xmath152 ( 5,50)(0,0)@xmath126 ( 90,2)(0,0)@xmath160 kww relaxation is also achieved with the complete set of equations ( [ eqn : stzdil : lin:1]-[eqn : stzdil : lin:3 ] ) . at time @xmath91 , @xmath166 , and no constraint is applied until time @xmath92 : @xmath140 ) = 0 $ ] . during the ramp , the deformation of the material is essentially elastic , no rearrangement occurs , hence , @xmath167 , while the imposed strain determines the initial value the stress : @xmath142 . during the ensuing relaxation , @xmath100 and , for a small deformation , @xmath168 . it results that , at all times , the quantity @xmath169 is constant : @xmath170 here , @xmath171 , but this expression may vary , depending on the preparation of a sample . to treat initial values of @xmath2 and @xmath64 in the greatest generality , the initial value for the difference @xmath172 is denoted @xmath173 ( and here , @xmath174 ) . at long times , the dynamics of @xmath64 and @xmath2 verify , @xmath175\left(\sigma-\delta\right)\\ & \simeq & -\left(\mu+{1\over\epsilon_0}\right ) \,e_0\,\left({e_1\,t\over\kappa}\right)^{-1/\kappa}\left(\sigma-\delta\right)\end{aligned}\ ] ] whence , @xmath176\ ] ] with , @xmath177 the evolution in time of the quantity @xmath172 is governed by the same equation as the dynamics of @xmath2 in the isotropic limit . therefore , all the previous discussion presented in the isotropic limit pertains to the general case , up to a constant term , which is fixed by the conservation of the quantity @xmath169 . the dynamics of @xmath64 and @xmath2 follow . in particular , the evolution of @xmath64 reads , @xmath178\over 1+\mu\epsilon_0 } \quad,\ ] ] and the relaxation modulus verifies , @xmath179\over 1+\mu\epsilon_0 } \quad.\ ] ] the isotropic expression for @xmath180 is recovered by taking @xmath181 . note that the relaxation of the stress could lead to a negative value if @xmath182 , which depends on the preparation of the sample . for @xmath116 ( _ i.e. _ @xmath115 ) , the stress undergoes a kww relaxation towards a non - vanishing value , which depends on the properties of the material and on the preparation of the sample . for @xmath123 power law relaxation is observed towards a non - vanishing value of the stress . in the current work , two origins are therefore identified for the existence of a non - vanishing stress at long times : the creation of mechanical anisotropy of the contact network ( jamming ) , or the existence of a ergodic to non - ergodic transition for the hopping motion in phase space ( freezing ) . recent observations by bonn _ et al _ indicate that for a given material , jamming could not be identified only to a loss of ergodicity , and that some amount of jamming could coexist with liquid - like behavior . @xcite such features are allowed in the current theory and show the importance of the stz mechanism for soft glassy materials . the validity of the linear approximation is checked by considering the long times relaxation of the rate of plastic deformation : @xmath183\ ] ] and the work of plastic deformation decays like @xmath13 since @xmath2 goes to a constant ; the non - linear term in equation ( [ eqn : stzdil : lin:3 ] ) is negligible at all times . when a constant strain rate @xmath184 is applied , the material is driven strongly out - of - equilibrium ; the response of the system involves non - linear terms . i study here the stationary states of a sheared material , in which , @xmath185 . in the first sections , i make a very primitive stability analysis of those stationary states , by assuming that @xmath185 is fixed at all times , and that the stress is given by the constitutive relation ( [ eqn : stzdil : lin:1 ] ) or ( [ eqn : stzdil : iso:1 ] ) ; this allows to decouple equation ( [ eqn : sigma:0 ] ) . the existence of an unstability in the complete set of equations accounts for the emergence of stick - slip motion in boundary lubrication . @xcite this question will be briefly discussed in section [ sec : hopf ] , but raises so many issues that it will be tackled thoroughly in another publication ; for the sake of completeness , the calculation of the hopf criterion is given in appendix [ app : hopf ] . in the limit @xmath186 ( no yield stress ) , the relation between shear rate and shear stress reads ( from ( [ eqn : stzdil : iso:1 ] ) ) : @xmath187 \quad.\ ] ] eliminating @xmath2 between equations ( [ eqn : stzdil : iso:1 ] ) and ( [ eqn : stzdil : iso:2 ] ) , yields : @xmath188 + \alpha { \dot\epsilon^2\over e_0}\ , \exp\left[{1\over\chi}\right ] \label{eqn : chidyn}\ ] ] which is positive iff , @xmath189 in the stationary flow , the relation between shear strain and volume reads , @xmath190 \quad.\ ] ] for future use , i introduce the critical rate of deformation , @xmath191 = 0.005 ( 110,60)(8,-8 ) ( 13,40)(0,0)@xmath192 ( 10,48)(0,0)@xmath184 ( 85,-13)(0,0)@xmath20 ( 10,-13 ) the bifurcation diagram for @xmath20 is presented figure [ fig : vf ] . if @xmath193 , then @xmath194 at all times : free - volume diverges , the system blows up . if @xmath195 , the equation for @xmath20 admits one solution : @xmath196 the existence of a stationary value for the free - volume indicates that for any finite rate of deformation , aging stops . in fact , this is not completely exact if a vanishingly small shear rate was applied , such that the associated steady free - volume is so small , that it is not accessible at reasonable timescales . as a consequence , for very small @xmath13 , the apparent viscosity saturates to a constant value this is a common feature of complex liquids ( see _ e.g. _ @xcite ) . in the steady state the value of the shear stress is , @xmath197 with , @xmath198 the system behaves like a power law fluid ; the exponent @xmath199 of the power law viscosity is related to the exponent @xmath111 of kww relaxation by , @xmath200 those two exponents are directly related to the slope @xmath65 of @xmath152 as a function of @xmath201 . the assumptions underlying the current approach do not permit to evaluate the activation volumes associated with density relaxation and shear motion . however , the single parameter @xmath65 determines two exponents associated with the linear and non - linear response of a strained material . the existence of such a relation constitutes a important result of the current theory , although the form given to this relation depends on non - generic features : how non - linear dilatancy is introduced in the equation of motion for free - volume ; how the update frequencies ( here taken constant ) are written . in a hard - sphere material , for example , collision frequencies depend on free - volume as @xmath152 , and the relation between @xmath2 and @xmath184 is expected to be modified by logarithmic corrections . molecular dynamics simulations of sheared fluid are consistent with power law rheologies for @xmath199 equal to 0.1 @xcite or 0.2 @xcite recent measurement of power law rheology have also been obtained on a colloidal glass , leading to values of @xmath199 ranging from 0.1 to 0.35 . @xcite a value of @xmath111 of order @xmath202 corresponds to @xmath203 , which seems in reasonable agreement with the currently available data . the theory of structural rearrangement described here , is expected to present some degree of universality , and hold , in a given form , for wide classes of materials . what this model shows is that there is an intimate relation between time - logarithmic density relaxation , kww relaxation , and power law rheology of a dense material , and that this relation may depend only on generic features for a given class of materials . therefore the existence of a relation between @xmath111 and @xmath199 could be tested by considering families of materials . the experimental or numerical evaluation of such a relation appears as an important mean of characterization of the coupling between two major modes of deformation . the constitutive equations are now ( [ eqn : stzdil : lin:1]-[eqn : stzdil : lin:3 ] ) . in the steady state , @xmath204 plugging expression ( [ eqn : stzdil : lin:1 ] ) for the plastic deformation in equation ( [ eqn : stzdil : lin:3 ] ) leads to the following relation between @xmath20 and @xmath2 in the steady state : @xmath205 = e_0e_1(^2 - 1_0 ) . @xmath205 using ( [ eqn : stzdil : lin:3 ] ) again , @xmath184 can then be written as a function of @xmath2 : @xmath206 for large @xmath2 or @xmath184 , the system behaves as a power law fluid , with the same exponent @xmath199 as in the isotropic limit . however , @xmath184 vanishes for a non - zero value of @xmath2 , which determines the yield stress : @xmath207 the following expression can also be obtained for @xmath184 as a function of @xmath20 : @xmath208 \over \sqrt{\alpha}\ ; \sqrt{e_0\alpha\,\exp\left[{\kappa/\chi}\right ] + e_1\mu_0\,\exp\left[{1/\chi}\right ] } } \quad,\ ] ] which is an increasing function of @xmath20 , and which has a maximum for @xmath209 : @xmath210 for @xmath211 , like in the isotropic case , the shear deformation leads to a constant increase of free - volume , and the system blows up . when a shear rate is suddenly imposed on a material , the transient dynamics often lead to a first rise of the stress before a dynamical yield stress is reached at which plastic deformations begin ; once plastic deformation occur , the stress relaxes towards its strain - rate dependent asymptotic value . @xcite these features are observed in sheared bulk solids @xcite but also in sheared thin films , @xcite granular materials @xcite or in solid friction . @xcite another generic property of transient regimes is that the dynamical yield stress depends on the age of the material : during aging , a material strengthens . these phenomena are captured by the proposed equations . strain softening and strengthening result essentially from the equation of motion for free - volume and are captured by the isotropic equations . figure [ fig : bump ] present some examples of transient dynamics obtained by numerical integration of equations ( [ eqn : sigma:0 ] ) , ( [ eqn : stzdil : iso:1]-[eqn : stzdil : iso:2 ] ) . parameters are set to @xmath212 . the initial value for the free - volume ( at time @xmath91 ) is set to @xmath213 , and the system relaxes without forcing during a time @xmath92 . at time @xmath92 , a constant strain rate is applied . to prevent the blow - up of the system , @xmath184 must be smaller than @xmath192 , which equals @xmath214 with the chosen parameters ; for this reason , the strain rate is set to @xmath215 . two different values of the parameter @xmath65 have been used to explore the two important cases when @xmath65 is larger or smaller than 1 . the stress is displayed as a function of the total time @xmath102 since preparation of the sample . = 0.005 ( 100,110)(5,0 ) ( 10,5 ) ( 5,95)(0,0)@xmath152 ( 5,50)(0,0)@xmath2 ( 90,2)(0,0)@xmath102 a dynamical yield stress clearly emerges for a sufficiently aged material , and increases with time . the yielding of the material is accompanied by a sudden rise of the free - volume allowing transformation rates to display non - vanishing values . these transients result from a rather complex interplay between the dynamics of free - volume and shear transformation . the complete study of this process may lead to many developments that would divert this work from the main questions that i would like to address , and i prefer to leave the discussion here . the stability analysis of equations ( [ eqn : stzdil : lin:1]-[eqn : stzdil : lin:3 ] ) and ( [ eqn : sigma:0 ] ) shows the existence of a hopf bifurcation which is responsible for oscillatory response to a given strain rate . this leads to some understanding of stick - slip motion in sheared lubricants @xcite or in granular materials . @xcite in this work , i mention this instability for the sake of completeness , but it raises many questions , and will require a more thorough consideration . the calculation of the hopf criterion of bifurcation is given in appendix [ app : hopf ] . in the isotropic limit , the steady plastic deformation is stable iff @xmath216 , with : @xmath217 ^ 2 \quad.\ ] ] for @xmath115 , @xmath218 : there is no hopf bifurcation . for @xmath117 , @xmath219 on the interval @xmath220 $ ] , and the steady state becomes unstable iff @xmath221 . this criterion is represented figure [ fig : hopf : iso ] . for driving strain rates above @xmath192 , the system looses stability , because there is no stationary solution for @xmath20 . below @xmath192 , the hopf criterion separates a domain of stability , above @xmath222 where the system reaches steady plastic deformation , and and domain below this curve where stick - slip motion is achieved . = 0.005 ( 110,60)(8,-8 ) ( 6,40)(0,0)@xmath222 ( 68,-13)(0,0)@xmath192 ( 85,-13)(0,0)@xmath184 ( 10,-13 ) the introduction of the variable @xmath64 in the set of equations ( [ eqn : sigma:0 ] ) , ( [ eqn : stzdil : lin:1]-[eqn : stzdil : lin:3 ] ) slightly modifies this picture ( see appendix [ app : hopf ] ) . two limit curves @xmath223 emerge , which delimit the unstable domain in the plane @xmath224 . these curves are displayed figure [ fig : hopf : lin ] along with the previous curves @xmath222 of the isotropic limit . the unstable domain is a subspace of what it is in the isotropic limit , and the system present a creeping zone , close to @xmath99 , when steady sliding is stable for very small values of the strain rate . = 0.005 ( 90,90)(0,0 ) ( 10,5 ) ( 5,75)(0,0)@xmath144 ( 70,2)(0,0)@xmath184 some examples of stick - slip motion are displayed figure [ fig : stickslip ] , with the set of parameters used in the phase diagram of figure [ fig : hopf : lin ] : @xmath225 , and @xmath226 in the isotropic limit ( solid lines ) and @xmath227 for the complete set of equations ( dashed lines ) . the fast bursts of plastic deformation are accompanied by sudden jumps in the free - volume . @xmath64 does not play a major role for the definition of the dynamical yield stress , as seen from the almost identical values of the maximum stress ; @xmath64 , however , strongly influences smaller values of the stress as it shifts upwards the unstable fixed point ; this also modifies the periodicity of stick - slip motion . = 0.005 ( 100,110)(5,0 ) ( 10,5 ) ( 5,95)(0,0)@xmath152 ( 5,50)(0,0)@xmath2 ( 90,2)(0,0)@xmath102 in a creep test , a material is tested by forcing a constant stress . since @xmath2 is constant , the elastic deformation is @xmath228 at all times , and the plastic deformation is determined by the constitutive equations . a small constant stress @xmath2 is applied at time @xmath229 . the plastic deformation of the material is determined by equations ( [ eqn : stzdil : iso:1]-[eqn : stzdil : iso:2 ] ) , for a fixed @xmath2 . at large times , the rate of plastic deformation reads @xmath230 this lead to a power law variation of the compliance , @xmath231 : @xmath232 if @xmath115 , or @xmath116 , the compliance diverges at long times ; if @xmath117 , or @xmath123 , the compliance saturates at long times : @xmath233 the system jams . this jamming is purely entropic , and does not result from a structural anisotropy of the force network . note that the asymptotic value of the compliance decays with the increasing age of the sample . the validity of the linear approximation must be checked by considering the evolution of the work of plastic deformation : @xmath234 the linear approximation is justified so long as this term is negligible , when compared to the first term in free - volume relaxation ( [ eqn : stzdil : iso:2 ] ) , which decays as @xmath235 . when @xmath117 , the work @xmath45 decays faster than @xmath235 , and therefore , the linear approximation is valid at all times ; when @xmath115 , however , this approximation fails when : @xmath236 or , @xmath237 the time @xmath238 is a typical time after which the linear approximation fails ; this time diverges when @xmath239 . it will appear clearer in the following that , when @xmath115 , free - volume presents a non - vanishing stationary value for any small applied stress . this asymptotic value is determined by the balance between density relaxation and shear induced dilatancy . the failure of the linear approximation , for @xmath115 , is due to the existence of a non - zero stationary value of @xmath20 for any applied stress . in the general case , the constitutive equations are ( [ eqn : stzdil : lin:1]-[eqn : stzdil : lin:3 ] ) . the constant applied stress deforms the material elastically , and at time @xmath240 , no plastic rearrangement has yet occurred . note that the initial value of the state variable @xmath64 depends on the preparation of the sample . in the experiments by cloitre _ et al _ , for example , the sample is prepared by applying a strong shear stress , above the yield stress during a rather long time . @xcite in this case , the initial value of @xmath64 is expected to be non - vanishing , and this leads to their observation of strain recovery . for the sake of generality , i will consider that the initial value of @xmath64 is non - vanishing : @xmath241 . in this section , i will assume that this value of @xmath64 is small enough so that the linear approximation is still valid . when plastic rearrangements occur in the material , they induce a structuration of the material , and the variable @xmath64 starts to evolve as @xmath168 . the plastic rate of deformation evolves as , @xmath242 and @xmath64 as , @xmath243 from this equation , the relaxation of the variable @xmath172 is easily obtained : @xmath244 \quad,\ ] ] with @xmath245 when @xmath115 , @xmath172 undergoes a kww relaxation towards 0 : the material relaxes towards a jammed state in which the stress is supported by structural anisotropy . when @xmath117 , @xmath172 saturates to a non - vanishing value : at long times , a fraction of the stress is supported by the bias between arrangements , measured by @xmath64 , but another fraction is supported by the loss of ergodicity resulting from the freezing of elementary shear processes . from equation @xmath168 , the compliance reads , @xmath246 ) \quad.\ ] ] the deformation relaxes to a constant value , which depends on the preparation of the sample ; the relaxation is either a kww or a power law relaxation depending on @xmath65 . note that the power law variation of the compliance that was found in the isotropic limit is recovered when @xmath247 ; for finite @xmath248 , this power law behavior shows up at short times , when @xmath64 is still far from saturation . in this section i start by considering the complete set of equations ( [ eqn : stzdil : lin:1]-[eqn : stzdil : lin:3 ] ) , and in particular , i recall how the dynamics of @xmath64 lead to jamming . then i show how this leads to an effective equation for free - volume which is valid in the isotropic limit or in the general case . a constant stress @xmath2 is applied . the dynamics of @xmath64 is governed by the following equation : @xmath249\ , \left(\sigma-\delta\right)\;\left(1-\mu_0\,\sigma\,\delta\right ) \quad . \label{eqn : deltadyn}\ ] ] since the normalized free - volume @xmath20 enters this equation only through a positive factor , the time - dependent value of @xmath20 does not change the stability analysis performed at any fixed free - volume . the bifurcation diagram for @xmath64 is drawn figure [ fig : delta ] . = 0.005 ( 110,60)(8,-8 ) ( 12,40)(0,0)@xmath64 ( 85,-13)(0,0)@xmath2 ( 10,-13 ) the dynamics of the variable @xmath64 leads to jamming for any stress smaller than the yield stress @xcite @xmath250 in which case , @xmath251 and @xmath99 . above the yield stress , the steady shear flow becomes a stable solution , in which case , @xmath252 and @xmath253\,\left(\sigma-{1\over\mu_0\,\sigma}\right ) \quad . \label{eqn : eps : creep}\ ] ] if @xmath254 , the systems jams , and the second term in the equation of motion for the free - volume vanishes . free - volume continues to relax logarithmically in time while the system does not deform . when @xmath255 , the system presents a non - zero plastic flow , and the evolution of free - volume is governed by the following asymptotic equation : @xmath188 + \alpha { e_0}\,\exp\left[-{1\over\chi}\right]\,\left(\sigma^2-{1\over\mu_0}\right ) \quad . \label{eqn : chi : creep}\ ] ] at the level of this equation , the jamming resulting from stz s shows up only through the term @xmath89 in equation ( [ eqn : chi : creep ] ) , which shifts the square of the applied stress . the stability analysis pertains essentially to the isotropic limit . from this point on , i will consider the response to a constant stress as governed by equations ( [ eqn : eps : creep ] ) and ( [ eqn : chi : creep ] ) ( which amounts to neglect the influence of @xmath64 during transients ) . i now study the free - volume dynamics which result from equation ( [ eqn : chi : creep ] ) . the material dilates iff @xmath256 , with : @xmath257}\ ] ] and contracts otherwise . the resulting bifurcation diagram for @xmath20 is drawn figure [ fig : chi:1 ] and [ fig : chi:2 ] for @xmath115 and @xmath117 respectively . for @xmath115 , free - volume dynamics admit a stationary solution for any @xmath258 $ ] , with , @xmath259 for values of the applied stress in the interval @xmath260 $ ] , the material undergoes steady plastic deformation : it is ductile . the shear deformation reads , @xmath261 and the strain rate behaves as a power of the stress for large stresses . when the applied stress varies in the interval , @xmath258 $ ] , the strain rate can display a whole interval of values @xmath262 \quad.\ ] ] for larger values of @xmath2 , above @xmath263 , free - volume diverges , leading to the break - up of the material . = 0.005 ( 110,60)(8,-8 ) ( 10,48)(0,0)@xmath2 ( 13,40)(0,0)@xmath263 ( 13,-1)(0,0)@xmath264 ( 85,-13)(0,0)@xmath20 ( 10,-13 ) the resulting dynamics for the compliance and for the free - volume are presented figure [ fig : compl : soft ] for different values of the driving stress @xmath2 . the first part of the dynamics is determined by the initial value of free - volume and the corresponding timescale . as easily seen on the bifurcation diagram ( [ fig : chi:1 ] ) , for small stresses , an initially large free - volume decays , and slows down the plastic deformation . this results in a plateau for the dynamics of the compliance ; a larger times , free - volume converges towards a constant value , and the dynamics of @xmath265 resumes a steady increase . these curves are to compare with measurements of particle diffusion : the particle diffusion in a dense material is expected to be dominated by collective processes ; the diffusive constant is then @xmath266 $ ] , and the average motion of a particle in a glassy medium is essentially governed by the same equation as @xmath265 . the features displayed of figure [ fig : compl : soft ] are amazingly similar to the curves found in several recent experiments . @xcite = 0.005 ( 100,110)(5,0 ) ( 10,5 ) ( 5,95)(0,0)@xmath152 ( 5,50)(0,0)@xmath267 ( 90,2)(0,0)@xmath153 if free - volume diverges , just before break - up , the current equations indicate that the strain rate should display a behavior of the form , @xmath268 however , this expression must be taken with caution , since the strongly out - of - equilibrium behavior of a material during break - up , when @xmath20 is large , is beyond the scope of the current theory . in particular , assumption has been made that the distribution of voids is close to an equilibrium state . during break - up , the distribution of free - volume is expected to display strong heterogeneities leading to the nucleation of fractures . for @xmath117 , and for any @xmath269 , @xmath270 : free - volume relaxes to 0 . in this case , the applied stress does not stop aging , shear deformation is vanishingly small : the material creeps . at long time , the linear term dominates free - volume relaxation : the calculation of linear response that has been presented in section [ sec : lin : iso ] is valid , and indicates that compliance saturates : the system jams . in general , part of this jamming is supported by the structuration of the material , which lowers the value of the stress that enters equation ( [ eqn : chi : creep ] ) ; the remaining stress is supported by entropic freezing of the rearrangement dynamics . the sign of free - volume relaxation can be inverted only for larger stresses , above @xmath271 ; it this case , the response depends sensitively on the value of @xmath2 and on the state of the material at the time @xmath92 when stress is applied : a larger stress is required to break an older material . the stationary state is unstable : when @xmath2 is large enough to invert the sign of free - volume dynamics , the material breaks up . in this case , there are two types of response : creep for low stress , and break - up for large values of the stress . the material is brittle . = 0.005 ( 110,60)(8,-8 ) ( 10,40)(0,0)@xmath2 ( 13,20)(0,0)@xmath263 ( 85,-13)(0,0)@xmath20 ( 10,-13 ) the resulting dynamics for free - volume and for the compliance are displayed figure [ fig : compl : hard ] . the saturation of the compliance is clearly seen for small stresses while the divergence occurs at a value at @xmath272 . the shape of these curves changes drastically in a very small interval of @xmath2 around @xmath271 because of long transient to escape from the unstable fixed point . these features are commonly observed in creep tests . @xcite = 0.005 ( 100,110)(5,0 ) ( 10,5 ) ( 5,95)(0,0)@xmath152 ( 5,50)(0,0)@xmath265 ( 90,2)(0,0)@xmath102 materials with @xmath115 are liable to display ductile behavior , but can break ; those with @xmath117 either creep or break - up , and could be classified as unconditionally brittle . the study of stz equations one a one - dimensional elasto - plastic decohesion has shown the importance of dynamics in ductile to brittle transition . @xcite the current analysis slightly contradicts those results , and indicates that two major types of materials can be identified , which display different qualitative behavior : this is in agreement with the common idea that some materials are ductile , some other brittle . in fact , for hard materials ( @xmath117 ) , the existence of a strong entropic crisis precludes dynamics of rearrangements to come into play : intrinsic properties of the material completely dominate the dynamical process of deformation . the existence of a dynamically driven transition between ductile and brittle behavior is still expected for soft materials ( @xmath115 ) for which the dynamics or arrangements can fully play its role . the interplay between dynamics and intrinsic glassy properties of material , as displayed in the current approach , seems a promising perspective . as far as fracture mechanics is concerned , this picture indicates that during the breaking of a hard glass , the motion of molecules in a material should be essentially localized around the surface of a crack , where there is some free - volume , whereas in the bulk , plastic deformation can occur only in places where a very large stress in concentrated . this is apparently related to a theoretical approach of fracture mechanics , recently introduced by brener and spatschek , @xcite where the dynamics of the fracture is controlled by diffusion of molecules along the surface . for soft glassy materials , on the contrary , the propagation of a fracture is expected to induce more easily observable plastic deformation in the bulk of the material . the current work relies on the introduction of dynamical equations for the intensive quantity associated with entropic fluctuations of molecular configurations . in this work , this quantity is identified to free - volume , but other physical pictures could be given . for example , free - volume is related to the average number of contacts per molecule ; in a foam , the deformation of the bubbles can be accounted for a effective free - volume that is determined by the packing of effective overlapping particles as those used in the simulations by durian , @xcite even though the relative proportions of a fluid mixture is kept constant ; in a colloidal suspension , the existence of electrostatic screening contributes to an effective size of the particle that depends on the packing structure . in fact , the assumption made in this work is rather general , and many situation can be expected to lead to a dynamically driven effective free - volume ; this may explain the ubiquity of glassy behavior . the theoretical approach to glassy materials can be summarized as : how ergodicity breaks ? or what ergodicity breaking is sufficient to account for slow modes of relaxation . the introduction of a whole distribution of timescales is a very general way to account for non - ergodicity . here i assume that a much weaker ergodicity breaking is sufficient , and this is directly supported by the observation that fluctuation - dissipation theorem holds in a weak ( but , not so weak ) form . this indicates that the underlying distributions of state variables should not ( in fact ) be too weird , and that an effective width accounts for its essential features . however , since the system is out - of - equilibrium , this width must , at least , become a dynamical quantity . the equation proposed appear as a very minimal first order closure of the dynamics of an underlying distribution of voids . the set of curves displayed in figure [ fig : relax3 ] is very reminiscent of the temperature dependency of relaxation spectra . @xcite obviously , the parameters of the theory should depend on temperature and other thermodynamical parameters , but here i purposely do not intend to address this question . there are ( at least ) two reasons for that : firstly , because glassy behavior is observed in many systems , in particular in suspensions or in granular materials , where thermodynamical temperature is not a relevant parameter . secondly , because it seems to me more interesting to extract all possible conclusions from the simple equations proposed without trying to incorporate them into a more specific setup . i do think that an effective dynamically driven free - volume should emerge from microscopic dynamics under minimal assumptions . therefore , it seems to me important to try to get some fundamental understanding of the consequences of such assumptions for a fixed set of parameters , without complicating the discussion . the dependency of the parameters of this theory on thermodynamic quantities is definitely an important issue to address in the future , but it first requires to have some intuition about the outcome of those equations . this work is intended to provide a grid of possible behavior displayed by simple equations , and i hope it can be helpful for further developments . finally , one major interest of the current work is , in my opinion , that is gives a unified picture where different mechanisms for jamming coexist . this coexistence of jamming with non - linear rheology has been observed experimentally , but could not be captured by earlier model of glassy rheology . moreover , it opens the way towards a description of glassy materials where thermodynamics ( that determine the parameters of constitutive equations ) and dynamics coexists . this work has primarily benefited from the support and encouragements of jean carlson and jim langer who received me among their people in the busy heights of ucsb ; i acknowledge the opportunity they gave me to pursue my ideas , and their continuous interest for my work . i have enjoyed useful discussions with them , and also with others : ralph archuleta , pascal favreau , anthony foglia , delphine gourdon , jacob israelachvili , daniel lavalle , jean - christophe nave . i also would like to thank daniel bonn for giving me early versions of his papers , which were helpful in the writing of this manuscript . this work was supported by the w. m. keck foundation , and the nsf grant no . dmr-9813752 , and epri / dod through the program on interactive complex networks . the system ( [ eqn : sigma:0 ] ) , ( [ eqn : stzdil : iso:1 ] ) , ( [ eqn : stzdil : iso:2 ] ) reduces to two equations for @xmath2 and @xmath20 , @xmath274\sigma\\ \dot\chi & = & -e_1 \exp\left[-{\kappa\over\chi}\right ] + \alpha\,e_0\,\exp\left[-{1\over\chi}\right]\,\sigma^2\end{aligned}\ ] ] and the jacobian of this dynamical system reads : @xmath275 ^ 2\cr 2\,e_0\,\alpha\,\sigma\,\phi & \left(-\kappa\,e_1\phi^\kappa + e_0\,\alpha \,{\sigma } ^2\,\phi\right)\log[\phi]^2\cr } \right ) \quad.\ ] ] in the steady state , the variable @xmath2 verifies , @xmath276 and @xmath277 takes on the value , @xmath278 for all @xmath279 , @xmath280 is a strictly increasing function of @xmath184 . in terms of @xmath281 , the eigenvalues @xmath282 of the jacobian verify : @xmath283 ^ 2 \right ) & + & \\ { \mu\,e_0\,e_1\,(\kappa+1)}\;\phi_{\rm s}^{\kappa+1}\,\ln\left[\phi_{\rm s}\right]^2 & = & 0 \quad,\end{aligned}\ ] ] and the hopf bifurcation is determined by a critical value of @xmath144 as a function of @xmath280 : @xmath284 ^ 2\ ] ] or , @xmath217 ^ 2 \quad.\ ] ] in the general case , the system ( [ eqn : sigma:0 ] ) , ( [ eqn : stzdil : lin:1]-[eqn : stzdil : lin:3 ] ) involves one additional variable ( @xmath64 ) , and the bifurcation analysis is somewhat more complicated . in the steady state , the relation between strain rate and free - volume can be rewritten as , @xmath285 since this expression is an strictly increasing function of @xmath277 , it is possible to carry out the analysis in terms of @xmath286 as an effective parameter . when the strain rate varies between @xmath287 and @xmath192 , @xmath277 varies on the interval @xmath288 $ ] . the values above @xmath214 are meaningless because there is no solution to the equations of motion . the hopf criterion can be written in the form , @xmath289 with : @xmath290 the quantity @xmath291 has the opposite sign of the real part of the complex eigenvalues of the jacobian : the steady state is stable , iff @xmath292 . for @xmath115 , and for any @xmath293 $ ] , @xmath294 is positive : the two values of @xmath144 , solutions of @xmath295 have the same sign . this sign can change only at a point where @xmath294 vanishes , that is @xmath296 or @xmath214 . the sign shared by these solutions on the interval @xmath293 $ ] can then be evaluated by taking the limit value of the coefficient @xmath297 close to 0 or 1 : it is positive . therefore , the equation @xmath295 has no real positive solution for @xmath115 : there is no hopf instability . it can be checked that there is no other instability , therefore , for @xmath115 , the system is stable . i am now considering the case when @xmath117 . to get a first idea of the phase diagram , in the parameter space @xmath298 , the quantity @xmath291 can be evaluated on the line @xmath299 : @xmath300 . on this line , the positiveness of @xmath291 is then equivalent to , @xmath301 which requires that : @xmath302 \quad.\ ] ] for @xmath303 the coefficient @xmath294 is negative , hence , equation @xmath295 has a single positive solution . on the interval @xmath304 $ ] this equation admits two solutions which are positive because @xmath297 is negative close to @xmath296 : the solution @xmath305 which vanishes at the point @xmath306 and the solution @xmath307 which exists on the whole interval @xmath288 $ ] . the unstability occurs for @xmath308 $ ] . at small shear rates , or small @xmath277 , the criterion @xmath291 is dominated by : @xmath309 whence , @xmath310 the solution @xmath307 can then be identified to : @xmath311 close to @xmath296 . the same expression was found in the isotropic limit , but here , the relation between @xmath277 and the shear rate , @xmath184 is different . the solution @xmath305 can be expanded as , @xmath312
this is the second in a series of papers in which we study high - resolution images of bl lacertae made at 15 ghz with the vlba , under the mojave program ( monitoring of jets in active galactic nuclei with vlba experiments , lister et al . , 2009 ) . in ( * ? ? ? * hereafter paper i ) we investigated a quasi - stationary bright radio feature ( component ) in the jet located 0.26 mas from the core , ( 0.34 pc , projected ) and identified it as a recollimation shock ( rcs ) . numerous components appear to emanate from this shock , or pass through it . they propagate superluminally downstream , and their tracks cluster around an axis that connects the core and the rcs . this behavior is highly similar to the results of numerical modeling @xcite , in which mhd waves or shocks are emitted by an rcs . in the simulations , the jet has a magnetic field that dominates the dynamics , and is in the form of a helix with a high pitch angle , @xmath10 . in bl lac the motions of the components are similar to those in the numerical models , and in addition the electric vector position angle ( evpa ) is longitudinal ; i.e. , parallel to the jet axis . for a jet dominated by helical field , this indicates that the toroidal component is substantial ( @xmath11 ) , a necessary condition for the comparison of the observations with the numerical simulations . hence , in paper i , we assumed that the superluminal components in bl lac are compressions in the beam established by slow- and/or fast- mode magnetosonic waves or shocks traveling downstream on a helical field . it has been common to assume that the evpa is perpendicular to the projection of the magnetic field vector b that is in the synchrotron emission region . this is correct in the frame of an optically - thin emission region , but may well be incorrect in the frame of the observer if the beam is moving relativistically @xcite . lyutikov , et al . show that if the jet is cylindrical and not resolved transversely , and if the b field has a helical form , then the evpa will be either longitudinal or perpendicular to the jet , depending on the pitch angle . this is partly seen in the polarization survey results of @xcite , where the bl lac objects tend to have longitudinal evpa in the inner jet , whereas the quasars have a broad distribution of evpa , relative to the jet direction . this suggests that in bl lacs the field may be helical , with pitch angles large enough to produce longitudinal evpa , although strong transverse shocks in a largely tangled field are also a possibility ( e.g. * ? ? ? the wide distribution of evpa values in quasars suggests that oblique shocks , rather than helical structures , might dominate the field order . however , a distribution of helical pitch angles could also explain the evpas in quasars , if symmetry is broken between the near and far sides of the jet . it has been suggested @xcite that this difference in the magnetic field is fundamental to the generic differences between quasars and bl lacs and , by inference , between fanaroff & riley class ii and i sources , respectively @xcite . bl lacs often show a bend in the jet , and the literature contains examples showing that in some cases the evpa stays longitudinal around the bend ; e.g. , 1803 + 784 ; @xcite , 1749 + 701 ; @xcite , and bl lac itself ; @xcite . in these examples the fractional polarization @xmath12 rises smoothly along the jet to values as high as @xmath13 . the field must be well - ordered for the polarization to be that high . in this paper we assume that the field is in a rather tight helix ( in the beam frame ) and that the moving patterns ( the transvese disturbances ) are alfvn waves propagating along the longitudinal component of the field . in a plasma dominated by the magnetic field , alfvn waves are transverse displacements of the field ( and , perforce , of the plasma ) , analogous to waves on a whip . the tension is provided by the magnetic field @xmath14 , and the wave velocity is proportional to the square root of the tension divided by the ( relativistic ) mass density . alfvn waves have been employed in various astronomical contexts , including the acceleration of cosmic rays @xcite , the solar wind @xcite , the jupiter - io system @xcite , turbulence in the ism @xcite , the bow shock of mars @xcite , and the solar atmosphere@xcite . in our case they are transverse waves on a relativistically - moving beam of plasma threaded with a helical magnetic field . the appropriate formulas for the phase speeds of the mhd waves are given in the appendix of paper i. changes in the ridge lines of bl lacs are also seen frequently . @xcite showed that in 1.4 years the bl lac object 0735 + 178 changed from having a `` staircase '' structure to being straight , and that there were prominent transverse motions . @xcite also studied 1803 + 784 and described various models that might explain the structure . @xcite studied the ridge line in 0836 + 710 at several frequencies and over a range of epochs . they showed that the ridge line corresponds to the maximum pressure in the jet . they discussed the concept of transverse velocity , and concluded that their measured transverse motions are likely to be caused by a `` moving wave pattern '' ; this was elaborated in @xcite . in our work here on bl lac we also see transverse motions , but their patterns move longitudinally and we identify them as alfvn waves . we calculate the resulting transverse velocity of the wave motion and show that it is non - relativistic . it has been more customary to discuss the fast radio components in a relativistic jet in hydrodynamic ( hd ) terms . we note here only a few examples of this . the shock - in - jet model @xcite was used by hughes , aller , & aller ( 1989a , 1989b , 1991 ) to develop models of several sources , including bl lac @xcite and 3c279 @xcite . @xcite recognized two threads of emission in 3c273 that they explained with kelvin - helmholtz instabilities , and this was developed more by @xcite . @xcite discussed the patterns and motions in 3c120 in terms of helical instability modes . in all these studies the magnetic field is needed of course for the synchrotron radiation , but it also is explicitly used to explain observed polarization changes as due to compression of the transverse components of magnetic field by the hd shock . but in these works the magnetic field has no dynamical role in the jet . on the contrary , in this paper , as in paper i , we assume that the dynamics in the jet are dominated by the magnetic field . the plan for this paper is as follows . in section [ sec : observations ] we briefly describe the observations . the definition of the ridge line of a jet is considered in section [ sec : ridgelines ] , and the transverse waves and their velocities , including the behavioral change in 2010 , are presented and discussed in section [ sec : waves ] . excitation of the waves by changes in the p.a . of the rcs is considered in section [ sec : excitation ] . in section [ xverse_mhd ] we identify the waves as alfvn waves , discuss their properties , and present simple models of the system . for bl lac @xmath15 , and the linear scale is @xmath16 . an apparent speed of @xmath17 corresponds to @xmath18 . for this study of bl lac we use 114 epochs of high - resolution observations made with the vlba at 15 ghz , between 1995.27 and 2012.98 . most of the observations ( 75/114 ) were made under the mojave program @xcite , a few were taken from our earlier 2-cm program on the vlba @xcite , and the rest were taken from the vlba archive . the data were all reduced by the mojave team , using standard calibration programs @xcite . following the reduction to fringe visibilities we calculated three main products at nearly every epoch : ( 1 ) : : an image , consisting of a large number of `` clean delta functions '' produced by the algorithm used for deconvolution , convolved with a `` median restoring beam '' , defined in section [ sec : ridgelines ] . ( 2 ) : : a model , consisting of a set of gaussian `` components '' found by model - fitting in the visibility plane ; each component has a centroid , an ellipticity , a size ( fwhm ) , and a flux density . the gaussians are circular when possible . the total set of components sums to the image , but in this paper we only use components that have been reliably measured at four or more epochs , have flux density @xmath19 , and can be tracked unambiguously from epoch to epoch . a typical epoch shows 4 - 6 of these `` robust '' components . the rcs is a permanent component and , together with the core , usually produces more than half of the total flux density from the jet . the centroids of the robust components for each epoch are plotted on the images in figure [ fig1 ] . + the centroid locations are measured relative to the core , which we take to be the bright spot at the north end of the source ; it usually is regarded as the optically - thick ( @xmath20 ) region of the jet . in principle , the core can move on the sky . we considered this in paper i , and concluded that any motions are less than @xmath21 in a few years , and they were ignored . our positional accuracy is conservatively estimated as @xmath22 , and in this paper we again ignore any possible core motions . ( 3 ) : : the ridge line , shown in figure 1 and discussed in section [ sec : ridgelines ] . the image , the components , and the ridge line are not independent , but each is advantageous when discussing different aspects of the source . in most cases the ridge line runs down the smallest gradient from the peak of the image , and the centroids of the components lie on the ridge line . however , when the jet has a sharp bend the algorithm can fail , as in figure [ fig1]c . this is discussed in section [ sec : ridgelines ] . the components move in a roughly radial direction , and plots of @xmath23 as well as the sky ( ra dec ) tracks are shown in paper i and in @xcite . the tracks cluster around an axis at @xmath24 and appear to emanate from a strong quasi - stationary component , c7 , that we identified as a recollimation shock ( rcs ) in paper i. the moving components have superluminal speeds ; the fastest has @xmath25 in units of the speed of light . we are dealing with moving patterns on the jet of bl lac , and in order to quantify them we first need to define the ridge line of a jet . at least four definitions have been used previously . @xcite used the line that connects the components at a single epoch , in studying 1803 + 784 . @xcite investigated three methods of finding the ridge line : at each radius making a transverse gaussian fit and connecting the maxima of the fits , using the geometrical center , and using the line of maximum emission . they found no significant differences among these procedures , for the case they studied , 0836 + 710 . they showed that the intensity ridge line is a robust structure , and that it corresponds to the pressure maximum in the jet . to quantify a ridge line we start with the image as in figure [ fig1 ] , which is the convolution of the `` clean delta functions '' with a smoothing beam . since we are comparing ridge lines from different epochs , we have used a constant `` median beam '' for smoothing , and not the individual ( `` native '' ) smoothing beams . the latter vary a little according to the observing circumstances for each epoch , and their use would effectively introduce `` instrumental errors '' into the ridge lines . the median beam is a gaussian with @xmath26 and @xmath27 each of the three parameters is the median of the corresponding parameters for all the epochs . the algorithm for the ridge line starts at the core , and at successive steps ( 0.1 mas ) down the image finds the midpoint , where the integral of the intensity across the jet , along a circular arc centered on the core , is equal on the two sides of the arc . the successive midpoints are then smoothed with a third - order spline . ridge lines are shown on the three images in figure [ fig1 ] . in figure [ fig1]a the bends in the jet are gradual and the algorithm works very well , as indeed would any of the methods mentioned above . in figure [ fig1]b there are two sharp bends and our algorithm makes a smooth line that misses the corners of the bends . in this case connecting the components would be better , if the modelling procedure actually put components at the corners . in figure [ fig1]c the jet appears to bifurcate , and our algorithm picks the west track . in this case a visual inspection of the image is required to see what is going on . in fact there is another problem with figure [ fig1]c . the image has a step to the east ( looking downstream ) about 1 mas from the core , where a short ew section connects two longer ns sections . since the restoring beam is nearly ns the details of this step can not be reconstructed . the calculated ridge line in figure [ fig1]c does not reproduce the step , but makes a smooth track . figure [ fig : ridge_all ] shows nearly all the ridge lines that we consider in this paper ; a few are not shown because they occur very close in time to another one . in all cases the rcs is located close to the semi - circle , drawn 0.25 mas from the core . successive panels are adjacent in time , although there is a 1-yr gap in the data between panels ( d ) and ( e ) . the only other substantial data gap is seen in panel ( a ) , from 1998.18 to 1999.04 . in figure [ fig : ridge_all ] the epochs are set nearly equally among the panels , with the separations picked to emphasize the various waves that are discussed below . it is important to establish the reliability of the ridge lines because our analysis rests on them , and some of the structures that we interpret as waves are smaller than the synthesized vlba beam . we first note that as with all vlbi our sampling of the @xmath28 plane is sparse , and different samplings can produce different ridge lines . to see how strong this effect is , we emulated an observation with missing antennas by analyzing a data set with and without one and two antennas , and we did this analysis both with the native restoring beams and the median restoring beam described above . the results for 2005 - 09 - 16 are shown in figure [ fig : nativevsmedian ] ; they are similar to the results we obtained for two other epochs . in figure [ fig : nativevsmedian]a we show two ridge lines , the solid one is calculated with the full data set and the dashed line is obtained when data from the sc and hn antennas are omitted . the latter calculation does not use many of the baselines , including the longest ones . the chief effect is a shift of the pattern downstream , by roughly 0.1 mas . this shift is not a statistical effect , but is mainly due to the different smoothing beams that were used for the two cases . we found that the differences in the ridge lines increased with increasing difference in the p.a.s of the smoothing beams . in figure [ fig : nativevsmedian]a the difference in p.a of the smoothing beams is 17 . in figure [ fig : nativevsmedian]b we used the median beam . in this case the curves are close with differences of typically @xmath29as out to 4 mas , where the surface brightness becomes low . beyond 4 mas the differences rise to @xmath30as . another way to investigate the reliability of the ridge lines is to examine pairs of ridge lines measured independently but close together in time . the full data set contains 10 pairs where the separation is no more than 10 days , and these are all shown in figure [ fig : pairs ] . they are calculated with the median restoring beam . note that the bottom three panels have a different vertical scale than the others . in general the comparison is very good within 4 mas of the core . panel ( i ) contains one ridge line that stops at 3.6 mas because the brightness at the ridge becomes too low ; this limit also can be seen in a few places in the other figures . panel ( i ) contains the only pair that has a continuous offset , @xmath31 . these data were taken during an exceptional flux outburst at 15 ghz in bl lac , seen in the mojave data ( unpublished ) , and roughly coincident with outbursts seen at shorter wavelengths @xcite . an extra coreshift leading to a position offset is expected with such an event @xcite . in any event , this pair appears to be different from the others , and we do not include it in the statistics . figure [ fig : histogram ] shows the histogram of separations between the paired ridge lines , after excluding those in panel ( i ) of figure [ fig : pairs ] . in forming the ridge lines a 3-pixel smoothing was used , and for the histogram we have used every third point . the median separation is @xmath32 . thus the repeatability of the ridge lines is accurate to about @xmath32 . the reliability also depends on the effect discussed in connection with figure [ fig1 ] , that the ridge - finding algorithm can smooth around a corner , and can be in error by perhaps @xmath33 . however , the error is roughly constant over short time spans , as in figure [ fig : pairs ] panel ( e ) where the sharp bend at @xmath34 is smoothed the same in the two curves . this smoothing will have little effect on calculations of wave velocity , which is our main quantitative use of the ridge lines . we ignore the smoothing in this paper . from this investigation we conclude that caution must be taken in interpreting the ridge lines , especially when comparing ridge lines obtained at different epochs , or with different frequencies . the details of the restoring beam can have a noticeable effect on the ridge line , and to avoid misinterpretation the restoring beam should be the same for all the ridge lines that are being intercompared . when considering these ridge lines it is important to keep the geometry in mind : the jet has a small angle to the line - of - sight ( los ) , and the foreshortening is about a factor of 10 ( paper i ) . also , the projected images in figure [ fig1 ] can hide three - dimensional motions . to work with skew and non - planar disturbances , we use the coordinate systems shown in figure [ fig : coords ] . east , north , and the los form the left - hand system ( x , y , z ) and the jet lies at angle @xmath35 from the los in the @xmath36 formed by the los and the mean jet axis . this plane is perpendicular to the sky plane and is at angle p.a . from the @xmath37 axis . the rotated system @xmath38 is used to describe transverse motions : @xmath39 is in the sagittal plane , @xmath40 is perpendicular to it , and @xmath41 is along the jet . by `` transverse motion '' we mean that a point on the beam has a motion in the @xmath42 plane : @xmath43 . the component @xmath44 lies in the sagittal plane and its projection on the sky is along the projection of the jet . this component therefore is not visible , although a bright feature moving in the @xmath39 direction might be seen as moving slowly along the jet . however , the @xmath45 component remains perpendicular to the los as @xmath35 or p.a . changes , and its full magnitude is always seen . thus a measured transverse motion is a lower limit . if the beam is relativistic then time compression of the forward motion must be added ; see section [ sec : transverse ] . ] some of the panels in figure [ fig : ridge_all ] show disturbances that appear to move down the jet , and at other epochs the jet is fairly quiet . we now consider several of the disturbances in detail , starting with the structures seen in figure [ fig : ridge_all ] , panel ( b ) . figure [ fig : ridge9900 ] is an expanded view of figure [ fig : ridge_all ] , panel ( b ) . it includes ridge lines for 14 consecutive epochs over a period of about 1.6 yr . beyond 1 mas the early epochs ( solid lines ) show the jet bending to the se . later epochs show the bend farther downstream , and at 2000.31 and later the jet bends to the sw before bending se . we anticipate a result from section [ sec : velocity ] and draw vector a at @xmath46 across the tracks . the intersections of vector a with the tracks are shown in the inset in figure [ fig : ridge9900 ] . the velocity implied by the line in the inset is close to @xmath17 or @xmath47 . the pattern on the ridge line is moving superluminally downstream at nearly constant velocity . we consider three possible explanations for this . * 1 ) * we see the projection of a conical pattern due to a ballistic flow from a swinging nozzle , like water from a hose . the argument against this is that line b in figure [ fig : ridge9900 ] is parallel to vector a and approximately tangent to the western crest ; this feature of the ridge lines is not radial from the core as it would be if it were a ballistic flow . in figure [ fig : ridge_all ] all the panels except ( a ) , ( b ) , and ( e ) show clearly that the excursions of the ridge lines are constrained to lie in a cylinder , not a cone . * 2 ) * the moving pattern is due to a helical kink instability that is advected downstream with the flow . in the kink the field would be stretched out and become largely parallel to the observed bends in the jet that , in this case , seem to be transverse waves @xcite . this would produce an evpa normal to the wave crest in figure [ fig : ridge9900 ] rather than longitudinal . but in bl lac the evpa tends to be longitudinal , even along the bends . in figure [ fig : pol ] we show the polarization image for 2005 - 09 - 23 , taken from the mojave website@xmath48 . similar polarization images for bl lac , at several wavelengths , are shown in osullivan & gabuzda ( 2009 , figure 19 ) for epoch 2006 - 07 - 02 . both of these epochs are part of the large wave d shown later in figure [ fig : ridge0506 ] . in these polarization images the evpa is nearly parallel to the jet out to about 5 mas and @xmath12 is high on the ridge , indicating that the magnetic field remains in a relatively tightly - coiled helix around the bend and is not nearly parallel to the axis , as it should be for an advected kink instability . wave d is the largest wave in the bl lac data , and seems to have the cleanest longitudinal polarization . at other epochs the evpa tends to be longitudinal , but can be off by up to 40 . we have only one epoch of polarization data for wave a , but that one does show an evpa that is tightly longitudinal in the bend . thus we believe that the evpa results preclude the identification of the structures seen in figure [ fig : ridge9900 ] as due to a kink instability . * 3 ) * the moving patterns are transverse mhd waves ; i.e. , alfvn waves . for this to be possible the plasma must be dynamically dominated by a helical magnetic field . this condition for the jet of a bl lac has been suggested many times ; see e.g. , @xcite , @xcite . note that we implicitly assumed the helical , strong - field case in discussing the kink instability , in the preceding paragraph , and we also assumed it in paper i. thus , we assume that the moving pattern under vector a in figure [ fig : ridge9900 ] is an alfvn wave , with velocity @xmath49 . in figure [ fig : ridge9900 ] a second wave is seen between @xmath50 and @xmath51 mas , where the ridge lines for epochs 2000.31 and later bend to the sw . the two waves in figure [ fig : ridge9900 ] can be thought of as one wave with a crest to the west . this wave is generated by a swing of the nozzle to the west followed by a swing back to the east about 2 years later , as discussed below in section [ sec : excitation ] . the 1999 - 2000 wave is displayed in a different form in figure [ fig : waves9901 ] , which shows the ridge lines from 1999.37 to 2001.97 . vertical spacing is proportional to epoch , and the axes have been rotated by 13 ; arrows at top show north and east . tick marks on the right are 0.1 mas apart . the dots show the points described later in section [ sec : velocity ] , where the slope changes , and the solid line a is a linear fit through the points , with speed @xmath52 . this wave is prominent until 2000.99 . in 2001.22 the structure has changed . there are alternate possibilities to explain this new structure , b. it may be a new wave , with the crests connected with line b ( drawn with the same slope as line a ) . in this case the wave must have been excited somehow far from the rcs . the fit of line b to the wave crests is poor and would be improved if acceleration were included , but there is not enough data for that . alternatively , structure b may simply be a relic of the trailing side of wave a , perhaps relativistically boosted by the changing geometry ( the bend ) seen in figure [ fig : ridge_all ] panel ( c ) . a third wave c is shown by the dashed line that again is drawn with the same slope . . the three lines are parallel and all have slope 0.92 mas @xmath53 . [ fig : waves9901],scaledwidth=49.0% ] panel ( c ) of figure [ fig : ridge_all ] shows the ridge lines projected on the sky for 2001 2002 . wave b from figure [ fig : waves9901 ] is seen as the bump to the east at @xmath51 mas , which moves downstream at succeeding epochs . the projected axis of the jet is curved at these epochs , and the possible acceleration noted above for wave b may simply be a relativistic effect inherent in the changing geometry . wave a in figure [ fig : waves9901 ] is barely visible in figure [ fig : ridge_all ] panel ( a ) as a gentle bump in 1999.04 , so it is first apparent in early 1999 at a distance @xmath54 mas from the core . this is reminiscent of the behavior of the components discussed in paper i ; figure 3 of that paper shows that most of the components first become visible near @xmath50 mas . wave c also appears to start near @xmath54 mas . in figure [ fig : ridge9900 ] the short arrow c shows an eastward swing of the inner jet between 2000.01 and 2000.31 . this is seen in figure [ fig : waves9901 ] in the ridge line for 2000.31 , which shows a new inner p.a . the effect of these p.a . swings on the beam is discussed in section [ sec : excitation ] . the different panels in figure [ fig : ridge_all ] show that the jet can be bent , and even when relatively straight , can lie at different p.a.s . hence there is no unique rotation angle for the ridge lines in a plot such as that in figure [ fig : waves9901 ] . the rotation angle used in figure [ fig : waves9901 ] was found by the velocity algorithm described in section [ sec : velocity ] for wave a. further examples of waves are shown in @xmath55 , omitting the extraneous ridge lines to avoid confusion . the wave motions are indicated by the arrows , which are propagation vectors derived in section [ sec : velocity ] . table [ tbl : waves ] lists the details for these waves . @xmath56 is the measured proper motion , @xmath0 is the apparent speed in units of @xmath1 , @xmath57 the wave speed in the coordinate frame of the galaxy , assuming @xmath58 , and p.a . is the projected direction of the propagation vector . the amplitude is an estimate of the projected distance ( in mas ) across the wave , perpendicular to the propagation vector . wave d is the largest such feature seen in the data . unfortunately , there was an 11-month data gap prior to 2005.71 , and the wave can not be seen at earlier times . the amplitudes of the larger waves appear to be comparable with the wavelength , as suggested for example by the inclination angle @xmath59 shown in figure [ fig : ridge0506 ] : @xmath60 . but this is an illusion caused by the foreshortening , which is approximately a factor of 10 ( paper i ) , so the deprojected value of @xmath59 is about 5 . note that this is a lower limit , since the transverse motion can have a component in the @xmath61 plane in figure [ fig : coords ] . figure [ fig : movie ] contains one frame of a movie of bl lac showing the jet motions and ridge line fits at 15 ghz . the full movie is available in the electronic version of this paper . ccccccccc epoch & n & @xmath56 & & @xmath62&@xmath63 & p.a . & & amplitude + & & ( mas y@xmath64 ) & & & & ( deg ) & & ( mas ) + a 1999.37 - 2000.99&14 & 0.92 & @xmath65 & 3.9 & 0.979 & @xmath66 & @xmath67 & 0.5 + d 2005.71 - 2006.86 & 5 & 1.25 & @xmath68 & 5.6 & 0.987 & @xmath69 & @xmath70 & 0.9 + e 2008.33 - 2008.88 & 8 & 3.01 & @xmath71 & 13.5 & 0.998 & @xmath72 & @xmath73 & 0.3 + f 2009.33 - 2009.96 & 6 & 1.11 & @xmath74 & 5.0 & 0.985 & @xmath75 & @xmath76 & 0.2 + notes . columns are as follows : ( 1 ) wave label , ( 2 ) inclusive range of epochs , ( 3 ) number of epochs , ( 4 ) apparent speed , ( 5 ) error , ( 6 ) apparent speed in units of @xmath1 , ( 7 ) speed in galaxy frame , assuming @xmath58 , ( 8) p.a . of the wave , ( 9 ) error , ( 10 ) estimated amplitude . mas , with a major axis position angle at @xmath77 , as indicated in the lower left corner of the frame . the fitted ridge line is shown as a dashed line in the image , and again as a solid line to the left of the image . these have also been linearly interpolated between the individual vlba epochs . the points of changing slope ( see section [ sec : velocity ] ) at individual vlba epochs are shown as the small symbols . at left the ridge lines are shown with different colors for the various waves . the yellow @xmath78 line on the right is advancing at the speed of light ( @xmath79 ) and is included for reference . the entire movie is available in the web version of this paper . [ fig : movie],scaledwidth=47.0% ] in figure [ fig : ridge_all ] panels ( g ) and ( h ) we see that by 2010 the earlier transverse wave activity in the jet has subsided , and that after 2010.5 the jet is well - aligned at @xmath80 with a weak wiggle . but the wiggle is not stationary . figure [ fig : ridge1013 ] shows the ridge lines plotted on axes rotated by 95 , and spaced proportionately to epoch . most of the ridge lines have a quasi - sinusoidal form . almost all the epochs show a negative peak in the inner jet , with a minimum near @xmath81 . this is a quasi - standing feature , of variable amplitude . at most epochs there is a positive peak near @xmath82 . this also is a quasi - standing feature , but less distinct than the inner one . what is causing the quasi - standing features ? the patterns can hardly be true standing waves because that requires a reflection region . a rotating helix would project as a traveling wave , as on a barber pole , so a simple barber - pole model is excluded . possible motions of the core are only about @xmath21 ( paper i ) , so any registration errors due to core motion are much smaller than the observed changes , which are up to @xmath33 . there is little indication of wave motion in figure [ fig : ridge1013 ] , at least not at the speeds seen in figure [ fig : ridge_all ] . it appears then , that during the period 2010 - 2013 , the jet was essentially straight but with a set of weak quasi - stationary patterns , with variable amplitude . we estimated the velocity of wave a in figure [ fig : ridge9900 ] in two independent ways . in the first we assume that there is a constant propagation vector , and we shift and superpose the ridge lines on a grid of ( @xmath56 , p.a . ) where @xmath56 is the speed of the wave and p.a . is its propagation direction . if the ridge lines form a simple wave , then the solution is found when the lines lie on top of each other . this is shown in figure [ fig : superpose ] , where a reasonable fit can be selected by eye . the result is @xmath83 at @xmath84 . this solution is somewhat subjective and the quoted errors do not have the usual statistical significance . as an alternative procedure to visually aligning the ridge lines , we developed a method of identifying a characteristic point on the wave , just downstream of the crest , where the wave amplitude has begun to decrease . define the slope of the ridge line as @xmath85 in pixels , where in figure [ fig : waves9901 ] , @xmath86 and @xmath37 are rotated ra and dec , and take the first downstream location where the slope exceeds @xmath87 . this point is marked with the dot @xmath88 on the ridge line for 2000.57 in figure [ fig : ridge9900 ] . the @xmath86 and @xmath37 positions vs time for these locations are then fit independently using the same methods as described in @xcite to extract a vector proper motion for this characteristic point on the wave . the two methods agree well and the analytic solution is @xmath89 at @xmath90 , and the apparent speed is @xmath91 . the propagation vector is shown in figure [ fig : ridge9900 ] and the speed and direction of the wave are listed in table [ tbl : waves ] . the table also includes @xmath63 the speed of the wave in the galaxy frame , assuming @xmath58 . this calculation assumes that the ridge lines lie in a plane ; i.e. , are not twisted . this is not neccessarily the case . rather , since the inner jet , near the accretion disk , may wobble in 3 dimensions , ( mckinney et al , 2013 ) it seems likely that the rcs may execute 3-dimensional motion and that the downstream jet will also . see section [ sec : excitation ] . note that the p.a . of the first and last propagation vectors in table [ tbl : waves ] ( -1670 , -1671 ) is the same ( to within the uncertainties ) as the p.a . of the axis ( -1666 ) defined in paper i as the line connecting the core with the mean position of the recollimation shock . in the context to be developed later , the jet acts as a whip being shaken rapidly at the rcs , and tension in the whip continually pulls it towards the mean pa . in table [ tbl : waves ] the speeds for the first , second and fourth waves are all similar at @xmath92 , but wave e ( 2008 ) is 3 times faster . wave e has @xmath93 , which is comparable to the speed for the fastest component in bl lac , @xmath94 , although the components speeds vary widely , from @xmath95 to @xmath96 @xcite . wave e is also distinguished by its polarization ; the evpa is transverse not longitudinal like the others . we defer further discussion of wave e to another paper . the ridge waves are relativistic transverse waves with apparent speeds @xmath0 from 3.9 to 13.5 times the speed of light , and we assume that they have a small amplitude . from the usual formula for apparent speed , @xmath97 and taking values of @xmath98 from table [ tbl : waves ] and using @xmath58 , we find @xmath99 for the speed of the waves in the frame of the host galaxy . we now discuss the jet motion in terms of the coordinate system @xmath100 shown in figure [ fig : coords ] . consider a transverse motion that is in the ( @xmath101 ) plane . let the beam contain a co - moving beacon that is at the origin and emits a pulse at time @xmath102 , where @xmath103 is in the coordinate frame of the galaxy . when @xmath104 yr the signal from the origin will have traveled @xmath105 down the z axis , towards the observer . also at @xmath104 the beacon has moved from the origin to the point @xmath106 where @xmath107 is the transverse speed , and @xmath108 is the longitudinal speed of the beam , both in the frame of the galaxy . at this point the beacon emits a second signal that also travels at the speed of light . in the z - direction , this signal trails the first one by ( @xmath109 ) years . the apparent transverse speed of the beacon in the direction perpendicular to the jet , in the galaxy frame , is then @xmath110 and is to be differentiated from the apparent speed @xmath111 commonly used in studies of superluminal motion , which is the apparent speed _ along _ the jet . note the close relation between equations [ vapp ] and [ vtrans ] . equation [ vtrans ] can be inverted to find @xmath112 , a lower limit to the transverse speed . for wave a in figure [ fig : ridge9900 ] we obtain an estimate for the transverse speed at @xmath113 mas by taking the transverse motion as 0.5 mas and the time interval as @xmath114 yr , giving @xmath115 and @xmath116 and , from equation [ vtrans ] with @xmath58 and @xmath117 ( paper i ) , @xmath118 . this is a model - dependent rough value , but it shows that the transverse speed is non - relativistic . this is necessary for consistency , since the derivation of the relativistic form of the mhd wave speeds shown in paper i assumes that the velocity perturbation is small . we suggested in paper i that component 7 is a recollimation shock , and that the fast components emanate from it . if this is correct , then the rcs should be a nozzle and its orientation should dictate the direction of the jet . in this section we investigate this possibility . we first note that it is not possible to make a detailed mapping between the p.a . of the rcs and the later wave shape , for two reasons . first , the algorithm for the ridge line smooths over 3 pixels ( 0.3 mas ) , and thus smooths over any sharp features in the advected pattern . the second reason is more speculative . our conjecture is that the wave is launched by plasma flowing through the nozzle and moving close to ballistically until its direction is changed by a swing in the p.a . of the nozzle . but magnetic tension in the jet continually pulls it towards the axis , and this means that it will bend , and that small - scale features will be stretched out and made smooth . we start by comparing the p.a . of the rcs with the p.a . of the downstream ridge line at @xmath50 mas . figure [ fig : pa ] shows the p.a . of the rcs measured at 15 ghz and at 43 ghz . the latter is calculated from data kindly provided by the boston university vlbi group . we used the result found in paper i , that the 15 ghz core is a blend of the first two 43 ghz components and that the 15 ghz component 7 is the rcs , as is the third 43 ghz component . we calculated the centroid of the first two 43 ghz components , to find an approximate position for the 15 ghz core , and then calculated the p.a . of the 43 ghz rcs from that centroid . the result is shown in figure [ fig : pa ] . we eliminated one discrepant point at 43 ghz , which was separated by about 20 from nearby 43 ghz points , and one discrepant point at 15 ghz . the correspondence between the two frequencies is generally good , especially after 2005.0 where the agreement is typically within 3 . this further justifies our claim ( paper i ) that the the location of this component is independent of frequency , and that it is a recollimation shock . figure [ fig : pa ] also contains the p.a . of the 15 ghz ridge line , close to @xmath119 1.0 mas . between 2005.0 and 2010.0 the ridge line p.a . lags the rcs pa , by roughly 0.6 to 1.5 yr . after 2010 the pa of both the rcs and the ridge line stabilizes , and the subsequent variations , with rms amplitude about 3 , may mainly be noise . prior to 2005.0 the variations are faster and more frequent and the lag is erratic . in places there appears to be no lag , but around 2000.0 and again around 2004.0 it is about 0.5 yr . thus it appears that the swinging in pa of the rcs is coupled to the transverse motions of the ridge line . when the rcs is swinging rapidly and strongly , as before 2005 , then so also is the ridge at 1 mas , with an irregular lag in p.a . that sometimes is about a half a year , and at other times is negligible . but when the rcs is swinging more slowly , as after 2005 , then the ridge at 1 mas is also swinging slowly , with a lag of about a year , and after 2010.0 they both are stable , with only small motions that may be dominated by measurement errors . we suggest that the large transverse waves on the ridge are excited by the swinging in p.a . of the rcs . consider wave a , seen in figure [ fig : ridge9900 ] . its crest lies near line b and moves downstream at @xmath120 . in 1999.37 the crest is at about @xmath121 mas and at @xmath120 would have been at the rcs ( @xmath122 mas ) around 1998.3 . this is in a data gap at 15 ghz , but at 43 ghz there was a peak in p.a . in mid- or late-1998 . given that in 1999 the time lag between the rcs and the ridge at 1 mas apparently was much less than 1 yr , the association between the peak in the rcs p.a . in 1998 and the crest of wave a is plausible . the fall in p.a . in 1999 and 2000 is seen as the short arrow c in figure [ fig : ridge9900 ] , and it corresponds to the upstream side of wave a. the downstream side is the advected rise in p.a . of the rcs from mid-1997 to the peak in mid - or late 1998 . the p.a . of the rcs fell from mid-1996 to mid-1997 , and we might expect to find a corresponding crest to the east on wave a , about 1 mas downstream of the main crest to the west . in fact several of the earliest ridge lines in figure [ fig : ridge9900 ] do show a minor crest to the east at about @xmath123 mas , which is 2 mas or 2 years at 0.92 mas y@xmath64 , downstream of the main crest to the west . a substantial acceleration in the wave speed would be needed for this to match . in any event , we can not speculate usefully on this because it takes place beyond 3 mas , where there is a general bend to the east at all epochs . we conclude that a plausible association can be made between the large swing west then east of the rcs between 1998.0 and 2000.1 , and wave a that is later seen on the ridgeline . a similar connection can be made for wave d , seen in figure [ fig : ridge0506 ] in 20052006 . it can plausibly be attributed to the large swing of the rcs to the east that began in 2004 and continued into 2005 . this wave does not have a crest as wave a does , but a crude analysis can be made as follows . assume that point @xmath124 on the 2005.71 ridge line is the advected beginning of the wave . with a speed of 1.25 mas y@xmath64 ( table [ tbl : waves ] ) this means that the swing to the east began around 2003.5 . this date is indicated on the abscissa in figure [ fig : pa ] . apart from one high point at 2004.1 the p.a . of the rcs falls gradually from 2003.1 until late 2004 , when it must fall abruptly to meet the first point after the data gap in 2005 . this also is seen in figure [ fig : ridge0506 ] ; the first four epochs have ridge lines that lie together and are straight at @xmath125 out to @xmath126 mas . this is consistent with the rcs p.a . being stationary from roughly 2004.7 to 2005.7 at @xmath127 . this is in a data gap , and this analysis suggests that the rcs p.a . was @xmath128 during most or all of the gap . we conclude that the large swing in p.a . of the rcs from mid-2003 until mid-2005 generated wave d , the largest wave in our data set . the p.a . of the rcs rose rapidly from 2005.7 to about 2006.5 , but the p.a . of the ridge line rose more slowly , and not as far . from 2007.0 to 2009.0 the p.a . of the inner jet was roughly constant at about -170 , while the p.a . of the rcs slowly dropped to the same value . we do not have a straightforward interpretation of this behavior . we also see in figure [ fig : ridge_all ] panels ( e ) , ( f ) and ( g ) that following the passage of wave d the jet slowly straightened out . the p.a . of the inner jet @xmath129 propagated as a low - amplitude wave , at roughly the same speed as the large waves , @xmath130 . as a further complication , during this slow straightening out of the jet we see two more low - amplitude waves . the high - speed wave e ( figure [ fig : ridge0808 ] ) has no obvious antecedent in the p.a . of the rcs . wave f ( figure [ fig : ridge0909 ] ) is seen a year after wave e , at the `` usual '' speed of 1.1 mas y@xmath64 . these waves together make a complex set of possibly twisted ridge lines , seen together in figure [ fig : ridge_all ] panel ( f ) . in section [ sec : jet_1013 ] we showed that the waves on the jet subsided in 2010 , and in 2010 - 2012 the ridge line had only a weak variable wiggle . during this time the p.a . of the rcs was essentially constant ; the variations seen in figure [ fig : pa ] may represent the errors in the measurements , which would be about @xmath131 . these variations in space and time have some regularities , as discussed in section [ sec : jet_1013 ] , but they do not appear to have a connection to the p.a . of the rcs . in paper i we saw that the component tracks all appear to come from or go through the rcs ( component 7 ) and that they lie in a window centered on @xmath132 @xcite . this now is understood in terms of the waves on the ridge lines , since the components all lie on a ridge . the jet is analagous to a whip with a fixed mean axis being shaken with small amplitudes , in various transverse directions . the whip will occupy a narrow cylinder centered on the axis , and in projection the cylinder becomes our window . in paper i we showed that the magnetic field in the jet of bl lac has a strong transverse component . we assumed that it has a helical form , and that it is likely that the field dominates the dynamics in the jet . this is the condition for the existence of mhd waves that propagate down the jet . we suggested that the moving synchrotron - emitting components are compressions set up by fast and/or slow magnetosonic waves , possibly shocks . now we introduce the third branch of mhd waves in the jet plasma , the alfvn wave , which is a transverse s ( shear ) wave , with the disturbance occurring normal to the propagation direction . in section [ sec : waves ] we showed that the moving patterns on the jet are transverse waves , and now we suggest that they are alfvn waves . the phase speed of a transverse alfvn wave is given by @xmath133 where @xmath134 is the relativistic scalar alfvn speed , given in equation a6 of paper i , and @xmath135 is the angle between the propagation direction and the magnetic field . the alfvn wave has similar propagation properties ( with respect to the magnetic field direction ) as the slow wave ; i.e. it moves along the field , but not at all normal to it ( @xmath136 ) . note that alfvn waves generally will not produce shocks in an ideal mhd plasma . we now discuss these waves in the jet and present simple models that allow us to estimate the pitch angle @xmath10 of the helix , which we define as the angle between the axis of the helix and the direction of the magnetic field when projected onto that axis . a simple relation exists for the relativistic phase speeds of the three mhd waves : @xmath137 where @xmath138 is the sound speed ( relative to the speed of light ) , and @xmath139 are the fast , slow , and transverse mhd wave speeds . equation [ relation ] may be readily verified from equation [ va ] combined with equations a1 and a2 of paper i. with this result , the three equations for the phase speeds , together with the definitions of the cusp and magnetosonic speeds in equations a3 and a4 in paper i , can be solved for the magnetosonic and alfvn speeds : @xmath140 @xmath141 finally , the propagation angle to the magnetic field @xmath135 can be found from equations [ va ] and [ alfven ] . in dealing with this system of equations we are helped with constraints on the mhd wave speeds : @xmath142 , also @xmath143 for an adiabatic sound wave in a relativistic gas . in addition , we adopt a constraint from the one - sidedness of bl lac , @xmath144 , where @xmath145 is the lorentz factor of the beam in the frame of the galaxy ; this gives a jet / counterjet intensity ratio of about @xmath146 for @xmath147 and a spectral index of @xmath148 @xcite . we assume that the three waves travel downstream in the beam frame and parallel to the jet axis . therefore , the propagation angle of all three waves is the pitch angle of the helix itself : @xmath149 . we do not , in fact , measure the wave speeds themselves but rather their apparent speeds in the frame of the galaxy . to relate these to their speeds in the beam frame we first use equation [ vapp ] and then the relativistic subtraction formula @xmath150 where the superscripts define the coordinate frame . we now have 5 input quantities to the calculation : @xmath151 , @xmath35 and @xmath145 , and with them we can calculate @xmath152 , @xmath153 , @xmath154 , @xmath10 , and the magnetosonic mach number defined as @xmath155 , where @xmath156 is the magnitude of the spatial component of the four - velocity and @xmath157 is the lorentz factor . to illustrate the relationships among the various waves we show in figure [ fig : banana ] ( the `` banana diagram '' ) the results for the specific configuration @xmath158 . these values correspond to the fastest superluminal component in bl lac ( paper i ) and to the apparent speeds of the transverse waves noted in section [ sec : waves ] above . the diagram contains quantities defined in the frame of the beam : sound speed and alfvn speed , @xmath10 the pitch angle of the helix , and @xmath159 the magnetosonic mach number . the diagram is bounded at the left and bottom by @xmath160 and @xmath161 . at the top , for @xmath162 , the boundary traces the curve @xmath163 , but for @xmath164 ( in this case ) , this curve sometimes ventures into a region where there are no solutions for @xmath10 . this region can be eliminated from the banana by continuing the curve for @xmath165 with one that satisfies the criterion @xmath166 at constant @xmath167 , as we have done here . inside the banana our conditions for magnetic dominance @xmath168 and @xmath169 are satisfied everywhere except in a thin quasi - horizontal region at top right , and in a thin quasi - vertical region at left . at the cusp at right @xmath170 , indicating a purely toroidal field and no propagating alfvn waves , regardless of the value of @xmath152 . the banana diagram is set on the plane defined by the lorentz factor of the beam and the apparent ( superluminal ) speed of the slow mhd wave , both measured in the galaxy frame . the location of the banana on this plane is set by the specific set of input parameters as on the top left . figure [ fig : banana ] shows that knowing the apparent speeds of the three mhd waves and the angle @xmath35 of the jet to the line of sight is not enough to completely determine the jet properties . we must either determine one more quantity or make an assumption about the jet system . we will make two different assumptions for the sound speed , each yielding a simple model . the cases are first , a cold jet , in which the plasma sound speed is negligible ; and the other assumes that @xmath171 . * model ( a ) : cold plasma * in paper i we investigated a model of the jet in which an observed slowly moving component with @xmath172 is due to a slow magnetosonic wave whose speed , relative to the jet plasma , is negligible : @xmath173 . this means that the plasma is cold and @xmath174 ( eqn . 4 ) . in this case the apparent slow component speed is the beam speed itself . with this speed for the beam , we then assumed that a fast component was due to a fast magnetosonic wave , and , from the observed apparent speed , we were able to deduce its speed on the jet . this model can be placed in figure [ fig : banana ] . the model uses @xmath175 , and is located at the dot marked `` * * a * * '' on the boundary of the diagram at @xmath176 . with @xmath177 and @xmath178 , the fast pattern speed is three times greater than the speed of the beam , when the speeds are measured by their lorentz factors . because we now also have a measurement of the apparent transverse alfvn wave propagation speed ( @xmath179 , a typical value from table 1 ) , we can extend this model to include computation of the total alfvn speed @xmath154 , the magnetosonic speed @xmath153 , and the magnetic field pitch angle @xmath10 . with @xmath180 negligible , in the galaxy frame we again have @xmath181 , @xmath182 , and now @xmath183 . then , using equation [ beta_sub ] , these become in the frame of the beam @xmath184 , @xmath185 , and @xmath186 , yielding @xmath187 a moderate helical magnetic field . since @xmath188 when @xmath189 , we also can calculate the magnetosonic mach number defined in equation [ beta_ms ] . this yields @xmath190 and qualifies this model as a trans - magnetosonic jet . * model ( b ) : hot plasma * the plasma hardly can be cold as in model ( a ) because the source is a powerful synchrotron emitter and the electron temperature is probably of order 100 mev ; the electron component of the plasma therefore is probably relativistic . on the other hand , the sound speed may or may not be near @xmath191 , depending on how heavily the plasma is contaminated with heavy , non - relativistic ions . for lack of further information , we choose @xmath192 . but , as seen in figure [ fig : banana ] , we still need another parameter to establish the solution . we took @xmath193 for model ( a ) to match a slow superluminal component , but if we do that now with @xmath192 it yields @xmath194 and @xmath195 . this value for @xmath196 is less than that typically found in radio beaming studies where @xmath197 @xcite . furthermore , the pitch angle @xmath198 is less than the one estimated from polarization analyses , @xmath165 . ( homan , private communication ) . to reconcile these values we drop the assumption that the superluminal component with @xmath199 is a slow mhd wave propagating downstream ; it might for example be a reverse mhd shock or wave traveling upstream in the beam frame and seen moving slowly downstream in the galaxy frame . instead , we choose @xmath5 , because it yields acceptable values for @xmath196 and @xmath10 , and matches the speed of a number of superluminal components . the final solution , seen at point @xmath88 in figure [ fig : banana ] , contains three quantities that are chosen to match observations , @xmath58 , @xmath200 , @xmath201 , and two quantities picked because they are plausible and give reasonable results , @xmath5 , and @xmath202 . the derived quantities are @xmath203 , @xmath204 , @xmath205 , @xmath206 , @xmath207 , @xmath208 , and @xmath209 . the slow and transverse mhd waves are non - relativistic ( @xmath210 ) in the frame of the beam , but the fast wave is mildly relativistic , with @xmath211 . note that relativistic addition to produce the observed speed is non - linear ; @xmath211 plus @xmath212 gives @xmath213 . this is discussed in paper i. with @xmath203 and @xmath214 we now calculate the doppler factor @xmath215 , which agrees closely with values in the literature @xcite . also , @xmath216 agrees with estimates from polarization analyses . thus we see that the model for bl lac with alfvn waves on a helical magnetic field is able to explain the moving transverse patterns on the jet of bl lac . it implies a modest lorentz factor for the actual plasma flow ( @xmath217 ) and explains the faster propagation of the components and the transverse disturbances as mhd sonic and alfvn waves , respectively . they are generated primarily at the site of the recollimation shock and propagate downstream on the helical field , each with a speed in the galaxy frame that is the relativistic sum of the wave speed in the beam frame and the plasma flow speed in the galaxy frame . the new model does , however , have a disadvantage . the magnetosonic mach number of 4.7 is rather high and in conflict with the original discussion in paper i on the generation of a collimation shock like c7 : the super - magnetosonic flow emanating from the black hole region should transition to a trans - magnetosonic flow ( @xmath218 ) after it passes through c7 . however , we see in figure [ fig : banana ] that models with both low mach number and high pitch angle are mutually exclusive : trans - magnetosonic models have @xmath219 , and models with @xmath220 have magnetosonic mach numbers of @xmath221 or more . in order to obtain a more sophisticated model that is compatible both with generation of a post - collimation - shock flow and the polarization observations , it is likely that one or more of the rather restrictive model assumptions in this paper will have to be relaxed . further insight into the propagation of an alfvn wave on a jet can be gained by examining the group velocity , which has only one value , @xmath222 , and is always directed along the magnetic field @xcite . an isolated wave packet will spiral down the jet along the helical magnetic field . a uniform disturbance across the jet will produce a ripple that moves along all the field lines ; i.e. across the jet . the net result is a jump or bend that propagates downstream with speed proportional to the cosine of the pitch angle . this has a close analogy to a transverse mechanical wave on a coiled spring , or slinky . in both cases there is longitudinal tension , provided for the jet by the magnetic field . figure [ fig : polars ] shows relativistic phase polar diagrams for the two models discussed above and identified in figure [ fig : banana ] . the diagrams show mhd wave phase speeds in 3-dimensional velocity space with the origin of each at the center of the diagram . each diagram was computed using the relativistic equations a1-a6 in paper i. all surfaces are axisymmetric about the horizontal magnetic axis . in each panel the dotted , solid , and broken lines show respectively the speed - of - light sphere ( unity in all directions ) , the two compressional mhd wave surfaces ( fast [ @xmath223 and slow [ @xmath167 ] ) , and the transverse alfvn wave surface ( @xmath225 ) . unlike the speed of light , the speeds of the mhd waves depend on the polar angle @xmath135 between the propagation and field directions . all three mhd modes are labeled in the left half of the diagrams . the arrows labeled in the right half of the diagrams show the three characteristic wave speeds : sound ( @xmath152 ) , alfvn ( @xmath154 ) , and magnetosonic ( @xmath153 ) , which values are realized along the field for the slow and alfvn modes and normal to the field for the fast mode . as mentioned earlier , the slow and alfvn waves can propagate skew to the field , but not normal to it . some of the relationships among the three types of mhd waves can be seen in figure [ fig : polars]b . the outer solid loop traces the fast magnetosonic mode , whose speed is a maximum @xmath153 at @xmath226 , and is the same as that of the alfvn wave ( dashed loop ) , @xmath154 , when @xmath227 , provided @xmath168 , where @xmath152 is the sound speed in the plasma . the propagation speed of the alfvn wave is proportional to @xmath228 and this also is approximately true for the slow magnetosonic wave , the inner loop . so far we have been discussing phase polar diagrams in a uniform magnetic field , and now address how this applies to a plasma jet with a helical field . figure [ fig : helicalfield ] shows a schematic diagram of a helical field jet with the properties of model ( b ) discussed above and in figures [ fig : banana ] and [ fig : polars ] . the helical field will have a pitch angle of @xmath229 , so the polar diagram in figure [ fig : polars ] will be rotated by that amount . the propagation direction of the mhd waves points downstream in our model , allowing us to read off the values of their propagation speeds from the polar diagram : @xmath230 , @xmath231 , and @xmath232 . if the helical field and plasma properties are uniform along the jet , the results will be the same everywhere , producing mhd waves with uniform velocities . b is rotated by the angle @xmath233 to align the dashed arrow with the helical field direction on the near side of the jet ( left ) . the propagation speeds of the three mhd waves along the jet axis then can be directly read off the polar diagram ( @xmath230 , @xmath234 , @xmath232 ) . for a uniform helical field one obtains the same results at any point ( e.g. , on the far side of the jet at right . ) [ fig : helicalfield],scaledwidth=50.0% ] however , there will be a longitudinal current that will cause the field strength and pitch angle to be functions of the radial coordinate @xmath235 . ( see the cut - away view of a plasma rope in figure 6.14 of @xcite , for a simple view of the radial variations in b and @xmath10 . ) but , there should be a cylindrical shell around the axis , covering a modest range of @xmath235 , in which the synchrotron emissivity into the direction of the observer is maximized . we assume that this shell is the dominant region and that the field strength and pitch angle there are the effective values that control the dynamics . ( see @xcite for a discussion of this point . thus , if this is the case , then our dynamical analysis of the waves and a polarization analysis of the emission should result in similar magnetic pitch angle estimates for the magnetic field . a preliminary polarization analysis ( using methods similar to those in @xcite and to be discussed in more detail in the next paper in this series ) produces estimates of at least 60- 70for the pitch angle . this is in agreement with our result for the hot model ( b ) , which gives @xmath236 . the jet of bl lac is highly variable and displays transverse patterns that propagate superluminally downstream on the ridge line . they are not ballistic , like water from a hose , but are constrained , like waves on a whip . the magnetic field is well - ordered with a strong transverse component that we assume is the toroidal part of a helical field . in @xcite we assumed that the helical field provided support for fast- and slow - mode mhd waves whose compressions we see as the superluminal components . we here assume that the moving transverse patterns are alfvn waves propagating on the longitudinal component of the magnetic field . the full set of ridge lines is shown in figure [ fig : ridge_all ] , and we show six examples of the alfvn waves in figures 7 , 9 , 10 , 11 , and 12 . a movie ( figure [ fig : movie ] ) provides assistance in studying the motions . the transverse wave activity died down in 2010 and the jet settled to a fixed position angle ( p.a . ) , with a mild wiggle . this wiggle was not stationary , but appeared to oscillate transversely , with amplitude about 0.4 mas . this mild wiggle persisted through the remaining data period , up to 2013.0 . although the transverse propagating alfvn waves were greatly reduced in 2010 - 2013 , the superluminal components , which we identified in paper i as mhd sonic waves , continued roughly as before . figure 2 in paper i shows that during this period they continued with about the same frequency and speed as earlier . furthermore , during the latter half of this period , from about 2011.4 to 2013.0 , bl lac was exceptionally active at shorter wavelengths @xcite , from 1 mm through gamma - rays . this general behavior can fit into our model . we have magnetosonic waves responsible for the superluminal components , and alfvn waves responsible for the moving transverse patterns . these are independent mhd modes , and can be separately excited . we suspect , however , that the increase in short - wavelength activity during the same period as the reduction in alfvn waves ( 20102013 ) is not a coincidence . the velocity of the transverse waves was established by finding characteristic points on the ridge lines where the slope changes , as well as by visual inspection of the delayed superposition of the ridge lines . three of the apparent velocities are near @xmath237 , and one is much faster , with @xmath238 . with @xmath58 and @xmath239 the speeds in the galaxy frame are approximately @xmath240 and in the beam frame @xmath241 . an alfvn wave displaces the jet in the transverse direction , and the observed motion can be converted into a transverse speed . for wave d , the largest wave we observed , the transverse speed , in the galaxy frame , is @xmath242 . this is a rough estimate but safely non - relativistic , and consistent with our assumption that the waves have low amplitude . the timing and direction of some of the the waves are correlated with the p.a . of the recollimation shock ( rcs ) , which swings over 25 in an irregular fashion . it appears that the waves are excited by the swinging of the rcs . this is analogous to exciting a wave on a whip by shaking it . in paper i ( figure 3 ) we saw that the ridge lines occupy a cylinder about 0.7 mas wide and 3 mas long , or 3 ly wide and 120 ly long when a deprojection factor of 10 is used . ( see also @xcite figure 1 . ) we now understand that this cylinder is formed by the transverse waves , whose axes generally are close to the source axis at @xmath243 . the width is set by the amplitude of the largest waves while the length is set by the general bend of the source to the se . we briefly describe the alfvn waves , and provide a method for calculating physical quantities in the jet in terms of the measured wave speeds . we investigate two simple models of the system ; in the first the plasma is cold and the sound speed @xmath161 . this gives results for the lorentz factor and the pitch angle that are in moderate disagreement with results from observations . the second model uses a hot plasma with @xmath244 , and assumes that the slow magnetosonic wave has apparent speed @xmath5 . this yields @xmath245 , pitch angle @xmath236 , alfvn speed @xmath246 , and magnetosonic mach number @xmath247 . this describes a plasma in which the helical magnetic field is strong with a dominant toroidal component . in our model the lorentz factor for the beam is approximately 4.5 and is smaller than the observed apparent speed of most of the transverse waves as well as the fast superluminal components discussed in paper i. another way to say this is that in most cases the pattern speed is greater than the beam speed . this comes about because the pattern traces a wave traveling downstream on the beam . we conclude that the rapid movements of the transverse patterns in the jet of bl lac can be described as alfvn waves excited at the rcs and propagating downstream on the longitudinal component of a helical magnetic field . the jet can be described as a relativistic , rapidly shaken whip . we suggest that other similar sources be investigated with these ideas in mind . we thank the referee for comments that have improved the manuscript , m. perucho for reading the manuscript and offering helpful suggestions , and the mojave team for comments on the manuscript , and for years of work in producing the data base that makes this work possible . tga acknowledges support by dfg project number os 177/2 - 1 . th was partly supported by the jenny and antti wihuri foundation and by the academy of finland project number 267324 ; ts was partly supported by the academy of finland project 274477 . yyk is partly supported by the russian foundation for basic research ( project 13 - 02 - 12103 ) , research program ofn-17 of the division of physics , russian academy of sciences , and the dynasty foundation . abp was supported by the `` non - stationary processes in the universe '' program of the presidium of the russian academy of sciences . the vlba is a facility of the national radio astronomy observatory , a facility of the national science foundation that is operated under cooperative agreement with associated universities , inc . the mojave program is supported under nasa - fermi grant nnx12a087 g . this study makes use of 43 ghz vlba data from the vlba - bu blazar monitoring program ( vlba - bu - blazar ; http://www.bu.edu/blazars/vlbaproject.html ) , funded by nasa through the fermi guest investigator program . part of this research was carried out at the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . this research has made use of nasa s astrophysics data system .
since einstein formulated special relativity in 1905 , the basic space - time symmetry has been that of the lorentz group . he established the energy - momentum relation which is valid for slow massive particles , high - speed massive particles , and massless particles . einstein formulated this relation initially for particles without internal space - time structures , but it is widely accepted that it is valid for all particles including particles with spin and/or internal space - time extension . in 1939 , wigner published his most fundamental paper dealing with internal space - time symmetries of relativistic particles @xcite . in this paper , wigner introduced the lorentz group to physics . furthermore , by introducing his `` little groups , '' wigner provided the framework for studying the internal space - time symmetries of relativistic particles . the scientific contents of this paper have not yet been fully recognized by the physics community . we are writing this report as a continuation of the work wigner initiated in this history - making paper . while particle physicists are still struggling to understand internal space - time symmetries of elementary particles , wigner s lorentz group is becoming useful to many other branches of physics . among them are optical sciences , both quantum and classical . in quantum optics , the coherent and squeezed states are representations of the lorentz group @xcite . recently , the lorentz group is becoming the fundamental language for classical ray optics . it is gratifying to note that optical components , such as lenses , polarizers , interferometers , lasers , and multi - layers can all be formulated in terms of the lorentz group which wigner formulated in his 1939 paper . classical ray optics is of course a very old subject , but we can not do new physics without measurements using optical instruments . indeed , classical ray optics constitutes wigner s frontier in physics . the word `` group theory '' sounds like abstract mathematics , but it is gratifying to note that wigner s little groups can be formulated in terms of two - by - two matrices , while classical ray optics is largely a physics of two - by - two matrices . the mathematical correspondence is straight - forward . in order to see this point clearly , let us start with the following classic example . the second - order differential equation @xmath0 is applicable to a driven harmonic oscillator with dissipation . this can also be used for studying an electronic circuits consisting of inductance , resistance , capacitance , and an alternator . thus , it is possible to study the oscillator system using an electronic circuit . likewise , an algebra of two - by - two matrices can serve as the scientific language for several different branches of physics , including special relativity , ray optics , and quantum optics . rccc + & & & + & massive , slow & covariance & massless , fast + & & & + energy- & & einstein s & + momentum & @xmath1 & @xmath2 & @xmath3 + & & & + internal & @xmath4 & & @xmath5 + space - time & & wigner s & + [ -1 mm ] symmetry & @xmath6 & little group & gauge trans . + & & & + relativistic & & lorentz - squeezed & + extended & quarks & harmonic & partons + [ -1 mm ] particles & & oscillators & + there are many physical systems which can be formulated in terms of two - by - two matrices . if we restrict that their determinant be one , there is a well established mathematical discipline called the group theory of @xmath7 . this aspect was noted in the study of lorentz transformations . in group theoretical terminology , the group @xmath7 is the universal covering group of the group of the lorentz group . in practical terms , to each two - by - two matrix , there corresponds one four - by - four matrix which performs a lorentz transformation in the four - dimensional minkowskian space . thus , if a physical system can be explained in terms of two - by - two matrices , it can be explained with the language of lorentz transformations . furthermore , the physical system based on two - by - two matrices can serve as an analogue computer for lorentz transformations . optical filters , polarizers , and interferometers deal with two independent optical rays . they superpose the beams , change the relative phase shift , and change relative amplitudes . the basic language here is called the jones matrix formalism , consisting of the two - by - two matrix representation of the @xmath7 group @xcite . the four - by - four mueller matrices are derivable from the two - by - two matrices of @xmath7 . for these two - beam systems , the poincar sphere serves as an effective language @xcite . since the two - beam system is described by the lorentz group , the poincar sphere is necessarily a representation of the lorentz group . we shall use this sphere to study the degree of coherence between the beams . para - axial lens optics can also be formulated in terms of two - by - two matrices , applicable to the two - component vector space consisting of the distance from the optical axis and the slope with respect to the axis . the lens and translation matrices are triangular , but they are basically representations of the @xmath8 group which is the real subgroup of the group @xmath7 @xcite . laser optics is basically multi - lens lens optics . however , the problem here is how to get a simple mathematical expression for the system of a large number of the same lens separateled by equal distance . here again , group theory simplifies calculations @xcite . in multi - layer optics , we deal with two optical rays moving in opposite directions . the standard language in this case is the s - matrix formalism @xcite . this is also a two - by - two matrix formalism . as in the case of laser cavities , the problem is the multiplication of a large number of matrix chains @xcite . it is shown in this report that the two - by - two representation of the six - parameter lorentz group is the underlying common scientific language for all of the instruments mentioned above . while the abstract group theoretical ideas make two - by - two matrix calculations more systematic and transparent in optics , optical instruments can act as analogue computers for lorentz transformations in special relativity . it is gratifying to note that special relativity and ray optics can be formulated as the physics of two - by - two matrices . in sec . [ lorentz ] , we explain how the lorentz group can be formulated in terms of four - by - four matrices . it is shown that the group can have six independent parameters . in sec . [ spinor ] , we explain how it is possible to formulate the lorentz group in terms of two - by - two matrices . it is shown that the four - by - four transformation matrices can be constructed from those those of the two - by - two matrices . in sec . [ wlittle ] , we discuss the historical significance of wigner s 1939 paper @xcite on the lorentz group and its application to the internal space - time symmetries of relativistic particles . in sec . [ polari ] , we present the basic building blocks for the two - by - two representation of the lorentz group in terms of the matrices commonly seen in ray optics . in secs . [ polari ] and [ interfer ] , we study the polarizations for the two - beam system . it is shown that both the jones and mueller matrices are representations of the lorentz group . the role of the stokes parameters is discussed in detail . it is shown in sec . [ poinc ] that the poincar sphere is a graphical representation of the poincar group , and serves as a device to describe the degree of coherence . in secs . [ olens ] , [ mlens ] , [ lcav ] , and [ mlayer ] , we discuss one - lens system , multi - lens system , laser cavities , and multi - layer optics , respectively . in all of these sections , the central scientific language is the lorentz group . in appendix [ sqosc ] , we expand the content of the third row of table [ einwig ] . it is noted that the covariant harmonic oscillator formalism can unify the quark model for slow hadrons with the parton model for ultra - relativistic quarks . it is then shown that the same oscillator formalism serves as the basic scientific language for squeezed states of light . in appendix [ euler ] , it is shown that the lie - group method , in terms of the generators , is not the only method in constructing group representations . for the rotation group and the three - parameter subgroups of the lorentz group , it is simpler to start with the minimum number of starter matrices . for instance , while there are three generators for the rotation group in the lie approach , we can construct the most general form of the rotation matrix from rotations around two directions , as goldstein constructed the euler angles @xcite . in appendix [ conju ] , it is noted that the four - by - four matrices are real , their two - by - two counterparts are complex . however , there is a three - parameter real subgroup called @xmath8 . it is shown that the complex subgroup @xmath9 is equivalent to @xmath8 through conjugate transformation . let us consider the space - time coordinates @xmath10 . then the rotation around the @xmath11 axis is performed by the four - by - four matrix @xmath12 this transformation is generated by @xmath13 likewise , we can write down the generators of rotations @xmath14 and @xmath15 around the @xmath16 and @xmath17 axes respectively . @xmath18 these three generators satisfy the closed set of commutations relations @xmath19 = i \epsilon_{ijk } j_{k } .\ ] ] this set of commutation relations is for the three - dimensional rotation group . the lorentz boost along the @xmath11 axis takes the form @xmath20 which is generated by @xmath21 likewise , we can write generators of boosts @xmath22 and @xmath23 along the @xmath16 and @xmath17 axes respectively , and they take the form @xmath24 these boost generators satisfy the commutation relations @xmath25 = i \epsilon_{ijk } k_{k } , \qquad \left[k_{i } , k_{j } \right ] = -i \epsilon_{ijk } j_{k } .\ ] ] indeed , the three rotation generators and the three boost generators satisfy the closed set of commutation relations given in eq.([rotcom ] ) and eq.([locom ] ) . these three commutation relations form the starting point of the lorentz group . the generators given in this section are four - by - four matrices , but they are not the only set satisfying the commutation relations . we can construct also six two - by - two matrices satisfying the same set of commutation relations . the group of transformations constructed from these two - by - two matrices is often called @xmath26 or the two - dimensional representation of the lorentz group . throughout the present paper , we used the two - by - two transformation matrices constructed from the generators of the @xmath26 group . in sec . [ lorentz ] , we started with four - by - four transformation matrices applicable to the four - dimensional space - time . we then ended up with a set of closed commutation relations for the six generators consisting of three rotation and three boost generators . these generators are in the form of four - by - four matrices . in this section , we shall see first that there is a set of two - by - two matrices satisfying the same set of commutation relations , constituting the two - by - two representations of the lorentz group . the representation so constructed is called the @xmath26 group , or the universal covering group of the lorentz group . the transformation matrices are applicable to two - component @xmath7 spinors . the algebraic property of this two - by - two representation is the same as that of the four - by - four representation . as we shall see in this paper , the spinors and the four - vectors correspond to jones vectors and stokes parameters respectively in polarization optics @xcite . the question then is whether we can construct the four - vector from the spinors . in the language of polarization optics , the question is whether it is possible to construct the coherency matrix @xcite from the jones vector . with this point in mind , let us start from the following form of the pauli spin matrices : @xmath27 these matrices are written in a different convention . here @xmath28 is imaginary , while @xmath29 is imaginary in the traditional notation . also in this convention , we can construct three rotation generators @xmath30 which satisfy the closed set of commutation relations @xmath31 = i \epsilon_{ijk } j_{k } .\ ] ] we can also construct three boost generators @xmath32 which satisfy the commutation relations @xmath33 = -i \epsilon_{ijk } j_{k } .\ ] ] the @xmath34 matrices alone do not form a closed set of commutation relations , and the rotation generators @xmath35 are needed to form a closed set : @xmath36 = i \epsilon_{ijk } k_{k } .\ ] ] the six matrices @xmath35 and @xmath34 form a closed set of commutation relations , and they are like the generators of the lorentz group applicable to the ( 3 + 1)-dimensional minkowski space . the group generated by the above six matrices is called @xmath26 consisting of all two - by - two complex matrices with unit determinant . let us write the two - by - two transformation matrix as @xmath37 this matrix has four complex elements with eight real parameters . however , the six generators are all traceless and , the determinant of the matrix has to be one . thus , it has six independent parameters . in order to construct four - vectors , we need two different spinor representations of the lorentz group . let us go to the commutation relations for the generators given in eqs.([comm1 ] ) , ( [ comm2 ] ) and ( [ comm3 ] ) . these commutators are not invariant under the sign change of the rotation generators @xmath35 , but are invariant under the sign change of the squeeze operators @xmath34 . thus , to each spinor representation , there is another representation with the squeeze generators with opposite sign . this allows us to construct another representation with the generators : @xmath38 we call this representation the `` dotted '' representation . if we write the transformation matrix @xmath39 of eq.([alpha ] ) in terms of the generators as @xmath40 then the transformation matrix in the dotted representation becomes @xmath41 in both of the above matrices , hermitian conjugation changes the direction of rotation . however , it does not change the direction of boosts . we can achieve this only by changing @xmath39 to @xmath42 , and we shall call this the `` dot '' conjugation . likewise , there are two different set of spinors . let us use write @xmath43 for the spinor in the undotted and dotted representations respectively . then the four - vectors are constructed as @xcite @xmath44 & { } & u\dot{v } = ( t + z ) , \quad v\dot{u } = -(t - z)\end{aligned}\ ] ] leading to the matrix @xmath45 where @xmath46 and @xmath47 are one if the spin is up , and are zero if the spin is down , while @xmath48 and @xmath49 are zero and one for the spin - up and spin - down cases . if the two - by - two matrix of eq . [ alpha ] is applicable to the column vector of eq.([uv ] ) , what is the transformation matrix applicable to the row vector @xmath50 from the right - hand side ? it is the transpose of the matrix applicable to the column vector @xmath50 . we can obtain this column vector from @xmath51 by applying to it the matrix @xmath52 this matrix also has the property @xmath53 where the superscript @xmath54 means the transpose of the matrix . the transformation matrix applicable to the column vector of eq.([dotcol ] ) is @xmath42 of eq.([eldot ] ) . thus the matrix applicable to the row vector @xmath50 in eq.([dotmat ] ) is @xmath55 this is precisely the hermitian conjugate of @xmath39 . let us now write the @xmath56 matrix of eq.([dotmat ] ) as @xmath57 where the set of variables @xmath58 is transformed like a four - vector under lorentz transformations . then the lorentz transformation on @xmath56 can be performed as @xmath59 where the transformation matrix @xmath39 is that of eq.([alpha ] ) . as we have seen in this section , the construction of four - vectors from the two - component spinors is not a trivial task , but has been discussed in the literature @xcite . likewise , it is possible to construct four - by - four lorentz transformation matrices from the two - by - two matrix of eq.([alpha ] ) @xcite . as we shall see in the present paper , the matrix of the form eq.([dotmat ] ) or eq.([matv ] ) appears in optics as the coherency matrix and the density matrix for two beam systems such as polarization optics and interferometers . in his 1939 paper , wigner introduced his little groups in order to study the internal space - time structures of relativistic particles . the little group is the maximal subgroup of the lorentz group which leaves the momentum of the particle invariant . the little groups , originally developed for studying relativistic symmetries of particles , are now becoming an important scientific language for ray optics as we shall see in this paper . let us first discuss the physical basis of wigner s little groups . if the speed of a particle is much smaller than that of light , energy - momentum relation is @xmath60 . if the speed is close to that of light , the relation is @xmath3 . these two different relations can be combined into one covariant formula @xmath61 . this aspect of einstein s @xmath62 is well known , as indicated in table [ einwig ] . in addition , particles have internal space - time variables . massive particles have spins while massless particles have their helicities and gauge variables . our first question is whether this aspect of space - time variables can be unified into one covariant concept . the answer to this question is yes . little group does the job , also as indicated table [ einwig ] . particles can also have space - time extensions . for instance , in the quark model , hadrons are bound states of quarks . however , the hadron appears as a collection of partons when it moves with speed close to the velocity of light . quarks and partons seem to have quite distinct properties . are they different manifestations of a single covariant entity ? this is one of the most pressing issues in high - energy particles physics . the third row of table [ einwig ] addresses this question . the mathematical framework of this program was developed by eugene wigner in 1939 @xcite . he constructed the maximal subgroups of the lorentz group whose transformations will leave the four - momentum of a given particle invariant . these groups are known as wigner s little groups . thus , the transformations of the little groups change the internal space - time variables of the particle , while leaving its momentum invariant . in his paper @xcite , wigner shows that , for each massive particle , there is a lorentz frame in which the particle is at rest . then the three - dimensional rotation group leaves its momentum invariant , while changing the direction of its spin . thus , the little group for this particular case is @xmath63 . in other lorentz frames , the little group is a lorentz - boosted @xmath63 group . for a massless particle , wigner notes that there are no lorentz frames in which the particle is at rest . the best we can do is to align its momentum to a given axis , or the @xmath11 axis . then , its momentum is invariant under rotations around the @xmath11 axis . in addition , wigner found two more transformations which leave the momentum invariant . the physics of these additional degrees was not explained in wigner s original paper @xcite , but their mathematics has been worked out . the physics of these degrees of freedom was later determined to be that of gauge transformations @xcite , and its complete understanding was not achieved until 1990 @xcite . the group of lorentz transformations consists of three boosts and three rotations . the rotations therefore constitute a subgroup of the lorentz group . if a massive particle is at rest , its four - momentum is invariant under rotations . thus the little group for a massive particle at rest is the three - dimensional rotation group . then what is affected by the rotation ? the answer to this question is very simple . the particle in general has its spin . the spin orientation is going to be affected by the rotation ! if we use the four - vector coordinate @xmath10 , the four - momentum vector for the particle at rest is @xmath64 , and the three - dimensional rotation group leaves this four - momentum invariant . this little group is generated by the three rotation generators given in eq.([j33 ] ) and eq.([j12 ] ) . they satisfy the commutation relations for the rotation group given eq.([rotcom ] ) . if the rest - particle is boosted along the @xmath11 direction , it up a non - zero momentum component along the same direction . the @xmath63 generators will also be boosted . the boost matrix takes the form of eq.([boostz ] ) , and the boosted generators will be @xmath65 and this boost will not change the commutation relation of eq.([rotcom ] ) . for a massless particle moving along the @xmath66 direction , wigner observed that the little group generated by the rotation generator around the @xmath11 axis , namely @xmath67 of eq.([j33 ] ) , and two other generators which take the form @xmath68 if we use @xmath34 for the boost generator along the i - th axis , these matrices can be written as @xmath69 with @xmath22 and @xmath23 given in eq.([k12 ] ) . the generators @xmath70 and @xmath71 satisfy the following set of commutation relations . @xmath72 = 0 , \qquad [ j_{3 } , n_{1 } ] = in_{2 } , \qquad [ j_{3 } , n_{2 } ] = -in_{1 } .\ ] ] in order to understand the mathematical basis of the above commutation relations , let us consider transformations on a two - dimensional plane with the @xmath73 coordinate system . we can then make rotations around the origin and translations along the @xmath16 and @xmath17 directions . if we write these generators as @xmath74 and @xmath75 respectively , they satisfy the commutation relations @xcite @xmath76 = 0 , \qquad [ l , p_{x } ] = ip_{y } , \qquad [ l , p_{y } ] = -ip_{x } .\ ] ] this is a closed set of commutation relations for the generators of the @xmath77 group , or the two - dimensional euclidean group . if we replace @xmath78 and @xmath71 of eq.([e2lcom ] ) by @xmath79 and @xmath75 , and @xmath67 by @xmath39 , the commutations relations for the generators of the @xmath77-like little group becomes those for the @xmath77-like little group . this is precisely why we say that the little group for massless particles are like @xmath77 . it is not difficult to associate the rotation generator @xmath67 with the helicity degree of freedom of the massless particle . then what physical variable is associated with the @xmath78 and @xmath71 generators ? indeed , wigner was the one who discovered the existence of these generators , but did not give any physical interpretation to these translation - like generators in his original paper @xcite . for this reason , for many years , only those representations with the zero - eigenvalues of the @xmath80 operators were thought to be physically meaningful representations @xcite . it was not until 1971 when janner and janssen reported that the transformations generated by these operators are gauge transformations @xcite . the role of this translation - like transformation has also been studied for spin-1/2 particles , and it was concluded that the polarization of neutrinos is due to gauge invariance @xcite . the @xmath63-like little group remains @xmath63-like when the particle is lorentz - boosted . then , what happens when the particle speed becomes the speed of light ? the energy - momentum relation @xmath61 become @xmath81 . is there then a limiting case of the @xmath63-like little group ? since those little groups are like the three - dimensional rotation group and the two - dimensional euclidean group respectively , we are first interested in whether @xmath77 can be obtained from @xmath63 . this will then give a clue as to how to obtain the @xmath77-like little group as a limiting case of @xmath63-like little group . with this point in mind , let us look into this geometrical problem . in 1953 , inn and wigner formulated this problem as the contraction of @xmath63 to @xmath77 @xcite . let us see what they did . we always associate the three - dimensional rotation group with a spherical surface . let us consider a circular area of radius one kilometer centered on the north pole of the earth . since the radius of the earth is more than 6,450 times longer , the circular region appears to be flat . thus , within this region , we use the @xmath77 symmetry group . the validity of this approximation depends on the ratio of the two radii . how about then the little groups which are isomorphic to @xmath63 and @xmath77 ? it is reasonable to expect that the @xmath77-like little group can be obtained as a limiting case of the @xmath63-like little group for massless particles . in 1981 , it was observed by bacry and chang @xcite and by ferrara and savoy @xcite that this limiting process is the lorentz boost to infinite - momentum frame . in 1983 , it was noted by han _ et al _ that the large - radius limit in the contraction of @xmath63 to @xmath77 corresponds to the infinite - momentum limit for the case of the @xmath63-like little group to the @xmath77-like little group for massless particles . they showed that transverse rotation generators become the generators of gauge transformations in the limit of infinite momentum @xcite . let us see how this happens . the @xmath67 operator of eq.([j33 ] ) , which generates rotations around the @xmath11 axis , is not affected by the boost conjugation of the @xmath82 matrix of eq.([boost ] ) . on the other hand , the @xmath14 and @xmath15 matrices become @xmath83 which are given in eq.([n1n2 ] ) . the generators @xmath78 and @xmath71 are the contracted @xmath15 and @xmath14 respectively in the infinite - momentum . in 1987 , kim and wigner studied this problem in more detail and showed that the little group for massless particles is the cylindrical group which is isomorphic to the @xmath77 group @xcite . this completes the second row in table [ einwig ] , where wigner s little group unifies the internal space - time symmetries of massive and massless particles . the transverse components of the rotation generators become generators of gauge transformations in the infinite - momentum limit . let us go back to table i given in sec . [ intro ] . as for the third row for relativistic extended particles , the most efficient approach is to construct representations of the little groups using the wave functions which can be lorentz - boosted . this means that we have to construct wave functions which are consistent with all known rules of quantum mechanics and special relativity . it is possible to construct harmonic oscillator wave functions which satisfy these conditions . we can then take the low - speed and high - speed limits of the covariant harmonic oscillator wave functions for the quark model and the parton model respectively . this aspect was extensively discussed in the literature @xcite , and is beyond the scope of the present report . however , it is important to note that the covariant harmonic oscillator formalism use for this purpose can serve as the fock space description for the squeezed state of light @xcite , we give a brief discussion of this aspect in appendix [ sqosc ] , entitled `` squeezed harmonic oscillators . '' in this section , we discussed wigner s little groups applicable to internal space - time symmetries of relativistic particles . however , as we shall see in this paper , these little groups play an important role in understanding ray optics . conversely , optical configurations in ray optics can serve as analogue computers for space - time symmetries in particle physics . let us consider two optical beams propagating along the @xmath11 axis . we are then led to the column vector : @xmath84 we can then achieve a phase shift between the beams by applying the two - by - two matrix : @xmath85 if we are interested in mixing up the two beams , we can apply @xmath86 to the column vector . if the amplitudes become changed either by attenuation or by reflection , we can use the matrix @xmath87 for the change . in this paper , we are dealing only with the relative amplitudes , or the ratio of the amplitudes . as we shall see in sec . [ interfer ] , the above two - by - two matrices have their corresponding four - by - four matrices respectively . repeated applications of these matrices lead to the form @xmath88 where the elements are in general complex numbers . the determinant of this matrix is one . thus , the matrix can have six independent parameters . this matrix takes the identical form as the two - by - two matrix given in eq.([alpha2 ] ) . however , the construction process is different . the lie - group generators are used for eq.([alpha2 ] ) , while eq.([alpha2 ] ) is constructed from repeated applications of the transformation matrices . the difference between these two methods is discussed in appendix [ euler ] . in either case , this matrix is the most general form of the matrices in the @xmath26 group , which is known to be the universal covering group for the six - parameter lorentz group . this means that , to each two - by - two matrix of @xmath26 , there corresponds one four - by - four matrix of the group of lorentz transformations applicable to the four - dimensional minkowski space @xcite . it is possible to construct explicitly the four - by - four lorentz transformation matrix from the parameters @xmath89 and @xmath90 . this expression is available in the literature @xcite , and we consider here only special cases . the four - by - four representation of the lorentz group can be constructed from the two - by - two representation of the @xmath26 group , which is known as the universal covering group of the six - parameter lorentz group . this aspect is discussed in sec . [ spinor ] . let us go back to eq.([alpha2 ] ) , the @xmath26 group represented by this matrix has many interesting subgroups . if the matrices are to be hermitian , then the subgroup is @xmath91 corresponding to the three - dimensional rotation group . if all the elements are real numbers , the group becomes the three - parameter @xmath8 group . this subgroup is equivalent to @xmath9 which is the primary scientific language for squeezed states of light @xcite . we can also consider the matrix of eq.([alpha2 ] ) when one of its off - diagonal elements vanishes . then , it takes the form @xmath92 where @xmath93 is a complex number with two real parameters . in 1939 @xcite , wigner observed this form as one of the subgroups of the lorentz group . he observed further that this group is isomorphic to the two - dimensional euclidean group , and that its four - by - four equivalent can explain the internal space - time symmetries of massless particles including the photons . in ray optics , we often have to deal with this type of triangular matrices , particularly in lens optics and stability problems in laser and multi - layer optics . in the language of mathematics , dealing with this form is called the iwasawa decomposition @xcite . if the jones matrix contains all the parameters for the polarized light beam , why do we need the mathematics in the four - dimensional space ? the answer to this question is well known . in addition to the basic parameter given by the jones vector , the stokes parameters give the degree of coherence between the two rays . let us go back to the jones spinor of eq.([jones ] ) , and construct the quantities : @xmath94 & { } & s_{21 } = < \psi_{2}^{}\psi_{1 } > , \qquad s_{22 } = < \psi_{2}^{}\psi_{2 } > .\end{aligned}\ ] ] then the stokes vector consists of @xmath95 & { } & s_{2 } = s_{12 } + s_{21 } , \quad s_{3 } = -i(s_{11 } + s_{22 } ) .\end{aligned}\ ] ] the four - component vector @xmath96 transforms like the space - time four - vector @xmath10 under lorentz transformations . the mueller matrix is therefore like the lorentz - transformation matrix . as in the case of special relativity , let us consider the quantity @xmath97 then @xmath98 is like the mass of the particle while the stokes four - vector is like the four - momentum . if @xmath99 , the two - beams are in a pure state . as @xmath98 increases , the system becomes mixed , and the entropy increases . if it reaches the value of @xmath100 , the system becomes completely random . it is gratifying to note that this mechanism can be formulated in terms of the four - momentum in particle physics @xcite . although we can borrow all the elegant mathematical identities of the two - by - two representations of the lorentz group , this formalism does not allow us to describe the loss of coherence within the interferometer system . in order to study this effect , we have to construct the coherency matrix : @xmath101 under the optical transformations discussed in this section , the coherency matrix is transformed as @xmath102 as in the case of the lorentz transformation given in eq.([ldag ] ) . using this formalisms based on the lorentz group , we can discuss the group theoretical property of polarization optics in detail @xcite . however , polarization optics is based on two independent beams propagating in the same direction . since interferometers are also based on the same optical system , we can continue our discussion in the following section on interferometers . typically , one beam is divided into two by a beam splitter . we can write the incoming beam as @xmath103 then , the beam splitter can be written in the form of a rotation matrix @xcite : @xmath104 which transforms the column vector of eq.([expo1 ] ) into @xmath105\exp{\left\{i(kz - \omega t)\right\ } } \cr -[\sin(\theta/2)]\exp{\left\{i(kz - \omega t)\right\ } } } .\ ] ] the first beam @xmath106 of eq.([expo1 ] ) is now split into @xmath106 and @xmath107 of eq.([expo2 ] ) . the intensity is conserved . if the rotation angle @xmath108 is -@xmath109 , the initial beam is divided into two beams of the same intensity and the same phase @xcite . these two beams go through two different optical path lengths , resulting in a phase difference . if the phase difference is @xmath110 , the phase shift matrix is @xmath111 when reflected from mirrors , or while going through beam splitters , there are intensity losses for both beams . the rate of loss is not the same for the beams . this results in the attenuation matrix of the form @xmath112 with @xmath113 . this attenuator matrix tells us that the electric fields are attenuated at two different rates . the exponential factor @xmath114 reduces both components at the same rate and does not affect the degree of polarization . the effect of polarization is solely determined by the squeeze matrix @xmath115 in the detector or in the beam synthesizer , the two beams undergo a superposition . this can be achieved by the rotation matrix like the one given in eq.([rot22 ] ) @xcite . for instance , if the angle @xmath108 is @xmath116 , the rotation matrix takes the form @xmath117 if this matrix is applied to the column vector of eq.([expo2 ] ) , the result is @xmath118 the upper and lower components show the interferences with negative and positive signs respectively . we have shown previously @xcite that the four - by - four transformation matrices applicable to the stokes parameters are like lorentz - transformation matrices applicable to the space - time minkowskian vector @xmath10 . this allows us to study space - time symmetries in terms of the stokes parameters which are applicable to interferometers . let us first see how the rotation matrix of eq.([rot22 ] ) is translated into the four - by - four formalism . in this case , @xmath119 the corresponding four - by - four matrix takes the form @xcite @xmath120 let us next see how the phase - shift matrix of eq.([shif22 ] ) is translated into this four - dimensional space . for this two - by - two matrix , @xmath121 for these values , the four - by - four transformation matrix takes the form @xcite @xmath122 for the squeeze matrix of eq.([sq22 ] ) , @xmath123 as a consequence , its four - by - four equivalent is @xmath124 if the above matrices are applied to the four - dimensional minkowskian space of @xmath10 , the above squeeze matrix will perform a lorentz boost along the @xmath11 or @xmath125 axis with @xmath100 as the time variable . the rotation matrix of eq.([rot44 ] ) will perform a rotation around the @xmath17 or @xmath5 axis , while the phase shifter of eq.([shif44 ] ) performs a rotation around the @xmath11 or the @xmath125 axis . matrix multiplications with @xmath126 and @xmath127 lead to the three - parameter group of rotation matrices applicable to the three - dimensional space of @xmath128 . the phase shifter @xmath127 of eq.([shif44 ] ) commutes with the squeeze matrix of eq.([sq44 ] ) , but the rotation matrix @xmath126 does not . this aspect of matrix algebra leads to many interesting mathematical identities which can be tested in laboratories . one of the interesting cases is that we can produce a rotation by performing three squeezes @xcite . another interesting case is a combination of squeeze and rotation matrices which will lead to a triangular matrix with unit diagonal elements . this aspect is known as the iwasawa decomposition and is discussed in detail in ref . @xcite . according to the definition of the density matrix @xcite , the coherency matrix of eq.([cocy ] ) is also the density matrix . in this section , we shall discuss the cohrency matrix as the density matrix . under the influence of the @xmath129 transformation given in eq.([lt22 ] ) , this coherency matrix is transformed as @xmath130 & { } & \hspace{5ex } = \pmatrix{\alpha & \beta \cr \gamma & \delta } \pmatrix{s_{11 } & s_{12 } \cr s_{21 } & s_{22 } } \pmatrix{\alpha^ { * } & \gamma^ { * } \cr \beta^ { * } & \delta^ { * } } , \end{aligned}\ ] ] where @xmath131 and @xmath129 are the density matrix and the transformation matrix given in eq.([cocy ] ) and eq.([lt22 ] ) respectively . according to the basic property of the lorentz group , these transformations do not change the determinant of the density matrix @xmath131 . transformations which do not change the determinant are called unimodular transformations . as we shall see in this section , the determinant for pure states is zero , while that of mixed states does not vanish . is there then a transformation matrix which will change this determinant within the lorentz group . the answer is no . this is the basic issue we would like to address in this section . if the phase difference between the two waves remains intact , the system is said to in a pure state , and the density matrix can be brought to the form @xmath132 through the transformation of eq.([trans22 ] ) with a suitable choice of the @xmath129 matrix . for the pure state , the stokes four - vector takes the form @xmath133 in order to study the symmetry properties of the density matrix , let us ask the following question . is there a group of transformation matrices which will leave the above density matrix invariant ? in answering this question , it is more convenient to use the stokes four - vector . the column vector of eq.([pure4 ] ) is invariant under the operation of the phase shifter @xmath127 of eq.([shif44 ] ) . in addition , it is invariant under the following two matrices : @xmath134 & { } & f_{2}(v ) = \pmatrix { 1 + v^{2}/2 & - v^{2}/2 & 0 & v \cr v^{2}/2 & 1 - v^{2}/2 & 0 & v \cr 0 & 0 & 1 & 0 \cr u & -v & 0 & 1 } .\end{aligned}\ ] ] these mathematical expressions were first discovered by wigner in 1939 @xcite in connection with the internal space - time symmetries of relativistic particles . they went through a stormy history , but it is gratifying to note that they serve a useful purpose for studying interferometers where each matrix corresponds to an operation which can be performed in laboratories . the @xmath135 and @xmath136 matrices commute with each other , and the multiplication of these leads to the form @xmath137 this matrix contains two parameters . let us go back to the phase - shift matrix of eq.([shif44 ] ) . this matrix also leaves the stokes vector of eq.([pure4 ] ) invariant . if we define the `` little group '' as the maximal subgroup of the lorentz group which leaves a stokes vector invariant , the little group for the stokes vector of eq.([pure4 ] ) consists of the transformation matrices given in eq.([shif44 ] ) and eq.([d44 ] ) . next , if the phase relation is completely random , and the first and second components have the same amplitude , the density matrix becomes @xmath138 here is the question : is there a two - by - two matrix which will transform the pure - state density matrix of eq.([pure22 ] ) into the impure - state matrix of eq.([imp22 ] ) ? the answer within the system of matrices of the form given in eq.([lt22 ] ) is no , because the determinant of the pure - state density matrix is zero while that of the impure - state matrix is @xmath139 . is there a way to deal with this problem ? this problem was addressed in ref . @xcite . in this section , we restrict ourselves to the unimodular transformation of eq.([trans22 ] ) which preserves the value of the determinant of the density matrix . the stokes four - vector corresponding to the above density matrix is @xmath140 this vector is invariant under both the rotation matrix of eq.([rot44 ] ) and the phase shift matrix of eq.([shif44 ] ) . repeated applications of these matrices lead to a three - parameter group of rotations applicable to the three - dimensional space of @xmath128 . not all the impure - state density matrices take the form of eq.([imp22 ] ) . in general , if they are brought to a diagonal form , the matrix takes the form @xmath141 and the corresponding stokes four - vector is @xmath142 with @xmath143 the matrix which transforms eq.([imp4 ] ) to eq.([impp4 ] ) is the squeeze matrix of eq.([sq44 ] ) . the question then is whether it is possible to transform the pure state of eq.([pure4 ] ) to the impure state of eq.([impp4 ] ) or to that of eq.([imp4 ] ) . in order to see the problem in terms of the two - by - two density matrix , let us go back to the pure - state density matrix of eq.([pure22 ] ) . under the rotation of eq.([rot22 ] ) , @xmath144 the pure - state density matrix becomes @xmath145 for the present case of two - by - two density matrices , the trace of the matrix is one for both pure and impure cases . the trace of the @xmath146 is one for the pure state , while it is less than one for impure states . the next question is whether there is a two - by - two matrix which will eliminate the off - diagonal elements of the above expression that will also lead to the expression of eq.([impp22 ] ) . in order to answer this question , let us note that the determinant of the density matrix vanishes for the pure state , while it is non - zero for impure states . the lorentz - like transformations of eq.([trans22 ] ) leave the determinant invariant . thus , it is not possible to transform a pure state into an impure state by means of the transformations from the six - parameter lorentz group . then is it possible to achieve this purpose using two - by - two matrices not belonging to this group . we do not know the answer to this question . we are thus forced to resort to four - by - four matrices applicable to the stokes four - vector . we are interested in a transformation which will change the density matrix of eq.([pure22 ] ) to that of eq.([imp22 ] ) . for this purpose , we can use the stokes four - vector consisting of the four elements of the density matrix . the question then is whether it is possible to find a transformation matrix which will transform the pure - state four - vector of eq.([pure4 ] ) to the impure - state four - vector of eq.([imp4 ] ) . mathematically , it is more convenient to ask whether the inverse of this process is possible : whether it is possible to transform the four - vector of eq.([imp4 ] ) to that of eq.([pure4 ] ) . this is known in mathematics as the contraction of the three - dimensional rotation group into the two - dimensional euclidean group @xcite . let us apply the squeeze matrix of eq.([sq44 ] ) to the four - vector of eq.([imp4 ] ) . this can be written as @xmath147 after an appropriate normalization , the right - hand side of the above equation becomes like the pure - state vector of eq.([pure4 ] ) in the limit of large @xmath148 , as @xmath149 becomes equal to @xmath150 in the infinite-@xmath148 limit . this transformation is from a mixed state to a pure or almost - pure state . since we are interested in the transformation from the pure state of eq.([pure4 ] ) to the impure state of eq.([imp4 ] ) , we have to consider an inverse of the above equation : @xmath151 however , the above equation does not start with the pure - state four - vector . if we apply the same matrix to the pure state matrix , the result is @xmath152 the resulting four - vector is proportional to the pure - state four - vector and is definitely not an impure - state four - vector . the inverse of the transformation of eq.([sqimp ] ) is not capable of bringing the pure - state vector into an impure - state vector . let us go back to eq.([sqimp ] ) , it is possible to bring an impure - state into a pure state only in the limit of infinite @xmath148 . otherwise , it is not possible . it is definitely not possible if we take into account experimental considerations . the story is different for the little groups . let us start with the rotation matrix of eq.([rot44 ] ) , and apply to this matrix the transformation matrix of eq.([sqimp ] ) . then @xmath153 & { } & \times \pmatrix{\cosh\eta & -\sinh\eta & 0 & 0 \cr -\sinh\eta & \cosh\eta & 0 & 0 \cr 0 & 0 & 1 & 0 \cr 0 & 0 & 0 & 1 } .\end{aligned}\ ] ] if @xmath148 is zero , the above expression becomes the rotation matrix of eq.([rot44 ] ) . if @xmath148 becomes infinite , it becomes the little - group matrix @xmath154 of eq.([d1d2 ] ) applicable to the pure state of eq.([pure4 ] ) . the details of this calculation for the case of lorentz transformations are given in the 1986 paper by han _ et al . _ we are then led to the question of whether one little - group transformation matrix can be transformed from the other . if we carry out the matrix algebra of eq.([3mats ] ) , the result is @xmath155 where @xmath156 if @xmath157 , the above expression becomes the rotation matrix of eq.([rot44 ] ) . if @xmath158 , it becomes the @xmath135 matrix of eq.([d1d2 ] ) . here we used the parameter @xmath159 instead of @xmath148 . in terms of this parameter , it is possible to make an analytic continuation from the pure state with @xmath158 to an impure state with @xmath160 including @xmath157 . on the other hand , we should keep in mind that the determinant of the density matrix is zero for the pure state , while it is non - zero for all impure states . for @xmath158 , the determinant vanishes , but it is nonzero and stays the same for all non - zero values of @xmath159 less than one and greater than or equal to zero . the analytic expression of eq.([anal ] ) hides this singular nature of the little group @xcite . in secs . [ polari ] , [ interfer ] , and [ little ] , it was noted that the stokes parameters form a four - vector in the minkowskian space . thus , it was possible to discuss the density matrix in terms of the four - vectors . the poincar sphere was originally constructed from polarization optics . therefore , it is also a representation of the lorentz group . the poincar sphere for various polarization states have been thoroughly discussed in a recent book by brosseau @xcite . what is interesting is that the poincar sphere has two radii . one of them is the maximum radius specified by @xmath100 , and the other radius is the length of the space - like components , namely @xmath161 according to the four - vector property of the stokes parameter , @xmath162 is invariant under lorentz transformations . on the other hand , @xmath99 for a fully coherent state , while it is non - zero for a partially coherent state . the system is totally incoherent if @xmath163 . however , the lorentz group can not handle this decoherence mechanism . we observe here that the pure state with @xmath99 corresponds to the @xmath77-like little group for massless particles while it corresponds to the @xmath63-like little group for non - zero values of @xmath98 . the transition of @xmath63 to @xmath77 is well known as the group contraction . however , the inverse transformation requires further study . in analyzing optical rays in para - axial lens optics , we start with the lens matrix : @xmath164 and the translation matrix @xmath165 assuming that the beam is propagating along the @xmath11 direction . then the one - lens system consists of @xmath166 if we perform the matrix multiplication , @xmath167 if we assert that the upper - right element be zero , then @xmath168 and the image is focussed , where @xmath169 and @xmath170 are the distance between the lens and object and between the lens and image respectively . they are in general different , but we shall assume for simplicity that they are the same : @xmath171 . we are doing this because this simplicity does not destroy the main point of our discussion , and because the case with two different values has been dealt with in the literature @xcite . under this assumption , we are left with @xmath172 the diagonal elements of this matrix are dimensionless . in order to make the off - diagonal elements dimensionless , we write this matrix as @xmath173 indeed , the matrix in the middle contains dimensionless elements . the negative sign in front is purely for convenience . we are then led to study the core matrix @xmath174 here , the important point is that the above matrices can be written in terms of transformations of the lorentz group . in the two - by - two matrix representation , the lorentz boost along the @xmath11 direction takes the form @xmath175 and the rotation around the @xmath17 axis can be written as @xmath176 and the boost along the @xmath16 axis takes the form @xmath177 then the core matrix of eq.([core ] ) can be written as @xmath178 or @xmath179 if @xmath180 . if @xmath16 is greater than 2 , the upper - right element of the core is positive and it can take the form @xmath181 or @xmath182 the expressions of eq.([phi11 ] ) and eq.([chi11 ] ) are a lorentz boosted rotation and a lorentz - boosted boost matrix along the @xmath16 direction respectively . these expressions play the key role in understanding wigner s little groups for relativistic particles . let us look at their explicit matrix representations given in eq.([phi22 ] ) and eq.([chi22 ] ) . the transition from eq.([phi22 ] ) to eq.([chi22 ] ) requires the upper right element going through zero . this can only be achieved through @xmath148 going to infinity . if we like to keep the lower - left element finite during this process , the angle @xmath110 and the boost parameter @xmath183 have to approach zero . the process of approaching the vanishing upper - right element is necessarily a singular transformation . this aspect plays the key role in unifying the internal space - time symmetries of massive and massless particles . this is like einstein s @xmath184 becoming @xmath81 in the limit of large momentum . on the other hand , the core matrix of eq.([core ] ) is an analytic function of the variable @xmath16 . thus , the lens matrix allows a parametrization which allows the transition from massive particle to massless particle analytically . the lens optics indeed serves as the analogue computer for this important transition in particle physics . > from the mathematical point of view , eq.([phi22 ] ) and eq.([chi22 ] ) represent circular and hyperbolic geometries , respectively . the transition from one to the other is not a trivial mathematical procedure . it requires a further investigation . let us go back to the core matrix of eq.([core ] ) . the @xmath16 parameter does not appear to be a parameter of lorentz transformations . however , the matrix can be written in terms of another set of lorentz transformations . this aspect has been discussed in the literature @xcite . let us consider a co - axial system of an arbitrary number of lens . their focal lengths are not necessarily the same , nor are their separations . we are then led to consider an arbitrary number of the lens matrix given in eq.([lens ] ) and an arbitrary number of translation matrix of eq.([trans ] ) . they are multiplied like @xmath185 where @xmath80 is the number of lenses . the easiest way to tackle this problem in to use the lie - algebra approach . let us start with the generators of the sp(2 ) group : @xmath186 since the generators are pure imaginary , the transformation matrices are real . on the other hand , the @xmath39 and @xmath54 matrices of eq.([lens ] ) and eq.([trans ] ) are generated by @xmath187 if we introduce the third matrix @xmath188 all three matrices form a closed set of commutation relations : @xmath189 = ix_{3 } , \qquad \left[x_{1 } , x_{3}\right ] = -ix_{1 } , \qquad \left[x_{2 } , x_{3}\right ] = ix_{2 } .\ ] ] thus , these generators also form a closed set of lie algebra generating real two - by - two matrices . what group would this generate ? the answer has to be @xmath8 . the truth is that the three generators given in eq.([shear ] ) can be written as linear combinations of the generators of the @xmath8 group given in eq.([sq11 ] ) @xcite . thus , the @xmath190 matrices given above can also act as the generators of the @xmath8 group , and the lens - system matrix given in eq.([lsystem ] ) is a three - parameter matrix of the form of eq.([alpha2 ] ) with real elements . the resulting real matrix is written as @xmath191 and is called the @xmath192 matrix . the question then is how many lenses are need to give the most general form of the @xmath192 matrix @xcite . this matrix has three independent parameter . according to bargmann @xcite , this three - parameter matrix can be decomposed into @xmath193 which can be written as the product of one symmetric matrix resulting from @xmath194 and one rotation matrix : @xmath195 & -\sin[(\beta + \alpha)/2 ] \cr \sin[(\beta + \alpha)/2 ] & \cos[(\beta + \alpha)/2 ] } .\ ] ] the expression in eq.([symm ] ) is a symmetric matrix , while that of eq.([ortho ] ) is an orthogonal one . we can then decompose each of these two matrices into the lens and translation matrices . the net result is that we do not need more than three lenses to describe the lens system consisting of an arbitrary number of lenses . the detailed calculations are given in ref . in a laser cavity , the optical ray makes round trips between two mirrors . one cycle is therefore equivalent to a two - lens system with two identical lenses and the same distance between the lenses . let us rewrite the matrix corresponding to the one - lens system given in eq.([core ] ) . @xmath196 then one complete cycle consists of @xmath197 . here , the cycle starts from the midpoint between the two mirrors @xcite , unlike the traditional approach to this problem where the cycle starts from the surface of one of the two mirrors @xcite . by choosing the midpoint , we can eliminate the auxiliary flat mirror between them needed in the traditional approach . this problem was discussed in detail in ref . @xcite . for @xmath80 cycles , the expression should be @xmath198 . this calculation , using the above expression , will not lead to a manageable form . however , we can resort to the expressions of eq.([phi11 ] ) and eq.([chi11 ] ) . then one cycle consists of @xmath199 [ b(\eta ) r(\phi/2 ) b(-\eta ) ] \nonumber \\[2ex ] & { } & \hspace{5 mm } = b(\eta ) r(\phi ) b(-\eta ) , \end{aligned}\ ] ] if the upper - right element is negative . if it is positive , the expression should be @xmath200 [ b(\eta ) x(\chi/2 ) b(-\eta ) ] \nonumber \\[2ex ] & { } & \hspace{5 mm } = b(\eta ) x(\chi ) b(-\eta ) .\end{aligned}\ ] ] if these expressions are repeated @xmath80 times , @xmath201 if the upper - right element is negative . if it is positive , the expression should be @xmath202 as @xmath80 becomes large , @xmath203 and @xmath204 become very large , the beam deviates from the laser cavity . thus , we have to restrict ourselves to the case given in eq.([stable ] ) . the core of the expression of eq.([stable ] ) is the rotation matrix @xmath205^{n } .\ ] ] this means that one complete cycle in the cavity corresponds to the rotation matrix @xmath206 . the rotation continues as the beam continues to repeat the cycle . let us go back to eq.([cycle ] ) . the expression corresponds to a lorentz boosted rotation , or bringing a moving particle to its rest frame , rotate , and boost back to the original momentum . the rotation associated with the momentum - preserving transformation is called the wigner s little - group rotation , which is related to the wigner rotation commonly mentioned in the literature @xcite . the most efficient way to study multi - layer optics is to use the s - matrix formalism @xcite . this formalism is also based on two - by - two matrices , and we can write @xmath207 where @xmath208 and @xmath209 are the incoming , reflected and transmitted beams . here we use the matrix notion @xmath192 for the two - by - two @xmath210 matrix . we consider a system of two different optical layers . for convenience , we start from the boundary from @xmath211 to @xmath212 . we can write the boundary matrix as @xcite @xmath213 taking into account both the transmission and reflection of the beam . the parameter @xmath148 is of course determined by the transmission and reflection coefficient at the surface . this problem was studied in depth by monzon _ et al . _ @xcite , and their research on this subject is still continuing . as the beam goes through the @xmath212 , the beam undergoes the phase shift represented by the matrix @xmath214 when the wave hits the surface of the second medium , the boundary matrix is @xmath215 which is the inverse of the matrix given in eq.([bd11 ] ) . within the second medium , we write the phase - shift matrix as @xmath216 then , when the wave hits the first medium from the second , we have to go back to eq.([bd11 ] ) . thus , the one cycle consists of @xmath217 & { } & \hspace{10 mm } \times \pmatrix{\cosh(\eta/2 ) & -\sinh(\eta/2 ) \cr -\sinh(\eta/2 ) & \cosh(\eta/2 ) } \pmatrix{e^{-i\phi_2/2 } & 0 \cr 0 & e^{i\phi_2/2 } } .\end{aligned}\ ] ] the above matrices contain complex numbers . however , it is possible to transform simultaneously @xmath218 to @xmath219 and transform @xmath220 to @xmath221 using a conjugate transformation , as is shown in appendix [ conju ] . it is also possible to transform these expressions back to their original forms . this transformation property has been discussed in detail in ref . @xcite , and also in appendix [ conju ] of the present paper . as a consequence , the matrix of eq.([m1 ] ) becomes @xmath222 & { } & \hspace{10mm}\times \pmatrix { e^{-\eta/2 } & 0 \cr 0 & e^{\eta/2 } } \pmatrix{\cos(\phi_2/2 ) & -\sin(\phi_2/2 ) \cr \sin(\phi_2/2 ) & \cos(\phi_2/2 ) } .\end{aligned}\ ] ] in the above expression , the first three matrices are of the same mathematical form as that of the core matrix for the one - lens system given in eq.([phi11 ] ) . the fourth matrix is an additional rotation matrix . this makes the mathematics of repetition more complicated , but this has been done @xcite . as a consequence the net result becomes @xmath223 or @xmath224 where the parameters @xmath225 and @xmath226 are to be determined from the input parameters @xmath227 and @xmath228 . detailed calculations are given in ref . @xcite . it is interesting to note that the lorentz group can serve as a computational device also in multi - layer optics . we have seen in this report that the lorentz group provides convenient calculational tools in many branches of ray optics . the reason is that ray optics is largely based on two - by - two matrices . these matrices also constitute the group @xmath26 which serves as the universal covering group of the lorentz group . the optical instruments discussed in this report are the fundamental components in optical circuits . in the world of electronics , electric circuits form the fabric of the system . in the future high - technology world , optical components will hold the key to technological advances . indeed , the lorentz group is the fundamental language for the new world . it is by now well known that the lorentz group is the basic language for quantum optics . coherent and squeezed states are representations of the lorentz group . it is challenging to see how the lorentz nature of the above - mentioned optical components will manifest itself in quantum world . the lorentz group was introduced to physics by einstein and wigner to understand the space - time symmetries of relativistic particles and the covariant world of electromagnetic fields . it is gratifying to note that the lorentz group can serve as the language common both to particle physics and optical sciences . the purpose of this appendix is to expand the third row of table [ einwig ] using the covariant formalism of harmonic oscillators . we explain first what this formalism does for covariant picture of relativistic extended particles . we then point out that this oscillator formalism can serve as the basic language for two - mode coherent states @xcite , namely squeezed states of light . the concept of localized probability distribution is the backbone of the present formulation of quantum mechanics . of course , we would like to have more deterministic form of dynamics , and efforts have been and are still being made along this direction . one of the most serious problems with this probabilistic interpretation is whether this concept of probability is consistent with special relativity . in a given lorentz frame , we know how to do quantum mechanics with a localized probability distribution . how would this distribution look to an observer in a different lorentz frame ? * would the probability distribution appear the same to this observer ? * if different , how is the probability distribution distorted ? * is the total probability conserved ? * the lorentz boost mixes up the spatial coordinate with the time variable . what role does the time - separation variable play in defining the boundary condition for localization and the probability distribution ? we can answer some or all of the above questions only if we construct covariant wave functions , namely wave functions which can be lorentz - boosted . it is easy to construct these wave functions if we know the answer . in the initial development of quantum mechanics , the harmonic oscillator played the pivotal role . thus , it is clear to us that if there is a wave function which can be lorentz - boosted , this has to be the harmonic oscillator wave function . until we construct wave functions which can be lorentz - transformed , we can not say that quantum mechanics is consistent with relativity . indeed , we should examine this problem before attempting to invent more definitive quantum mechanics . since the hadron , in the quark model , is a bound - state of quarks @xcite , feynman , kislinger , and ravndal , in 1971 @xcite , raised the following question in connection with the quark model for hadrons : the hadronic spectrum can indeed be explained in terms of the degeneracy of three - dimensional harmonic oscillator wave functions in the hadronic rest frame ; however , what happens when the hadron moves ? indeed , et al . _ wrote a lorentz - invariant differential equation whose solutions can become non - relativistic wave functions if the time - separation variable can be ignored . this lorentz - invariant differential equation is a four - dimensional partial differential equation , with many different solutions depending on separation of variables and boundary conditions . there is a set of normalizable solutions which can serve as a representation space for wigner s little group for massive particles @xcite . we can start with this set of solutions and give physical interpretations , especially to the time - separation variable . finally , is the covariance of the oscillator wave function consistent with what we observe in the real world . here again , feynman plays the key role . while the quark model can be fit into the oscillator scheme in the hadronic rest frame , feynman in 1969 came up with the idea of partons @xcite . according to feynman s parton model , the hadron consists of an infinite number of partons when it moves with a velocity close to that of light . quarks and partons are believed to be the same particles , but their properties are totally different . while the quarks inside the hadron interact coherently with external signals , partons interact incoherently . if the partons are lorentz - boosted quarks , does the lorentz boost destroy the coherence ? before 1964 @xcite , the hydrogen atom was used for illustrating bound states . these days , we use hadrons which are bound states of quarks . let us use the simplest hadron consisting of two quarks bound together with an attractive force , and consider their space - time positions @xmath229 and @xmath230 , and use the variables @xmath231 the four - vector @xmath232 specifies where the hadron is located in space and time , while the variable @xmath16 measures the space - time separation between the quarks . according to einstein , this space - time separation contains a time - like component which actively participates as can be seen from @xmath233 when the hadron is boosted along the @xmath11 direction . in terms of the light - cone variables defined as @xcite @xmath234 the boost transformation of eq.([boostm ] ) takes the form @xmath235 the @xmath46 variable becomes expanded while the @xmath48 variable becomes contracted . does this time - separation variable exist when the hadron is at rest ? yes , according to einstein . in the present form of quantum mechanics , we pretend not to know anything about this variable . indeed , this variable belongs to feynman s rest of the universe . in this report , we shall see the role of this time - separation variable in the decoherence mechanism . also in the present form of quantum mechanics , there is an uncertainty relation between the time and energy variables . however , there are no known time - like excitations . unlike heisenberg s uncertainty relation applicable to position and momentum , the time and energy separation variables are c - numbers , and we are not allowed to write down the commutation relation between them . how does this space - time asymmetry fit into the world of covariance @xcite . the answer is that wigner s @xmath63-like little group is not a lorentz - invariant symmetry , but is a covariant symmetry @xcite . it has been shown that the time - energy uncertainty relation applicable to the time - separation variable fits perfectly into the @xmath63-like symmetry of massive relativistic particles @xcite . the c - number time - energy uncertainty relation allows us to write down a time distribution function without excitations @xcite . if we use gaussian forms for both space and time distributions , we can start with the expression @xmath236 for the ground - state wave function . what do feynman _ et al . _ say about this oscillator wave function ? in their classic 1971 paper @xcite , feynman _ _ start with the following lorentz - invariant differential equation . @xmath237 this partial differential equation has many different solutions depending on the choice of separable variables and boundary conditions . _ insist on lorentz - invariant solutions which are not normalizable . on the other hand , if we insist on normalization , the ground - state wave function takes the form of eq.([ground ] ) . it is then possible to construct a representation of the poincar group from the solutions of the above differential equation @xcite . if the system is boosted , the wave function becomes @xmath238 this wave function becomes eq.([ground ] ) if @xmath148 becomes zero . the transition from eq.([ground ] ) to eq.([eta ] ) is a squeeze transformation . the wave function of eq.([ground ] ) is distributed within a circular region in the @xmath239 plane , and thus in the @xmath240 plane . on the other hand , the wave function of eq.([eta ] ) is distributed in an elliptic region with the light - cone axes as the major and minor axes respectively . if @xmath148 becomes very large , the wave function becomes concentrated along one of the light - cone axes . indeed , the form given in eq.([eta ] ) is a lorentz - squeezed wave function . this lorentz - squeeze mechanism is illustrated in fig . [ ellipse ] . there are many different solutions of the lorentz invariant differential equation of eq.([osceq ] ) . the solution given in eq.([eta ] ) is not lorentz invariant but is covariant . it is normalizable in the @xmath241 variable , as well as in the space - separation variable @xmath11 . how can we extract probability interpretation from this covariant wave function ? this issue has been discussed thoroughly in the literature @xcite . another pressing problem in physics is that hadrons , like the proton , can be regarded as quantum bound states of quarks when they move slowly . however , they appear like collections of partons when they move with velocity very close to that of light @xcite . since the quarks and partons have quite different properties , it is a challenging problem in modern physics to show that they are the same but appear different depending on the observer s lorentz frame . it has been shown that this lorentz - squeezed wave function can explain the quark model and the parton model as two different manifestations of one covariant entity @xcite . thus , the covariant harmonic oscillator can occupy the third row of table [ einwig ] . the lorentz - squeezed wave function can be written as @xmath242\right\ } .\ ] ] as was discussed in the literature for several different purposes @xcite , this wave function can be expanded as @xmath243 this is an expansion in terms of the oscillations along two different directions . if those directions are @xmath244 and @xmath245 , we can write eq.([expan ] ) as @xmath246 the mathematics of harmonic oscillators can be translated into second - quantized fock space . in the fock space , this expression describes the two - photon coherent state discussed first by yuen in 1976 @xcite . the group @xmath26 consists of two - by - two unimodular matrices whose elements are complex . there are therefore six independent parameters , and thus six generators of the lie algebra . this group is locally isomorphic to the six - parameter lorentz group or @xmath247 applicable to the minkowskian space of three space - like directions and one time - like direction . we can construct representations of the group starting from the lie algebra given in secs . [ lorentz ] and [ spinor ] . we can construct the representation by using a method similar to what goldstein did for the three - dimensional rotation group in terms of the euler angles @xcite . there are three - generators for the rotation group , but goldstein starts with rotations around the @xmath11 and @xmath16 directions . rotations around the @xmath17 axis and the most general form for the rotation matrix can be constructed from repeated applications of those two starting matrices . let us call this type of approach the `` euler construction . '' in constructing the lorentz group , we observe first that we need rotations around two different directions . as for the boost , we need boosts along one given direction since boots along other directions can be achieved by rotations . there are three basic advantages of this approach . first , the number of `` starter '' matrices is less than the number of generators . for example , we need only two starters for the three - parameter rotation group . in our case , we started with two matrices for the three - parameter group @xmath8 and also for @xmath9 . second , each starter matrix takes a simple form and has its own physical interpretation . the third advantage can be stated in the following way . repeated applications of the starter matrices will lead to a very complicated expression . however , the complicated expression can be decomposed into the minimum number of starter matrices . for example , this number is three for the three - dimensional rotation group . this number is also three for @xmath91 and @xmath8 . we call this the euler decomposition . the present paper is based on both the euler construction and the euler decomposition . among the several useful euler decompositions , the iwasawa decomposition plays an important role in the lorentz group . we have seen in this paper what the decomposition does to the two - by - two matrices of @xmath8 , but it has been an interesting subject since iwasawa s first publication on this subject @xcite . it is beyond the scope of this paper to present a historical review of the subject . however , we would like to point out that there are areas of physics where this important mathematical theorem was totally overlooked . for instance , in particle theory , wigner s little groups dictate the internal space - time symmetries of massive and massless particles which are locally isomorphic to @xmath63 and @xmath77 respectively @xcite . the little group is the maximal subgroup of the lorentz group whose transformations do not change the four - momentum of a given particle @xcite . the @xmath77-like subgroup for massless particles is locally isomorphic to the subgroup of @xmath26 which can be started from one of the matrices in eq.([shear1 ] ) and the diagonal matrix of eq.([ps11 ] ) . thus there was an underlying iwasawa decomposition while the @xmath77-like subgroup was decomposed into rotation and boost matrices @xcite , but the authors did not know this . in optics , there are two - by - two matrices with one vanishing off - diagonal element . it was generally known that this has something to do with the iwasawa effect , but simon and mukunda @xcite and han _ et al . _ @xcite started treating the iwasawa decomposition as the main issue in their papers on polarized light . in para - axial lens optics , the matrices of the form given in eq.([shear1 ] ) are the starters , and repeated applications of those two starters will lead to the most general form of @xmath8 matrices . it had been a challenging problem since 1985 @xcite to write the most general two - by - two matrix in lens optics in terms the minimum number of those starter matrices . this problem has been solved recently @xcite , and the central issue in the problem was the iwasawa decomposition . the core matrix of eq.([core ] ) contains the chain of the matrices @xmath248 the lorentz group allows us to simplify this expression under certain conditions . for this purpose , we transform the above expression into a more convenient form , by taking the conjugate of each of the matrices with @xmath249 then @xmath250 leads to @xmath251 in this way , we have converted @xmath252 of eq.([su11 ] ) into a real matrix , but it is not simple enough . after multiplication , the matrix of eq.([abcd ] ) will take the form @xmath259 where @xmath260 and @xmath261 are real numbers . if @xmath82 and @xmath131 vanish , this matrix will become diagonal , and the problem will become too simple . if , on the other hand , only one of these two elements become zero , we will achieve a substantial mathematical simplification and will be encouraged to look for physical circumstances which will lead to this simplification . let us summarize . we started in this section with the matrix representation @xmath252 given in eq.([su11 ] ) . this form can be transformed into the @xmath56 matrix of eq.([abcd ] ) through the conjugate transformation @xmath262 where @xmath131 is given in eq.([ccc ] ) . conversely , we can recover the @xmath252 representation by @xmath263 for calculational purposes , the @xmath56 representation is much easier because we are dealing with real numbers . on the other hand , the @xmath252 representation is of the form for the s - matrix we intend to compute . it is gratifying to see that they are equivalent . let us go back to eq.([abcd ] ) and consider the case where the angles @xmath110 and @xmath226 satisfy the following constraints . @xmath264 thus @xmath265 then in terms of @xmath108 , we can reduce the matrix of eq.([abcd ] ) to the form @xmath266 thus the matrix takes a surprisingly simple form if the parameters @xmath108 and @xmath148 satisfy the constraint @xmath267then the matrix becomes @xmath268 this aspect of the lorentz group is known as the iwasawa decomposition @xcite , and has been discussed in the optics literature @xcite . matrices of this form are not so strange in optics . in para - axial lens optics , the translation and lens matrices are written as @xmath269 respectively . these matrices have the following interesting mathematical property @xcite , @xmath270 and @xmath271 we note that the multiplication is commutative , and the parameter becomes additive . these matrices convert multiplication into addition , as logarithmic functions do . s. ferrara and c. savoy , in _ supergravity 1981 _ , s. ferrara and j. g. taylor eds . ( cambridge univ . press , cambridge , 1982 ) , p. 151 . see also p. kwon and m. villasante , j. math . * 29 * , 560 ( 1988 ) ; _ ibid . _ * 30 * , 201 ( 1989 ) . b. c. sanders and a. mann , group 22 , proceedings of the 22nd international colloquium on group theoretical methods in physics , s. p. cornel _ et al_. eds . ( international press , boston , 1999 ) . see pp 474 - 478 . j. j. monzn and l. l. snchez - soto , am . * 64 * , 156 ( 1996 ) , j. j. monzn and l. l. snchez - soto , j. opt . a * 17 * , 1475 ( 2000 ) ; j. j. monzn , t. yonte , l. l. snchez - soto , and j. carinena , j. opt . am . a , * 19 * , 985 ( 2002 ) . r. p. feynman , _ the behavior of hadron collisions at extreme energies _ , in _ high energy collisions _ , proceedings of the third international conference , stony brook , new york , edited by c. n. yang _ et al . _ , pages 237 - 249 ( gordon and breach , new york , 1969 ) .
all radio signals from cosmic sources are distorted by the plasma turbulence in the interstellar medium ( ism ) . understanding of this turbulence is therefore essential for the proper interpretation of astronomical radio observations . the properties and characteristics of this turbulence can best be studied by observing point - like radio sources , where the results are not influenced by the extended structure of the source , but instead are directly attributable to the effect of the ism itself . pulsars are such sources . dispersion and scattering affect radio emission from pulsars . whereas dispersion in the plasma column introduces delays in arrival time that depend upon frequency and results in smearing of the pulse , scattering by density inhomogeneities causes angular broadening , pulse broadening , intensity modulation or scintillation , and distortion of radio spectra in the form of diffraction patterns . the scattering effects have already been studied extensively theoretically ( see , e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and observationally with ground vlbi of sgra@xmath3 @xcite and pulsars ( see , e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , as well as with ground - space vlbi of psr b0329 + 54 ( _ halca _ , * ? ? ? * ) and the quasar 3c273 ( _ radioastron _ , * ? ? ? whereas the vsop pulsar observations were done at a relatively high frequency of 1.7 ghz and with baselines of @xmath425,000 km and less , ground - space vlbi with _ radioastron _ allows observations at one - fifth the frequency , where propagation effects are expected to be much stronger , and with baselines @xmath5 times longer @xcite . such observations can resolve the scatter - broadened image of a pulsar and reveal new information about the scattering medium @xcite . in this paper , we study the scattered image of the pulsar b0329 + 54 with _ radioastron_. we demonstrate that the pulsar is detected on baselines that fully resolve the scattering disk . the interferometric visibility on these long baselines takes the form of random phase and amplitude variations that vary randomly with observing frequency and time . in the fourier - conjugate domain of delay and fringe rate , the visibility forms a localized , extended region around the origin , composed of many random spikes . we characterize the shape of this region using averages and correlation functions . we argue theoretically that its extent in delay is given by the average envelope of the impulse - response function of interstellar scattering , sometimes called the pulse - broadening function . we find that the observed distribution is well - fit by a model that is derived from an impulse - response function that has two different exponential scales . we discuss possible origins of the two scales . llllr 2012 nov 26 through 29 & 1 hr / day & gb & rcp@xmath6lcp & 570 s + 2014 jan 1 and 2 & 12 hr & wb , kl & rcp & 1170 s [ tab : diary ] our fundamental observable is the interferometric visibility @xmath7 . in the domain of frequency @xmath8 , this is the product of electric fields at two antennas @xmath9 and @xmath10 : @xmath11 this representation of the visibility is known as the cross spectrum , or cross - power spectrum . because electric fields at the antennas are complex and different , @xmath12 is complex . usually visibility is averaged over multiple accumulations of the spectrum , to reduce noise from background and the noiselike electric field of the source . the second argument @xmath13 allows for the possibility that the visibility changes in time , as it does for a scintillating source , over times longer than the time to accumulate a single spectrum . such a spectrum that changes in time is known as a `` dynamic spectrum '' @xcite . the correlator used to analyze our data , as discussed in sections [ sec : obs ] and [ sec : red ] , calculates @xmath14 @xcite . hereafter , we omit the baseline subscript indicating baseline @xmath15 in this paper , except in sections of the appendix where the baseline is important . in various domains , and functions derived from it . the fundamental observable is the visibility in the domain of frequency @xmath8 and time @xmath13 , @xmath16 ; this is known as the cross - power spectrum , or cross spectrum . an inverse fourier transform of @xmath8 to delay @xmath17 leads to the visibility @xmath18 ; this is the cross - correlation function of electric fields in the time domain ( see equation [ eq : vtau_def ] ) . a forward fourier transform of @xmath13 to fringe rate @xmath19 leads to @xmath20 . a forward transform of @xmath17 back to @xmath8 produces @xmath21 , and an inverse fourier transform of @xmath19 to @xmath13 returns to @xmath16 the square modulus of @xmath18 is @xmath22 . the cross - correlation function in @xmath17 of @xmath23 for right- and @xmath24 for left - circular polarization is @xmath25 . we denote the fourier transform by @xmath26 , and quantities in the domain of frequency @xmath8 by the accent @xmath27 . [ fig : ftcircle],scaledwidth=48.0% ] under the assumptions that the source is pointlike , and that we can ignore background and source noise , the impulse - response function of interstellar scattering @xmath28 determines the visibility of the source . a single delta - function impulse of electric field at the source is received as a function @xmath29 of time @xmath30 at the observer . here , @xmath30 is fourier - conjugate to @xmath8 and varies at the nyquist rate . the visibility is the product of fourier transforms of @xmath28 at the two antennas : @xmath31 where @xmath32 is the fourier transform of @xmath29 . we denote the typical duration of @xmath29 as @xmath33 , the broadening time for a sharp pulse . within this time span , @xmath29 has a complicated amplitude and phase . the function @xmath29 changes over longer times , as the line of sight shifts with motions of source , observer , and medium . this change takes place on a timescale @xmath34 , and over a spatial scale @xmath35 . the shorter and longer timescales @xmath33 and @xmath36 lead to our use of dual time variables : @xmath30 , of up to a few times @xmath37 and fourier - conjugate to @xmath8 ; and @xmath13 , of a fraction of @xmath36 or more and fourier - conjugate to @xmath19 . this duality is commonly expressed via the `` dynamic spectrum '' ( see section [ sec : appimpluseresponsefunction ] ) . if the scattering material remains nearly at rest while the line of sight travels through it at velocity @xmath38 , then one spatial dimension in the observer plane maps into time , and @xmath39 the averaged square modulus of @xmath28 is the pulse - broadening function @xmath40 . here , the subscripted angular brackets @xmath41 indicate an average over realizations of the scattering . this function is the average observed intensity for a single sharp pulse emitted at the source . an average over time is usually assumed to approximate the desired average over an ensemble of statistically - identical realizations of scattering . we derive a number of representations of the visibility and quantities derived from it , and show that these provide straightforward means to extract the impulse - response function . these functions are summarized in figure [ fig : ftcircle ] , and discussed briefly here , and in detail in section [ sec : impulse - response_visibility ] of the appendix . in particular , visibility in the domain of delay @xmath17 and time @xmath13 is @xmath18 . this is the correlation function of electric field at the two antennas @xmath9 and @xmath10 ( equation [ eq : vtau_def ] ) , and is the inverse fourier transform of @xmath42 from @xmath8 to @xmath17 . we are also concerned with the square modulus of @xmath18 ( see section [ sec : acf_visibility ] ) : @xmath43 we calculate @xmath44 for right- and left - circular polarizations separately , and then correlate them in delay @xmath17 to form @xmath45 , the cross - correlation between polarizations : @xmath46 here , @xmath45 is the correlation of a single measurement of @xmath23 and @xmath24 , and @xmath47 is the number of samples in @xmath23 and @xmath24 . when averaged over many realizations of the scattering material , @xmath48 is related to the statistics of the pulse - broadening function @xmath49 . most commonly , the average over many realizations of scattering material is approximated by averaging over a time much longer than @xmath34 ; for this reason we omit the time argument for @xmath50 . equivalently , evaluation of @xmath51 at the fringe rate @xmath52 of the maximum magnitude of @xmath45 yields the same time average . for this theoretical discussion , @xmath53 ; for practical observations , instrumental factors can offset the fringe rate from zero , so that @xmath52 provides the most reliable time average . for a baseline that extends much further than the scale of scattering @xmath35 ( see equation [ eq : k_result ] ) : @xmath54 here , we introduce the symbol @xmath55 to indicate convolution , and denote the time - reverse of @xmath49 as @xmath56 . our analysis method differs somewhat from @xcite , who used structure functions of intensity , visibility , and visibility squared to study scattering of pulsar b0950 + 08 on an extremely long baseline to _ radioastron_. the two methods are closely related theoretically . structure functions are particularly valuable when the characteristic bandwidth approaches the instrumental bandwidth , and can be extended to cases where the signal - to - noise ratio is low , as they discuss . lcc 2012 nov 26 & 60 & 60 + 2012 nov 27 & 90 & 60 + 2012 nov 28 & 175 & 60 + 2012 nov 29 & 235 & 60 + 2014 jan 1 & 20 & 60 + 2014 jan 2 & 70 & 100 + 2014 jan 2 & 90 & 120 [ tab : baselines ] the observations were made in two sessions : the first for one hour each on the four successive days november 26 to 29 , 2012 , and the second for a total of 12 hours on the two days january 1 and 2 , 2014 . the first session used the 10-m _ radioastron _ space radio telescope ( ra ) together with the 110-m robert c. byrd green bank telescope ( gb ) . the second session used the ra together with the @xmath0-m westerbork synthesis radio telescope ( wb ) , and the 64-m kalyazin radio telescopes ( kl ) . both right ( rcp ) and left circular polarizations ( lcp ) were recorded in november 2012 , and only one polarization channel ( rcp ) was recorded in january 2014 . because of an ra peculiarity at 324 mhz , the 316332 mhz observing band was recorded as a single upper sideband , with one - bit digitization at the ra and with two - bit digitization at the gb , wb , and kl . science data from the ra were transmitted in real time to the telemetry station in pushchino @xcite and then recorded with the _ radioastron _ data recorder ( rdr ) . this type of recorder was also used at the kl , while the mk5b recording system was used at the gb and wb . table [ tab : diary ] summarizes the observations . the data were transferred via internet to the astro space center ( asc ) in moscow and then processed with the asc correlator with gating and dedispersion applied @xcite . to determine the phase of the gate in the pulsar period , the average pulse profile was computed for every station by integrating the autocorrelation spectra obtained from the asc correlator . the autocorrelation spectra @xmath57 are the square modulus of electric field at a single antenna . in november 2012 the projected baselines to the space radio telescope were about 60 , 90 , 175 , and 235 thousand kilometers for the four consecutive days , respectively . data were recorded in 570-second scans , with 30-second gaps between scans . in january 2014 the projected baselines were about 20 , 70 , and 90 thousand kilometers during the 12-hour session . data were recorded in 1170-second scans . the ra operated only during three sets of scans of 60 , 100 and 120 min each , with large gaps in between caused by thermal constraints on the spacecraft . the auto - level ( agc ) , phase cal , and noise diode were turned off during our observations to avoid interference with pulses from the pulsar . table [ tab : baselines ] gives parameters of the earth - space baselines observed . all of the recorded data were correlated with the asc correlator using 4096 channels for the november 2012 session and 2048 channels for the january 2014 session , with gating and dedispersion activated . the on - pulse window was centered on the main component of the average profile , with a width of 5 ms in the november 2012 session and 8 ms in the january 2014 session . these compare with a 7-ms pulse width at 50% of the peak flux density @xcite . the off - pulse window was offset from the main pulse by half a period and had the same width as the on - pulse window . the correlator output was always sampled synchronously with the pulsar period of 0.714 s ( single pulse mode ) . we used ephemerides computed with the program tempo for the earth center @xcite . the results of the correlation were tabulated as cross power spectra , @xmath16 , written in standard fits format . rrrrrcc nov 2012 & @xmath58 & @xmath59 & @xmath60 & @xmath61 & @xmath62 & @xmath63 + jan 2014 & @xmath64 & @xmath65 & @xmath66 & @xmath67 & @xmath68 & [ tab : scatt_params ] using autocorrelation spectra at gb , kl , and wb , we measured the scintillation time @xmath36 and bandwidth @xmath69 . the results are given in table [ tab : scatt_params ] . our analysis using interferometric data , for which the noise baseline is absent and the spectral resolution was higher , is more accurate for the constants @xmath70 and @xmath71 as discussed below , so we quote those values in table [ tab : scatt_params ] . the asc correlator calculates the cross - power spectrum , @xmath16 , as discussed in sections [ sec : theory_intro ] and [ sec : visibility ] . the resolution of the resulting cross - power spectra is 3.906 khz for the 2012 observations and 7.812 khz for the 2014 observations . because the scintillation bandwidth was comparable to the channel bandwidth for the 2014 observations , as shown in table [ tab : scatt_params ] , and because the single recorded polarization at that epoch prevented us from correlating polarizations to form @xmath45 , as discussed in section [ sec : two_scales ] , we focus our analysis and interpretation on the 2012 observations . we investigated the scattering of the pulsar from the visibility in the delay - fringe - rate domain , @xmath20 . we studied the statistics of visibility @xmath72 as a function of delay , fringe rate , and baseline length . if there were no scattering material between the pulsar and the observer , we would expect for @xmath73 one spike at zero delay and fringe rate with magnitude that remains constant as a function of baseline length , and with width equal to the inverse of the observed bandwidth in delay , and the inverse of the scan length in fringe rate . scattering material in between changes this picture . first we expect the spike at zero delay and fringe rate to decrease in magnitude with increasing baseline length , perhaps to the point where it would become invisible . second , we expect additional spikes to appear around the spike at zero delay and fringe rate . the distribution of these spikes give us invaluable information about the statistics of the scattering material . as we discuss in this section , we fitted models to the distribution of visibility , as measured by the correlation function @xmath45 , and thus derived scintillation parameters that describe the impulse - response function for propagation along the line of sight from the pulsar . we also computed the maximum visibility as a function of projected baseline length , as we discuss in detail in a separate paper ( paper ii : popov et al . , in preparation ) . for strong single pulses the visibility in the cross spectra , @xmath16 , had signal - to - noise ratios sufficiently large for a useful analysis . however , we decided to analyze the data from the time series of multiple pulses . fourier transform of the cross spectrum , @xmath16 , to the delay / fringe - rate domain yields @xmath74 and concentrates the signal into a central region , and thus provides a high signal - to - noise ratio . the sampling rate of individual cross spectra in the time series was the pulse period of 0.714 s , as noted in section [ sec : corr ] . the time span of cross spectra used to form @xmath72 varied , ranging from 71.4 s to 570 s , depending upon the application . in figure [ fig : frrt_delay ] we display the magnitude of the visibility in the delay / fringe - rate domain , @xmath73 , for a 500-s time span . the data were obtained on 29 november 2012 in the rcp channel for a projected @xmath75 gb - ra baseline . the cross spectra , @xmath16 , from which we obtained @xmath73 were sampled with 4096 spectral channels across the 16-mhz band , at the pulsar period of 0.714 s ; consequently , the resolution was 0.03125 @xmath76s in delay , and 2 mhz in fringe rate . as figure [ fig : frrt_delay ] shows , no dominant central spike is visible at zero delay and fringe rate , as would be expected for an unresolved source . our long baseline interferometer completely resolves the scattering disk . instead we see a distribution of spikes around zero delay and fringe rate that is concentrated in a relatively limited region of the delay - fringe rate domain . the locations of the various spikes appear to be random . because the scattering disk is completely resolved on our long baseline , we conclude that the spikes are a consequence of random reinforcement or cancellation of paths to the different locations of the two telescopes , and hence interferometer phase . in figure [ fig : frrt_delay ] , the distribution of the magnitude of visibility is relatively broad along the delay axis and relatively narrow along the fringe rate axis . the extent is limited in delay to about the inverse of the scintillation bandwidth , @xmath77 ; and in fringe rate to about the inverse of the diffractive timescale @xmath78 . within this region , the visibility shows many narrow , discrete spikes . if statistics of the random phase and amplitude of scintillation are gaussian , and the phases of the fourier transform randomize the different sums that comprise the visibility in the delay - fringe rate domain , then the square modulus of @xmath72 should be drawn from an exponential distribution , multiplied by the envelope defined by the deterministic part of the impulse - response function , as discussed in the appendix . along the delay axis , @xmath79 takes the general form suggested by figures [ fig : frrt_delay ] and [ fig : delay_fine_structure ] : a narrow spike surrounded by a broad distribution . we found that the central spike takes the form of a sinc function in both delay and fringe rate coordinates , as expected for uniform visibility across a square passband @xcite . the widths are somewhat larger than values expected from observing bandwidth of 16 mhz and time span of 71.4 s , of @xmath80 ns and @xmath81 mhz respectively , probably because of the non - uniformity of receiver bandpasses and pulse - to - pulse intensity variations , respectively . the broader part of the distribution takes an exponential form along the fringe - rate axis in this case ; more generally , the form can be complicated , particularly over times longer than 600 s. traveling ionospheric disturbances may affect the time behavior of our 92-cm observations ; in particular , they may be responsible for the 20 to 25 mhz width of the narrow component in fringe rate , as noted in table [ tab : scatt_params ] . we do not analyze the broader distribution in fringe rate further in this paper ; we will discuss this distribution , and the influence of traveling ionospheric disturbances , in a separate publication ( paper iii , popov et al . in preparation ) . because of the relatively small optical path length of the ionosphere , even at @xmath82 cm , they can not affect the cross spectrum @xcite . the distribution of the magnitude of the visibility in delay / fringe - rate domain changes with baseline length . figure [ fig : delay_fine_structure ] displays cross - sections through the maximum of the distribution of magnitude for a range of baseline lengths , as a function of delay . the maxima lie near zero fringe rate , as expected . under the plausible and usual assumption that the correct fringe rate lies at the fringe rate , @xmath83 , where the distribution peaks , the cross - section represents the visibility averaged over the time span of the sample : @xmath84 the top panel of figure [ fig : delay_fine_structure ] shows this cross - section through figure [ fig : frrt_delay ] . the next lower panel shows the cross - section for the slightly shorter kl - ra baseline . the three lower plots give the equivalent cross - sections for 10 times and 100 times shorter projected baselines . these three short - baseline cross - sections are qualitatively different from the long baseline cross sections : the visibility has a central spike resulting from the component of the cross - spectrum that has a constant phase over frequency , as well as the broad distribution from the component that has a varying phase over frequency . the central spike is strongest for the shortest baseline and weaker for the next longer baselines , as expected based on the results of sections [ sec : visibility ] and [ sec : acf_visibility ] . at very long baselines the central spike is absent even after averaging the visibility over the whole observing period , and only the broad component is present . as expected from figure [ fig : frrt_delay ] , in the delay / time domain the broad component appears as spikes distributed over a range of about @xmath85s in delay . these spikes keep their position in delay for the scintillation time of about 100 to 115 s , as listed in table [ tab : scatt_params ] . the character of the broad component changes with baseline length as well : mean and mean square visibility are the same for short and long baselines ; but excursions to small and large visibilities are more common for a long baseline ( * ? ? ? * eq . 12 ) . averages of the visibility , and averages of the correlation function of visibility , extract the parameters of the broad and narrow components of visibility . such averages approximate the statistical averages discussed in sections [ sec : theory_intro ] and [ sec : impulse - response_visibility ] . they seek to reduce noise from the observing system and emission of the source , as well as variations from the finite number of scintillations sampled , while preserving the statistics of scintillation . the averages and correlation functions allow the inference of parameters of the impulse - response function of propagation from the statistics of visibility . the mean square modulus of visibility , @xmath86 , provides useful and simple characterization of visibility . to approximate the average over realizations of scattering @xmath41 , we average over many samples in time @xmath13 and over bins in delay @xmath17 . we realize the average over time by evaluating @xmath20 at the fringe rate of maximum amplitude @xmath52 , as discussed in section [ sec : theory_intro ] . we also average over 16 lags in delay @xmath17 . the resulting average shows a broad component surrounding the origin ; on shorter baselines , it shows a spike at the origin . the broad distribution samples the properties of the fine structure seen in figures [ fig : frrt_delay ] and [ fig : delay_fine_structure ] , and the spike to those seen on the shorter baselines in figure [ fig : delay_fine_structure ] . we argue in section [ sec : impulse - response_visibility ] that the spike in @xmath87 is related to the average visibility , and the broad component to the impulse - response function . figure [ fig : cross - section_fr - dl ] shows an example of the broad component of @xmath87 . this is estimated as @xmath88 , by selecting the peak fringe rate @xmath52 to average in time for each of 6 scans , averaging the results for the scans , and averaging over 16 lags of delay to smooth the data . these averaging procedures serve to approximate the average over an ensemble of realizations of scattering . background noise adds complex , zero - mean noise to @xmath72 , with uniform variance at all lags ; this adds a constant offset to the average @xmath89 . along the delay axis , at the fringe rate @xmath52 where the magnitude of visibility peaks , close to zero mhz . the visibilities for the gb - ra baseline on 2012 nov 28 at 21:40 ut are shown as open circles . the visibilities were computed by an inverse fourier transform of the spectra , @xmath42 , over 71.4 s time spans , and then by averaging over 6 observing scans , each 570 s long . they were then further averaged in delay , over 16 points or @xmath90 @xmath76s , to smooth fluctuations . the dashed horizontal line shows the offset contributed by background noise . the solid gray line shows the reconstructed form given by equation [ eq : ctau_2exp_form ] , offset by the noise level , with parameters taken from the fit shown in figure [ fig : ktwoscaleexamples ] . the light dashed curve shows only the narrow component of the two - exponential model . units of visibility are correlator units.,scaledwidth=47.0% ] using equation [ eq : krldef ] , we estimated @xmath92 , the averaged cross - correlation function between the square modulus of right - circular polarized ( rcp ) and of left - circular polarized ( lcp ) of visibility in the delay domain . ( note that @xmath93 is not the correlation function of the average @xmath94 , but rather the average of the correlation function @xmath95 . ) because the background noise in the two circular polarizations is uncorrelated , they do not contribute an offset to @xmath50 . this allows us to follow the effects of the impulse - response function to much lower levels than for @xmath96 . the correlation function @xmath50 is thus less subject to effects of noise , and is more sensitive to the broad component of the distribution , than @xmath96 . to compute an estimate of @xmath92 , we calculated the squared sum of real and imaginary components of @xmath97 , the inverse fourier transform of the cross - power spectrum . we formed these for each strong pulse , and normalized them by the autocorrelation functions at each antenna . from these we formed the un - averaged correlation function @xmath98 we then averaged @xmath45 over 570-sec scans to form @xmath99 . averaging in the time domain approximated an average over realizations of the impulse - response function for the scattering medium . each 570-sec scan included 100 to 250 strong pulses , yielding one averaged sample of @xmath100 for each scan . we obtained 22 measurements in total , with 6 samples of @xmath101 for november 26 , 28 , 29 observing sessions . we obtained only 4 such samples for november 27 because of no significant detections of @xmath7 for two scans on that date . examination of the averaged cross correlation function , @xmath102 , revealed a spike at the origin and two exponential scales for the broad component , a large one and a small one . figure [ fig : ktwoscaleexamples ] shows an example . the spike at the origin arises from the fine structure of scintillation in the broad component of visibility , as seen in figures [ fig : frrt_delay ] and [ fig : delay_fine_structure ] . this structure is identical in right- and left - circular polarizations , so its correlation leads to the spike . the two exponential scales are apparent as the slopes of the steeper and narrower parts of the distribution . we see these two scales even for single pulses , which are strong enough to show the two - scale structure . we did not observe these scales without doubt in spectra from single - dish observations , because the resulting correlation functions are more subject to noise , gain fluctuations , and interference . the scales are both present for @xmath96 , but the longer scale is seen more clearly in @xmath50 ( as comparison of equations [ eq : ctau_2exp_form ] through [ eq : k_2exponential_model_params ] shows ) . we formalized the two exponential scales seen for @xmath103 with a model fit . the model assumed a pulse - broadening function with two exponential scales . under this assumption , a short pulse appears at the observer with average shape : @xmath104 the pulse rises rapidly , and falls as the sum of the two exponentials . the assumed form for @xmath49 leads to predictions for the forms of @xmath105 and @xmath106 , as discussed in section [ sec : appendix2exponentials ] . for @xmath105 , we expect a cusp at the origin , and two exponentials with scales @xmath107 and @xmath108 and different weights on either side . for @xmath106 , correlation smooths the cusp at the origin , producing a smooth peak , with the same exponential scales appearing to either side . figure [ fig : ktwoscaleexamples ] shows the best - fitting model of this form for the data shown there . this model has parameters : @xmath109 the model reproduces the two scales , and the smooth peak , well . the model also predicts the magnitude of the spike accurately , with zero average visibility @xmath110 . the model shown in figure [ fig : cross - section_fr - dl ] shows the model for @xmath44 , reconstructed using equation [ eq : ctau_2exp_form ] with parameters from the fit to figure [ fig : ktwoscaleexamples ] . the two scales appear in the model , although the offset from noise contributes at large delay @xmath17 . as the figure shows , a single exponential does not fit the model well : the narrow component is satisfactory at small @xmath17 , but falls well under the data at larger @xmath17 . a high - winged function such as a lorentzian can fit @xmath44 well , but the rounded peak leads to a very wide peak for @xmath91 that can not match the data , and the inversion to a @xmath111 that remains finite , and is zero for @xmath112 as causality demands , is problematic . the best - fitting scales and the magnitudes of the two contributions varied from scan to scan , but in a manner that was consistent with our finite sample of the scintillation pattern , and the inhomogeneous averaging of pulses with different intensities . we show a histogram of the results of our fits to 570-sec intervals in figure [ fig : scaledistributions ] . on 26 to 29 november 2012 , the shorter scale averaged to @xmath113 , and the longer scale to @xmath114 . the scales had a relative power of @xmath115 . on a long baseline that fully resolves the scattering disk , as figure [ fig : delay_fine_structure ] shows , we observe multiple sharp spikes in the visibility @xmath20 is a consequence of the variation of the amplitude and phase of visibility . ( see also paper ii , popov et al . in preparation ) . the characteristic region of that variation , @xmath116 , reflects the product of the inverses of the scintillation bandwidth @xmath117 and the scintillation timescale @xmath118 . these quantities are the width in time of the impulse - response function , and the time for the impulse - response function to change as the line of sight to the observer moves through the scattering material . detailed examination of the correlation function of visibility @xmath119 reveals the presence of two characteristic , exponential scales . both scales are visible in the single - pulse correlation functions of right and left circular polarization , as well as in the correlation function @xmath120 averaged over 570 s shown in figure [ fig : ktwoscaleexamples ] . for an assumed screen distance of half the pulsar distance of @xmath121 kpc @xcite , the two scales correspond to diffractive scales of : @xmath122 the diffractive scale is the lateral distance at the screen where phases decorrelate by a radian @xcite . the refractive scale gives the scale of the scattering disk : @xmath123 in contrast , @xcite measured angular broadening for psr b0329 + 54 of @xmath124 mas at @xmath125 mhz , where @xmath126 is the full width of the scattered image at half the maxium intensity . this corresponds to a refractive scale of @xmath127 . this upper limit is somewhat smaller than the values obtained from our observations , even if one takes into account the facts that the larger scale contains only 0.38 of the power of the shorter one , and that the scattering material may be somewhat closer to the pulsar than to the observer . @xcite studied the scattering properties of psr b0329 + 54 in detail , using single - antenna observations at 102 mhz , 610 mhz , 5 ghz , and 10.6 ghz to form structure functions of the scintillation in time and frequency on a wide range of scales . they concluded that the scattering material has a power - law spatial spectrum with index @xmath128 , marginally consistent with the value of @xmath129 expected for a kolmogorov spectrum , with an outer scale of @xmath130 . using vlbi , @xcite observed psr b0329 + 54 at 2.3 ghz and set limits on the separation on the emission regions corresponding to different components of the pulse profile . @xcite observed psr b0329 + 54 at @xmath131 ghz with ground - space baselines to halca , and found that the source varied strongly with time . they ascribed this variation to scintillation , with the scintillation bandwidth comparable to the observing bandwidth at their observing frequency . self - calibration with timespans less than the scintillation time returned a pointlike image , as expected . @xcite analyzed these data , including ground - ground baselines . they studied both single - antenna autocorrelation functions @xmath132 and cross - correlation functions @xmath133 . they detected two timescales for the scintillation pattern , of 20 min and 1 min . they found that the properties of scattering could not be explained by a single , thin screen , and further that velocities indicated relative motions within the scattering medium . @xcite had previously observed two coexisting scales of scattering for psr b0329 + 54 . they found scintillation bandwidths of @xmath134 hz and @xmath135 hz , measured as the @xmath136 point of the correlation function of intensity at an observing frequency of 102 mhz , using the large cophase array of puschino observatory . the ratio of these scales , @xmath137 is larger than the ratio of @xmath138 that we observe . scaled to our observing frequency of 324 mhz , using the @xmath139 scaling appropriate for a kolmogorov spectrum , and converting from @xmath140 to @xmath17 using the uncertainty relation @xmath141 , we find that these values correspond to 1.3 and 8 @xmath76s , respectively , about a factor of 3 smaller that the scales we observe . of course , interpolation over a factor of 3 in observing frequency and the different observing techniques may introduce biases , and scattering parameters likely vary over the years between the two measurements . two scales of scattering have also been observed for other pulsars @xcite . two scales of scattering may be a consequence of a variety of factors . non - gaussian statistics of scattering can produce multiple scales , although this usually appears as a continuum of scales rather than two different individual scales , as in a power - law distribution or a levy flight @xcite . a kolmogorov model for scattering in a thin screen does not fit as well as our model based upon a two - exponential impulse - response function , or even as one based upon one exponential . a model with two discrete scales appears to fit our data better . one explanation is anisotropic scattering . this can produce two scales , corresponding to the major and minor axes of the scattering disk , as discussed in section [ sec : anisotropic ] of the appendix . the ratio of the scales of @xmath142 corresponds to the parameter @xmath143 , and an axial ratio of @xmath144 . in a simple model for anisotropic scattering in a thin screen , we expect the ratio of power in the scales to be approximately @xmath145 , as shown in section [ sec : anisotropic ] of the appendix . this compares well with our observed ratio of @xmath146 . however , our observations for ground - space baselines at a variety of orientations do not show anisotropy . a variety of models , involving material with varying anisotropy distributed along the line of sight , and strong anisotropy that slips between our long baselines , might match our data . a second explanation is the complicated structure observed within dynamic spectra : most commonly observed as `` scintillation arcs '' @xcite . recently , it has been suggested that this structure arises from interference among subimages , resulting from refraction by interstellar reconnection sheets @xcite . this complicated structure produces time and frequency variations on a wide range of scales . of course , we are considering very long baselines , where the scintillation - arc patterns should be completely uncorrelated between antennas . this may lead to blurring , resulting in a 2-scale correlation function without particularly strong structure corresponding to the discrete arcs seen on shorter baselines @xcite . we do not see any direct evidence of scintillation arcs , as such . the magnitude of the visibility shows a featureless decline with increase of either of the 2 dimensions @xmath147 and @xmath148 . the gb autocorrelation functions do not show scintillation arcs either , for our observations . we made vlbi observations of psr b0329 + 54 with _ radioastron _ at 324 mhz on projected baselines of up to 235,000 km . our goal was to investigate scattering by the interstellar medium . these properties affect radio observations of all celestial sources . while the results of such observations are in general influenced by the convolution of source structure with the scattering processes , pulsars are virtually point - like sources and signatures in the observational results can be directly related to the scattering properties of the interstellar medium . on long baselines , in the domain of delay @xmath17 and fringe - rate @xmath19 , the correlation function of visibility @xmath72 is a collection of narrow spikes , located within a region defined by the inverses of the scintillation bandwidth @xmath117 and the scintillation timescale @xmath118 . for shorter baselines , a sharp spike at the center of this region represents the average visibility ; on long baselines where the average visibility drops to near zero , this spike is absent . the mean square visibility , @xmath149 , is well fit with a smooth model , indicating that the visibility spikes are the result of random interference of many scattered rays . to form a quantity less subject to effects of noise , we convolve the mean - square left- and right - circular polarized visibility to form @xmath150 . the average correlation function @xmath50 shows two exponentials with different characteristic timescales . the forms of @xmath151 and @xmath152 are well fit with a simple model , that assumes that the average pulse - broadening function @xmath49 is the sum of two exponentials with different timescales . on 2012 nov 26 to 29 , the shorter timescale was @xmath153 , and the longer timescale was @xmath154 , with the longer - scale exponential containing approximately 0.38 times the power of the shorter - scale exponential . this double exponential may arise from anisotropic scattering ; or from scattered radiation at large angle , perhaps corresponding to the subimages seen in single - dish and shorter - baseline observations . further investigation of the properties of the image of the scattered pulsar on long and short baselines , using these data , will help to clarify the origin of the two scales . the radioastron project is led by the astro space center of the lebedev physical institute of the russian academy of sciences and the lavochkin scientific and production association under a contract with the russian federal space agency , in collaboration with partner organizations in russia and other countries . the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . this study was supported by russian foundation for basic research grant 13 - 02 - 00460 and basic research program p-7 of the presidium of the russian academy of sciences . c.r.g . acknowledges support of the us national science foundation ( ast-1008865 ) . n.b . was supported by nserc . we thank the referee for constructive comments which helped to improve the manuscript . _ facilities : _ . under general assumptions , refraction and scattering convolve the electric field of a source with an impulse - response function @xmath28 . this function varies with position in the observer plane , decorrelating over some lateral scale ; and with time , as the line of sight to the source moves with respect to the scattering material , and as the scattering material evolves . the task of this section is to relate the impulse - response function to the statistics of visibility , as given by the functions @xmath44 and @xmath91 introduced in section [ sec : theory_intro ] above . the visibility @xmath155 is the conjugated product of electric fields at two antennas ( equation [ eq : vdef ] ) . we usually omit the subscripts indicating baseline @xmath15 on @xmath7 , unless they are important for the immediate argument . we denote the fourier transform from the time or delay domain ( @xmath13 or @xmath17 ) to the frequency or fringe rate domain ( @xmath8 or @xmath19 ) by @xmath156 , and its inverse by @xmath157 . we accent symbols with tilde `` @xmath27 '' to denote quantities that depend on observing frequency @xmath8 , and the same symbols without accent for the fourier - conjugate domain of delay @xmath17 or time @xmath30 . we assume that the variables describing time and frequency @xmath158 , are discrete . they range from @xmath159 to @xmath160 , where @xmath47 is the number of samples in the time or frequency span . for @xmath17 and @xmath30 one sample is the inverse of the nyquist rate , and they can span the time to accumulate a single realization of the spectrum ; for @xmath13 and its fourier conjugate @xmath19 one sample is the averaging time for one spectrum , and they can span one observation . our convention for normalization of the fourier transform is that a function @xmath161 normalized to unit area in the delay domain has value unity at zero frequency : @xmath162 . conversely , if @xmath163 is normalized to unit area in the frequency domain , @xmath164 . this is the `` @xmath165 '' convention of wolfram mathematica @xcite . with this convention , parseval s theorem takes the form : @xmath166 as noted above , the observed electric field of a pulsar @xmath167 is the convolution of the electric field emitted at the source with a kernel @xmath28 that depends on scattering : @xmath168 where we introduce the symbol @xmath55 for convolution . the kernel @xmath28 is the impulse - response function ; in other words , if the pulsar emits a sharp spike , then the observed electric field of the pulse is simply a copy of @xmath28 . because of this convolution , @xmath28 is also known as the propagation kernel ; it is also known as the green s function , and the s matrix ( * ? ? ? * and references therein ) . both @xmath169 and @xmath28 vary at the nyquist rate : the inverse of the total observed bandwidth . usually , we assume that the intrinsic electric field of the source is white noise at the nyquist rate : it is drawn from a gaussian distribution in the complex plane at each instant @xcite . the impulse - response function extends over a time span of a few times @xmath33 , representing the time over which a sharp pulse at the source would be received . it is zero outside this relatively narrow time window . if the statistics of the scattering material are stationary , the characteristic shape and scales of @xmath28 will remain fixed , while details of amplitude and phase vary on the timescale @xmath34 ( equation [ eq : t_sc ] ) . an average of the squared electric field over many impulses emitted by the source over times longer than @xmath34 will reveal the characteristic form . one simple model form for @xmath28 that includes deterministic and random parts is the product of a non - varying envelope @xmath170 , and a random function @xmath171 , that varies rapidly with @xmath30 during the course of each pulse : @xmath172 both @xmath171 and @xmath173 span a few times @xmath33 . over that time span , @xmath171 varies wildly , and randomly ; however , it exhibits nearly the same form for the next pulse . over the longer timescale @xmath174 , the form of the random function @xmath171 changes slowly . for typical observations of a pulsar , such as those described in this paper , @xmath33 is some fraction of the width of one pulse , or a few microseconds ; whereas @xmath34 is many pulsar periods , or many seconds . such situations , where a convolution may have a slowly - varying kernel , are commonly treated as `` dynamic spectra '' ( see * ? ? ? the intensity received by an observer for an electric - field impulse at the source , averaged over many such impulses with different realizations of the scattering material , is the square modulus of the deterministic part of @xmath28 , which we call @xmath49 : @xmath175 here , the subscripted angular brackets @xmath176 indicate a statistical average over realizations of the scattering medium , for example as approximated by an average over pulses spanning a time greater than @xmath34 . often , @xmath49 is called the pulse - broadening function . so that propagation kernel leaves the intensity of the source unchanged , when averaged over time , we set : @xmath177 for strong scattering , as is observed for most pulsars at most wavelengths , we expect that many different paths , with random amplitude and phase , will contribute to the received pulse at each instant @xmath30 . therefore , we expect that the random part @xmath171 will have the statistics of a random walk at each instant : the observed electric field will be drawn from a circularly gaussian distribution in the complex plane , with zero mean . on the other hand , the deterministic part @xmath173 sets the standard deviation of @xmath28 ; it reflects how many paths , and with what strength , contribute at each delay . this model for scintillation is closely related to the amplitude - modulated - noise ( amn ) model for pulsar emission @xcite . in this model , the electric field emitted by the pulsar is the product of noise , drawn from a zero - mean gaussian distribution in the complex plane , with a more - slowly varying envelope that determines the standard deviation of the noise at each instant . we suppose that the random part of the propagation kernel is completely uncorrelated in time , at the nyquist rate , within its span of a few @xmath33 . then , at a location `` @xmath9 , '' @xmath178 on the other hand , @xmath179 is nearly the same for each emitted pulse ; it changes only over the longer timescale @xmath34 . the question of how this slower variation of @xmath171 with time depends upon baseline length is much more complicated , and we discuss it briefly below . however , if the lateral separation of the two stations @xmath9 and @xmath10 is much greater than the scale of the scattering pattern , then the random parts of @xmath28 for the two stations , @xmath179 and @xmath180 , are completely uncorrelated . as the previous discussion shows , the impulse - response function involves three timescales : @xmath171 changes at the nyquist rate ; @xmath173 varies over the typical span of the impulse - response function @xmath33 ; and the time for the random variations of @xmath171 to change is @xmath34 . the dynamic cross - power spectrum provides a useful description for these different variations @xcite . a single sample of the cross - power spectrum , when averaged over time less than @xmath34 , has the characteristic scale @xmath181 , resulting from the finite span of the impulse - response function and the uncertainty principle . the time variation of the cross - power spectrum over times @xmath182 captures the changes of @xmath171 . visibility @xmath183 in the domain of frequency @xmath8 and time @xmath13 is the product of the fourier transforms of electric fields at stations a and b ( see equation [ eq : vdef ] ) : @xmath184 we suppose that each sample of the cross - spectrum is averaged over many realizations of the source electric field @xmath169 , over a time short compared with @xmath34 . this reduces noise from the source and backgrounds . one may represent the visibility in four domains , linked by fourier transforms of frequency @xmath8 to delay @xmath17 , and time @xmath13 to fringe rate @xmath19 , as figure [ fig : ftcircle ] illustrates . in this paper we are particularly concerned with @xmath185 , visibility in the domain of delay @xmath17 and time @xmath13 . this is the fourier transform of @xmath186 . the convolution theorem for fourier transforms shows that @xmath97 is the cross - correlation function of electric fields in the time domain : @xmath187 here , @xmath30 indexes individual samples of electric field , over a short interval near the index time of the measurement of the cross - power spectrum , @xmath13 . visibility in the delay - rate domain , conjugate to the frequency - time domain , takes the form : @xmath188 \right ] \label{eq : vtauftvnu}\end{aligned}\ ] ] searches for interference fringes are often conducted in this domain : because absolute calibration of delay and fringe rate are usually impossible for very - long baseline interferometry , the peak of @xmath73 can be used to determine them @xcite . visibility depends on the separation @xmath189 of stations @xmath9 and @xmath10 , as well as on delay and rate , or time and frequency . from equations [ eq : g_conv ] and [ eq : vtau_def ] , and the assumption that the electric field of the source @xmath169 is a stationary random variable without correlation in time , we find that visibility in the delay domain is the cross - correlation function of @xmath28 at the two stations : @xmath190 this leads to the expected form of @xmath191 : a spike at @xmath192 , with average magnitude equal to the average correlation @xmath193 ; and a broad component of width @xmath33 , with random amplitude , and phase variations that increase with baseline length , corresponding to the random character of @xmath171 and its decorrelation with increasing baseline . equations [ eq : normalize_gra ] and [ eq : normalize_gd ] show that @xmath194 for @xmath195 , and @xmath196 for uncorrelated @xmath179 and @xmath180 . figures [ fig : frrt_delay ] and [ fig : delay_fine_structure ] show examples . on intermediate baselines , the time structure of the correlation of @xmath171 is more complicated , in a way that depends on the geometrical distribution of the paths that contribute to @xmath171 . for scattering material concentrated in a thin screen , for example , the shortest - length paths result in small @xmath30 in the impulse - response function , and also tend to appear at small angles at the observer . thus , at small delays correlation is high even for rather long baselines ; whereas at long delays correlation is poor even for shorter baselines . thus , correlation between antennas should decrease at later times @xmath30 within @xmath171 . this correlation is imprecise for scattering material distributed along the line of sight , where many deflections along the line of sight lead to a large time lag @xmath30 , but little or no angular deflection at the observer s interferometer . moreover , in the frequency domain , dynamic single - dish spectra can show slants and complicated patterns @xcite , suggesting complicated correlations of time and delay in the observer plane . analysis of the visibility is thus easiest on very short baselines and very long ones . equations [ eq : vtauftvnu ] and [ eq : vabtau_as_g ] provide the relation of the visibility in the frequency - time domain to the impulse - response function : @xmath197 averaging the visibility over many scintillations yields the average visibility , @xmath193 : @xmath198 the average of @xmath191 over many realizations of scintillations leaves the delta - function at the origin that corresponds to the average visibility : @xmath199 here , @xmath200 is the kronecker delta - function , with value 1 if @xmath192 and 0 otherwise . the secondary spectrum @xmath201 may be defined as the square modulus of @xmath72 . for zero baselines @xmath16 is real , so that @xmath202 , and our and their expressions are identical . their expression includes phase information in an elegant way for their short baseline , where departures from zero phase are small . for observations on long baselines , their expression is impractical because identification of the origin of @xmath203 is not possible , as figure [ fig : frrt_delay ] shows . consequently , the pair @xmath203 and @xmath204 can not be combined reliably . they also use the accent @xmath27 for visibility in the delay domain , although not for @xmath44 ; whereas we use the accent for quantities in the frequency domain of @xmath8 . ] : @xmath205 this function provides information similar to @xmath73 , shown in figure [ fig : frrt_delay ] , but is easier to deal with statistically . because @xmath206 is the square modulus of the complex visibility @xmath20 , background noise adds noise to @xmath44 , except for the central lag , where it may contribute a constant offset . similarly , self - noise will add noise with an envelope that follows the average form of @xmath44 . this behavior is in contrast to that of correlation functions of single - dish quantities such as the intensity , where noise can contribute to the mean correlation function . our long baselines fully resolve the scattered image . the phases of scintillation elements in @xmath16 appear to be random , and the phases of @xmath201 show no discernible patterns . an inverse fourier transform from fringe rate @xmath19 to time @xmath13 leads to @xmath22 : @xmath207 .\end{aligned}\ ] ] evaluated at the fringe rate @xmath52 of its peak magnitude , the secondary spectrum @xmath208 is a time average , that approximates an average over realizations of the scintillation pattern of @xmath22 : @xmath209 the autocorrelation function of @xmath44 is @xmath91 : @xmath210 note that @xmath22 appears without averaging in this expression . conveniently , the correlation @xmath211 between the secondary spectra @xmath44 in the right and left circular polarizations eliminates some effects of noise and interference , as noted in section [ sec : two_scales ] . the behaviors of @xmath44 and @xmath91 are simplest to describe on very short baselines , where @xmath212 for all @xmath213 , and @xmath194 ; and for very long baselines , where the random parts of @xmath214 and @xmath215 are completely uncorrelated , so that @xmath110 . if the correlation of the random parts of the propagation kernels @xmath193 is constant , then the ensemble - average values of these correlation functions are : @xmath216 again , @xmath200 is the kronecker delta - function , with value 1 if @xmath192 and 0 otherwise ; and similarly for @xmath217 . the time - reversed pulse - broadening function is @xmath218 , given by @xmath219 . the constant @xmath220 is the mean square of @xmath49 : @xmath221 . we consider a situation where the impulse - response function is the sum of two exponentials with different time constants @xmath107 and @xmath108 , with a rapid rise from @xmath222 at @xmath223 . as we discuss below , this may result in a variety of circumstances . we parametrize the impulse - response function : @xmath224 in the text , we also make use of the inverse scales @xmath225 and @xmath226 ; these can provide better physical insight . the autocorrelation of @xmath49 provides the form for @xmath227 , the mean square visibility in the delay - time domain , as given by equation [ eq : c_result ] . for this impulse - response function , under the assumption that the baseline is so long that @xmath110 , this takes the form : @xmath228 the autocorrelation function @xmath91 of the mean square visibility for this impulse - response function then takes the form : @xmath229 where again @xmath230 is the kronecker delta function , and : @xmath231 thus , the two exponential scales @xmath232 appear again , in @xmath91 , as do the weights @xmath233 , @xmath234 . an observer sees an anisotropic distribution of radiation from a screen at distance @xmath235 from the observer . the probability of receiving radiation from the screen at position @xmath236 is : @xmath237 we suppose without loss of generality that @xmath238 . if the source is at infinite distance beyond the screen , the delay @xmath17 along this path ( ignoring any contribution from the screen ) is : @xmath239 where @xmath240 is the speed of light . if the source is at distance @xmath241 beyond the screen , then @xmath235 is replaced by @xmath242 in this and subsequent equations . we convert the distribution of angles in equation [ eq : anis_angle_dist ] to coordinates @xmath243 , where @xmath244 . the resulting distribution of @xmath245 is : @xmath246 where @xmath247 parametrizes the anisotropy . we integrate over @xmath248 to find the pulse - broadening function : @xmath249 thus , @xmath250 here , @xmath251 is the regular modified bessel function of order 0 , and @xmath252 . note that for @xmath253 , this distribution becomes the familiar exponential distribution , with scale @xmath71 , as expected for @xmath254 . @xcite present a similar expression . in the case of @xmath255 , at small @xmath17 the distribution has the limit : @xmath256 at large values , @xmath257 so that @xmath258 at a particular scale @xmath17 , the logarithmic derivative of @xmath259 is @xmath260 , although the coefficient depends weakly on @xmath17 . thus , @xmath259 exhibits two exponential scales : @xmath71 at large @xmath17 , and @xmath261 at small @xmath17 . the relative strength of the scales is about @xmath262 near the larger scale @xmath71 , with the larger scale being the weaker .
active galactic nuclei ( agns ) are often believed to play a significant role in galaxy evolution . as the observed scaling relations between black hole mass and global properties of their host galaxies seem to require self - regulation between black hole growth and galaxy evolution , theoretical models often include agn feedback as a core component of galaxy evolution ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? nonetheless , understanding how black holes and galaxies coevolve over hubble time is observationally limited due to the scarcity of direct measurements of scaling relations at high redshift @xcite , and the nature of agn feedback and the coupling of the agn energy output with the interstellar matter ( ism ) is a subject of active research in this field . agn gas outflows are often observed over galaxy scales @xcite , hence , outflows may function as an effective channel of agn feedback . by connecting nuclear activity with the larger scale ism and star formation ( sf ) , agn outflows may push gas out of host galaxies and/or heat up the ism , leading to suppression of sf @xcite . on the other hand , outflows may trigger sf by compressing the ism @xcite . currently , the evidence of negative agn feedback on sf is yet to be conclusive ( e.g. , * ? ? ? * ) , while detections of positive feedback on sf have been reported by several observational works based on individual objects ( e.g. , * ? ? ? * ; * ? ? ? thus , it is of importance to investigate the true nature of agn feedback and understand the overall effect of agns through gas outflows . the direct connection between agn activity and sf seems complex as various observations have been presented in the literature . for typical agns , a positive correlation has been reported between agn luminosity , probed by either optical or x - ray , and sf luminosity ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , indicating an average scaling between agn activity and sfr . in contrast , at high redshift in particular , x - ray agns seem to show enhanced sfr @xcite . since the sfr of agn host galaxies is difficult to measure and there may be systematic differecne among various sfr indicators based on uv , optical , ir signatures ( e.g. * ? ? ? * ; * ? ? ? * ) , more detailed investigation is required to reveal the nature of agn and sf connection . in comparing with non - agn galaxies , a recent study by @xcite reported that x - ray and optical agns tend to have lower specific star formation rate ( ssfr ) compared to star - forming galaxies . a similar work by @xcite showed that radio and optical agns have on average lower ssfr than star - forming galaxies , while mid - infrared selected agns show comparable or slightly higher ssfr than star - forming galaxies . these results based on the present - day galaxies may suggest that agn feedback suppresses sf . however , the physical mechanism for connecting agn and sf is yet to be clear , requiring further detailed studies of the link between outflows and sf . to understand the role of agn gas outflows in galaxy evolution , it is important to investigate how common and strong these outflows are so that agns as a population , instead of individual objects , can be used for connecting outflows with agn energetics and galaxy evolution . while the demography of ionized gas outflows in agns was limited to relatively small samples in the past , more robust results became available recently based on statistical samples of agns from large surveys . to perform a statistical investigation of gas outflows and their connection to agn activity , we are performing a series of studies using a large sample of type 2 agns from sloan digital sky survey ( sdss ) . in the first of this series , ( * ? ? ? * here after , paper 1 ) presented a census of ionized gas outflows based on the [ ] @xmath05007 kinematics , by analyzing a sample of 39,000 present - day type 2 agns . at least 50% of agns over the full luminosity range explored in that study , shows kinematic signatures of gas outflows . this outflow fraction is considered as a lower limit for low - luminosity agns since their outflow indicator ( i.e. , a wing component of the [ ] line ) is relatively weak and the gravitational potential of their host galaxy dominates in forming the emission line profile . in fact , for high - luminosity agns , the outflow fraction is over 90% , indicating that the majority of energetic agns show outflows . the steep increase of the outflow fraction with eddington ratio indicates a close link between outflow kinematics and agn accretion . in contrast , we find no strong correlation between radio luminosity and ionized gas kinematics , indicating that the outflows are not directly connected to radio activity for most agns . in the second of this series , presented 3-d biconical outflow models combined with a thin dust plane , in order to constrain the intrinsic physical nature of gas outflows . our models successfully reproduced the observed emission line profiles and the distribution of the velocity - velocity dispersion ( vvd ) diagram measured from the [ ] line . based on the monte carlo simulations of the vvd distribution , we constrained the physical parameters of the outflows , including launching velocity , opening angle , and dust extinction . in this paper , we mainly compare agns and sf galaxies in terms of the [ ] kinematics and their link to sf , in order to investigate the role of gas outflows as agn feedback . throughout the paper , we use the cosmological parameters as @xmath3 km s@xmath4 mpc@xmath4 , @xmath5 , and @xmath6 . 110,000 galaxies at z@xmath20.3 , with well - defined emission - lines . , scaledwidth=38.0% ] we extend our previous sample of agns in paper i to all emission line galaxies at z@xmath2@xmath10.3 , by including star - forming galaxies . the initial sample of 235,922 emission line galaxies is selected from the sdss data release 7 @xcite , using the mpa - jhu catalogue , with signal - to - noise ratio ( s / n ) @xmath7 10 in the continuum , and s / n @xmath7 3 in the four emission lines , h@xmath8 , [ ] @xmath05007 , h@xmath9 , and [ ] @xmath06584 , which are used to classify each object in the diagnostic diagram ( see fig . 1 ) . using this initial sample , we applied a further constraint to select objects with well - defined emission line profiles with the amplitude - to - noise ( a / n ) ratio larger than 5 for the [ ] and h@xmath9 lines . as a result , we obtained a total of 112,726 emission line galaxies , which are classified into three groups : agns , composite objects , and star - forming galaxies as listed in table 1 , based on the classification scheme by @xcite . in the first paper of this series , we presented the detailed analysis on the pure agn and composite samples ( paper i ) . in this work , we mainly compare agns with sf galaxies in terms of [ ] kinematics , and investigate how gas outflows are related to star - formation . cccc redshift & agn & composite & s - f + ( 1)&(2)&(3)&(4 ) + 0.003 @xmath2z@xmath2 0.1 & 14923 ( 6733 ) & 14714 ( 4700 ) & 58928 ( 23999 ) + 0.1 @xmath2z@xmath2 0.2 & 7382 ( 4495 ) & 5614 ( 1857 ) & 10019 ( 2967 ) + 0.2 @xmath2z@xmath2 0.3 & 447 ( 289 ) & 384 ( 147 ) & 315 ( 86 ) + 0.003 @xmath2z@xmath2 0.3 & 22752 ( 11517 ) & 20709 ( 6704 ) & 69265 ( 27052 ) + we derive the properties of each galaxy in the sample in order to compare them with ionized gas kinematics , as we presented in detail for agns in paper 1 . here , we briefly summarize sample properties . first , we adopt the stellar velocity dispersion and stellar mass from the mpa - jhu catalogue to investigate the host galaxy mass scale and gravitational potential . in the following analysis , we include stellar velocity dispersion ( @xmath10 ) value down to 30 km s@xmath4 , which is slightly lower than the instrumental resolution of the sdss spectra while we exclude unreliably values ( @xmath10 @xmath11 420 km s@xmath4 ) as recommended by the sdss catalogue . since stellar velocity dispersions are more uncertain for lower mass galaxies due to the spectral resolution limit as well as the lower s / n , we will also use the stellar mass ( m@xmath12 ) estimates as a tracer of the host galaxy gravitational potential . as shown in fig 2 , the sample galaxies follow the correlation between stellar mass and stellar velocity dispersion as expected from fiber - jackson relation although the scatter becomes larger for lower mass galaxies due to measurement uncertainties . note that three groups have different mass ranges . while the stellar mass of the most agn host galaxies ranges from 10@xmath13 to 10@xmath14 m@xmath15 , that of sf galaxies lies between @xmath110@xmath16 to 10@xmath17 m@xmath15 . thus , when we compare agns with sf galaxies , we will use stellar mass fixed subsamples , to account for the difference of their gravitational potential . in addition to the host galaxy properties , we also determine the properties of black holes . for black hole mass estimates , we use the updated scaling relations , e.g. , black hole mass - stellar velocity dispersion ( m@xmath18-@xmath19 ) relation based on the reverberation - mapped agns from @xcite ( see also * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and black hole mass - stellar mass relation . as we pointed out in paper i , the range of black hole mass significantly changes depending on the adopted scaling relation . we adopted the black hole mass estimates based on the black hole mass - stellar mass relation to be consistent with paper 1 , while the choice of the black hole mass estimates do not change the main results presented in this paper . x= 0.04 dex and @xmath20y=0.01 dex . , scaledwidth=48.0% ] we used the strong [ ] line at 5007 line as a tracer of the ionized gas outflows , by measuring the velocity shift from the systemic velocity and the velocity dispersion compared to the stellar velocity dispersion , as we discussed in paper i. here we briefly summarize the procedure for completeness . using the sdss spectra , we first fitted stellar absorption lines with a series of stellar population models , which were constructed with the penalized pixel - fitting code @xcite based on simple stellar population models with solar metallicity and various ages ( 60 myrs to 12.6 gyrs ) @xcite . based on the stellar absorption fit , we determined the systemic velocity , which will be used for calculating the velocity shift of [ ] . by subtracting the best - fit stellar population model , we generated pure emission line spectra , on which we fitted the [ ] line with one or two gaussian components using the mpfit code ( markwardt 2009 ) . only if the [ ] line profile shows a prominent wing component , i.e. , the peak of the second component is a factor of three larger than the noise level ( i.e. , a / n ratio @xmath11 3 ) , we adopted the 2nd gaussian component , in order to avoid unreliable detection of wing features . once we obtained the best fit model for [ ] , we calculated the first and second moments of the line profile using the best - fit model as @xmath21 @xmath22}}]^2 = { \int \lambda^2 f_\lambda d\lambda \over \int f_\lambda d\lambda } - \lambda_0 ^ 2 , \end{aligned}\ ] ] where @xmath23 is the flux at each wavelength . then , we calculated the velocity shift ( v@xmath24 ) of the first moment of [ ] with respect to the systemic velocity , which is determined from stellar absorption lines , and the velocity dispersion ( @xmath25 ) of [ ] from the second moment . we determined the uncertainty of v@xmath24 and @xmath25 by performing monte carlo simulations , which generate the distribution of measurements based on the 100 mock spectra generated by randomizing flux with flux error at each wavelength . we adopted the 1 @xmath26 dispersion of the distribution as the measurement uncertainty , respectively for v@xmath24 and @xmath25 . when we examined the fractional error of @xmath27}}$ ] of the sample , the distribution shows a gaussian distribution with a mean value of @xmath28 in log scale ( -0.89@xmath290.44 for agns , -0.74@xmath290.39 for composite objects , and -0.79@xmath290.54 for sf galaxies ) , which corresponds to @xmath116% uncertainty . to avoid uncertain measurements , we excluded 7902 objects , that have a large uncertainty ( i.e. fractional error @xmath11 1 ) , from the sample ( 7.0% ) when we used the measurements of the [ ] kinematics in the analysis . in the case of v@xmath24 , the mean uncertainty is @xmath30 km s@xmath4 for agns , @xmath31 km s@xmath4 for composite objects , and @xmath32 km s@xmath4 for sf galaxies , indicating that very weak velocity shifts ( i.e. , @xmath2 30 km s@xmath4 ) are difficult to detect . note that the quoted errors are the r.m.s from the distribution of errors of individual objects . the measured [ ] velocity dispersions were corrected for the wavelength - dependent instrumental resolution of the sdss spectroscopy , which is close to 55 - 60 km s@xmath4 at the observed wavelength of [ ] . a number of objects has relatively narrow [ ] , for which the measured @xmath27}}$ ] is smaller than the instrumental resolution . we excluded 1872 objects with very low velocity dispersion ( i.e. , @xmath27}}$ ] @xmath2 30 km s@xmath4 ) or unresolved [ ] . in total we excluded 8.7% of the sample , which have either very small velocity dispersion ( i.e. , @xmath27}}$ ] @xmath2 30 km s@xmath4 ) or large fractional errors ( @xmath111 ) . note that including these objects does not change our conclusions in the following analysis . -to - stellar velocity dispersions as a function of [ ] luminosity ( top ) , respectively for agns ( red ) , composite objects ( green ) , and sf galaxies ( blue ) . we divide the sample into two groups : [ ] fitted with single gaussian ( left ) and double gaussian ( right ) . the ratio of the agn sample is systematically higher than that of sf galaxies , indicating that the non - gravitational component , i.e. , outflows is dominant in agns , particularly at high luminosity regime . the bottom panels show the mean ratio of [ ] divided by stellar mass ( in units of km s@xmath4 m@xmath33 ) as a function of [ ] luminosity . , title="fig : " ] + velocity dispersion with [ ] luminosity , respectively for agns ( left ) , composite objects ( middle ) , and sf galaxies ( right ) , using the entire sample ( top ) and for the stellar mass - limited sample ( bottom ) . the color represent the number density in each bin with @xmath20x= 0.03 dex and @xmath20y=0.01 dex.,scaledwidth=45.0% ] to investigate the signature of gas outflows , we compare gas and stellar kinematics using [ ] velocity dispersion ( @xmath25 ) and stellar velocity dispersion ( @xmath10 ) in figure 3 . we find a clear difference between agn sample and sf galaxies . agns with larger @xmath25 have larger @xmath10 , indicating that the host galaxy gravitational potential plays a role in broadening the gas emission lines . however , the relation is not linear as @xmath25 is larger than @xmath10 particularly for high luminosity agns . as we discussed in paper i , we interpret the ratio of @xmath25 to @xmath10 larger than unity as an indication of non - gravitational kinematic components , i.e. outflows . in contrast , sf galaxies show almost one - to - one relationship between @xmath25 and @xmath10 , suggesting that the broadening of gas emission line can be entirely accounted by the virial motion due to the galaxy gravitational potential . note that the non - linear relation between @xmath25 and @xmath10 for the agn sample is mainly due to the wing component in the [ ] line profile , which is not well constrained when fwhm is used to represent the velocity of [ ] line ( see paper i for detailed discussion ) . in the case of composite objects , @xmath25 and @xmath10 are linearly correlated each other . although agn outflows may be present in composite objects , the outflow velocity is relatively low due to the average low luminosity ( see the lack of high [ ] velocity agns in figure 3 ) , hence the non - gravitational kinematic component can be easily diluted by the dominant virial component . in figure 4 , we investigate the ratio between @xmath25 and @xmath10 as a function of [ ] luminosity . here we separate the emission line galaxies , depending on the presence of a wing component in [ ] ( i.e. , single gaussian vs. double gaussian ) . regardless of agn activity , most objects without a wing component show the @xmath25 /@xmath10 ratio close to unity except for the agns in the highest luminosity bin , indicating that [ ] gas kinematics of these objects are mostly due to the host galaxy gravitational potential . in contrast , among the objects with a wing component in [ ] ( right panels in figure 4 ) , there is a systemic difference of the @xmath25 /@xmath10 ratio between agns and sf galaxies . for agns , @xmath25 is much larger than @xmath10 ( by 0.13 dex on average ) , indicating the presence of non - virial kinematics , while for sf galaxies , @xmath25 is comparable to @xmath10 except for the highest luminosity bin ( the average ratio log ( @xmath25 /@xmath10 ) = 0.01 ) . note that we detect a wing component in [ ] , which is fitted by the second gaussian component , for @xmath139% of sf galaxies . however , the @xmath25 values are similar to stellar velocity dispersions . instead of @xmath10 , we can use stellar mass to normalize gas outflow kinematics by the host galaxy gravitational potential ( bottom panels of figure 4 ) . we find qualitatively same trends that the broadening of [ ] in agns can not be entirely accounted by the gravitational potential while the broadening of [ ] in sf galaxies are mainly due to the host galaxy potential . the average difference between agns and sf galaxies is 0.16 dex . note that even among the objects with single gaussian [ ] , we clearly detect the systematic difference of the ratio between agns and sf galaxies ( the average difference is 0.11 dex ) . we investigate how the [ ] kinematics are related to [ ] luminosity for agns and sf galaxies in figure 5 . there is an increasing trend of @xmath25 with increasing luminosity in agns although the trend is very broad . the increasing trend is also present among composite objects while the the luminosity range of [ ] is an order of magnitude narrower . these trends show that gas outflows are stronger in agns with higher luminosity , indicating agn energetics are responsible for the outflow kinematics . in contrast , we find no clear increasing trend among sf galaxies . on average the @xmath25 is lower in sf galaxies while the range of @xmath25 is as broad as that of agns . since the average stellar mass of agns and sf galaxies is very different as we pointed out in s.2.2 , we plot the same comparison after limiting stellar mass between 10@xmath34 and 10@xmath35 ( bottom panels in fig . 5 ) . again , we find that agns and sf galaxies show different trends , suggesting that these different trends are not due to the difference in mass scales between agn and sf galaxy samples . in figure 6 , we combine the velocity and velocity dispersion of [ ] to investigate the difference of gas kinematics between agns and sf galaxies . we find a dramatic contrast between agns and sf galaxies in terms of the range of gas velocities . while sf galaxies are concentrated at the lower center with a limited range of velocity shift ( @xmath36v@xmath24@xmath36 @xmath2 100 km s@xmath4 ) and velocity dispersion ( @xmath25 @xmath2 @xmath1150 km s@xmath4 ) , agns show a much larger distribution , including agns with extreme kinematics ( i.e. , @xmath36v@xmath24@xmath36 @xmath11 200 km s@xmath4 and @xmath25 @xmath37400 km s@xmath4 ) . the average @xmath25 is @xmath38 , @xmath39 , and @xmath40 km s@xmath4 , respectively for agns , composite objects , and sf galaxies , indicating that agns have more than a factor of 2 broader [ ] than sf galaxies . also , the dispersion of the @xmath25 distribution is significantly different with a factor of @xmath13 broader distribution in agns . note that the @xmath25 distribution is skewed toward high values , indicating the presence of extreme kinematics in the agn sample . in the case of v@xmath24 , the average value is @xmath41 , @xmath42 , and @xmath43 km s@xmath4 , respectively for agns , composite objects , and sf galaxies , indicating that the average v@xmath24 is similar among 3 groups . this is mainly due to the fact that the majority of objects have relatively low velocity shift , which can not be well constrained for given v@xmath24 uncertainties ( see s.2.3 ) . however , we clearly detect that the distribution of v@xmath24 of agns is a factor of @xmath13 broader than that of sf galaxies , suggesting the influence of gas outflows in agns . considering the fact that galaxy bulge gravitational potential can also broaden the emission line , of which velocity dispersion is expected to correlate with stellar mass , we plot subsamples of agns and sf galaxies in the same stellar mass range , 10 @xmath2 log m@xmath12 @xmath2 10.5 ( bottom panels in figure 6 ) . in this case , the range of velocity and velocity dispersion decrease , however the strong difference in the vvd distributions still remains . the mean @xmath25 is @xmath44 , @xmath45 , and @xmath46 , respectively for agns , composite objects , and sf galaxies , showing a factor of @xmath12 difference in terms of the mean value and dispersion of the distribution between agns and sf galaxies . the distribution of v@xmath24 is similar to that of the total sample , with the mean v@xmath24 = @xmath47 , @xmath48 , @xmath49 , respectively for agns , composite objects , and sf galaxies . + in figure 7 we present the vvd diagram as a function of [ ] luminosity , respectively for agns ( top ) , composite objects ( middle ) , and sf galaxies ( bottom ) . in agns we find the vvd diagram dramatically expands toward higher values with increasing [ ] luminosity , indicating the gas outflows are driven by agn energetics . composite objects in general follow the same trend although at the highest luminosity bin , the number of objects is relatively small . in contrast , sf galaxies do not show any significant change of the vvd diagram as a function of [ ] luminosity , suggesting that photoionization mechanism of the [ ] line is not directly responsible for the [ ] kinematics . as we already showed in figure 4 , the gas kinematics are mainly governed by the gravitational potential of the host galaxy . + if we investigate the vvd diagram as a function of stellar mass , instead of the [ ] luminosity , we find no dramatic expansion of the diagram . the vvd distribution is already broad even at low mass regime . other than the increase of the sample size , agns , composite objects and sf galaxies show similar distributions , suggesting that stellar mass is not directly related to the expansion of the vvd diagram . in this section we investigate how the outflow kinematics are related to star formation in the host galaxies . since the star formation rate ( sfr ) of agn host galaxies is difficult to directly measure due to the contamination of agn emission to sfr indicators , i.e. , the h@xmath9 line , uv continuum , etc ( e.g. , * ? ? ? * ) , we use the sfr based on ir luminosity by adopting the catalogue of @xcite , which provides the ir luminosity of @xmath1330,000 sdss galaxies and the estimated sfr converted from ir luminosity based on the conversion equation by @xcite , @xmath50 in their work @xcite determined ir luminosity from artificial neural network techniques , which were tested and trained based on the herschel - sdss strip 82 matched sample of 1136 galaxies . as recommended by @xcite , we used reliable estimates of sfr using @xmath51 @xmath2 0.1 ( see also * ? ? ? * ) . in this case , a total of 30,352 objects ( 26.9% ) is excluded from our agn+composite+sf sample in table 1 . more specifically , 10,316 objects from pure agns ( 45.3% ) , 4,710 objects from composite objects ( 22.7% ) , and 15,326 objects from sf galaxies ( 22.1% ) were removed due to the @xmath51 @xmath2 0.1 criterion . since we loose a large fraction of agns , we tried to use a less conservative criterion , i.e. , @xmath51 @xmath2 0.3 , in order to include more objects with less secure sfr estimates . in this case , 12,042 objects ( 10.7% of the total sample ) were excluded ( 3,375 , 1,781 , and 6,886 objects respectively from pure agns , composite objects , and sf galaxies ) . the results based on the sample with @xmath51 @xmath2 0.3 remained qualitatively the same . note that the ir - based sfr represents the average sfr over @xmath110@xmath16 years as the ir emission is reradiated by cold dust around ob stars . in addition to ir - based sfr , we also used the d@xmath52 provided by the sdss catalogue for a consistency check ( see section 4.3 ) . in figure 8 , we present the specific star formation rate ( ssfr ) , by dividing the estimated sfr by stellar mass , as a function of the outflow velocities . to account for the difference of stellar mass range between agn and sf galaxies , we present the comparison in each stellar mass bin . here , the outflow velocities as expressed as @xmath53}'$ ] , are calculated by adding the velocity and velocity dispersion of [ ] in quadrature , which can be understood as the corrected velocity dispersion of [ ] accounting for the projection effect . then , the outflow velocities are normalized by stellar velocity dispersion in order to show the non - gravitational ( i.e. , outflow ) kinematics . in the case of sf galaxies , the ssfr does not change along the stellar mass bin , showing that non - agn galaxies form a star forming sequence with a constant ssfr within a factor of @xmath12 scatter ( see table 2 ) . in each stellar mass bin , the distribution of log ssfr is centered at the mean value of @xmath54 yrs@xmath4 , which is averaged over all sf galaxies . with respect to the normalized gas kinematics , we find no correlation , indicating that the kinematics of ionized gas are not related to sf . as expected , sfr is not affected by gas kinematics in sf galaxies since gas kinematics are mainly due to the galaxy gravitational potential rather than agn outflows . note that at lower stellar mass bin , stellar velocity dispersion is on average more uncertain , broadening the distribution of gas - to - stellar velocity dispersion ratio . in contrast , we find that agns and composite objects have on average lower ssfr than sf galaxies . the mean log ssfr is @xmath55 and @xmath56 , respectively for pure agns and composite objects , showing a factor of 4 and 3 lower sfr than non - agn galaxies at fixed stellar mass . moreover , we find that the ssfr of agn sample tend to increase with increasing outflow velocities . in the 10.5@xmath2 log m@xmath15/m@xmath15 @xmath2 11 bin , for example , there is a clear increasing trend of ssfr with increasing outflow velocities in agns , while the distribution is almost round in sf galaxies . this trend implies a close connection between agn outflows and star formation in the host galaxies . interestingly , agns with higher outflow velocities have on average higher ssfr than agns with lower outflow velocities , implying that outflows do not instantly suppress sf , although the time scale of sf quenching has to be carefully investigated . since agns with strong outflows show higher sfr than agns with weak outflows , this may indicate a close link between outflows and sf . we find no evidence of enhanced sf in agns . even for the agns with extreme outflow velocities , ssfr is not very high and rather it is comparable to that of sf galaxies ( dashed line ) , suggesting that host galaxies of extreme agn outflows have regular sfr compared to sf galaxies . in contrast , ssfr decreases with decreasing outflow velocities . for example , agns with no or weak outflows ( i.e. , when @xmath57 ) , ssfr is almost an order of magnitude lower than sf galaxies or agns with strong outflows . in other words , agns with strong outflows have similar sfr compared to sf galaxies , while agns with weak / no outflows have much lower sfr than non - agn galaxies . thus , the 0.6 dex mean difference of ssfr between agn and sf galaxy samples is mainly due to the population of agns with weak [ ] outflows . the stellar velocity dispersion measured from sdss spectra may not be the best parameter to represent the virial motion caused by the gravitational potential of the host galaxies , particularly for lower mass galaxies , since the spectral resolution of the sdss spectroscopy is limited , and the effect of the rotational broadening and the projection due to the random inclination of the stellar disk to the line - of - sight can be significant for rotation - dominant galaxies . considering these effects , we use stellar mass , instead of @xmath10 , to normalize the gas kinematics . assuming the fiber - jackson relation , we use stellar mass to the 1/4 power ( m@xmath58 ) as a proxy for stellar velocity dispersion representing the virial motion ( bottom panel in fig . 9 ) . we see qualitatively similar distribution of ssfr with the normalized outflow kinematics . except for the agns with lowest ssfr , there is a general trend that agns with higher outflow velocities have higher ssfr , while the ssfr of agns with no outflows have much lower ssfr . /@xmath10@xmath11 0.3 ; top ) , weak outflows ( 0 @xmath2 log @xmath59/@xmath10@xmath2 0.3 ; middle ) , and no outflows ( log @xmath59/@xmath10@xmath2 0 ; bottom ) . in the left panels , agns and sf galaxies in a fixed stellar mass bin ( 10.5 @xmath2 log m@xmath15 @xmath2 11 ) are presented . the relative ssfr is calculated with respect to the mean ssfr of all sf galaxies . , title="fig : " ] /@xmath10@xmath11 0.3 ; top ) , weak outflows ( 0 @xmath2 log @xmath59/@xmath10@xmath2 0.3 ; middle ) , and no outflows ( log @xmath59/@xmath10@xmath2 0 ; bottom ) . in the left panels , agns and sf galaxies in a fixed stellar mass bin ( 10.5 @xmath2 log m@xmath15 @xmath2 11 ) are presented . the relative ssfr is calculated with respect to the mean ssfr of all sf galaxies . , title="fig : " ] /l@xmath60 @xmath11 - 2 ; top ) , intermediate eddington ratio ( -3 @xmath2 log l@xmath61/l@xmath60 @xmath2 - 2 ; middle ) , and low eddington ratio ( -3.5@xmath2 log l@xmath61/l@xmath60 @xmath2 - 3 ; bottom ) . in the left panels , agns and sf galaxies in a fixed stellar mass bin ( 10.5 @xmath2 log m@xmath15 @xmath2 11 ) are presented . the relative ssfr is calculated with respect to the mean ssfr of all sf galaxies . note that the eddington ratio estimates based on the [ ] luminosity may have systematic uncertainties due to the contribution from the sf region to the observed [ ] luminosity . , title="fig : " ] /l@xmath60 @xmath11 - 2 ; top ) , intermediate eddington ratio ( -3 @xmath2 log l@xmath61/l@xmath60 @xmath2 - 2 ; middle ) , and low eddington ratio ( -3.5@xmath2 log l@xmath61/l@xmath60 @xmath2 - 3 ; bottom ) . in the left panels , agns and sf galaxies in a fixed stellar mass bin ( 10.5 @xmath2 log m@xmath15 @xmath2 11 ) are presented . the relative ssfr is calculated with respect to the mean ssfr of all sf galaxies . note that the eddington ratio estimates based on the [ ] luminosity may have systematic uncertainties due to the contribution from the sf region to the observed [ ] luminosity . , title="fig : " ] to further investigate the difference of ssfr between agns and sf galaxies , we compare the distribution of ssfr between agns ( pure agn + composite objects ; red ) , and sf galaxies ( blue ) in figure 9 . here we divide the sample into 3 groups : strong ( i.e. , log @xmath53}'$]/@xmath10@xmath11 0.3 ; top panel ) , intermediate ( i.e. , 0@xmath2 log @xmath53}'$]/@xmath10@xmath2 0.3 ; middle panel ) , and weak outflows ( i.e. , log @xmath53}'$]/@xmath10@xmath20 ; bottom panel ) . while the distribution of ssfr is broad , agns with intermediate and weak outflows have lower ssfr by @xmath62 and @xmath63 dex , respectively , compared to the mean ssfr of sf galaxies . in contrast , agns with strong outflows have similar ssfr compared to sf galaxies ( see table 2 for details ) . considering the stellar mass difference between agns and sf galaxy sample , we compare the ssfr distribution at fixed stellar bass bin , e.g. , at @xmath64 log m@xmath12 @xmath65 ( right panel in figure 9 ) . we find a similar trend of decreasing mean ssfr with decreasing outflow velocities . since the distribution of ssfr is very broad ( @xmath11 dex in fwhm ) , large ranges of agn outflow velocities and sfr are present in the sample . however , it is clear that there is strong trend of decreasing ssfr for agns with no outflows . although it is previously found that optical and x - ray agns have on average lower ssfr than sf galaxies ( e.g. * ? ? ? * ; * ? ? ? * ) , a physical mechanism for explaining the trend was not clearly identified . we try to interpret these results by connecting agn outflows with sf in the following scenario . when gas is supplied , both agn and sf are triggered as we typically observe the correlation between x - ray and ir luminosities ( e.g. * ? ? ? however , it would take the order of a dynamical time for outflows to impact on the ism over galactic scales . once the outflows start playing a role in impacting on the ism , we may observe the decrease of both sfr and agn activity . after that , as the agn activity decreases , the outflow kinematics become weaker and finally disappear . in this scenario , we link agns with strong outflows hosted by galaxies with regular sfr and agns with no outflows hosted by galaxies with lower sfr as a evolutionary sequence . perhaps , we are seeing for the first time , the physical link between agn outflows and the delayed suppression of sf . we further investigate whether ssfr is related with agn energetics since outflow velocities and agn luminosity show a broad correlation , albeit with a large scatter . in figure 10 , we present the distribution of ssfr , after dividing agns into three eddington ratio bins . note that since we can not detect a very weak [ ] line , the eddington ratio of the agn sample ranges from 10@xmath66 to 1 while most agns are located within 10@xmath67 @xmath2 log l / l@xmath60 @xmath2 -0.5 ( see figure 5 in paper 1 ) . we find a similar trend that low eddington agns have clearly lower ssfr than sf galaxies , while high eddington agns have comparable ssfr with respect to sf galaxies . while the difference of ssfr distribution is negligible between high eddington agns and sf galaxies with a small negative offset of @xmath68 , low eddington agns clearly show much lower ssfr by @xmath69 dex , indicating that sfr is much lower in the host galaxies of low eddington agns . the same trend is detected when we limit the sample within the stellar mass bin 10.5@xmath2 log m@xmath12/m@xmath15 @xmath2 11 ( right panel in figure 10 ) . combing the trend of ssfr with outflow velocities and eddington ratios , it seems that more energetic agns with strong outflows and high eddington accretion tend to be hosted by typical sf galaxies , while low - eddington agns with weak / no outflows have relatively quiescent host galaxies . in figure 11 , we investigate how host galaxy properties are related with agn outflows and ssfr . first , we adopt the galaxy morphology information from galaxy zoo ( 1st col . in figure 11 ; * ? ? ? * ) . since each galaxy is classified as elliptical , spiral , or uncertain object , we only present ellipticals and spirals by excluding uncertain objects . in agn sample spiral galaxies generally follow the broad trend between ssfr and outflow velocities while elliptical galaxies are located separately at a much lower ssfr region . these elliptical galaxies would not be included in our emission line galaxy sample if agn was not present since they have no or weak emission lines from star forming region . however , by excluding these elliptical galaxies , we still see that the ssfr of agns ranges down to much lower level compared to that of sf galaxies . considering the potential contamination of elliptical host galaxies in the agn sample , we investigated the ssfr distribution using only spiral galaxies . we find that the difference of mean ssfr between agns and sf galaxies and the decreasing ssfr with decreasing outflow velocities and eddington ratios remain the same , indicating that the contamination of elliptical galaxies is not mainly responsible for the much lower ssfr of agns with no outflows . second , we color - code agns based on the 4000 - break strength d@xmath52 as a sfr indicator @xcite , which is adopted from the mpa - jhu catalogue ( 2nd col . in figure 11 ) . as expected ssfr correlates with d@xmath52 in agns , while sf galaxies have overall lower values of d@xmath52 . the decreasing trend of d@xmath52 with increasing outflow velocities confirms a consistency between ir - based sfr and d@xmath52-based sfr . third , we also utilize the uv - to - optical color as a sfr indicator . by matching our sdss sample with galex general release 6/7 including ais and mis data @xcite , we obtained nuv - r color for 38,105 objects . as shown in figure 11 ( 3rd col . ) , we find qualitatively the same trend although the size of the matched sample is relatively small . finally , we present eddington ratios in the 4th col . in figure 11 , showing a clear increasing trend with both ssfr and outflow velocities , albeit with considerable scatter . these findings indicate that a clear trend , although not bimodal , is present between high eddington , strong outflow agns in sf galaxies and low eddington , no outflow agns in weakly star - forming galaxies . ccccc & agn+composite & agns & composite & s - f + & & all mass + ( 1)&(2)&(3)&(4)&(5 ) + 0.3 @xmath2 log @xmath25/@xmath10 & @xmath70 & @xmath71 & @xmath72 & @xmath73 + 0 @xmath2 log @xmath25/@xmath10@xmath2 0.3 & @xmath74 & @xmath75 & @xmath76 & @xmath77 + log @xmath25/@xmath10@xmath2 0 & @xmath78 & @xmath79 & @xmath80 & @xmath81 + -2 @xmath2log l@xmath61/l@xmath60 & @xmath82 & @xmath83 & @xmath84 & + -3 @xmath2 log l@xmath61/l@xmath60 @xmath2 -2 & @xmath85 & @xmath86 & @xmath87 & @xmath88 + log l@xmath61/l@xmath60 @xmath2 -3 & @xmath89 & @xmath90 & @xmath91 & + & & 10.5 @xmath2 log m@xmath15 @xmath2 11 + 0.3 @xmath2 log @xmath25/@xmath10 & @xmath92 & @xmath93 & @xmath94 & @xmath95 + 0 @xmath2 log @xmath25/@xmath10@xmath2 0.3 & @xmath96 & @xmath97 & @xmath98 & @xmath99 + log @xmath25/@xmath10@xmath2 0 & @xmath100 & @xmath101 & @xmath102 & @xmath103 + -2 @xmath2log l@xmath61/l@xmath60 & @xmath104 & @xmath105 & @xmath106 & + -3 @xmath2 log l@xmath61/l@xmath60 @xmath2 -2 & @xmath107 & @xmath86 & @xmath108 & @xmath109 + log l@xmath61/l@xmath60 @xmath2 -3 & @xmath110 & @xmath111 & @xmath91 & + & & 10 @xmath2 log m@xmath15 @xmath2 10.5 + 0.3 @xmath2 log @xmath25/@xmath10 & @xmath112 & @xmath113 & @xmath114 & @xmath115 + 0 @xmath2 log @xmath25/@xmath10@xmath2 0.3 & @xmath116 & @xmath117 & @xmath118 & @xmath119 + log @xmath25/@xmath10@xmath2 0 & @xmath120 & @xmath121 & @xmath122 & @xmath123 + -2 @xmath2log l@xmath61/l@xmath60 & @xmath124 & @xmath125 & @xmath126 & + -3 @xmath2 log l@xmath61/l@xmath60 @xmath2 -2 & @xmath127 & @xmath128 & @xmath129 & @xmath130 + log l@xmath61/l@xmath60 @xmath2 -3 & @xmath131 & @xmath132 & @xmath133 & + we reported that optical type 2 agns have on average lower ssfr compared to sf galaxies as previously found by @xcite and @xcite based on x - ray , optical , and radio agns . however , there is a large range of ssfr among agns , and we discovered a link between agn outflows and ssfr . while agns with powerful outflows have comparable ssfr with respect to sf galaxies , agns with weak or no outflows tend to have much lower ssfr . moreover , there is a clear trend that the mean ssfr decreases with decreasing eddington ratios . combining these two trends , we see a distribution from agns with strong outflow , high eddington accretion , and regular sfr to agns with no outflow , low eddington accretion , and much lower sfr . the trend between eddington ratio and ssfr naturally explains the observed agn - star formation luminosity relation in the local universe @xcite . we may interpret these trends as a evolutionary sequence i.e. , when there is gas supply , both agn and sf are triggered . agns with high - eddington accretion develop strong outflows while sf is on - going as in regular sf galaxies . when the outflows start impacting on the ism after the order of a dynamical time , the effect of agn feedback is observable as the sfr decreases while agn becomes weaker and no strong outflows are visible . note that the sfr based on ir luminosity traces the reradiation by cold dust around ob stars and averaged over @xmath110@xmath16 yrs . also , the eddington ratio was calculated based on [ ] luminosity , which does not reflect the short time scale flickering of accretion disk activity . rather , [ ] luminosity is averaged over longer time scale . if the outflow can effectively push the gas supply and/or prevent the cooling of the ism , it is naturally expected that both sfr and agn accretion will decrease . the distribution of agns from strong outflow , high eddington , and regular sfr to no outflow , low eddington , and low sfr may be understood as the evolutionary sequence caused by delayed agn feedback . alternatively , the same evolutionary sequence can be interpreted as the consequence of depletion of gas supply . when gas is supplied , both agn and sf are triggered . however , once gas is depleted , both agn activity and sf decreases . thus , we naturally expect that agns with strong outflow , high eddington , and regular sfr become agns with no outflow , low eddington , and low sfr . in this case , no agn feedback is invoked and simply gas depletion causes the transition . instead of evolutionary sequence , it is also possible that intrinsic gas content varies among galaxies at given stellar mass . if the amount of gas is intrinsically large , agns may tend to be high - eddington with strong outflows and host galaxies show regular sfr as in sf galaxies . on the other hand , if gas content is intrinsically lower , sfr is lower than that of galaxies in star - forming sequence , and agns have relatively low eddington and weak outflows . for given our data set , it is not clear which scenario best explains the observed trend . in particular , in the evolutionary sequence scenario , it is very difficult to confirm whether agn feedback causes the transition or the transition is natural outcome of gas depletion . in the case of the 3rd scenario , we may be able to investigate the difference of intrinsic gas content based on co observations by selecting agns with strong and no outflow for given stellar mass and morphology . the decreasing trend of ssfr with decreasing eddington ratio down to 10@xmath67 shown in figure 10 can be interpreted in two different ways : selection bias and intrinsic nature . we first examine whether the lower ssfr of low eddington agns is due to the selection bias . since we classify agns based on the emission line flux ratios in the bpt diagram , if agns produce weak emission lines , which can be over - shined by the emission produced by strong star formation , the combined observed emission lines are likely to be classified as sf galaxies rather than agns . thus , it is possible that we may miss very weak black hole activity present in strongly sf galaxies due to the limitation of the optical bpt classification . note that we do not consider very weak black hole activity with extremely low eddington ratios ( @xmath13410@xmath66 ) ( e.g. , * ? ? ? * ; * ? ? ? * ) since these very weak activities are not generally considered as agn population and the feedback energy from these objects is expected to be negligible in the context of galaxy evolution . if we focus on low eddington ratio agns between 10@xmath66 and 10@xmath135 , it is not clear why we should expect these agns are dominantly hosted by highly sf galaxies , while we expect that very weak black hole activity is present in most galaxies . by selecting sf galaxies , @xcite reported that the mean sf luminosity correlates with the mean agn x - ray luminosity down to eddington ratio @xmath136 ( see also * ? ? ? * ) , although it is extremely difficult to investigate the distribution of eddington ratios below @xmath136 ( see e.g. * ? ? ? * ) . for investigating whether our optical agn sample misses low eddington agns hosted by strongly sf galaxies , x - ray agn sample is very useful since optical bpt - based classification can be avoided . in fact , @xcite reported that their swift - bat agn sample shows on average lower ssfr compared to sf galaxies , indicating the same trend as we found in our sample . we independently investigate this issue by selecting x - ray agns from swift - bat 70 month catalogue @xcite after matching with our sdss sample . although the number of x - ray agns , for which we can determine stellar mass and eddington ratio , is relatively small , these 59 x - ray agns show a consistent trend that agns with low eddington ratio tend to show lower ssfr than agns with high eddington ratio . for example , since the eddington ratio ( log l@xmath61/l@xmath60 ) of this matched sample ranges from -3 to 0 , we divide them into 3 groups . we find that the d@xmath52-based ssfr of high eddington agns ( log l@xmath137/l@xmath138 @xmath11 -1 ) is higher by 0.13 and 0.83 dex , respectively , than that of intermediate ( -2@xmath2 log l@xmath137/l@xmath138 @xmath2 - 1 ) and low eddington agns ( log l@xmath137/l@xmath138 @xmath2 -2 ) , although the distribution of ssfr in each group is very large with rms of 0.7 to 0.8 dex due to presumably small number statistics . note that we were not able to use ir - based sfr since ir luminosity taken from @xcite is available only for 20 out of 59 objects . although we see a consistent trend between eddington ratios and ssfr based on the x - ray sample , we can not firmly conclude whether the lack of low - eddington agns in sf galaxies in our optical sample is due to selection effect or intrinsic nature since the size of the matched sample is very small and the x - ray flux limit of the swift - bat sample is still shallow ( down to l@xmath137/l@xmath138 @xmath2@xmath1 - 3 ) . in investigating the trend between ssfr and eddington ratio , it is important to reliably estimate agn bolometric luminosity . in the case of optically - identified type 2 agns , the [ ] luminosity is typically used for calculating eddington ratios as in our study ( e.g. * ? ? ? * ; * ? ? ? however , the bolometric correction of [ ] is difficult to determine due to various effects , including dust extinction and the dependency on the ionization parameter @xcite . also , star forming region can contribute to the [ ] flux observed through an aperture larger than the size of the narrow - line region . therefore , it is important to decompose the agn and sf fraction in the observed [ ] luminosity ( e.g. , * ? ? ? * ) . by simply assuming the maximum agn fraction in the [ ] luminosity of sf galaxies as 1% , 10% , and 50% , we simulate the distribution of ssfr as in figure 10 in order to test the effect of the contribution of non - detected agns in sf galaxies we find that the case of 1% and 10% agn luminosity does not change the decreasing trend of the ssfr as a function of eddington ratios down to 10@xmath139 . when we assume 50% of [ ] in sf galaxies is originated from agns , we still see the same trend while the difference of ssfr between high and low eddington agns becomes smaller . due to the aforementioned issues and uncertainty of eddington ratio , it is clear that more detailed investigations are required to confirm the trend of the decreasing ssfr with decreasing eddington ratios . in particular , a detailed comparison between x - ray and optical agns , for example , using the optical follow - up survey of the swift - bat agns @xcite , will be very useful to constrain the nature of host galaxies of low eddington agns , which is beyond the scope of this paper . since we have adopted the estimated sfr based on ir from @xcite , it is important to discuss whether the main results of this paper is affected by the uncertainty of ir - based sfr . as shown in figure 11 , other sfr indicators , i.e. , d@xmath52 and uv - to - optical color show consistent results . while sf galaxies have overall lower values of d@xmath1404000 , there is a clear spread of d@xmath52 among agns , which broadly correlates with outflow velocities . also , the uv - to - optical color shows a similar trend with agn outflow velocities . these results suggest that although the ir - based sfr is not the best sf indicator , there is a clear trend of ssfr with agn outflow velocities , while sf galaxies do not show such a trend . using a large sample of @xmath1110,000 type 2 agns out to z@xmath10.3 , we investigated the kinematics of the ionized gas outflows , demography of outflows in agns and sf galaxies , and the connection between outflows and sf . we summarize the main results as follows . @xmath141 gas and stellar velocity dispersions are comparable to each other in sf galaxies , indicating that he gravitational potential of host galaxies determines are mainly responsible for gas and stellar kinematics . @xmath141 in contrast , agns show much larger [ ] velocity dispersion than stellar velocity dispersion , indicating that strong non - gravitational kinematics , i.e. , outflows , are present . the kinematic component of outflows is comparable to or stronger than the virial motion caused by the gravitational potential . @xmath141 the fraction of agns with outflows steeply increases with agn luminosity and eddington ratio . in particular , the majority of luminous agns presents strong non - gravitational kinematics in the [ ] profile . @xmath141 we find a dramatic difference of the outflow signatures between agns and star - forming galaxies . the distribution in the [ ] velocity - velocity dispersion diagram dramatically expands toward large values with increasing agn luminosity , implying that the outflows are agn - driven , while that of sf galaxies show no significant change as a function of [ ] luminosity . @xmath141 the ssfr of non - outflow agns is much lower than that of strong outflow agns , while the ssfr of strong outflow agns is comparable to that of sf galaxies . we interpret this trend as a result of delayed agn feedback as it takes the order of a dynamical time for the outflows to suppress star formation . alternatively , gas depletion or intrinsic difference of gas content may cause the trend . we thank the anonymous referee for the suggestions , which were useful to clarify the results and interpretation . we thank sara ellison for her helpful comments on the ir - based sfr . support for this work was provided by the national research foundation of korea grant funded by the korea government ( no . 2016r1a2b3011457 ) . baumgartner , w. h. , tueller , j. , markwardt , c. b. , et al . 2013 , , 207 , 19 bennert , v. n. , auger , m. w. , treu , t. , woo , j .- h . , & malkan , m. a. 2011 , , 726 , 59 barbosa , f. k. b. , storchi - bergmann , t. , cid fernandes , r. , winge , c. , & schmitt , h. 2009 , , 396 , 2 baldwin , j. a. , phillips , m. m. , & terlevich , r. 1981 , pasp , 93 , 5 boroson , t. 2005 , , 130 , 381 brinchmann , j. , charlot , s. , white , s. d. m. , et al . 2004 , , 351 , 1151 cappellari , m. , & emsellem , e. 2004 , pasp , 116 , 138 carniani , s. , marconi , a. , maiolino , r. , et al . 2016 , , 591 , a28 chen , c .- t . j. , hickox , r. c. , alberts , s. , et al . 2013 , , 773 , 3 choi , y .- y . , woo , j .- h . , & park , c. 2009 , , 699 , 1679 ciotti , l. , & ostriker , j. p. 2007 , , 665 , 1038 crenshaw , d. m. , kraemer , s. b. 2000 , , 532 , l101 crenshaw , d. m. , kraemer , s. b. , hutchings , j. b. , et al . 2000 , , 120 , 1731 crenshaw , d. m. , kraemer , s. b. , & george , i. m. 2003 , , 41 , 117 crenshaw , d. m. , schmitt , h. r. , kraemer , s. b. , mushotzky , r. f. , & dunn , j. p. 2010 , , 708 , 419 croton , d. j. , et al . 2006 , , 365 , 11 das , v. , et al . 2005 , , 130 , 945 diamond - stanic , a. m. , & rieke , g. h. 2012 , , 746 , 168 di matteo , t. , springel , v. , & hernquist , l. 2005 , , 433 , 604 degraf , c. , di matteo , t. , treu , t. , et al . 2015 , , 454 , 913 dubois , y. , gavazzi , r. , reirani , s , & silk , j. 2013 , , 433 , 3297 dubois , y. , peirani , s. , pichon , c. , et al . 2016 , , 463 , 3948 ellison , s. l. , teimoorinia , h. , rosario , d. j. , & mendel , j. t. 2016 , , 455 , 370 ellison , s. l. , teimoorinia , h. , rosario , d. j. , & mendel , j. t. 2016 , , 458 , l34 faber , s. m. , & jackson , r. e. 1976 , , 204 , 668 fabian , a. 2012 , ara&a , 50 , 455 fischer , t. c. , crenshaw , d. m. , kraemer , s. b. , schmitt , h. r. , & trippe , m. l. 2010 , , 140 , 577 fischer , t. c. , crenshaw , d. m. , kraemer , s. b. , & schmitt , h. r. 2013 , , 209 , 1 gallo , e. , treu , t. , jacob , j. , et al . 2008 , , 680 , 154 - 168 greene , j. e. , & ho , l. c. 2005 , , 627 , 721 , d. , & mathur , s. 2004 , , 606 , l41 harrison , c. m. , alexander , d. m. , mullaney , j. r. , & swinbank , a. m. 2014 , , 441 , 3306 heckman , t. m. , miley , g. k. , & green , r. f. 1984 , , 281 , 525 heckman , t. m. 2004 , , 613 , 109 hopkins , p. f. , hernquist , l. , cox , t. j. , & kere , d. 2008 , , 175 , 356 hopkins , p. f. , torrey , p. , faucher - gigure , c .- a . , quataert , e. , & murray , n. 2016 , , 458 , 816 husemann , b. , jahnke , k. , snchez , s. f. , et al . 2014 , , 443 , 755 husemann , b. , scharwchter , j. , bennert , v. n. , et al . 2016 , , 594 , a44 kennicutt , r. c. , jr . 1998 , , 36 , 189 kewley , l. j. , groves , b. , kauffmann , g. , & heckman , t. 2006 , , 372 , 961 komossa , s. , & xu , d. 2007 , , 667 , l33 komossa , s. , xu , d. , zhou , h. , storchi - bergmann , t. , & binette , l. 2008 , , 680 , 926 kormendy , j. , & ho , l. c. 2013 , ara&a , 51 , 511 koss , m. , et al . , submitted lamassa , s. m. , heckman , t. m. , ptak , a. , & urry , c. m. 2013 , , 765 , l33 lintott , c. et al . 2011 , , 410 , 166 liu , g. , zakamska , n. l. , greene , j. e. , nesvadba , n. p. h. , & liu , x. 2013 , , 436 , 2576 markwardt , c. b. 2009 , aspc , 411 , 251 marconi , a. , & hunt , l. k. 2003 , , 589 , l21 matsuoka , k. , & woo , j .- h . 2015 , , 807 , 28 mauch , t. , & sadler , e. m. 2007 , , 375 , 931 miller , b. p. , gallo , e. , greene , j. e. , et al . 2015 , , 799 , 98 morrissey , p. 2007 , apjs , 173 , 682 mullaney , j. r. , alexander , d. m. , aird , j. , et al . 2015 , , 453 , l83 mullaney , j. r. , alexander , d. m. , fine , s. , et al . 2013 , , 433 , 622 mullaney , j. r. , pannella , m. , daddi , e. , et al . 2012 , , 419 , 95 nelson , c. h. , & whittle , m. 1995 , , 99 , 67 nelson , c. h. , & whittle , m. 1996 , , 465 , 96 nesvadba , n. p. h. , lehnert , m. d. , eisenhauer , f. , et al . 2006 , , 650 , 693 netzer , h. 2009 , mnras , 399 , 1907 park , d. , kelly , b. c. , woo , j .- h . , & treu , t. 2012 , , 203 , 6 park , d. , woo , j .- h . , bennert , v. n. , et al . 2015 , , 799 , 164 peng , c. y. 2007 , , 671 , 1098 rafferty , d. a. , brandt , w. n. , alexander , d. m. , et al . 2011 , , 742 , 3 reeves , j. n. , obrien , p. t. , & ward , m. j. 2003 , , 593 , l65 rosario , d. j. , santini , p. , lutz , d. , et al . 2012 , , 545 , a45 ruiz , j. , et al . 2001 , , 122 , 2961 ruiz , j. r. , crenshaw , d. m. , kraemer , s. b. , et al . 2005 , , 129 , 73 salviander , s. , & shields , g. a. 2013 , , 764 , 80 santini , p. , rosario , d. j. , shao , l. , et al . 2012 , , 540 , a109 snchez - blzquez , p. , et al . 2006 , , 371 , 703 shankar , f. , weinberg , d. h. , & miralda - escud , j. 2013 , , 428 , 421 shimizu , t. t. , mushotzky , r. f. , melndez , m. , koss , m. , & rosario , d. j. 2015 , , 452 , 1841 silk , j. , & rees , m. j. 1998 , , 331 , l1 treu , t. , woo , j .- h . , malkan , m. a. , & blandford , r. d. 2007 , , 667 , 117 veilleux s. , & osterbrock d. e. , 1987 , , 63 , 295 villar - martn , m. , humphrey , a. , delgado , r. g. , colina , l. , & arribas , s. 2011 , , 418 , 2032 villar - martn , m. , arribas , s. , emonts , b. , et al . 2016 , , 460 , 130 wang , j. , mao , y. f. , & wei , j. y. 2011 , , 741 , 50 white , r. l. , becker , r. h. , helfand , d. j. , & gregg , m. d. 1997 , , 475 , 479 woo , j .- h . , treu , t. , malkan , m. a. , & blandford , r. d. 2006 , , 645 , 900 woo , j .- h . , treu , t. , malkan , m. a. , & blandford , r. d. 2008 , , 681 , 925 woo , j .- h . , treu , t. , barth , a. j. , et al . 2010 , , 716 , 269 woo , j .- h . , kim , j. h. , imanishi , m. , & park , d. 2012 , , 143 , 49 woo , j .- h . , et al . 2015 , , 801 , 38 woo , j .- h . , bae , h .- j . , son , d. , & karouzos , m. 2016 , , 817 , 108 , t. , barth , a. j. , greene , j. e. , et al . 2011 , , 739 , 28 zhang , k. , dong , x .- b . , wang , t .- g . , & gaskell , c. m. 2011 , 737 , 71
in the absence of an applied bias voltage , a mesoscopic conductor can sustain a dc current component if two or more parameters of the device ( for example gate voltages ) are periodically modulated in time . pumping _ is said to be _ adiabatic _ if the parameter variation is slow on the scale defined by the dwell time of the electrons . in this case the pumped charge depends on the size and the shape of the pumping cycle but not on its detailed time evolution , i.e. it is of geometric nature . @xcite the idea of electron pumping dates back to a work of thouless . @xcite the first experiment on electron pumping in single electron devices was performed by pothier _ et al . _ @xcite since then much theoretical and experimental work has been devoted to electron pumping . @xcite for non - interacting systems the theory of adiabatic pumping , formulated by brouwer , @xcite is based on the generalization @xcite of the scattering approach for quantum transport to time - dependent phenomena . this formulation has been applied to study several aspects of pumping in non - interacting systems , such as the study of noise and decoherence , @xcite the role of discrete symmetries , @xcite the possibility of spin pumping , @xcite and the effect of superconducting elements and andreev reflection . @xcite a diagrammatic approach was used in refs . to calculate the pumped charge through a noninteracting system using random matrix theory in the limit of a large number of channels in the leads . furthermore several works investigate pumping by surface acoustic waves both theoretically @xcite and experimentally . @xcite pumping in interacting systems has been studied much less so far . quantum pumping through a luttinger liquid has been discussed . @xcite in quantum dots pumping has been studied in the limit of weak interaction , @xcite in the kondo regime , @xcite as well as in the coulomb - blockade regime . @xcite in refs . the pumped charge through an open quantum dot is computed by means of the bosonization technique . aono , in ref . uses the keldysh green s function formalism complemented by the assumption that the dot retarded green s function takes the non interacting form ( this holds true in the non - interacting limit as well as in the kondo regime ) . the authors of ref . integrate numerically the master equation ( in the born - markov approximation ) for the reduced density matrix of a double - dot pump . recently a general approach to pumping through interacting quantum dots has been put forward by relating the pumped charge to the instantaneous retarded green s functions of the quantum dot . @xcite in this paper we study adiabatic pumping through interacting quantum dots for temperatures much above the kondo temperature but much below the level spacing in the dot . in this case a perturbative expansion in the tunnel coupling between the dot and the leads is justified . moreover we can restrict ourselves to consider only one level in the dot with a strong local repulsion in the case of double occupancy . we aim at the understanding of the influence of coulomb interaction on the pumping characteristics . in order to achieve this , we extend a diagrammatic real - time technique @xcite that has been developed to describe non - equilibrium dc transport through an interacting quantum dot . as compared to the formalism in our recent work , @xcite the perturbative approach presented here , although limited to weak tunnel - coupling strengths , is more transparent in identifying the physical origin of the various contributions to the pumped charge . in particular it is straightforward to relate the pumped current to the dynamics of the average charge of the dot . in this work we calculate the leading- and next - to - leading - order contribution of the perturbation expansion in the tunnel coupling to the pumped charge per pumping cycle . we distinguish two cases : pumping by changing periodically either the gate voltage and one tunnel barrier or varying both left and right tunnel barrier . considering the first case , and furthermore taking into account only lowest - order tunneling processes associated with sequential tunneling , the adiabatic pump works analogously to a peristaltic pump . the next - order correction turns out to be only due to a time - dependent renormalization of the dot - level position that is induced by the combination of coulomb interaction and tunnel coupling to the leads . remarkably this effect is not masked by other higher - order transport processes such as cotunneling . the situation is even more dramatic for pumping with the tunnel barriers . in this case , the lowest - order tunneling processes do not give rise to any pumping . the dominant pumping mechanism is , then , of higher order , namely pumping by making use of the time - dependent level renormalization . as a consequence , the gate - voltage dependence of the pumped charge provides a transparent experimental access to probe quantum - fluctuation effects . the paper is organized as follows : section [ sec_model ] introduces the model of the quantum dot . the time evolution of the system is described by a generalized master equation in section [ sec_master ] and an adiabatic expansion is performed and applied to the current pumped through the dot in section [ sec_current ] . the expansion in the tunnel coupling is further discussed in section [ sec_tunnel ] . the explicit evaluation of the formulae obtained up to here is done in section [ sec_rules ] using a diagrammatic technique . in section [ sec_results ] results for the pumped current and the pumped charge are presented and discussed . we consider a single - level quantum dot with on - site coulomb interaction coupled to two non - interacting leads . the system is described by the hamiltonian @xmath0 where @xmath1 , @xmath2 , and @xmath3 describe , respectively , the left ( l ) and right ( r ) leads , the dot , and tunneling between dot and leads , and are given by [ hamiltonians ] @xmath4 \label{eqn_hamilton_tunnel}.\end{aligned}\ ] ] in eqs . ( [ hamiltonians ] ) , @xmath5 ( @xmath6 ) is the fermionic annihilation ( creation ) operator for an electron with spin @xmath7 , momentum @xmath8 , energy @xmath9 in lead @xmath10 ; @xmath11 ( @xmath12 ) is the fermionic annihilation ( creation ) operator for an electron with spin @xmath13 in the dot ; and @xmath14 is the number operator for the dot electrons with spin @xmath13 . the coulomb interaction on the dot is described by the on - site energy @xmath15 associated with double occupation . the leads are assumed to be in thermal equilibrium with the same chemical potential and to have flat bands with constant density of states @xmath16 . by periodically changing at least two of the system parameters , a dc current can be pumped through the dot . we choose the level position of the dot @xmath17 and the tunnel matrix elements @xmath18 to be time - dependent . we only allow for the modulus , but not the phase , of @xmath18 to vary in time , since a time - dependent phase would correspond to a bias voltage . we define the time - dependent intrinsic line width @xmath19 , the total intrinsic line width @xmath20 , as well as @xmath21 and @xmath22 . to keep all formulae transparent , we set @xmath23 throughout the paper . as described above we consider an interacting quantum dot coupled to non - interacting leads . since the leads act as baths , it is convenient to trace out the degrees of freedom of the non - interacting lead states to arrive at an effective description of the reduced system . in the limit of temperature much lower than the level spacing of the dot , only one level will contribute to transport . therefore the hilbert space for the dot is four dimensional : the quantum dot can be empty , singly occupied with a spin - up or a spin - down electron or doubly occupied . these states , labeled by @xmath24 , have energy @xmath25 , @xmath26 and @xmath27 , respectively . in the following we use a matrix notation in the four dimensional hilbert space of the dot , with boldface symbols for vectors and matrices . the probabilities to find the dot in the respective state are @xmath28 . the starting point of our analysis is the generalized master equation for the time evolution of the system , @xmath29 where the matrix elements @xmath30 of the kernel @xmath31 describe the transition from a state @xmath32 at time @xmath33 to a state @xmath34 at time @xmath35 . for the system considered here , eq . ( 2 ) defines the most general kinetic equation for the dot probabilities without any approximation . off - diagonal matrix elements of the reduced density matrix for the quantum dot , that correspond to _ real _ superposition of different states @xmath36 , do not couple to the diagonal ones since @xmath34 and @xmath32 differ by a conserved quantum number , particle number or spin . therefore , off - diagonal elements of the reduced density matrix do not enter any transport quantity . nevertheless , quantum fluctuation effects involving _ virtual _ intermediate states of the quantum dot in higher - order processes such as cotunneling , are fully taken into account by eq . ( 2 ) by properly evaluating the kernel @xmath37 . @xcite our goal is to describe the response of the system to slow periodic variations of the system parameters @xmath38 with frequency @xmath39 . after waiting long enough , such that any memory of the initial dot - state distribution @xmath40 has died out , the dynamics of the system is fully determined by the explicit time dependence of the system parameters the latter enters eq . ( [ eqn_determine_p ] ) in two ways , namely by the kernel @xmath31 being a functional of the system parameters @xmath41 with @xmath42 $ ] , and by the non - markovian structure , i.e. , the time derivative of @xmath43 at time @xmath35 depends on @xmath44 at an earlier time @xmath33 at which the system parameters had different values . in the adiabatic regime it is possible to simplify considerably the form of the master equation by performing an _ adiabatic expansion _ , i.e. , an expansion in the pumping frequency @xmath39 , assuming that @xmath39 is small as compared to both the energy scales that determine the decay time of the kernel and the time integral of the kernel ( which sets the time scale of the system s response to the parameter s change ) . the zeroth - order , _ instantaneous _ , term of the adiabatic expansion corresponds to freezing the value of all system parameters , @xmath38 , at time @xmath35 , which corresponds to solving a time - independent problem . to obtain the first - order correction we need to systematically collect all contributions linear in the pumping frequency or , equivalently , linear in the time derivative of @xmath45 at time @xmath35 . for this , we perform a taylor expansion of @xmath46 around @xmath35 up to linear order , @xmath47\ , .\ ] ] furthermore , we perform an adiabatic expansion of the kernel @xmath31 itself . the zeroth - order term , @xmath48 , is indicated with the superscript @xmath49 for _ instantaneous _ and the subscript @xmath35 to emphasize that the system parameters @xmath50 are frozen at time @xmath35 . it depends only on the time difference @xmath51 , and only parametrically on @xmath35 through @xmath38 . the first - order term is obtained by linearizing the time dependence of all parameters @xmath41 with respect to the final time @xmath35 , i.e. , @xmath52 , and retaining only linear terms in time derivatives . this linear correction to the kernel is indicated by the superscript @xmath53 for _ adiabatic _ , @xmath54 finally , we need to perform an adiabatic expansion for the occupation probabilities in the dot , @xmath55 the instantaneous probabilities @xmath56 are the solution of the time - independent problem with all parameter values fixed at time @xmath35 . they are obtained from the master eq . ( [ master_appr ] ) in the stationary limit , @xmath57 together with the normalization condition @xmath58 , where @xmath59 , and we have introduced the laplace transform @xmath60 to define @xmath61 . the first adiabatic correction can be obtained from eq . ( [ master_appr ] ) , using eqs . ( [ w_exp ] ) and ( [ p_exp ] ) . we find @xmath62 where again @xmath63 is the laplace transform at zero frequency and @xmath64 . once @xmath63 are evaluated and the instantaneous probabilities @xmath65 are known from eq . ( [ eqn_instantaneous_p ] ) , the adiabatic corrections @xmath66 are obtained from eq . ( [ eqn_ad_expansion_p ] ) together with the normalization condition @xmath67 . the charge @xmath68 pumped in one cycle @xmath69 is related to the time - dependent current @xmath70 flowing through the left barrier by @xmath71 by accounting for the time evolution of the system before time @xmath35 in a similar way as done for the dot - state probabilities above , we can express the current into the left lead as @xmath72 where @xmath73 , and @xmath74 includes all processes associated with transitions where the number of electrons ( with charge @xmath75 ) entering reservoir @xmath76 minus the ones leaving it equals @xmath77 . it is straightforward to perform an adiabatic expansion for eq . ( [ eqn_current ] ) in the same way as for the master equation . the instantaneous or zeroth - order level of the adiabatic expansion is sufficient to describe the dc current that is driven through the quantum dot by an applied transport voltage . @xcite it is , furthermore , sufficient for modeling _ , i.e. , the generation of a dc current component by applying an ac transport voltage and appropriately changing some system parameter in time . in the absence of any dc or ac transport voltage , as considered in the present paper , the instantaneous part of the current vanishes . in order to describe _ pumping _ , one needs to compute the first - order adiabatic correction of the current . using for eq . ( [ eqn_current ] ) the same procedure as for the master equation , we find the adiabatic part of the current to be @xmath78 .\ ] ] alongside with the adiabatic expansion we perform a perturbation expansion in powers of the tunnel coupling strength @xmath79 for both the instantaneous and the adiabatic correction of the kernel @xmath37 , the probabilities @xmath80 , and the current @xmath81 . we indicate the order of the perturbation expansion in @xmath79 by adding a superscript , i.e. , @xmath82 for the instantaneous contribution to the kernel , and similarly for @xmath83 . the expansion of the instantaneous probabilities begins in zeroth order in @xmath79 , @xmath84 , in order to be able to fulfill the normalization condition @xmath85 . by expanding eq . ( [ eqn_instantaneous_p ] ) in powers of @xmath79 , we find that the instantaneous probabilities should fulfill the two equations , [ instprob ] @xmath86 together with the normalization conditions @xmath87 and @xmath88 . as discussed above , the instantaneous part of the current vanishes in all order in @xmath79 due to the absence of an applied transport voltage . in order to determine the adiabatic corrections to the probabilities , we expand also eq . ( [ eqn_ad_expansion_p ] ) in powers of @xmath79 , [ adprob ] @xmath89 with the normalization conditions @xmath90 and @xmath91 . we emphasize that , in order to properly match the powers of @xmath79 on the left and right hand side of eq . ( [ adprob ] ) , one has to start the expansion of the adiabatic correction of the probabilities in _ minus first _ order in @xmath79 , @xmath92 . at first glance , such an expansion might look divergent for the weak - coupling limit , @xmath93 . however , in the validity range of the adiabatic expansion everything remains well defined : the adiabaticity condition requires that the energy scale defined by the pumping frequency @xmath39 is much smaller than the tunnel - coupling strength @xmath79 . since the time derivative on the left hand side of eq . ( [ adprob ] ) introduces a factor @xmath39 , we see that @xmath94 scales with @xmath95 , which is always much smaller than @xmath96 in the adiabatic limit . the perturbation expansion of the adiabatically pumped current is derived from eq . ( [ eqn_current_ad ] ) and reads [ expcur ] @xmath97 \ , .\end{aligned}\ ] ] we see that the lowest - order contribution to the pumped current starts in zeroth order in the tunnel coupling strength @xmath79 , as it consists in a product of a first- and a minus - first order term in the tunneling coupling , but scales linearly with the pumping frequency @xmath39 . this contrasts with the dc current driven by a finite bias voltage , for which the lowest - order contribution is linear in @xmath79 . certain properties of pumping can be derived by a closer inspection of the perturbative expansion of the master equation . to zeroth order in @xmath79 the pumped current is nonzero only if @xmath98 is non vanishing , which , according to eq . ( [ adprob_a ] ) requires that the zeroth - order instantaneous probabilities @xmath99 depends on time . however the latter are simply determined by the boltzmann factors of the corresponding state energies : @xmath100 where @xmath101 is the energy related to the dot state @xmath34 , @xmath102 is the inverse temperature and @xmath103 the partition function . in particular , the probabilities @xmath104 are independent of the tunnel couplings . as a consequence , in order to have a non - vanishing zeroth - order pumped current @xmath105 , one of the pumping parameters has to be the level position . when pumping with the two barrier heights , @xmath105 vanishes . in order to evaluate the kernel @xmath106 of the master equation eq . ( [ eqn_determine_p ] ) we use the diagrammatic perturbation approach to transport through interacting quantum dots developed in ref . . while in ref . the diagrammatic language was derived for dc transport with time - independent system parameters , we generalize the approach in this section to account for the adiabatic time dependence of the external parameters . we start with deriving the master equation from a very general point of view in order to relate its kernel to a set of diagrams to be evaluated . in general , the ( time - dependent ) transport properties are governed by the time evolution of the reduced density matrix of the dot obtained after tracing out the degrees of freedom of the non - interacting lead electrons . since the leads are non - interacting and in thermal equilibrium , the lead electrons can be integrated out making use of wick s theorem , i.e. contracting pairs of creation and annihilation operators @xmath107 ( @xmath108 ) . furthermore , since in the case of pumping there is no voltage applied between left and right lead , the occupation of electronic states in both leads is described by the same fermi distribution function @xmath109 . the time evolution of the reduced density matrix is related to the propagator @xmath110 by @xmath111 , width=240 ] contributions to this propagator can be depicted as diagrams on the keldysh contour , where contractions of fermion operators of the leads are indicated as tunneling lines . an example is shown in fig . [ fig_red_density ] . the propagator @xmath110 can be expressed in terms of its irreducible part @xmath112 by means of a dyson equation : @xmath113 the irreducible diagram part @xmath112 is defined as the sum over all diagrams in which any vertical cut crosses at least one tunneling line ( see fig . [ fig_block_example ] as an example ) . + . [ fig_block_example],width=259 ] performing the time derivative of eq . ( [ eqn_propagator ] ) , plugging in eq . ( [ eqn_reduced_self ] ) , and shifting the lower bound of the remaining time integral to minus infinity , we obtain the generalized master equation of eq . ( [ eqn_determine_p ] ) . in a similar way we proceed for the current @xmath114 of particles flowing into reservoir @xmath76 , which is given by @xmath115 \ , .\ ] ] it can also be expressed in terms of diagrams . they contain a vertex at time @xmath35 that has the same structure as the tunneling vertices in fig . [ fig_red_density ] , and can be attached to the upper or the lower propagator line . we find eq . ( [ eqn_current ] ) where @xmath116 has the following properties . the terms @xmath74 are given by all diagrams for which the number of tunneling lines with reservoir index @xmath76 running from the upper to the lower propagator minus the number of those with reservoir index @xmath76 running in the opposite direction equals @xmath77 . in the following we summarize the diagrammatic rules for the kernel @xmath117 ( as described in ref . ) and discuss additional rules for the evaluation of its adiabatic expansion . examples for the application of the rules given below are shown in appendix [ ap_example ] . we start with rules for the laplace transform of @xmath118 in @xmath119-th order in the tunnel coupling . this will directly lead us to the desired objects @xmath120 and @xmath121 . + ( 1 ) draw all topologically different diagrams with @xmath119 directed tunneling lines connecting pairs of vertices containing lead electron operators . assign a reservoir index @xmath76 , an energy @xmath122 and a spin index @xmath13 to each of these lines . assign states @xmath34 and the corresponding energies @xmath123 to each element of the keldysh contour connecting two vertices . furthermore , draw an external line from the upper leftmost beginning of a dot propagator to the upper rightmost end of a dot propagator that carries the ( imaginary ) energy @xmath124 . + ( 2 ) for each time segment between two adjacent vertices ( independent on whether they are on the same or on opposite branches of the keldysh contour ) write a resolvent @xmath125 where @xmath126 is the difference of left going minus right going energies ( including energies of tunneling lines and the external line - the positive imaginary part of @xmath127 will keep all resolvents regularized ) . + ( 3 ) each vertex containing a dot operator @xmath128 gives rise to a matrix element @xmath129 where @xmath34 ( @xmath32 ) is the dot state entering ( leaving ) the vertex with respect to the keldysh contour . + ( 4 ) the contribution of a tunneling line of reservoir @xmath76 is @xmath130 if the line is going backward with respect to the closed time path and @xmath131 $ ] if it is going forward . + ( 5 ) the overall prefactor is given by @xmath132 where @xmath133 is the total number of vertices on the backward propagator and @xmath134 the number of crossings of tunneling lines . + ( 6 ) integrate over the energies of tunneling lines and sum over reservoir and spin indices . + to derive the rules for the adiabatic corrections @xmath135 we first analyze how the time - dependent parameters enter the expression of the kernel @xmath136 . the time - dependent variables for which the adiabatic expansion has to be performed are @xmath137 and @xmath138 , where the first one only appears in the product @xmath139 associated with a tunneling line , and the latter only in the isolated - dot propagator @xmath140 for each segment between adjacent vertices . while for the instantaneous kernels all parameters were taken at time @xmath35 , now we perform a series expansion around the same time @xmath35 and keep all contributions linear in a time derivative of the pumping parameters , @xmath141 . \label{firstorder_eps_arb}\end{aligned}\ ] ] the factors @xmath142 or @xmath143 can be included in the diagrammatic rules in the following way : introduce an additional external frequency line with the imaginary energy @xmath144 from the vertex at @xmath145 to the rightmost vertex at @xmath35 ( or the imaginary energy @xmath146 from the beginning of a dot propagator line at @xmath147 to the rightmost upper end of a dot propagator line at @xmath35 ) , performing the first derivative with respect to @xmath148 ( or second derivative with respect to @xmath149 ) then set @xmath150 and @xmath151 . the external frequency lines are drawn as dotted lines in fig . [ adiabaticdiagrams ] . is needed for the evaluation of @xmath152 ; @xmath153 , @xmath154 , and @xmath155 are needed for the contributions to @xmath156 due to both @xmath157 and @xmath158 ; @xmath159 does not contribute and can be omitted ; and @xmath160 is additionally needed for the evaluation of the contribution to @xmath156 due to @xmath158.,width=240 ] the rules to compute the contribution to the adiabatic corrections @xmath135 due to the time - dependence of @xmath161 read : + ( 7a ) add to all diagrams needed for @xmath162 additional external frequency lines between any vertex @xmath145 and the right corner of the diagram and assign to them an ( imaginary ) energy @xmath163 . note that an eventual external frequency line between two right corners of a diagram does not contribute and can always be omitted . + ( 7b ) follow the rules ( 1 ) to ( 6 ) taking into account the extra lines . + ( 7c ) perform a first derivative with respect to @xmath148 and multiply it by the factor @xmath164 . sum all the contributions obtained in this way . + ( 7d ) set all the external frequencies @xmath148 and @xmath165 to @xmath166 . the contribution to the adiabatic correction @xmath135 due to the time - dependence of the level position can be computed in a similar way : + ( 8a ) in addition to the external frequency lines added according to rule ( 7a ) , put one more external frequency line from the left corner of the diagram with no vertex to the right corner . + ( 8b ) follow the rules ( 1 ) to ( 6 ) taking into account the extra lines . + ( 8c ) perform a second derivative with respect to @xmath148 and multiply by @xmath167 , where @xmath34 ( @xmath32 ) is the dot state entering ( leaving ) the vertex of the external frequency line at @xmath145 with respect to the keldysh contour . the term @xmath168 ( @xmath169 ) is omitted if the segment associated with @xmath101 ( @xmath170 ) does not belong to the diagram . sum all the contributions obtained in this way . + ( 8d ) set the external frequencies @xmath148 and @xmath165 to @xmath166 . in this section we show the results for the pumped current and the pumped charge through a single - level quantum dot . as no bias voltage is applied , the only contribution to the pumped current arises from the adiabatic correction , and hence we drop the superscript @xmath53 for the current . as we discuss below , the properties of the pumped charge , in the regime discussed in this paper , can be understood to a large extent in terms of the time - dependence of the occupation of the quantum dot , @xmath171 . for this reason , we first want to discuss the perturbation expansion of the instantaneous average charge occupation . the contribution to zeroth order in @xmath79 turns out to be determined by the boltzmann factors of the energies associated with the states of the isolated quantum dot . this yields @xmath172 the first - order correction accounts for quantum fluctuations due to tunneling from and to the leads . there are two qualitatively different effects which are due to tunneling and correspondingly we present the results for the first - order corrections as a sum of two contributions @xmath173 , where the contribution to broadening is the sum of the contributions of the two leads @xmath174 . first , the dot levels acquire a finite life - time broadening due to the coupling to lead @xmath76 , which is accounted for by @xmath175 where @xmath176 is the derivative of @xmath177 , and @xmath178 is the digamma function . the first term accounts for the broadening of the resonance at @xmath179 between empty and singly - occupied dot ; it has a prefactor @xmath180 , when the dot is empty , and is zero when the dot is doubly occupied . the second term accounts for the broadening of the resonance at @xmath181 between singly- and doubly - occupied dot ( this contribution is zero if the dot is empty ) . in the case that the dot is singly occupied both terms contribute with a prefactor @xmath96 . second , the combination of tunneling and charging energy gives rise to a renormalization of the level position , @xmath182 with @xmath183 as it is expected from the poor man s scaling analysis . @xcite the level renormalization is positive when the level is in the vicinity of the fermi energy of the leads and negative for @xmath181 being close to the fermi energy . this means that the distance between the two resonances from empty to singly - occupied dot and from singly- to doubly - occupied dot is effectively decreased . the changes to the instantaneous average occupation due to the level renormalization reads @xmath184 the sum of eqs . ( [ eqn_n_broad ] ) and ( [ eqn_n_renormalization ] ) is directly found from the correction in first order @xmath79 to the occupation probability , which is shown explicitly in appendix [ ap_results ] . we remark that the above interpretation of the two terms is in full agreement with known exact results for the non - interacting case . for @xmath185 ( and flat density of states in the leads ) , the level renormalization vanishes . the spectral density is then equal to the breit - wigner function and its expansion in zeroth and first order in @xmath79 leads to the non - interacting result of @xmath186 and @xmath187 . we now proceed with solving eqs . ( [ instprob]a ) , ( [ adprob]a ) and ( [ expcur]a ) for arbitrary interaction @xmath15 to get the zeroth - order adiabatically pumped current . the result of the diagrammatic approach explained above can be written in the form @xmath188 this suggests the following interpretation . as the dot occupation is changed in time by varying the pumping parameters ( one of them must be the level position since @xmath189 is independent of the tunnel - coupling strengths ) , the charge moves in and out of the quantum dot generating a current from / into the leads . the contributions flowing through barrier @xmath76 split weighted by the time - dependent relative tunnel couplings @xmath190 . by means of eqs . ( [ instprob]b ) , ( [ adprob]b ) and ( [ expcur]b ) , one finds the first - order-@xmath79 contribution to the current , @xmath191 again , we have written the result in such a form that an identification of the pumping mechanism is straightforward . the first term of eq . ( [ eq_second_current ] ) contains the contribution due to the correction of the average dot occupation induced by the tunnel coupling to the left lead . intuitively the finite - life - time broadening due to the coupling to the left lead is associated with tunnel processes of electrons through the left barrier . any change in the life - time broadening due to coupling to the left lead will , therefore , result in a current through the left barrier only . as a result , this first term contains a total time derivative , and as parameters are periodically changing in time , it will not lead to a net pumped charge after the full pumping cycle . we conclude that changing the life - time broadening of the dot level does not contribute to adiabatic pumping . the second term has the same structure as the zeroth - order contribution , eq . ( [ eq_second_current ] ) . it can be understood as the correction term introduced by renormalizing the position of the dot level , which may be time dependent via time - dependent tunnel couplings or a time - dependent gate voltage . again , the charge transferred in / out of the quantum dot splits into two currents to or from both leads with relative weight @xmath190 . now , even if the dot level is constant and only both the tunnel couplings @xmath192 are varying in time , a finite charge can be pumped by means of level renormalization . it is useful to compare these findings with a perturbation expansion of the dc current driven by a dc transport voltage . in lowest ( first ) order , current is carried by sequential - tunneling processes . a systematic calculation of the second - order linear conductance @xcite shows that quantum fluctuation due to tunneling give rise to three different types of correction terms . the first one , which dominates the linear conductance in the coulomb - blockade regime away from resonance , is due to cotunneling . one way to depict cotunneling is to understand it as transport through the finite - life - time broadened dot level . it would , thus , correspond to the first term of eq . ( [ eq_second_current ] ) . close to resonance , however , there are two more corrections to the sequential - tunneling linear conductance . they can be cast as sequential tunneling but with renormalized level position , as discussed above , or with renormalized tunnel coupling strength . for the dc current , all these three contributions are present at the same time , which makes it challenging to identify them separately in an experiment . for the adiabatically pumped charge , where correction terms associated with a renormalization of the tunnel couplings and level - broadening effects vanish , the situation is distinctively different . studying adiabatic pumping is , therefore , a convenient tool to access the energy - level renormalization . this is most dramatic in the case when the zeroth - order pumped current is zero , i.e. , when pumping is done by changing both tunnel couplings . in this case , the _ dominant _ contribution to the pumped charge is due to time - dependent level renormalization . when writing the pumped charge , we report as indices , in the following , the particular choice of pumping parameters it refers to . for example if the pumping fields are @xmath193 and @xmath179 , we indicate the charge as @xmath194 . we now concentrate on weak pumping . we write the time - dependent parameters in the form @xmath195 and @xmath196 ( with @xmath197 and @xmath198 having zero time average ) and expand the current up to bilinear response in the time - dependent part of the parameters . choosing @xmath179 and @xmath193 as pumping parameters we obtain up to first order in @xmath79 : @xmath199,\ ] ] where the prefactor @xmath200 characterizes the amplitudes of the pumping parameters as well as their relative phase : @xmath201 it was already pointed out for a noninteracting system in ref . that in the limit of weak , adiabatic pumping the pumped charge is proportional to the surface enclosed in parameter space during one pumping cycle , which in this case is equal to @xmath200 . furthermore , @xmath202 is the instantaneous occupation of the dot computed with the time - dependent parameters taken at their time - average value , and @xmath203 . the first term inside the brackets is the zeroth - order-@xmath79 contribution to the charge and , therefore , it is the dominant one . it has two peaks as a function of the average level position , which are located , in the limit @xmath204 , at @xmath205 and @xmath206 . the second term is first order in @xmath79 , stems from level renormalization , and vanishes in the non - interacting case . the first - order-@xmath207 correction tends to decrease the distance between the two resonances . in units of @xmath208 as a function of the time - average level position in units of @xmath209 for different values of @xmath15 . pumping parameters are @xmath179 and @xmath193 . the temperature is @xmath210 . [ fig_charge_level ] ] in fig . [ fig_charge_level ] the pumped charge of eq . ( [ eq_charge0_ufinite ] ) is shown in units of @xmath208 as a function of the time - average value of the level position , for different strengths of the interaction . the two peaks are directly related to transitions between singly - occupied and empty dot and between doubly- and singly - occupied dot . the shift by @xmath211 of the peak positions is of combinatoric origin and reflects the fact that the probability of single occupation of the dot is increased due to the two spin states which lead to the same occupation number . both peaks contribute with the same sign , as expected since the underlying pumping mechanism is the same for both resonances : in both cases the dot filling increases ( decreases ) when the level position is decreased ( increased ) . we now focus our attention on pumping with the two tunnel - coupling strengths @xmath193 and @xmath212 . in this case , there is no zeroth - order-@xmath79 contribution to the pumped charge , as discussed above . the contribution to first order in @xmath79 reads @xmath213 where @xmath214 accounts for the pumping - parameter amplitudes and their relative phase as discussed in detail before for the quantity @xmath215 . in units of @xmath216 as a function of the time - average level position in units of @xmath209 for different values of @xmath15 . pumping parameters are @xmath193 and @xmath212 . the temperature is @xmath210 . [ fig_secondorder],width=3 ] the result for the pumped charge as a function of the level position is shown in fig . [ fig_secondorder ] . the solid line shows the result for very small interaction . as expected , it tends to zero , because the level renormalization vanishes . in the presence of interaction two peaks appear , which separate for increasing @xmath15 . the two peaks are related to the two resonances at the level positions . they contribute with opposite sign . this reflects the opposite sign of the level renormalization for the two resonances . we remind that the first non - vanishing contribution of the perturbation expansion to the charge pumped through the dot by periodic change of the tunnel barriers is uniquely due to the effect of level renormalization . the height of the peaks increases for increasing @xmath15 , growing logarithmically for large @xmath15 . eventually , this increase will be cut - off by the bandwidths @xmath217 , which we here chose to be infinite . the different sign of the pumped charge for the two resonances could serve as a signature to distinguish level - renormalization - induced pumping from parasitic peristaltic pumping due to cross capacitances of the gates modulating the tunnel couplings to the quantum dot . we have presented a perturbative approach in tunneling to adiabatic pumping through interacting quantum dots . in particular , a general diagrammatic technique to perform the adiabatic expansion has been developed . this technique has been applied to compute the pumped charge through a single - level quantum dot at temperatures much higher than the kondo temperature . two pumping schemes have been considered : pumping with the level position and one tunnel barrier , and pumping with the two barriers . when pumping with the level position and one tunnel barrier , the dominant mechanism of the adiabatic pump works analogously to a peristaltic pump . the next - order correction is related to the level renormalization induced by the interplay of coulomb repulsion and electron tunneling . the situation is far more interesting for the case of pumping with the two barriers . with this pumping scheme there is no pumping in lowest order in the tunnel coupling , and the first non - vanishing contribution is due to the time - dependent level renormalization . hence , we have demonstrated the importance of level - renormalization effects in pumping through interacting quantum dots . in particular , our results suggest that adiabatic pumping can be used to gain experimental access to the level renormalization in quantum dots . we acknowledge financial support by the european community via grants rtnnano and miur - prin , the dfg via sfb491 , and the nsf under grant no . phy99 - 0794 . in this section we show how to apply the diagrammatic rules for the matrix element @xmath218 . + we start with the instantaneous term to lowest order in the tunnel coupling , @xmath219 . the corresponding diagrams are shown in fig . [ fig_ad_example ] . two topologically different diagrams contribute , and each of them has to be summed over the spin index @xmath13 and the lead index @xmath76 , and to be integrated over @xmath122 . we obtain : @xmath220 \ , . \nonumber\ ] ] by letting @xmath221 and making use of @xmath222 , where @xmath223 indicates cauchy s principal value , we get @xmath224 part of the adiabatic correction to the matrix element @xmath225 . all appearing reservoir and spin indices @xmath76 , @xmath13 are to be summed over and the energy @xmath122 is to be integrated over . [ fig_ad_example ] ] for the adiabatic correction we need to introduce additional external frequency lines according to the rules 7a and 8a ( see fig.[fig_ad_example ] ) . the additional line of rule 8a , going from the left corner of the diagram with no vertex to the right corner , does not contribute in this case and we have omitted drawing it . the evaluation of these diagrams leads us to the result @xmath226 we now calculate the second - order-@xmath79 contribution to the same matrix element of the instantaneous kernel . all diagrams contributing to the matrix element @xmath227 are depicted in fig . [ fig_inst_example2 ] . we sum over all appearing indices @xmath228 . the variables @xmath229 denotes the opposite spin of @xmath13 . as an example , we report the result for the sum of the first diagram of the first and the second line ( after setting @xmath221 ) : @xmath230 \ , . \end{aligned}\ ] ] to obtain the full second - order contribution @xmath227 we need to evaluate the remaining diagrams in fig . [ fig_inst_example2 ] along the same lines as discussed in this appendix . some intermediate results were not presented in the main part of this article as they were lengthy or not immediately necessary for the interpretation of the pumped current . here we discuss in detail the first corrections to the occupation probabilities . the corrections to the occupation probabilities are used for the evaluation of eq . ( [ expcur ] ) but do not appear directly in the results for the pumped current . we find for @xmath231 , the first order adiabatic correction in minus first order in @xmath79 : @xmath232}.\nonumber\ ] ] the adiabatic correction of the occupation probability is proportional to the time - derivative of @xmath233 , and is therefore an eigenvector of the matrix @xmath234 . the sign of this correction depends on the sign of the time derivative of @xmath138 . the first - order-@xmath79 correction to the instantaneous occupation probability is : @xmath235 it consists of a part due to level renormalization ( first row ) and a part due to level broadening ( second and third row ) . the correction due to level renormalization affects all four probabilities in the same functional way . + the correction due to broadening has two contributions . the first one is related to the broadening due to fluctuations between empty and singly - occupied dot . it is zero in the case that the dot is doubly occupied and largest when the dot is empty . the second contribution is related to the broadening due to fluctuations between singly- and doubly - occupied dot . it is zero when the dot is empty and largest when the dot is doubly occupied . b. l. altshuler and l. i. glazman , science * 283 * , 1864 ( 1999 ) ; y. makhlin and a. d. mirlin , phys . lett . * 87 * , 276803 ( 2001 ) ; h .- q . zhou , s. y. cho , and r. h. mckenzie , phys . lett . * 91 * , 186803 ( 2003 ) . i. l. aleiner , b. l. altshuler , and a. kamenev , phys . b * 62 * , 10373 ( 2000 ) . m. moskalets and m. bttiker , phys . b * 66 * , 035306 ( 2002 ) . m. moskalets and m. bttiker , phys . b * 64 * , 201305 ( 2001 ) . n. e. fletcher , j. ebbecke , t. j. b. m. janssen , f. j. ahlers , m. pepper , h. e. beere , and d. a. ritchie , phys . b * 68 * , 245310 ( 2003 ) ; j. ebbecke , n. e. fletcher , t. j. b. m. janssen , f. j. ahlers , m. pepper , h. e. beere , and d. a. ritchie , appl . phys . lett . * 84 * , 4319 ( 2004 ) . e. cota , r. aguado , and g. platero , phys . 94 * , 107202 ( 2005 ) ; e. cota , r. aguado , and g. platero , phys . 94 * , 229901(e ) ( 2005 ) . r. citro , n. andrei , and q. niu , phys . b * 68 * , 165312 ( 2003 ) . j. knig , h. schoeller , and g. schn , phys . lett . * 76 * , 1715 ( 1996 ) ; j. knig , j. schmid , h. schoeller , and g. schn , phys . rev . b * 54 * , 16820 ( 1996 ) ; h. schoeller , in _ mesoscopic electron transport _ , edited by l.l . sohn , l.p . kouwenhoven , and g. schn ( kluwer , dodrecht , 1997 ) ; j. knig , _ quantum fluctuations in the single - electron transistor _ ( shaker , aachen , 1999 ) .
let @xmath0 be a coxeter system . there is a canonical real representation of @xmath1 which is called the _ geometric representation _ of @xmath0 and which is known to be faithful by the work of tits in the 1960s ( a proof can be found in @xcite . in the proof he considers a convex cone @xmath2 in the dual module which is stabilized under the action of @xmath1 . by means of the standard generating set @xmath3 he defines an open simplicial cone @xmath4 which turns out to be a prefundamental domain for the action of @xmath1 on @xmath2 . it is now the standard terminology to call the cone @xmath2 the _ tits cone _ and @xmath4 the _ ( open ) fundamental chamber_. the conjugates of the fundamental generators act as linear reflections on the module and its dual . their reflection hyperplanes yield a simplicial decomposition of @xmath2 and thus provide a simplicial complex which is the _ coxeter complex _ of @xmath0 . the geometric representation of @xmath0 , the geometry of the tits cone , and the properties of the coxeter complex are fundamental tools for the investigation of coxeter groups . coxeter systems play an important role in various branches of mathematics . in combinatorics and geometric group theory they are a rich source of interesting phenomena ( @xcite , @xcite ) . in representation theory and the theory of algebraic groups the _ crystallographic _ coxeter systems occur as _ weyl groups _ of several structures . there , they play the role of basic invariants , often called the _ type _ of the algebraic object under consideration . in @xcite and @xcite _ weyl groupoids _ and their root systems have been introduced as basic invariants of nichols algebras . their definition emerged from earlier work in @xcite and @xcite , they are natural generalizations of weyl groups . weyl groupoids of finite type are those which generalize finite weyl groups . these have been studied intensively in @xcite , @xcite , @xcite , @xcite , where a classification has been obtained . this classification is considerably harder than the classification of the finite weyl groups . it was observed by the first two authors in 2009 that the final outcome of this classification in rank 3 is intimately related to grnbaum s list of simplicial arrangements in @xcite . there is an obvious explanation of this connection : to each weyl groupoid one can associate a tits cone which is the whole space if and only if the weyl groupoid is of finite type . based on these considerations , _ crystallographic _ arrangements have been introduced in @xcite , where it is shown that their classification is a consequence of the classification of finite weyl groupoids . in view of the importance of the tits cone and the coxeter complex for the understanding of coxeter systems it is natural to investigate their analogues in the context of weyl groupoids . although it is intuitively clear what has to be done in order to generalize these notions , there are instances where things have to be modified or some extra argument is needed . our intention is to provide the basic theory of the tits cone and the coxeter complex of a weyl groupoid . this paper deals with the combinatorial aspects of this project and therefore the crystallographic condition does nt play a role at all . hence , here we deal with _ tits arrangements _ rather than with weyl groupoids . the basic strategy for setting up the framework is to start with a simplicial arrangement @xmath5 on a convex open cone @xmath2 and to investigate the abstract simplicial complex @xmath6 associated with it . this is a _ gated chamber complex _ and has therefore a natural type function . we call a simplicial arrangement on a convex open cone a tits arrangement if @xmath7 is a thin chamber complex and introduce the notion of a root system of a tits arrangement . given a simplicial arrangement of rank @xmath8 on an open convex cone , there are two canonical procedures to produce simplicial arrangements of smaller rank . we would like to point out that this paper is meant to be a contribution to the foundations of the theory of weyl groupoids of arbitrary type . the concepts and ideas are at least folklore and several of the results for which we give proofs are well established in the literature . the only exception is probably our systematic use of gated chamber complexes at some places . our principal goal here is to provide a fairly complete account of the basic theory of the tits cone of a weyl groupoid by developing the corresponding notions to the extent that is needed for just this purpose . therefore , we have decided to include short proofs for standard facts and refer to other sources only when we need a more elaborate result . we are not able to give a systematic account of the origins of the concepts and ideas which play a role in this paper . here are some comments based on the best of our knowledge : 1 . simplicial arrangements were first introduced and studied by melchior @xcite and subsequently by grnbaum @xcite . shortly afterwards , simplicial arrangements attracted attention in the seminal work of deligne @xcite : they are a natural context to study the @xmath9 property of complements of reflection arrangements , since the set of reflection hyperplanes of a finite coxeter group is a simplicial arrangement . they further appeared as examples or counterexamples to conjectures on arrangements . 2 . we do not know where arrangements of hyperplanes on convex cones were considered for the first time . the concept seems most natural and they are mentioned in @xcite without further reference . of course , our definition of a simplicial arrangement on an open convex cone is inspired by the tits cone of a coxeter system . the fact that arrangements of hyperplanes provide interesting examples of gated sets in metric spaces appears in @xcite for the first time . at least in the simplicial case it was observed much earlier @xcite . the observation that there is a natural link between root systems and simplicial arrangements is quite natural . we already mentioned that it was our starting point to investigate the tits - cone of a weyl groupoid . but it also appears in dyer s work on rootoids @xcite , @xcite . it is conceivable that the observation was made much earlier by other people and is hidden somewhere in the literature . the paper is organized as follows . in section [ sec : ex ] we will provide some basic examples for our objects of interest . these will be used for later reference , since they are either basic examples or counterexamples for the properties which we introduce . like in example [ exm : thinf ] , the geometric representation of a coxeter group is a prominent sample for simplicial arrangements . in section [ sec : hyp ] we fix notation and develop the notion of a hyperplane arrangement on an open convex cone in a real vector space @xmath10 . we introduce the common substructures for these objects , i. e. subarrangements and restrictions . furthermore we define the chamber graph of an arrangement and show that parabolic subsets of the set of chambers are gated subsets with respect to the canonical metric . in section [ sec : tits ] we introduce simplicial arrangements and tits arrangements . we add additional combinatorial structure to a tits arrangement by associating to the set of hyperplanes a set of roots , linear forms which define the hyperplanes . with respect to the study of nichols algebras root systems with additional properties will be interesting , however in this paper we deal with roots systems as very general objects . we also associate to a simplicial arrangement a canonical simplicial complex . the main results regarding this complex are proven in the appendix . in section [ subarr ] we consider subarrangements and restrictions of simplicial arrangements and tits arrangements . we give criteria when the substructures of a simplicial / tits arrangement is again a simplicial / tits arrangement , and in the case of a tits arrangement we describe canonical root systems . the appendix provides proofs for the properties of the poset associated to a simplicial hyperplane arrangement which are stated in section 4 . while most of these properties are quite intuitive , a rigid proof can be tedious . appendix [ app : simpcomp ] recalls the basic definitions of simplicial complexes as we need them . in appendix [ app : s ] we show that the poset @xmath11 associated to a simplicial hyperplane arrangement @xmath12 is indeed a simplicial complex . the results of this section are summarized in proposition [ ssimpcomp ] . in particular , we provide an equivalent definition for simplicial cones , which has an implicit simplicial structure . appendix [ app : scc ] provides the remaining properties of @xmath13 . the first part recalls the definitions of chamber complexes and type functions . in the second part it is shown that @xmath13 is a gated chamber complex with a type function , a collection of the results can be found in proposition [ prop - sproperties ] . we also show that the notions of being spheric and thin , which we introduced for simplicial hyperplane arrangements before , coincide with the classical notions for chamber complexes . * acknowledgements . * part of the results were achieved during a mini - workshop on nichols algebras and weyl groupoids at the mathematisches forschungsinstitut oberwolfach in october 2012 , and during meetings in kaiserslautern , hannover , and giessen supported by the deutsche forschungsgemeinschaft within the priority programme 1388 . one of the authors was partly funded by a scholarship of the justus - liebig - universitt giessen . consider the following setting . let @xmath14 then @xmath15 consists of two connected components @xmath16 let @xmath17 , @xmath18 be the dual basis in @xmath19 to @xmath20 , then we can write @xmath21 we will later define objects which can be written in this way , i. e. as intersections of half spaces , as simplicial cones . both @xmath22 and @xmath23 are cones with exactly two bounding hyperplanes , for @xmath22 these are @xmath24 and @xmath25 , for @xmath23 these are @xmath26 and @xmath27 . note that in this example only one of the bounding hyperplanes , namely @xmath26 , of @xmath22 , @xmath23 meets @xmath2 , while the other meets @xmath28 , but not @xmath2 itself . . ] [ exm : notthin ] let @xmath10 and @xmath2 be as before . define for @xmath29 the lines @xmath30 then the connected components of @xmath31 are again simplicial cones for suitable linear forms . the number of connected components is not finite in this case , and every component has bounding hyperplanes which meet @xmath2 ( see figure [ fig : thinf ] ) . . ] [ exm : thinf ] let @xmath32 and @xmath33 then @xmath2 is a convex open cone . consider the universal coxeter group @xmath1 in 3 generators , i. e. @xmath34 the geometric representation ( see ( * ? ? ? * chapter 5.3 ) ) of @xmath1 yields a set of reflection hyperplanes @xmath35 , which meet @xmath2 after choosing a suitable basis . we obtain the picture in figure [ fig : univ3 ] by intersecting @xmath2 with a hyperplane parallel to @xmath36 , which corresponds to the beltrami - klein model of hyperbolic 2-space . . ] the connected components of @xmath37 are again cones which carry a simplicial structure . however , the vertices of the simplices are not contained in @xmath2 , but in its boundary @xmath38 . [ exm : univ3 ] throughout this paper , all topological properties are with respect to the standard topology of @xmath39 , unless stated otherwise . let @xmath40 . a subset @xmath41 is called a _ cone _ , if @xmath42 for all @xmath43 , @xmath44 . for a subset @xmath45 we call @xmath46 the _ cone over @xmath47_. let @xmath35 be a set of linear hyperplanes in @xmath40 , and @xmath2 an open convex cone . we say that _ @xmath35 is locally finite in @xmath2 _ , if for every @xmath48 there exists a neighborhood @xmath49 of @xmath50 , such that @xmath51 is a finite set . a _ hyperplane arrangement ( of rank @xmath8 ) _ is a pair @xmath12 , where @xmath2 is a convex open cone in @xmath10 , and @xmath35 is a ( possibly infinite ) set of linear hyperplanes such that 1 . @xmath52 for all @xmath53 , 2 . @xmath35 is locally finite in @xmath2 . if @xmath2 is unambiguous from the context , we also call the set @xmath35 a hyperplane arrangement . let @xmath54 . then the _ support of @xmath47 _ is defined as @xmath55 if @xmath56 is a singleton , we write @xmath57 instead of @xmath58 , and we omit the index @xmath35 , if @xmath35 is unambiguous from the context . in this paper we call the set @xmath59 the _ section of @xmath47 ( in @xmath35)_. again , we will omit @xmath35 when there is no danger of confusion . the connected components of @xmath37 are called _ chambers _ , denoted with @xmath60 or just @xmath61 , if @xmath62 is unambiguous . let @xmath63 . define the _ walls of @xmath64 _ as @xmath65 define the _ radical of @xmath35 _ as @xmath66 . we call the arrangement _ non - degenerate _ , if @xmath67 , and _ degenerate _ otherwise . a hyperplane arrangement is _ thin _ , if @xmath68 for all @xmath69 . 1 . by construction , the chambers @xmath70 are open sets . 2 . in our notation , if @xmath62 is a hyperplane arrangement , @xmath35 being locally finite in @xmath2 implies : for every point @xmath48 there exists a neighborhood @xmath49 of @xmath50 such that @xmath71 is finite . let @xmath12 be a hyperplane arrangement . then for every point @xmath48 there exists a neighborhood @xmath72 such that @xmath73 . furthermore the set @xmath74 is finite for every compact set @xmath75 . [ compfin ] let @xmath48 , in particular there exists an open neighborhood @xmath76 of @xmath50 such that @xmath77 is finite . by taking the smallest open @xmath78-ball contained in @xmath76 and centred at @xmath50 , we can assume @xmath79 . let @xmath80 , with @xmath81 . let @xmath82 , then @xmath83 , and @xmath84 is an open subset such that @xmath85 . since @xmath77 is finite , @xmath86 is finite . we can therefore repeat this process finitely many times until we find an open ball @xmath87 such that @xmath88 . the second assertion is a consequence of the first : let @xmath47 be compact , and for @xmath89 let @xmath72 denote an open subset such that @xmath90 . then @xmath91 as @xmath92 . but the @xmath72 are open and @xmath47 is compact , therefore there exists a finite set @xmath93 , @xmath94 , such that @xmath95 consequently , if @xmath96 , then there exists an index @xmath97 such that @xmath98 and @xmath99 . hence @xmath100 and @xmath74 is finite . we want to say more about hyperplane arrangements arising as @xmath101 for points @xmath102 . let @xmath103 be a hyperplane arrangement and let @xmath102 . if @xmath104 , we call @xmath105 a _ subarrangement of @xmath62_. define @xmath106 , we call @xmath107 the _ induced arrangement at @xmath50 _ or the _ parabolic subarrangement at @xmath50_. a _ parabolic subarrangement of @xmath62 _ is a subarrangement @xmath108 of @xmath62 , such that @xmath109 for some @xmath102 . furthermore set @xmath110 . let @xmath111 be a connected metric space . define the segment between @xmath112 to be @xmath113 when referring to the metric of @xmath39 , we will use the more common notion of the _ interval _ between two points @xmath50 and @xmath114 : @xmath115 & : = \sigma(x , y ) = \{\lambda x + ( 1 - \lambda ) y \in { { \mathbb{r}}}^r \mid \lambda \in [ 0,1 ] \},\\ ( x , y ) & : = \sigma(x , y)\setminus \{x , y\ } = \{\lambda x + ( 1 - \lambda ) y \in { { \mathbb{r}}}^r \mid \lambda \in ( 0,1 ) \}.\end{aligned}\ ] ] note that the intervals @xmath116 $ ] , @xmath117 can be defined analogously . note that every hyperplane @xmath118 separates @xmath10 into half - spaces . one way to describe the half - spaces uses linear forms . choose an arbitrary linear form @xmath119 such that @xmath120 . then @xmath121 and @xmath122 ( @xmath123 and @xmath124 ) are the two _ open ( closed ) half - spaces bounded by @xmath118_. for an arbitrary subset @xmath75 , if @xmath47 is contained in one open half space bounded by @xmath118 , we denote this particular half - space by @xmath125 . in this case we write @xmath126 for the unique half - space not containing @xmath47 . by definition every chamber @xmath64 is contained in a unique half space of @xmath53 , therefore the sets @xmath127 exist for all @xmath53 , @xmath128 . let @xmath129 be chambers , we say that @xmath53 _ separates _ @xmath64 and @xmath26 , if @xmath130 . we also say that two closed chambers @xmath131 are _ separated _ by @xmath53 , if @xmath118 separates @xmath64 and @xmath26 . we denote by @xmath132 the set of hyperplanes separating @xmath129 . if @xmath133 , @xmath134 , then @xmath135 . [ starsep ] assume @xmath136 , then the half - space @xmath137 is well defined and unique . as a consequence we have @xmath138 , and @xmath118 does not separate @xmath64 and @xmath26 . the pair @xmath139 is a hyperplane arrangement with chambers corresponding to @xmath140 . [ lem : subarr ] the set @xmath107 is locally finite in @xmath2 since @xmath35 is locally finite in @xmath2 , and every @xmath135 meets @xmath2 . let @xmath141 denote the connected components of @xmath142 . assume @xmath143 are both contained in @xmath144 . then there exists a hyperplane @xmath53 such that @xmath22 , @xmath23 are contained in different half - spaces with respect to @xmath118 , in contradiction to lemma [ starsep ] . since @xmath145 , every chamber in @xmath140 is therefore contained in a unique chamber in @xmath141 . likewise , every chamber in @xmath146 contains a chamber in @xmath147 , which completes the proof . we note that it is always possible to mod out the radical of an arrangement to obtain a non - degenerate one . assume that @xmath62 is a hyperplane arrangement , set @xmath148 and 1 . @xmath149 , 2 . @xmath150 , 3 . @xmath151 , 4 . @xmath152 . note that the set @xmath107 is clearly finite if @xmath48 . the set @xmath153 is an open convex cone in @xmath154 . the set @xmath155 is open as the image of an open set and @xmath156 is open . furthermore , if @xmath157 $ ] is an interval in @xmath155 , then there exist @xmath158 with @xmath159 , @xmath160 . the interval @xmath161 $ ] is contained in @xmath2 as @xmath2 is convex , and we find @xmath162 ) = [ y , z]$ ] . finally , @xmath153 is a cone , since for @xmath163 we find @xmath164 with @xmath165 . since @xmath2 is a cone , for every @xmath166 we have @xmath167 , so @xmath168 . we gather basic properties of @xmath169 . the pair @xmath170 is a non - degenerate hyperplane arrangement with chambers @xmath171 . the chambers are furthermore in one to one correspondence to @xmath70 . [ lem : red ] the set @xmath169 is a set of hyperplanes in @xmath154 , since @xmath172 for all @xmath53 . let @xmath173 with @xmath174 for @xmath53 , then @xmath175 , since @xmath52 . we show that @xmath169 is locally finite in @xmath153 . let @xmath176 , then there exists an @xmath164 such that @xmath177 . since @xmath35 is locally finite in @xmath2 , there exists a neighborhood @xmath178 containing @xmath179 such that @xmath180 is finite . as @xmath156 maps open sets to open sets , @xmath181 is a neighborhood of @xmath114 in @xmath153 . now let @xmath182 for some @xmath183 . let @xmath173 with @xmath184 and assume @xmath185 . then there exists @xmath186 with @xmath187 and @xmath188 with @xmath189 , thus @xmath190 for some @xmath191 . hence @xmath192 as well , and we can conclude that @xmath193 . we have thus established for @xmath53 @xmath194 therefore @xmath195 is finite . for the chambers , note that @xmath196 is a connected component of @xmath197 , in particular , as @xmath198 , @xmath199 , furthermore for a connected component @xmath146 of @xmath197 , @xmath200 is a connected component of @xmath37 . this completes the proof . let @xmath62 be a hyperplane arrangement , @xmath102 . then set @xmath201 , @xmath202 , @xmath203 , @xmath204 , @xmath205 . the pair @xmath206 is a non - degenerate hyperplane arrangement with chambers @xmath207 . [ cor : redsub ] this follows from lemma [ lem : subarr ] and lemma [ lem : red ] , since @xmath208 . the hyperplane arrangement in example [ exm : notthin ] is degenerate , as @xmath209 . reducing this arrangement yields the non degenerate and thin hyperplane arrangement @xmath210 . [ exm : deg ] we now consider the structure given by the chambers and their adjacency . let @xmath62 be a hyperplane arrangement . we call two chambers @xmath211 _ adjacent _ , if @xmath212 is a hyperplane . if @xmath129 are are adjacent and @xmath213 , then we also call @xmath129 to be _ adjacent by @xmath118_. define the _ chamber graph _ @xmath214 to be the simplicial graph with vertex set @xmath70 , @xmath215 is an edge if and only if @xmath64 and @xmath26 are adjacent . we call a path in @xmath216 connecting @xmath217 a _ gallery from @xmath64 to @xmath26_. we say a gallery from @xmath64 to @xmath26 is _ minimal _ , if it is of length @xmath218 , where @xmath219 is the distance in @xmath216 . let @xmath220 , @xmath221 , @xmath222 . then @xmath223)$ ] , and @xmath224 is finite . [ finsep ] as @xmath2 is convex , the line @xmath225 $ ] is contained in @xmath2 , and is compact . therefore , as @xmath35 is locally finite , by lemma [ compfin ] @xmath226)$ ] is finite . let @xmath53 and assume @xmath227 $ ] is empty , then @xmath228)$ ] is well - defined and @xmath229 ) = d_h(l)$ ] , so @xmath118 does not separate @xmath64 and @xmath26 . we can conclude @xmath230)$ ] . on the other hand , let @xmath231)$ ] . since @xmath221 , @xmath222 , the interval @xmath225 $ ] is not contained in @xmath118 and neither @xmath232 nor @xmath233 is in @xmath118 . assume @xmath234 , since @xmath235 is convex , we obtain @xmath225 \subset d_h(\{v\})$ ] , contradicting our assumption . therefore @xmath236 . this yields @xmath226 ) \subseteq s(k , l)$ ] . two chambers @xmath220 are adjacent by @xmath118 if and only if @xmath237 . in this situation @xmath26 is the unique chamber adjacent to @xmath64 by @xmath118 . [ adjsep ] assume first that @xmath129 are adjacent by @xmath118 . so @xmath238 . note that @xmath239 by definition . since @xmath240 , the set @xmath241 is a convex cone spanning @xmath118 . take a point @xmath50 in the interior of @xmath242 . for a suitable @xmath243 we find @xmath244 and @xmath245 . the hyperplane @xmath118 separates @xmath246 into two connected components , say @xmath247 and @xmath248 . since @xmath64 is open , @xmath249 , w. l. o. g. @xmath64 meets @xmath247 . since @xmath245 , indeed we find @xmath250 . assuming that @xmath26 meets @xmath247 implies that @xmath251 , but then @xmath252 generates @xmath10 , contradicting the assumption of @xmath64 and @xmath26 being adjacent . thus @xmath253 , and @xmath118 separates @xmath64 and @xmath26 . assume @xmath254 such that @xmath255 separates @xmath64 and @xmath26 . then @xmath256 and @xmath257 , in particular @xmath258 . this implies @xmath259 , therefore @xmath260 . for the other implication assume that @xmath118 is the unique hyperplane separating @xmath64 and @xmath26 . in this case , note that @xmath64 and @xmath26 are distinct . if @xmath118 is not in @xmath261 , @xmath64 and @xmath26 are on the same side of every wall of @xmath64 , thus equal , a contradiction . therefore @xmath239 . consider @xmath221 and @xmath222 . the interval @xmath225 $ ] meets only @xmath118 by lemma [ finsep ] . therefore @xmath225 \cap h = \{p\}$ ] , with @xmath262 . since @xmath35 is locally finite , by lemma [ compfin ] we find a neighborhood @xmath76 of @xmath263 such that @xmath264 . again @xmath76 is separated by @xmath118 into two connected components , one contained in @xmath64 , one in @xmath26 . therefore @xmath265 is contained in @xmath266 and this set generates @xmath118 . thus @xmath64 and @xmath26 are adjacent . for the second statement assume @xmath267 is a chamber adjacent to @xmath64 by @xmath118 . then we established @xmath268 . if @xmath269 , there exists a hyperplane @xmath255 separating @xmath26 and @xmath267 , and we can assume @xmath270 and @xmath271 . then @xmath255 separates @xmath64 and @xmath267 , which implies @xmath259 . this contradicts the fact that @xmath270 . let @xmath69 , @xmath272 . then there exists a chamber @xmath26 adjacent to @xmath64 by @xmath118 . [ lem : adjex ] take a point @xmath50 in the interior of @xmath273 with respect to @xmath118 and a neighborhood @xmath76 of @xmath50 contained in @xmath2 with @xmath264 . since @xmath53 , @xmath274 is contained in a chamber @xmath26 , which is adjacent to @xmath64 by @xmath118 . in the case where @xmath62 is not thin , we need to be able to handle hyperplanes which are not in @xmath35 but occur as a wall of a chamber . assume @xmath69 , @xmath240 and @xmath275 . then @xmath276 . in particular for all @xmath220 the set @xmath127 is well defined and @xmath277 holds . [ boundwalls ] assume @xmath278 , then @xmath118 separates @xmath2 into two half spaces . as @xmath118 is a wall of @xmath64 and @xmath275 , we find that @xmath64 is not a connected component of @xmath279 . therefore @xmath276 , which proves our claim . let @xmath217 . then there exists a minimal gallery of length @xmath224 connecting @xmath64 and @xmath26 . [ sepisdis ] by lemma [ finsep ] @xmath280 is finite , so let @xmath281 . for @xmath282 , we have @xmath277 for every @xmath283 and all @xmath284 by lemma [ boundwalls ] . this implies that @xmath64 and @xmath26 are the same connected component . let @xmath285 , then we obtain the statement from lemma [ adjsep ] . we now use induction on @xmath286 , let @xmath287 . we can sort @xmath280 in a way , such that @xmath288 . by lemma [ lem : adjex ] and lemma [ adjsep ] there exists a unique chamber @xmath146 adjacent to @xmath64 such that @xmath289 . we will show @xmath290 , then the statement of the lemma follows by induction . since we have @xmath291 , we have @xmath292 for @xmath293 . assuming that there exists an additional @xmath254 , @xmath294 for @xmath295 with the property that @xmath255 separates @xmath146 and @xmath26 , provides a contradiction as @xmath255 also separates @xmath64 and @xmath26 . by induction this yields a gallery of length @xmath286 from @xmath64 to @xmath26 . we show that this gallery is also minimal . for this purpose let @xmath296 be an arbitrary gallery connecting @xmath129 and let for @xmath297 the hyperplane @xmath298 such that @xmath299 . note that @xmath300 is unique with that property by lemma [ adjsep ] . assume @xmath134 , then there exists an index @xmath301 such that @xmath118 separates @xmath302 and @xmath26 but not @xmath303 and @xmath26 . so @xmath304 and @xmath305 , @xmath118 separates @xmath306 and @xmath303 . we obtain @xmath307 , as @xmath308 are adjacent by @xmath300 and @xmath300 is unique with that property by lemma [ adjsep ] . thus @xmath309 and this yields @xmath310 . above we constructed a gallery of length @xmath224 , hence the gallery is already minimal . the graph @xmath216 is connected . [ cor : gammacon ] let @xmath311 be a connected metric space . let @xmath312 and @xmath313 . a point @xmath314 is called a _ gate of @xmath50 to @xmath315 _ or _ the projection of @xmath50 on @xmath315 _ if @xmath316 for all @xmath317 . the gate of @xmath50 to @xmath315 is uniquely determined , if it exists , and we will then denote it by @xmath318 . the set @xmath315 is called gated if every @xmath312 has a gate to @xmath315 . let @xmath102 . define the graph @xmath319 , where @xmath320 . then @xmath321 is a subgraph of @xmath216 and is called the _ parabolic subgraph at @xmath50_. a subgraph @xmath322 of @xmath216 is called _ parabolic _ , if it is a parabolic subgraph at @xmath50 for some @xmath102 . the parabolic subgraphs and subarrangements correspond to the residues in the case where @xmath35 is a hyperplane arrangement coming from a coxeter group . parabolic subgraphs are connected . [ kxcon ] the graph @xmath321 corresponds to the chamber graph of the hyperplane arrangement @xmath139 , and is therefore connected by corollary [ cor : gammacon ] . let @xmath102 , @xmath69 . then there exists a unique chamber @xmath323 , such that @xmath324 for all @xmath135 . [ starunique ] we prove existence first . let @xmath69 and consider the intersection @xmath325 then @xmath3 is a chamber of @xmath326 , since it is a non - empty connected component of @xmath2 . by lemma [ lem : subarr ] @xmath3 corresponds to a unique chamber @xmath327 in @xmath140 , and @xmath328 holds . so let @xmath329 be another chamber in @xmath140 . then there exists an element @xmath53 such that @xmath118 separates @xmath327 and @xmath330 . by lemma [ starsep ] we obtain @xmath135 , but this implies @xmath331 . hence @xmath327 is unique with the property that @xmath324 for all @xmath135 . for @xmath332 the set @xmath140 are gated subsets of @xmath333.[stargate ] let @xmath102 and @xmath321 be a parabolic subgraph . let @xmath69 . by lemma [ starunique ] there exists a unique chamber @xmath334 such that @xmath335 for all @xmath135 . we will prove that @xmath336 is a gate of @xmath64 to @xmath140 . so let @xmath337 . if @xmath53 separates @xmath336 and @xmath26 , we immediately get by lemma [ starsep ] that @xmath135 . on the other hand we get that if @xmath254 separates @xmath336 and @xmath64 , by construction of @xmath336 we find @xmath338 . now assume @xmath339 , we find that since @xmath135 , @xmath118 does not separate @xmath336 and @xmath64 and therefore must separate @xmath336 and @xmath26 . assume on the other hand that @xmath340 , then it can not separate @xmath26 and @xmath336 , and therefore must separate @xmath336 and @xmath64 . summarized this yields @xmath341 we thus find for all @xmath337 that @xmath342 . so @xmath336 is indeed a gate for @xmath64 to @xmath321 . we will now consider the structure induced on hyperplanes @xmath118 by the elements in @xmath35 . let @xmath62 be a simplicial arrangement . let @xmath343 be a hyperplane in @xmath10 , such that @xmath52 . set @xmath344 this is a set of hyperplanes in @xmath118 which have non empty intersection with @xmath345 , if @xmath345 is not empty itself . we also define the connected components of @xmath346 as @xmath347 . the pair @xmath348 is a hyperplane arrangement of rank @xmath349 with chambers @xmath347 . by definition for @xmath350 we find @xmath351 . the set @xmath345 is an open convex cone in @xmath118 , which is isomorphic to @xmath352 . let @xmath353 , @xmath76 a neighborhood of @xmath50 such that @xmath77 is finite . then @xmath354 is also finite , hence @xmath355 is locally finite in @xmath345 . by definition @xmath347 are the chambers of this arrangement . in the case where @xmath53 , we need some basic relations between chambers in @xmath70 and in @xmath347 . assume @xmath53 and let @xmath69 . if @xmath240 , then @xmath273 is the closure of a chamber in @xmath347 . for every @xmath356 there exists exactly two chambers @xmath357 , such that @xmath358 and @xmath359 for @xmath360 . [ indfaces ] for the first statement , @xmath273 is contained in the closure of a chamber @xmath146 of @xmath348 , since its interior with respect to @xmath118 does not meet any hyperplane . let @xmath361 . then @xmath362 for some chamber let @xmath364 . assume @xmath365 , then @xmath366 , while @xmath273 is either contained in @xmath255 or in @xmath367 . a contradiction to @xmath368 . thus we find @xmath369 . since @xmath370 , there exists a wall of @xmath64 separating @xmath64 and @xmath50 . a contradiction , hence @xmath371 . for the second statement , take @xmath368 , then @xmath372 and @xmath373 . choose a neighborhood @xmath76 of @xmath47 such that @xmath264 , which is possible by lemma [ compfin ] . the hyperplane @xmath118 separates @xmath76 into two connected components , which are each contained in chambers @xmath374 . by definition @xmath22 and @xmath23 are adjacent by @xmath118 , and by lemma [ adjsep ] there can not exist another chamber being adjacent to @xmath22 or @xmath23 by @xmath118 . in this section we introduce simplicial cones and collect some basic properties . as before , throughout this section let @xmath40 . let @xmath119 be a linear form , then @xmath375 let @xmath87 be a basis of @xmath19 , then the _ open simplicial cone ( associated to @xmath87 ) _ is @xmath376 with notation as above we find @xmath377 let @xmath87 be a basis of @xmath19 . we can then define the _ closed simplicial cone ( associated to @xmath87 ) _ as @xmath378 we say a cone is _ simplicial _ if it is open simplicial or closed simplicial . a simplicial cone can also be defined using bases of @xmath10 . let @xmath4 be a basis of @xmath10 , then the open simplicial cone associated to @xmath4 is @xmath379 and the closed simplicial cone associated to @xmath4 is @xmath380 both concepts are equivalent , and it is immediate from the definition that if @xmath381 and @xmath382 are bases , then @xmath383 if and only if @xmath87 is , up to positive scalar multiples and permutation , dual to @xmath4 . a simplicial cone @xmath64 associated to @xmath87 carries a natural structure of a simplex , to be precise : @xmath384 is a poset with respect to inclusion , which is isomorphic to @xmath385 with inverse inclusion . if @xmath4 is the basis of @xmath10 dual to @xmath87 , we find @xmath386 to be the set of all convex combinations of subsets of @xmath4 , and @xmath386 is also isomorphic to @xmath387 . moreover , @xmath388 is the vertex set of the simplex @xmath386 . for a simplicial cone @xmath64 , we denote with @xmath389 a basis of @xmath19 such that @xmath390 . [ rem : scs ] an open or closed simplicial cone is convex and has non - empty interior . [ cone : simpprop1 ] let @xmath64 be an open simplicial cone and @xmath389 a basis associated to @xmath64 . then @xmath64 is convex and open as an intersection of convex and open subsets . furthermore let @xmath4 be dual to @xmath391 , then we obtain that @xmath392 , hence @xmath64 is not empty and thus has non - empty interior . if @xmath2 is a closed simplicial cone , it contains the closure of an open simplicial cone , and has non - empty interior . the common notation for cones introduces _ properness _ of a cone @xmath64 as the property of having non - empty interior and being closed , convex , and pointed , the latter meaning that @xmath393 . thus all closed simplicial cones are proper . in our context being proper is not of interest , the cones we are dealing with are either convex and open or already simplicial . in the following let @xmath40 and @xmath394 be an open convex cone . in this section we establish a notion of tits arrangements on @xmath2 . the interesting cases the reader may think of are @xmath395 , or a half - space @xmath396 for some @xmath119 . we can now define our main objects of interest . let @xmath394 be a convex open cone and @xmath35 a set of linear hyperplanes in @xmath10 . we call a hyperplane arrangement @xmath12 a _ simplicial arrangement ( of rank @xmath8 ) _ , if every @xmath397 is an open simplicial cone . the cone @xmath2 is the _ tits cone _ of the arrangement . a simplicial arrangement is a _ tits arrangement _ , if it is thin . 1 . the definition of `` thin '' requires that all possible walls are already contained in @xmath35 . if we do not require this , a bounding hyperplane may arise as a bounding hyperplane of @xmath2 itself . this happens in example [ exm : notthin ] , which is not thin . consider also the case where @xmath2 itself is an open simplicial cone . even the empty set then satisfies that the one chamber , which is @xmath2 itself , is a simplicial cone . however , it has no walls in @xmath35 , hence it is not thin . 2 . from the definition it follows that in dimensions @xmath398 and @xmath399 , there are very few possibilities for simplicial arrangements and tits arrangements . + in dimension @xmath398 , the empty set is the only possible tits arrangement . + in the case @xmath400 , we find @xmath401 or @xmath402 or @xmath403 . in the first case , @xmath404 is a tits arrangement and in the other two cases , @xmath404 is a simplicial arrangement , but not thin any more . 3 . later in this section we introduce the notion of @xmath405-spherical arrangements , which is a refinement of being thin . 4 . in the case where @xmath395 , the property of @xmath35 being locally finite is equivalent to @xmath35 being finite , as @xmath398 is contained in every hyperplane . furthermore @xmath395 is the only case where @xmath406 , as the cone over every neighborhood of @xmath398 is already @xmath39 . \(i ) consider example [ exm : notthin ] , there @xmath22 and @xmath23 are simplicial cones , with @xmath407 and @xmath408 . the set @xmath409 is therefore a simplicial arrangement of rank 2 in @xmath2 , which is finite and hence also locally finite . but @xmath410 is not thin , as @xmath411 note that this particular example can be turned into a thin arrangement by factoring out the radical ( cp . example [ exm : deg ] ) . however , one can easily find examples of non - degenerate simplicial arrangements which are not thin . \(ii ) in example [ exm : thinf ] , the set @xmath412 of hyperplanes is not finite , but it is locally finite in @xmath2 . if @xmath413 , we can not find a neighborhood @xmath76 of @xmath414 which meets only finitely many hyperplanes if and only if @xmath415 or @xmath416 , and hence @xmath417 . a chamber @xmath64 satisfies one of the following : 1 . @xmath418 for some @xmath419 , or 2 . @xmath420 for some @xmath419 . therefore , @xmath421 is a tits arrangement of rank 2 . likewise , the hyperplane arrangement in example [ exm : univ3 ] is a tits arrangement of rank 3 in @xmath2 . the vertices of the simplicial cones are contained in @xmath38 , but all walls meet @xmath2 . \(iii ) figure ( [ extits ] ) shows a tits arrangement coming from the weyl groupoid of an infinite dimensional nichols algebra . tits arrangements are non - degenerate . [ lem : titsnd ] let @xmath69 , @xmath87 a basis of @xmath19 such that @xmath422 . since @xmath68 and @xmath87 is a basis , we find @xmath423 . some of the `` classical '' cases are the following choices for @xmath2 . a simplicial arrangement @xmath62 is _ spherical _ , if @xmath395 . we say @xmath62 is _ affine _ , if @xmath424 for some @xmath425 . for an affine arrangement we call @xmath426 the _ imaginary root _ of the arrangement . the tits cone of a tits arrangement @xmath62 resembles the tits cone for compact hyperbolic coxeter groups . the geometric representation of an irreducible spherical or affine coxeter group is a prototype of spherical or affine tits arrangements . let @xmath40 , a _ root system _ is a set @xmath427 such that 1 . @xmath428 , 2 . @xmath429 for all @xmath430 , 3 . there exists a tits arrangement @xmath62 such that @xmath431 . if @xmath432 is a root system , @xmath62 as in 3 ) , we say that the tits arrangement @xmath12 is _ associated to @xmath432_. let @xmath432 be a root system . we call a map @xmath433 a _ reductor of @xmath432 _ , if for all @xmath430 1 . @xmath434 for some @xmath435 , 2 . @xmath436 . a root system @xmath432 is _ reduced _ , if @xmath437 is a reductor . given @xmath432 and a reductor @xmath438 , when no ambiguity can occur , we denote @xmath439 . we note some immediate consequences of this definition . let @xmath432 be a root system and @xmath62 be a tits arrangement associated to @xmath432 , @xmath438 a reductor of @xmath432 . then 1 . @xmath440 is a reduced root system and @xmath62 is associated to @xmath432 . 2 . @xmath432 is reduced if and only if @xmath441 for all @xmath442 . if @xmath432 is reduced , @xmath437 is the only reductor of @xmath432 . for i ) , as @xmath443 , we find that @xmath444 , so @xmath445 is a root system . due to the properties of @xmath438 , we find that @xmath446 , so @xmath437 is a reductor of @xmath432 . for the second statement assume @xmath441 for all @xmath430 , then the identity is a reductor . assume that the identity is a reductor , then for @xmath447 with @xmath448 , @xmath449 , we find that @xmath450 . for the third statement assume that @xmath432 is reduced , so @xmath437 is a reductor of @xmath451 . let @xmath430 , then @xmath441 . if @xmath438 is a reductor of @xmath432 , @xmath452 , as @xmath453 is positive , we find @xmath454 and @xmath455 . in the case where for @xmath430 the set @xmath456 is finite , the canonical choice for @xmath438 is such that @xmath457 is minimal in @xmath458 ; with respect to an arbitrary scalar product . the notion of a reductor of @xmath432 is very general , the intention is to be able to reduce a root system even in the case where for @xmath430 the set @xmath456 does not have a shortest or longest element . in this most general setting , the existence of a reductor requires the axiom of choice . furthermore we find that , given a tits arrangement @xmath62 , root systems always exist . let @xmath12 be a tits arrangement , @xmath459 the unit sphere in @xmath39 with respect to the standard metric associated to the standard scalar product @xmath460 . let @xmath461 then @xmath462 is a reduced root system for @xmath35 . furthermore every reduced root system @xmath463 associated to @xmath12 is of the form @xmath464 , and every such set is a reduced root system associated to @xmath12 . [ rsx ] as @xmath53 is @xmath349-dimensional , @xmath465 , so @xmath466 for some vector @xmath467 , with @xmath468 . so @xmath451 is a root system associated to @xmath35 . let @xmath469 for some @xmath449 , then @xmath470 . as the map @xmath471 is bijective , we find @xmath472 . therefore @xmath432 is reduced . for the second statement note that the hyperplane @xmath283 determines @xmath473 uniquely up to a scalar . so any @xmath463 of the given form satisfies @xmath474 . furthermore it is reduced , as @xmath475 for @xmath430 . let @xmath12 be a tits arrangement associated to the root system @xmath432 , and fix a reductor @xmath438 of @xmath432 . let @xmath64 be a chamber . the _ root basis of @xmath64 _ is the set @xmath476 if @xmath477 is a tits arrangement associated to the root system @xmath432 , @xmath69 , then @xmath478 also , as a simplicial cone @xmath479 has exactly @xmath8 walls , the set @xmath480 is a basis of @xmath19 . furthermore , @xmath481 . the following proposition is crucial for the theory and motivates the notion of root bases . let @xmath62 be a tits arrangement associated to @xmath432 , @xmath64 a chamber . then @xmath482 . in other words , every root is a non - negative or non - positive linear combination of @xmath480 . [ posorneg ] the proof works exactly as in the spherical case , see ( * ? ? ? * lemma 2.2 ) . it is useful to identify tits arrangements which basically only differ by the choice of basis , therefore we make the following definition . let @xmath62 , @xmath483 be tits arrangements associated to the root systems @xmath432 and @xmath463 respectively . then @xmath62 and @xmath484 are called _ combinatorially equivalent _ , if there exists an @xmath485 such that @xmath486 , @xmath487 , @xmath488 . here @xmath489 denotes the dual action of @xmath490 on @xmath19 , defined by @xmath491 . let @xmath12 be a simplicial arrangement . the set of chambers @xmath61 gives rise to a poset @xmath492 with set - wise inclusion giving a poset - structure . note that we do not require any of these intersections to be in @xmath2 . by construction they are contained in the closure of @xmath2 , as every @xmath64 is an open subset in @xmath2 . we will at this point just note that @xmath13 is a simplicial complex . this will be elaborated in appendix [ app : s ] . the poset @xmath493 is a simplicial complex . [ ssimpcomp ] the complex @xmath493 is furthermore a chamber complex , which justifies the notion of chambers and is shown in appendix [ app : scc ] . we have a slight ambiguity of notation at this point , as a chamber in the simplicial complex @xmath493 is the closure of a chamber in @xmath70 . for the readers convenience , a chamber will always be an element @xmath69 , while we will refer to a chamber in @xmath493 as a _ closed chamber _ , written with the usual notation @xmath494 . we show in appendix [ app : scc ] that the closed chambers are indeed chambers in a classical sense . 1 . depending on @xmath2 , the above mentioned simplicial complex can also be seen as canonically isomorphic to the simplicial decomposition of certain objects , arising by intersection with the respective simplicial cones . in case @xmath395 , @xmath13 corresponds to a simplicial decomposition of the sphere @xmath495 . if @xmath396 for @xmath119 , we find @xmath13 to be a decomposition of the affine space @xmath496 , which we identify with the set @xmath497 . if @xmath2 is the light cone and @xmath35 is ch - like as defined in definition [ def : spherical ] below , we can find a corresponding decomposition of @xmath498 . 2 . in the literature ( see ( * ? ? ? * chapter v , 1 ) ) the simplicial complex associated to a finite or affine simplicial hyperplane arrangement is defined in a slightly different manner . let @xmath10 be a euclidean space , @xmath35 a locally finite set of ( possibly affine ) hyperplanes . define for @xmath499 , @xmath500 the set @xmath501 to be the halfspace with respect to @xmath118 containing @xmath232 . set @xmath502 then @xmath503 is an equivalence relation , its classes are called facets . facets correspond to simplices , and form a poset with respect to the inclusion @xmath504 . + one obtains immediately that every point in @xmath10 is contained in a unique facet . however , when the space is not the entire euclidean space but a convex open cone , it is desirable to consider some of the points in the boundary , as they contribute to the simplicial structure of @xmath13 . therefore , we prefer our approach before the classical one . 3 . the above approach could also be used to define being @xmath405-spherical ( see below ) for hyperplane arrangements , which are not necessarily simplicial . using the simplicial complex @xmath13 , it is now possible to refine the notion of being thin for a simplicial arrangement . a simplicial arrangement @xmath62 of rank @xmath8 is called _ @xmath405-spherical _ for @xmath505 if every simplex @xmath3 of @xmath493 , such that @xmath506 , meets @xmath2 . we say @xmath12 is _ ch - like _ if it is @xmath349-spherical . [ def : spherical ] immediate from the definition are the following : 1 . a simplicial hyperplane arrangement of rank @xmath8 is @xmath398-spherical , as @xmath69 is constructed as an open subset of @xmath2 . the hyperplane arrangement @xmath62 is spherical if and only if it is @xmath8-spherical . @xmath12 is thin and therefore a tits arrangement if and only if it is @xmath399-spherical . 4 . as @xmath493 is a simplicial complex w. r. t. inclusion , being @xmath405-spherical implies being @xmath507-spherical for @xmath508 . 5 . examples of ch - like arrangements are all arrangements belonging to affine weyl groups , where the affine @xmath349-plane is embedded into a real vector space of dimension @xmath8 , as well as all arrangements belonging to compact hyperbolic coxeter groups , where @xmath2 is the light cone . the notion ch - like is inspired by exactly this property for groups , i. e. being * * c*ompact * h*yperbolic . 6 . as a generalization of ( 5 ) , take a coxeter system @xmath0 of finite rank . then @xmath1 is said to be @xmath405-spherical if every rank @xmath405 subset of @xmath3 generates a finite coxeter group . the geometric representation of @xmath1 then yields a hyperplane arrangement which is @xmath405-spherical in the way defined above . therefore , being @xmath405-spherical can be seen as a generalization of the respective property of coxeter groups . an equivalent condition for @xmath62 to be @xmath405-spherical , which we will use often , is that every @xmath509-simplex meets @xmath2 . this uses just the fact that simplices of codimension @xmath405 are exactly @xmath509-simplices . the arrangement in example [ exm : notthin ] yields the simplicial complex @xmath13 , with maximal elements @xmath22 , @xmath23 , vertices @xmath510 , @xmath511 , @xmath512 , and the set @xmath513 as the minimal element . thus @xmath410 is @xmath398-spherical , but not @xmath399-spherical , as it is not thin . the arrangement in example [ exm : thinf ] is thin and therefore @xmath399-spherical and ch - like . the vertices are the rays @xmath514 , @xmath515 , the minimal element , which has codimension @xmath516 , is @xmath513 . since @xmath517 , it is not @xmath516-spherical and in particular not spherical . in example [ exm : univ3 ] , the arrangement @xmath12 is thin and therefore 1-spherical , but since all vertices are contained in @xmath38 , it is not 2-spherical and thus not ch - like . note that the group @xmath1 from this example is not compact hyperbolic . an important observation is the fact that the cone @xmath2 can be reconstructed from the chambers . for a simplicial hyperplane arrangement @xmath12 we find @xmath518 furthermore @xmath62 is ch - like if and only if @xmath519 holds . [ tchambers ] as @xmath520 , the inclusion @xmath521 holds . if @xmath48 , either @xmath522 for some @xmath69 , or @xmath50 is contained in a finite number of hyperplanes , thus in a simplex in @xmath13 and also in the closure of a chamber , and therefore @xmath523 . so let @xmath524 . assume further that @xmath50 is not contained in any simplex in @xmath13 , else the statement follows immediately as above . therefore @xmath525 , and as @xmath2 is convex , this means @xmath50 is in the boundary of @xmath2 . so let @xmath526 be the open @xmath527-ball with center @xmath50 , then @xmath528 . as @xmath76 is open , @xmath529 is open again . the set @xmath530 can not be empty , else @xmath531 is not contained in a chamber , in a contradiction to the construction of chambers . so let @xmath532 be the set of chambers @xmath64 with @xmath533 . if @xmath50 is not contained in the closure of @xmath534 , we find an @xmath535 such that the open ball @xmath246 does not intersect any @xmath536 . but @xmath537 must again meet some chambers @xmath64 , which are then also in @xmath532 , a contradiction . so @xmath523 and equality holds . for the second statement , if @xmath62 is ch - like , note that every @xmath48 is contained in some simplex @xmath538 , so @xmath539 holds . now if @xmath540 for some @xmath69 , @xmath50 is contained in some simplex @xmath249 . now @xmath35 is ch - like , so @xmath541 meets @xmath2 . boundaries of simplices are simplices , hence the intersection @xmath542 is again a simplex in @xmath13 , being ch - like yields that this intersection is empty , which proves the other inclusion . the other direction of the second statement is immediate from the definition of being @xmath349-spherical . for a tits arrangement @xmath62 it is possible to show that we can also describe @xmath2 as the convex closure of @xmath543 , or alternatively as @xmath544.\ ] ] we will require the existence of a type function of @xmath13 ( for the definition , see definition [ def : numb ] ) , which is given by the following proposition , proven in appendix [ app : scc ] . let @xmath62 be a simplicial arrangement . the complex @xmath545 is a chamber complex of rank @xmath8 with @xmath546 the complex @xmath13 is gated and strongly connected . furthermore there exists a type function @xmath547 of @xmath13 , where @xmath548 . the complex @xmath13 is thin if and only if @xmath12 is thin , and @xmath13 is spherical if and only if @xmath62 is spherical . [ prop - sproperties ] we will now take a closer look at the relations between the bases of adjacent chambers . again the proof follows ( ? ? ? * lemma 2.8 ) closely . let @xmath62 be a tits arrangement associated to @xmath432 , and let @xmath217 be adjacent chambers . assume @xmath549 for @xmath550 . if @xmath551 , then either 1 . @xmath552 or 2 . @xmath553 . holds . [ adjcham1 ] by lemma [ posorneg ] we can assume @xmath554 is either as in case ii ) or @xmath555 with @xmath556 . using that for arbitrary @xmath557 we have @xmath558 , we get @xmath559 . we also observe that @xmath560 , therefore we get @xmath561 . by choice of @xmath17 and since the elements in @xmath480 , @xmath562 are reduced , we find @xmath563 and @xmath552 . assume for @xmath69 that @xmath480 is indexed in some way , i.e. @xmath564 . for any set @xmath565 , define the map @xmath566 by @xmath567 . set @xmath568 . for every simplex @xmath249 there exists a description of the form @xmath569 for some @xmath570 by remark [ rem : scs ] , which gives an index set @xmath571 . finally this gives rise to a type function of @xmath494 in @xmath493 , by taking the map @xmath572 . by theorem [ numbext ] the map @xmath573 yields a unique type function @xmath574 of the whole simplicial complex @xmath13 . so let @xmath363 be another chamber , then the restriction @xmath575 is a type function of @xmath576 as well . assume @xmath577 , this yields a second type function of @xmath576 in the same way we acquired a type function of @xmath494 before , @xmath578 we now call the indexing of @xmath562 _ compatible with @xmath480 _ , if @xmath579 . note that since the type function @xmath574 is unique , there is a unique indexing of @xmath562 compatible with @xmath480 . the existence of the type function and lemma [ adjcham1 ] allow us to state the following lemma , which gives another characterization root bases of adjacent chambers . assume that @xmath12 is a tits arrangement associated to @xmath432 . let @xmath220 be adjacent chambers and choose a indexing @xmath564 . let the indexing of @xmath577 be compatible with @xmath480 . assume @xmath580 for some @xmath508 . then @xmath581 . [ adjcham15 ] if @xmath582 , this is immediate from lemma [ adjcham1 ] . so let @xmath583 and assume w.l.o.g . @xmath584 . consider the type function @xmath585 of @xmath494 and @xmath586 of @xmath576 as restrictions of the unique type function @xmath574 . being compatible yields that @xmath587 . then @xmath588 holds and we get @xmath589 . we conclude that @xmath590 , and as @xmath591 we find @xmath592 . then we find @xmath593 , so @xmath594 is a linear combination of @xmath17 and @xmath595 , which proves our claim . we will note more properties about parabolic subarrangements of tits arrangements , in particular we will give suitable root systems associated to these subarrangements . remember from corollary [ cor : redsub ] that @xmath206 is a hyperplane arrangement . assume @xmath62 is a tits arrangement associated to @xmath432 . let @xmath102 , define @xmath596 . for @xmath597 set @xmath598 . with notation as above , let @xmath597 , then @xmath599 . in particular , @xmath600 is a basis of @xmath601 . [ indeqdim ] the inclusion @xmath602 holds due to @xmath603 . now the space @xmath604 is a subspace of @xmath10 and @xmath605 is a face of @xmath494 containing @xmath50 . assume @xmath541 is a face of @xmath605 containing @xmath50 . then @xmath541 is also a face of @xmath494 and has the structure @xmath606 for some @xmath607 . then by definition @xmath608 for @xmath297 , and @xmath609 , so @xmath605 is a minimal face of @xmath494 containing @xmath50 . the inclusion @xmath602 implies @xmath610 , therefore the set @xmath611 is also a face of @xmath494 , furthermore it contains @xmath50 by definition . we can conclude @xmath612 , which yields the equality . the second claim follows since @xmath613 and the elements in @xmath480 are linearly independent . note that @xmath614 defines a map @xmath615 by @xmath616 , which is well defined since @xmath617 . we will identify this map with @xmath618 and think of @xmath619 as a subset of @xmath620 . most of the time we are interested only in @xmath619 and its combinatorial properties , therefore it does not matter whether we consider these in @xmath620 or in @xmath19 . but formally the transition to @xmath621 is necessary to obtain a root system in the strict sense . we will consider @xmath622 as a topological space with respect to the topology in @xmath623 , which comes from the standard topology on @xmath39 . assume @xmath12 is a tits arrangement associated to @xmath432 , @xmath102 . the hyperplane arrangement @xmath206 is simplicial . the chambers of this arrangement correspond to @xmath140 . moreover , if @xmath48 , then @xmath206 is spherical . if @xmath624 , @xmath625 , and @xmath12 is @xmath405-spherical for some @xmath626 , then @xmath206 is @xmath627-spherical . in particular @xmath628 is a tits arrangement associated to @xmath619 . [ indatpoint ] assume @xmath625 , so @xmath629 and @xmath630 . let @xmath597 , @xmath631 denote the standard epimorphism . let @xmath141 be the connected components of @xmath632 . the set @xmath633 is a basis for @xmath622 by lemma [ indeqdim ] and we find @xmath634 by definition . denote this intersection by @xmath146 . first we show that @xmath635 . assume there exists @xmath636 such that there exist @xmath637 with @xmath638 , @xmath639 . then we find @xmath640 with the properties : @xmath641 , @xmath642 , @xmath643 for all @xmath644 . for any @xmath645 and @xmath644 we find @xmath646 and @xmath647 as @xmath648 . so for @xmath649 the points @xmath650 , @xmath651 still satisfy the above inequalities . now let @xmath652 , then @xmath653 . so choosing @xmath649 large enough we find that the points @xmath654 , @xmath655 are close enough to @xmath50 to satisfy @xmath656 , @xmath657 and @xmath658 for all @xmath659 . so @xmath660 , in contradiction to the simplicial structure of @xmath13 . by its definition , @xmath146 is a simplicial cone , the arrangement @xmath206 therefore is simplicial . it remains to determine for which @xmath405 the pair @xmath206 is @xmath405-spherical . in the case where @xmath48 , @xmath661 and @xmath662 is spherical . construct a simplex @xmath663 in the following way : since @xmath64 is a simplicial cone , there exists an closed simplex @xmath3 with the property @xmath664 ( for details , see remark [ rem : chse ] ) . let @xmath665 denote the minimal face of @xmath3 such that @xmath666 , and let @xmath667 be the vertex set of @xmath665 . then the vertices @xmath668 span a face of @xmath3 , denote this face by @xmath663 . the vertices @xmath669 are linearly independent : assume @xmath670 , then @xmath671 . furthermore note that @xmath672 is spanned by @xmath667 . so we get a linear combination of the form @xmath673 . linear independence of @xmath674 yields @xmath675 for all @xmath676 . this also gives us @xmath677 . assuming inequality , we find more than @xmath678 vertices in @xmath672 , a contradiction to @xmath630 . so @xmath679 spans indeed an @xmath680-simplex in @xmath622 . we show @xmath681 . in general for every @xmath682 there exists a unique @xmath659 such that @xmath683 by remark [ rem : scs ] , all other @xmath684 satisfy @xmath685 . since for @xmath686 the vector @xmath232 is not contained in any proper face of @xmath494 containing @xmath50 , we find a unique @xmath644 such that @xmath683 . let @xmath687 , then @xmath688 with @xmath689 for all @xmath686 . then @xmath690 for all @xmath691 , @xmath692 yields @xmath693 . therefore @xmath694 . so assume @xmath695 , then @xmath696 and there exists a @xmath697 such that @xmath698 . now there exists a unique @xmath644 such that @xmath699 , so we find @xmath700 . therefore @xmath701 holds , and therefore @xmath702 . we can conclude @xmath681 . now assume @xmath624 and @xmath35 is @xmath405-spherical . let @xmath663 be as above , then @xmath703 is isomorphic as a simplicial complex to the closed chamber @xmath704 and all simplices contained therein . a face of @xmath663 meets @xmath2 if and only if a face of @xmath705 meets @xmath621 . now @xmath663 is an @xmath680-simplex , as it is spanned by @xmath706 vertices , likewise @xmath665 is an @xmath707-simplex . let @xmath708 be a face of @xmath663 , and assume @xmath709 is an @xmath710-simplex . then @xmath711 generate an @xmath712-simplex @xmath713 of @xmath13 . as @xmath35 is @xmath405-spherical , @xmath713 therefore meets @xmath2 if @xmath714 is satisfied , or equivalently @xmath715 . under this condition also @xmath716 meets @xmath621 , we can conclude that @xmath662 is @xmath405-spherical . since @xmath662 will not be more than @xmath706-spherical , since it is an arrangement of rank @xmath706 , we find that @xmath662 is @xmath717-spherical . so as @xmath62 is thin , @xmath206 is a tits hyperplane arrangement with root system @xmath619 . since chambers @xmath597 and in @xmath635 are uniquely determined by the sets @xmath600 and @xmath718 we find that @xmath156 induces a bijection between @xmath147 and @xmath146 . 1 . the tits arrangement @xmath206 is not so much dependent on the point @xmath50 as on the subspace spanned by @xmath50 , as one gets the same arrangement for every @xmath719 , @xmath720 . this hints to the fact that one can see a simplicial arrangement as a simplicial complex in projective space . however , we will not elaborate this further . 2 . it may be worth mentioning that the chambers @xmath140 and the underlying simplicial complex correspond to the star of the smallest simplex in @xmath13 containing @xmath50 . 3 . its possible to show that for a simplicial hyperplane arrangement @xmath62 , the arrangement @xmath206 is also simplicial . however , the existence of a root system simplifies the proof . 4 . when @xmath624 , @xmath628 may become @xmath721-spherical for some @xmath722 . the reason for this is that @xmath35 not being @xmath721-spherical does not imply that every @xmath723-simplex contained in a chamber in @xmath140 does not meet @xmath2 . note that the above statements make sense if @xmath724 , this occurs if and only if either @xmath48 is in the interior of a chamber , or @xmath624 does not meet any hyperplane @xmath53 . however , in this case we have @xmath725 and the induced arrangement is the empty arrangement . this is not a problem , since in this case @xmath726 and @xmath727 , but this case is somewhat trivial . therefore the requirement @xmath728 is quite natural to make , and we will assume this from now on . another trivial case occurring can be @xmath729 , in which case @xmath730 , @xmath731 and @xmath732 . we can also give an exact criterion to when @xmath662 is a spherical arrangement : assume @xmath62 is a tits arrangement associated to @xmath432 with rank @xmath733 . let @xmath102 . then @xmath107 and @xmath619 are finite if and only if @xmath48 . in particular , a simplicial arrangement is finite if and only if it is spherical . [ axfin ] if @xmath48 , @xmath107 , @xmath619 are finite by proposition [ indatpoint ] . so assume @xmath107 , @xmath619 are finite and let @xmath102 . then also @xmath140 is a finite set , so let @xmath734 and by lemma [ kxcon ] we find @xmath735 such that @xmath736 is maximal . note that any minimal gallery between chambers in @xmath140 is already in @xmath140 , since for @xmath53 with @xmath737 we obtain @xmath738 by lemma [ starsep ] . let @xmath625 , then the @xmath706 adjacent chambers @xmath739 of @xmath146 exist , since @xmath35 is thin , and are all closer to @xmath64 than to @xmath146 . let @xmath740 be adjacent by @xmath300 to @xmath64 , we can conclude @xmath741 for @xmath742 . let @xmath76 be an open ball with center @xmath50 , and let @xmath743 , @xmath744 , then @xmath745 and @xmath746 are open as well and contained in @xmath2 . take @xmath747 , so @xmath748 and @xmath749 is in @xmath746 . we find @xmath750 , which shows our first statement . the last statement is obtained by taking @xmath751 . we finish this section with two observations . the root system @xmath432 is reduced if and only if @xmath619 is reduced for every @xmath48 . this follows immediately since @xmath619 is constructed as a subset of @xmath432 and @xmath752 , note that for @xmath614 we find @xmath753 . let @xmath62 be a tits arrangement associated to @xmath432 . let @xmath102 with @xmath728 . let @xmath754 be adjacent by @xmath17 , and @xmath755 , @xmath756 indexed compatible with @xmath480 . then @xmath757 is a bijection . [ restmap ] we know by lemma [ indeqdim ] that @xmath758 . since @xmath759 consist of linear independent vectors , we get @xmath760 . it remains to show that the map is well defined . for @xmath761 we find @xmath762 . we have @xmath763 and @xmath608 . now @xmath764 for some @xmath765 by lemma [ adjcham15 ] , so @xmath766 and we are done . in this section , we will discuss how a tits arrangement @xmath62 induces a tits arrangement on hyperplanes in @xmath35 . in the classical theory of hyperplane arrangements this is also called the restriction of an arrangement [ cp . @xcite ] . let @xmath62 be a tits arrangement associated to @xmath432 , and let @xmath118 be a hyperplane . define @xmath767 and set @xmath768 we can also define the connected components of @xmath346 as @xmath347 . note that in the case @xmath769 there exists no hyperplane which is not in the arrangement . in the case @xmath770 the set @xmath355 is just the point @xmath513 or empty . our statements will remain true in these cases , but most of the time they will be empty . in particular we will examine the case where @xmath53 , since otherwise we will not necessarily see an induced simplicial complex . an interesting special case occurs for affine arrangements with imaginary root @xmath426 and @xmath771 , since in many cases this yields as @xmath772 a root system associated to a spherical tits arrangement . with notation as above , we find @xmath773 . this follows immediately from the definition . the elements @xmath356 are simplicial cones in @xmath774 . [ indsimp ] if @xmath769 , the statement is empty , so let @xmath775 . lemma [ indfaces ] yields that for @xmath776 the set @xmath777 is actually an maximal face of some chamber @xmath69 . this is a simplicial cone by remark [ rem : scs ] . the following observation is immediate from a geometric point of view , but necessary to point out : let @xmath343 be an arbitrary hyperplane , and let @xmath430 . then we find @xmath778 , @xmath779 and @xmath780 . [ indroots ] first note that @xmath781 by definition of @xmath782 . the equalities follow immediately by considering the cases @xmath648 or @xmath783 . the above lemma immediately yields the following . @xmath784 . let @xmath62 be a tits arrangement associated to @xmath432 , @xmath69 , @xmath240 . let @xmath785 . then 1 . @xmath786 for a unique @xmath356 , 2 . @xmath787 and @xmath788 for @xmath789 . 3 . @xmath790 . [ indbases ] part i ) is clear by the definition of @xmath70 and @xmath347 , since @xmath791 is a unique maximal face of @xmath792 . for the second statement assume @xmath793 for @xmath794 . we use that @xmath795 is a maximal face of @xmath494 . the maximal faces of @xmath795 are exactly the sets of the form @xmath796 for @xmath797 , by lemma [ indroots ] we obtain that the faces can also be written as @xmath798 for @xmath789 . as the maximal face @xmath798 spans a hyperplane in @xmath118 contained in @xmath799 , we conclude @xmath787 for @xmath789 and ii ) holds . assertion iii ) is a direct consequence of lemma [ indroots ] . let @xmath62 be a @xmath405-spherical simplicial arrangement , @xmath800 and @xmath53 . then @xmath801 is a @xmath507-spherical simplicial arrangement . if @xmath348 is a tits arrangement associated to @xmath432 , @xmath772 is a root system for @xmath802 . [ indarrange ] note that @xmath772 does not contain @xmath398 by definition , and if @xmath803 , we find @xmath804 with @xmath805 , so @xmath806 and @xmath807 . by lemma [ indroots ] we know that @xmath784 and by definition we have @xmath808 . furthermore we know that the connected components in @xmath347 are simplicial cones by corollary [ indsimp ] . let @xmath356 and @xmath69 such that @xmath146 is a face of @xmath64 . let @xmath794 such that @xmath809 . by ii ) in lemma [ indbases ] we find @xmath810 , and together with iii ) in lemma [ indbases ] we find reduced roots @xmath811 with the property @xmath812 . hence @xmath355 is a simplicial arrangement in @xmath345 , and @xmath813 holds . so if @xmath355 is thin , @xmath772 is a root system for @xmath355 . now assume @xmath541 is an @xmath706-simplex in the simplicial complex @xmath814 associated to the simplicial arrangement @xmath802 . since this is a subset of @xmath493 , @xmath541 meets @xmath345 if and only if @xmath541 meets @xmath2 , and therefore @xmath801 is @xmath507-spherical if @xmath62 is @xmath405-spherical . assume that @xmath348 is a tits arrangement associated to @xmath772 . let @xmath69 such that @xmath815 . then the set @xmath816 is a basis of @xmath817 . [ projbasis ] by proposition [ indarrange ] the set @xmath718 is a basis of @xmath817 , by lemma [ indbases ] iii ) we find that the elements in @xmath818 are non zero scalar multiples of @xmath718 . if @xmath12 is a simplicial arrangement , we call @xmath819 for @xmath53 as above the _ induced simplicial hyperplane arrangement ( by @xmath62 ) on @xmath118 _ or the _ restriction of @xmath62 to @xmath118_. the notation in this chapter is mostly taken from ( * ? ? ? * appendix a ) and @xcite . we recall the notion of partially ordered sets : let @xmath820 be a poset . for @xmath821 we write @xmath822 let @xmath820 , @xmath823 be posets . morphism of posets _ is a map @xmath824 , such that for @xmath825 we have @xmath826 it is called an isomorphism if it is bijective and @xmath827 is a morphism as well . [ def : simpcomp ] there are two different approaches to define what a simplex is , since we want to use both , we will introduce them here . a _ simplex _ can be seen as a poset @xmath828 isomorphic to @xmath829 for some set @xmath830 , where @xmath831 denotes the set - wise inclusion . simplicial complex _ is then a poset @xmath832 such that 1 . @xmath833 is a simplex for all @xmath834 , 2 . @xmath835 have a unique greatest lower bound , denoted by @xmath836 . a _ vertex _ of @xmath837 is an element @xmath838 such that @xmath839 and @xmath840 imply @xmath841 . another way to define a simplicial complex is to take a set @xmath830 , and let @xmath842 . then @xmath843 is a poset . it is a simplicial complex if furthermore for @xmath834 also @xmath844 . the set of vertices corresponds to the set @xmath830 . the first definition makes it easier to describe the star of a simplex , which will often be useful . so let @xmath832 be a simplicial complex . for @xmath845 we say that @xmath846 _ is a face of _ @xmath847 if @xmath848 . we will write @xmath849 if @xmath848 and @xmath850 . due to the second property , there exists a unique minimal element in @xmath837 which is denoted by @xmath851 . let @xmath834 , then the _ rank of _ @xmath846 , @xmath852 is the cardinality of the set of vertices contained in @xmath846 . we define the rank of @xmath837 , @xmath853 . for @xmath834 define the _ star of @xmath846 _ as @xmath854 . this is again a simplicial complex with minimal element @xmath846 . a _ chamber _ of @xmath837 is a maximal element in @xmath837 , we will denote the set of chambers as @xmath855 or @xmath61 , if @xmath837 is unambiguous . let @xmath856 be a map between simplicial complexes @xmath837 and @xmath857 . then @xmath618 is called a _ morphism of simplicial complexes _ if it is a morphism of posets and furthermore @xmath858 is an isomorphism for all @xmath859 . a subcomplex @xmath857 of @xmath837 is a subset of @xmath837 such that the inclusion @xmath860 is a morphism of simplicial complexes . for @xmath848 , the _ codimension of @xmath846 in @xmath847 _ is the rank of @xmath847 in @xmath861 , denoted by @xmath862 . we say that @xmath846 is a _ maximal face _ of @xmath847 , if @xmath863 . for simplices occurring as subsets of @xmath864 it is convenient to consider the dimension of a simplex rather than the rank . we will denote an @xmath286-dimensional simplex simply as an @xmath286-simplex . therefore , an @xmath286-simplex will be of rank @xmath865 . fix a simplicial arrangement @xmath62 . we will show that @xmath493 is actually a simplicial complex and furthermore a chamber complex with set of chambers @xmath70 . we already showed in remark [ rem : scs ] that the simplicial structure on a closed chamber @xmath494 is induced from the simplicial structure of @xmath866 , where @xmath3 is an open simplex such that @xmath867 . recall the notion of the convex hull of a set . let @xmath111 be a connected metric space . for an arbitrary subset @xmath45 the convex hull of @xmath47 is the smallest convex set @xmath868 , such that @xmath869 . for a different approach , define the segment between @xmath870 to be @xmath871 this can be used to define @xmath872 we can then recursively define @xmath873 then the convex hull of @xmath47 is the set @xmath874 . for @xmath40 and a linearly independent set @xmath875 the convex hull of @xmath47 can be more easily described as @xmath876 in this setting , we will refer to the set @xmath877 as the _ open convex hull of @xmath47_. if @xmath875 is linearly independent , then its convex hull @xmath3 is exactly a simplex of rank @xmath349 , with vertex set @xmath878 . the simplicial structure coincides with the poset @xmath879 . moreover , if @xmath4 is a basis of @xmath39 , and @xmath3 is the open convex hull of @xmath4 , the cone @xmath880 coincides with @xmath881 as defined in remark [ rem : scs ] . if @xmath541 is a face of @xmath3 , there exists a subset @xmath882 such that @xmath541 is the convex hull of @xmath883 . then @xmath884 is a face @xmath494 in the simplicial complex @xmath885 . therefore @xmath387 , @xmath3 , and @xmath886 are all isomorphic simplices , via the isomorphisms @xmath887 [ rem : chse ] let @xmath888 , @xmath889 be bases of @xmath10 such that @xmath890 . then , up to permutation , @xmath891 for some @xmath892 for all @xmath893 . the same holds for two bases @xmath894 of @xmath19 such that @xmath895 . [ equalcones ] let @xmath3 be open convex hull of @xmath4 . . one can choose positive scalars @xmath897 such that @xmath898 . then @xmath899 with @xmath900 , @xmath901 . let @xmath902 , we find @xmath903 . hence @xmath904 with @xmath905 and @xmath906 . therefore the matrix @xmath907 describing the base change is non - negative , and the same holds for its inverse @xmath908 . it is well known ( see theorem 4.6 in chapter 3 of @xcite for example ) that the inverse of a non - negative matrix is non - negative if and only if the matrix is monomial . adding the fact that the sum of every column adds up to @xmath399 , we get that @xmath907 and @xmath908 are already permutation matrices . the statement for @xmath909 just follows from remark [ rem : scs ] by considering the dual basis . for the following statement recall from [ rem : scs ] that for a basis @xmath87 of @xmath19 the cone @xmath910 is given by @xmath911 for a basis @xmath4 of @xmath10 the cone @xmath881 is given by @xmath912 let @xmath913 for a basis @xmath4 of @xmath10 and a basis @xmath87 of @xmath19 . let @xmath914 . then @xmath915 for all @xmath916 if and only if @xmath917 . likewise @xmath918 for all @xmath916 if and only if @xmath919.[purelc ] let @xmath920 . by remark [ rem : scs ] we find that @xmath87 is dual to @xmath4 up to positive scalar multiples . denote with @xmath921 the dual to @xmath916 . so @xmath922 . applying this to @xmath4 yields @xmath923 for all @xmath916 . this immediately yields both equivalences . let @xmath538 . for every @xmath53 , either @xmath242 or @xmath541 is contained in a unique closed half space of @xmath118 , denoted by @xmath924 . furthermore @xmath925 , and if @xmath926 for some @xmath69 , then @xmath927 . [ facintwal ] in case @xmath242 , there is nothing to show , so assume @xmath928 . then the first statement is an immediate consequence of corollary [ purelc ] and the fact that the elements @xmath64 are defined as connected components of @xmath929 . let @xmath119 such that @xmath930 and @xmath931 . let @xmath541 be the convex hull of the vertices @xmath932 , where @xmath800 as we assume @xmath928 . for the same reason we can assume @xmath933 for some @xmath934 , and without loss of generality we can assume @xmath935 . then @xmath936 is the convex hull of @xmath937 . by remark [ rem : scs ] @xmath885 is a simplicial complex , hence @xmath936 is a simplex in @xmath13 contained in @xmath885 . _ proof of proposition [ ssimpcomp ] . _ let @xmath538 , then @xmath938 for some @xmath69 , @xmath939 . in particular @xmath940 carries the structure of a simplex , since by remark [ rem : scs ] @xmath885 is a simplex . so let @xmath941 , we have to show @xmath942 . the intersection @xmath943 is not empty , as it contains @xmath944 . assume @xmath945 , for @xmath946 , @xmath947 . in the case @xmath948 there is nothing to show , as @xmath949 is a simplicial complex by remark [ rem : scs ] . so from now on let @xmath950 . consider the case that @xmath951 , we find @xmath952 . the set @xmath795 can be written as @xmath953 by remark [ rem : scs ] . by lemma [ facintwal ] the intersection @xmath954 for @xmath955 is either @xmath541 , in case that @xmath956 contains @xmath541 , or equals @xmath936 . in both cases , it is again a simplex in @xmath885 . we can conclude that @xmath957 for every @xmath538 . now let @xmath958 , then @xmath959 is a simplex in @xmath885 by lemma [ facintwal ] . but @xmath943 can be written as @xmath960 , using the previous part of the proof shows our claim . in this section we recall definitions and basic facts regarding chamber complexes and type functions following @xcite . note that the first property is always satisfied if @xmath966 is finite . in this case it is easy to see ( cp . * , p.15 ) ) that every chamber has the same rank . a consequence from 2 ) is that every element in a gallery is again a chamber . for two chambers @xmath967 and @xmath859 with @xmath968 and @xmath969 we have @xmath970 ( cp . this allows us to define the _ corank of @xmath315 _ as @xmath971 for any chamber @xmath4 containing @xmath315 . we call @xmath4 and @xmath965 _ adjacent _ , if @xmath972 . a type function @xmath574 with index set @xmath565 induces a weak type function @xmath979 . conversely , we show in lemma [ wnumisnum ] that a weak type function @xmath980 on a strongly connected chamber complex gives rise to a type function @xmath574 such that @xmath981 . let @xmath837 be a chamber complex , such that all sets @xmath985 are gated for @xmath986 with @xmath987 . let @xmath4 be a chamber and @xmath574 be type function of @xmath988 , there exists a unique weak type function @xmath989 such that @xmath990 . [ gsimpwn ] by definition every maximal element in @xmath13 is of the form @xmath494 for some @xmath69 . assume @xmath995 for @xmath996 , then @xmath494 is contained in some @xmath53 by remark [ rem : scs ] , which contradicts to the definition of @xmath70 . this proves @xmath994 . two closed chambers @xmath1005 are adjacent , if @xmath1006 . with respect to lemma [ adjsep ] we find that @xmath129 are adjacent if and only if @xmath1007 is a unique minimal gallery from @xmath64 to @xmath26 . the complex @xmath13 is a chamber complex , since every simplex is contained in a chamber and two chambers @xmath217 are connected by a gallery of length @xmath224 by lemma [ sepisdis ] and @xmath224 is finite by lemma [ finsep ] . for two elements @xmath217 we can therefore define the distance @xmath1011 as the length of a minimal gallery connecting @xmath129 . let @xmath541 be a simplex in @xmath13 , and consider the simplicial complex @xmath1012 with chambers @xmath1013 . let @xmath1014 and assume @xmath1015 , so @xmath1016 . the fact that @xmath1014 implies @xmath1017 . we need to show that there exists a gallery in @xmath1018 from @xmath64 to @xmath26 , which we do by induction on @xmath1011 . for @xmath1019 we have that @xmath129 are adjacent . so let @xmath1020 and assume @xmath946 with the properties that @xmath982 are adjacent , @xmath1021 and @xmath287 . then @xmath1022 by lemma [ starsep ] , and @xmath1023 implies @xmath1024 . in particular we get @xmath1025 and by induction there exists a gallery from @xmath146 to @xmath26 in @xmath1013 , so we are done . let @xmath538 , then there exists @xmath69 and @xmath1027 such that @xmath1028 . in particular , @xmath1029 . a point @xmath50 as above exists , since @xmath541 has nonempty interior with respect to @xmath1030 . let @xmath837 be a strongly connected chamber complex . if @xmath837 is weakly numbered , it is already numbered . in particular , if @xmath1031 is a weak type function , there exists a type function @xmath574 such that @xmath1032 for all @xmath1033 . [ wnumisnum ] assume @xmath986 , and let @xmath129 be chambers with @xmath1038 . since @xmath837 is strongly connected , @xmath1018 is connected , and we find @xmath1039 . thus we find a gallery @xmath1040 from @xmath64 to @xmath26 with all @xmath1041 , so in particular @xmath1042 for all @xmath1043 . also @xmath1044 and @xmath302 are adjacent for @xmath1045 , therefore @xmath1046 and inductively we obtain @xmath1047 . this allows us to define @xmath1048 for every simplex @xmath541 and every chamber @xmath64 containing @xmath541 . by definition @xmath574 coincides with @xmath1049 on all simplices @xmath663 contained in @xmath64 , in particular @xmath1050 is a type function and thus a morphism of chamber complexes . the construction of the weak type function is actually quite simple . begin with a chamber @xmath64 and consider a type function @xmath574 of @xmath949 . let @xmath26 be adjacent to @xmath64 such that @xmath1051 . then set @xmath1052 . let @xmath1053 be the unique index such that @xmath1054 , then @xmath574 maps the vertex not contained in @xmath541 to @xmath97 , so @xmath1055 must map the vertex @xmath232 in @xmath1056 not contained in @xmath541 to @xmath97 as well . so as every simplex @xmath1057 is either contained in @xmath541 or contains @xmath232 , if @xmath3 is contained in @xmath541 then @xmath1058 is already defined , if it contains @xmath232 set @xmath1059 . one can check that @xmath1055 is a morphism of chamber complexes , and furthermore @xmath1055 is the only possible type function of @xmath1056 satisfying @xmath1052 . in this way we can inductively construct type functions for all chambers with arbitrary distance to @xmath64 . this construction works always , however being well defined arises as a problem : given a chamber @xmath26 with @xmath1060 , there may be two chambers @xmath1061 with @xmath1062 and @xmath1061 adjacent to @xmath26 . then @xmath26 has induced type functions from @xmath22 as well as from @xmath23 . now theorem [ numbext ] yields that these two induced type functions coincide , and thus the method gives us a weak type function of @xmath13 . we say that a complex @xmath837 is _ spherical _ if it contains a pair of chambers @xmath982 such that @xmath1063 for all panels @xmath1064 containing @xmath64 . the chamber @xmath146 with such a property is called _ opposite _ to @xmath64 . thin : since @xmath1067 has two connected components , the complex @xmath13 is meager . it is thin if and only if for every chamber @xmath4 and every wall @xmath1068 there exists a chamber @xmath1069 which is @xmath118-adjacent to @xmath4 . in this case @xmath1070 , since @xmath2 is convex . then @xmath1071 , therefore @xmath118 meets @xmath2 and is contained in @xmath35 . let @xmath13 be spherical . by definition we find two opposite chambers @xmath4 and @xmath1073 . as @xmath13 is meagre by construction , we also know @xmath1074 . in particular @xmath1075 , and @xmath70 is finite as well as @xmath35 . assume @xmath4 has no @xmath97-adjacent chamber for an @xmath1053 and let @xmath1064 be the @xmath97-panel containing @xmath4 , then @xmath1076 for all @xmath1077 , a contradiction . thus @xmath13 is also thin . by our previous argument therefore @xmath62 is thin . let @xmath1078 . since @xmath2 can not be written as a finite union of hyperplanes , we can assume that @xmath81 for all @xmath53 . thus take a neighborhood @xmath76 of @xmath50 and consider the chambers intersecting @xmath76 . by taking a smaller @xmath76 we can also assume that only a single chamber @xmath965 intersects @xmath76 . thus there exists a wall of @xmath965 not meeting @xmath2 , a contradiction to @xmath35 being thin . hence @xmath38 is empty and @xmath1072 holds . a. berman and r. j. plemmons , _ nonnegative matrices in the mathematical sciences _ , academic press [ harcourt brace jovanovich , publishers ] , new york - london , 1979 , computer science and applied mathematics . a. bjrner , m. las vergnas , b. sturmfels , n. white , and g. m. ziegler , _ oriented matroids _ , second ed . , encyclopedia of mathematics and its applications , vol . 46 , cambridge university press , cambridge , 1999 . b. grnbaum , _ arrangements of hyperplanes _ , proceedings of the second louisiana conference on combinatorics , graph theory and computing ( louisiana state univ . , baton rouge , la . , 1971 ) ( baton rouge , la . ) , louisiana state univ . , 1971 , pp . 41106 . l. paris , _ lectures on artin groups and the @xmath9 conjecture _ , groups of exceptional type , coxeter groups and related geometries , springer proc . math . stat . , vol . 82 , springer , new delhi , 2014 , pp . 239257 . to3em , _ uvres / collected works . i , ii , iii , iv _ , heritage of european mathematics , european mathematical society ( ems ) , zrich , 2013 , edited by francis buekenhout , bernhard matthias mhlherr , jean - pierre tignol and hendrik van maldeghem .
there are strong correlations among the measurable physical properties of galaxies . the classification of galaxies along the visual morphological sequence described by @xcite correlates well with the dominance of their central bulge , their surface brightnesses , and their colors . these properties also correlate with other properties , such as metallicity , emission - line strength , luminosity in visual bands , neutral gas content , and the winding angle of the spiral structure ( for a review , see @xcite ) . the surface brightnesses of giant galaxies classified morphologically as elliptical are known to be strongly correlated with their sizes ( @xcite ) . galaxy colors ( at least of morphologically elliptical galaxies ) are known to be strongly correlated with galaxy luminosity ( @xcite ) . the gravitational mass of a galaxy is closely related to the luminosity and other galaxy properties . these galaxy relations manifest themselves in the tully - fisher relation for spiral galaxies ( @xcite ) and the fundamental plane for morphologically elliptical galaxies ( @xcite ) ; @xcite review these dynamical relations . furthermore , the environment of a galaxy is related to its type , in the sense that early - type galaxies are found in denser regions than late - type galaxies , as first noted by @xcite and as found by numerous subsequent investigators ( @xcite ) . in short , different physical properties of galaxies are closely related to each other . in order to understand theoretically how galaxies formed and acquired their current properties , it is first necessary to characterize the observed distribution of galaxy properties in a comprehensive manner . such a characterization is a primary goal of the sloan digital sky survey ( sdss ; @xcite ) . the sdss has already created the largest sample to date ( 144,609 galaxies ) of luminous galaxies with well - measured photometric and spectroscopic properties . as a first step in understanding the joint distribution of galaxy properties , we present here a study of the number density and luminosity density distributions of galaxy colors , profiles , luminosities , surface brightnesses , and local densities . our purpose is to measure properties related to quantities that cosmological gasdynamical simulations ( @xcite ; @xcite ; @xcite ) or semi - analytic models ( @xcite ; @xcite ) can predict or will soon be able to predict , such as galaxy ages , sizes , stellar masses , and degrees of concentration . the paper is organized as follows . section [ data ] briefly describes the spectroscopic sample of the sdss . section [ properties ] describes how we measure the set of properties studied here . section [ results ] shows the number density and luminosity density distributions of these properties of galaxies , and provides simple fits for the dependence on luminosity of the properties of galaxies with nearly exponential radial profiles and those with more concentrated radial profiles . section [ conclusion ] summarizes and outlines future lines of research . the sdss ( @xcite ) is producing imaging and spectroscopic surveys over @xmath8 steradians in the northern galactic cap . a dedicated wide - field 2.5 m telescope ( siegmund et al . , in preparation ) at apache point observatory , sunspot , new mexico , images the sky in five bands between 3000 and 10000 ( @xmath9 , @xmath10 , @xmath11 , @xmath1 , @xmath12 ; @xcite ) using a drift - scanning , mosaic ccd camera ( @xcite ) , detecting objects to a flux limit of @xmath13 mags . the photometric quality of the observations are tracked using an automatic photometricity monitor ( @xcite ) . an automatic image - processing pipeline known as photo ( lupton et al . in preparation ) processes the images and produces a catalog . the astrometric calibration is also performed by an automatic pipeline ( @xcite ) . one of the goals is to spectroscopically observe 900,000 galaxies , ( down to @xmath14 mags ) , 100,000 luminous red galaxies ( @xcite ) , and 100,000 quasars ( @xcite , @xcite ) . this spectroscopic follow up uses two digital spectrographs on the same telescope as the imaging camera . many of the details of the galaxy survey are described in the galaxy target selection paper ( @xcite ) . other aspects of the survey are described in the early data release paper ( edr ; @xcite ) . the survey has begun in earnest , and to date has obtained about 30% of its intended data . the sdss images are reduced and catalogs are produced by the sdss pipeline photo , which measures the sky background and the seeing conditions , and detects and measures objects . the photometry used here for the bulk of these objects was the same as that used when the objects were targeted . however , for those objects which were in the edr photometric catalog , we used the better calibrations and photometry from the edr . the versions of the sdss pipeline photo used for the reductions of the data used here ranged from v5_0 to v5_2 . the treatment of relatively small galaxies , which account for most of our sample , did not change substantially throughout these versions . the astrometric calibration is also performed by an automatic pipeline which obtains absolute positions to better than 0.1 arcsec ( @xcite ) . the magnitudes are calibrated to a standard star network ( @xcite ) approximately in the ab system . there are small differences between the system output by the sdss pipelines and a true ab system , amounting to @xmath15 in the @xmath9 , @xmath10 , @xmath11 , @xmath1 , and @xmath12 bands . because these were discovered at a relatively late date in the preparation of this manuscript , we have not self - consistently included these shifts in our results ( that is , by recalculating @xmath16-corrections based on the revised colors ) . instead we have applied them _ a posteriori _ to our results in the @xmath17 , @xmath18 , @xmath19 , @xmath20 , and @xmath21 bands . object fluxes are determined several different ways by photo , as described in @xcite . the primary measure of flux used for galaxies is the sdss petrosian magnitude , a modified version of the quantity proposed by @xcite . the essential feature of petrosian magnitudes is that in the absence of seeing they measure a constant fraction of a galaxy s light regardless of distance ( or size ) . they are described in greater detail below ( and also by @xcite and @xcite ) . as measures of the galaxy profile shape , photo also calculates the radius containing half the petrosian flux ( @xmath22 ) and that containing 90% of the petrosian flux ( @xmath23 ) . as described in @xcite , the bulk of the fibers in the sdss are allocated to three samples : the main sample galaxies , the luminous red galaxies , and quasars . here we will concern ourselves solely with the main sample galaxies . there are three main criteria for their selection : @xmath24 here , @xmath25 is an estimate of the magnitude using the local psf as a weighted aperture , @xmath26 is an estimate of the magnitude using the better of a de vaucouleurs and an exponential fit to the image ( accounting for seeing ) , @xmath27 is a modified form of the magnitude described by @xcite , and @xmath28 is the half - light surface brightness , defined as the average surface brightness within the radius which contains half of the petrosian flux . all of these quantities are reddening - corrected using the dust maps of @xcite . in practice , the values of the target selection parameters vary across the survey in a well - understood way , but for the bulk of the area , they are : @xmath29 , @xmath30 mag , and @xmath31 mag in 1 arcsec@xmath32 . objects near the spectroscopic flux limit are nearly five magnitudes brighter than the photometric limit ; that is , the fluxes are measured at signal - to - noise ratio of a few hundred . fibers are assigned to a set of circular tiles with a field of view @xmath33 deg in radius by an automatic tiling pipeline ( @xcite ) . the targets are observed by a multi - fiber spectrograph capable of taking 640 spectra simultaneously ( 48 of the fibers are , in normal sdss operations , devoted to sky fibers and spectrophotometric standards ) . an automatic data processing pipeline spec2d wavelength- and flux - calibrates the spectra and outputs a one - dimensional spectrum for each object ( schlegel , et al . , in preparation ) . the resulting spectra have a resolution of @xmath34 , cover roughly 3800 to 9000 , and typically have signal - to - noise per pixel around 10 . this spectrum is then analyzed by another pipeline ( spectro1d ; subbarao , et al . , in preparation ) which classifies the spectrum and determines the redshift of the object . the redshifts used in the current analysis are determined independently using a separate pipeline originally designed specifically for bright stars(specbs ; schlegel , et al . , in preparation ) whose results are nearly identical for the main galaxy sample to the `` official '' sdss redshift determination ( over 99% of the objects obtain nearly identical redshifts in each pipeline ) . as of april 2002 , the sdss had imaged and targeted 2,873 deg@xmath32 of sky and taken spectra of approximately 350,000 objects over @xmath35 deg@xmath32 of that area . the completeness over this area is roughly 91% ; most of the missing galaxies ( 7% of the total ) are lost due to the fact that two fibers on the same tile can not be placed more closely than 55@xmath36 , about 1% are lost because they are not able to be assigned a fiber , and about 1% can not have a redshift determined . from these data , we created a well - defined sample for calculating large - scale structure and galaxy property statistics , known as large - scale structure sample10 . sample10 consists of all of the photometry for all of the targets over that area ( as extracted from the internal sdss operational database ) , all of the spectroscopic results ( as output from specbs ) , and , most significantly , a description of the angular window function of the survey and the flux and surface brightness limits used for galaxies in each area of the sky . for most of the area , the same version of the analysis software used to create the target list was used in this sample . however , for the area covered by the early data release ( edr ; @xcite ) we used the version of the analysis software used for that data release , since it was substantially better than the early versions of the software used to target that area . for photo , the most important piece of analysis software run on the data , the versions used for the photometry range from v5_0 to v5_2 . the region covered by this sample is similar to , but not exactly , the region which will be released in the sdss data release 1 ( dr1 ) , scheduled for january 1 , 2003 ( which will use photo v5_3 , a newer version of the software which among other things improves the handling of large galaxies ) . for the purposes of this paper , we further restrict sample10 to galaxies in the redshift range @xmath37 , the absolute magnitude range @xmath38 ( assuming a hubble constant @xmath39 km@xmath40 s@xmath40 mpc@xmath40 and @xmath41 ) , and to galaxies with @xmath42 . to calculate the distance modulus and the comoving volume we assume a model with @xmath43 and @xmath44 . these cuts restrict our sample to 144,609 galaxies . figure [ emg_imagr50 ] shows the distributions of @xmath1-band apparent magnitude ( corrected for galactic extinction ) and petrosian half - light radius for this sample of galaxies . the roll - off at the brightest and faintest magnitudes occurs because we have selected the sample in the @xmath11-band . here we describe the set of galaxy properties whose joint distribution we intend to measure , and how those properties are measured by the sdss . these properties are colors , luminosities , profile shapes , surface brightnesses , and local densities . to measure the galaxy fluxes , we rely on the sdss petrosian magnitudes . the essential feature of petrosian magnitudes is that in the absence of seeing they measure a constant fraction of a galaxy s light regardless of distance ( or size ) . more specifically , we define the `` petrosian ratio '' @xmath45 at a radius @xmath11 from the center of an object to be the ratio of the local surface brightness averaged over an annulus at @xmath11 to the mean surface brightness within @xmath11 : @xmath46}{\left . \int_0^r dr ' 2\pi r ' i(r ' ) \right/ [ \pi r^2]},\ ] ] where @xmath47 is the azimuthally averaged surface brightness profile and @xmath48 , @xmath49 define the annulus . the sdss has adopted @xmath50 and @xmath51 . the petrosian radius @xmath52 is the radius at which the @xmath45 falls below a threshhold value @xmath53 , set to 0.2 for the sdss . the petrosian flux is defined as the flux within a circular aperture with a radius equal to @xmath54 , where @xmath55 for the sdss . petrosian magnitudes are described in greater detail by @xcite and @xcite . while sdss petrosian magnitudes contain over 99% of the flux within an exponential profile , they contain only around 80% of the flux within a de vaucouleurs profile in the absence of seeing ; the fraction is higher in the presence of seeing , as shown by @xcite . for the sdss , @xmath52 is defined in the @xmath11-band and used to set the aperture for all the bands , such that the colors represent a measure of spectral energy distribution of the galaxy over a well - defined region of the galaxy . here we constrain the spectral energy distribution of the galaxies using the broad - band colors between the @xmath9 , @xmath10 , @xmath11 , @xmath1 , and @xmath12 petrosian magnitudes of the sdss . a given set of observer - frame galaxy colors does not correspond uniquely to a particular sed , so we can not infer with certainty the colors in any other frame . however , we can use our knowledge of the form of galaxy seds based on other data to make an attempt in order to interpolate between the observed - frame colors to a different frame by fitting a reasonable model sed to the galaxy colors . to do so , we use the routines in kcorrect v1_11 , described by @xcite , which uses a similar method to the method of @xcite for calculating photometric redshifts . this method is demonstrably better than using the spectra ( which suffer from the aperture biases associated with small - fiber spectroscopy , the inaccuracy of spectrophotometry , and the fact that we do not take spectra in the @xmath9 and @xmath12 bands ; _ cf . _ @xcite ) and better than simply interpolating the magnitudes between bands ( which does not account well for the 4000 break ) . the essence of the method is that each set of observed magnitudes in the five bands is fit by a linear combination of four templates . the coefficients in the linear combination , and the templates themselves , are optimized to reproduce the observed magnitudes . for each galaxy , we now have a model for the sed . we use this sed to calculate the correction from the observed band to fixed frame bandpasses with shapes corresponding the the restframe coverage of galaxies at @xmath56 , the median redshift of the sample . the usual choice is to correct the colors to the @xmath57 frame ; however , this choice is not optimal , because most galaxies in our sample are not at @xmath57 . figure [ response ] shows the responses of the unshifted sdss and the sdss system shifted to @xmath58 . we refer to bands in this system as @xmath17 , @xmath18 , @xmath19 , @xmath20 , and @xmath21 , following the nomenclature of @xcite . we note in passing that the @xmath16-correction for a galaxy _ at _ redshift @xmath56 to this system is @xmath59 for all bands independent of galaxy sed . as a test of the validity of the colors , figure [ emg_colors ] shows the distribution of galaxy colors as a function of redshift in a volume - limited sample in the range @xmath60 and @xmath61 . the greyscale is the conditional distribution of color on redshift . the middle line in each panel is the median , the outer lines are the 10% and 90% quantiles . the distribution of galaxy colors is nearly constant with redshift , indicating that the @xmath16-corrections are relatively sensible . note that we expect some change in these distributions given that galaxies evolve ; in fact there is a hint of such evolution in the @xmath62 plot . we use these estimated petrosian @xmath63 , @xmath64 , @xmath65 and @xmath66 colors as our measures of the sed shape for each galaxy . as a measure of luminosity , we use the @xmath1-band flux converted to @xmath20-band luminosity luminosity using the @xmath16-correction and the standard cosmological formulae for the distance modulus , assuming @xmath67 and @xmath68 ( @xcite ) . one measure of the morphology of galaxies is the radial dependence of their surface brightness . for example , if one expresses this radial dependence in the form ( suggested by @xcite ) : @xmath69,\ ] ] galaxies with nearly exponential profiles ( srsic index @xmath70 ) tend to be considered `` late - type , '' while galaxies with nearly de vaucouleurs profiles ( srsic index @xmath71 ) tend to be considered `` early - type . '' for this reason , we use measures of the azimuthally - averaged radial profiles of the galaxies in our analysis . although @xcite and @xcite have shown that the petrosian inverse concentration index @xmath72 is reasonably well correlated with eyeball morphological classifications for nearby , large galaxies , this parameter is too sensitive to the effects of seeing to be of great use for the bulk of the sdss main sample . the top panel of figure [ emg_seeing ] shows the dependence of inverse concentration parameter on seeing for a volume - limited subsample , limited to galaxies with @xmath73 and @xmath74 , demonstrating this dependence . this seeing dependence means that the same galaxy , observed under different conditions , can have considerably different measured concentrations . even more importantly , it implies that the same galaxy , observed at different _ distances _ , will have different measured concentrations . photo fits two - dimensional models to the galaxy images which account for the effects of seeing , but they are limited to the cases of @xmath75 and @xmath76 , and in addition ( because they were designed to study faint galaxy colors ) for the versions of photo used here ( before v5_3 ) they only performed the fit out to a relatively small radius compared to the size of an sdss spectroscopic target galaxy . thus , for our work here we have decided to fit the parameters in equation ( [ sersic ] ) to the radial profile of each galaxy ( accounting for the effects of seeing ) . in order to do so , we find the parameters @xmath77 , @xmath78 , and @xmath79 in equation ( [ sersic ] ) which minimize @xmath80 with respect to the observed @xmath1-band radial profile and errors ( as expressed by the profmean and proferr parameters in the sdss catalog ) . when performing the fit , we convolve with the axisymmetric double gaussian fit to the psf determined by the sdss photometric pipeline ( lupton et al . , in preparation ) . the resulting distribution of @xmath80 per degree of freedom peaks at approximately unity , but has a much larger tail than expected for gaussian errors ; this suggests that a more detailed model ( either for the seeing or for galaxy profiles ) which better fits the data is possible to develop . after performing the fit , we can calculate the petrosian inverse concentration parameter @xmath81 based on the srsic index @xmath79 . the top panel of figure [ emg_seeing ] shows this inverse concentration parameter as a function of seeing for the volume - limited subsample . our measure of galaxy profile shape is less seeing - dependent than the raw estimate @xmath81 . we have fit srsic profiles to the profiles of all bands ; the results of the @xmath10 , @xmath11 , @xmath1 , and @xmath12 profiles are all fairly consistent , while the @xmath9 band results tend to be much less concentrated than the others . it is worth examining how our axisymmetric srsic fit is affected by the non - axisymmetry of galaxies , and how it is related to other measures of profile shape , such as the petrosian concentration parameter and the bulge - to - total ( @xmath82 ) ratio . in the top panels of figure [ test_nfit_plot ] , we investigate how our srsic fits behave for galaxies with srsic profiles and galaxies which consist of a de vaucouleurs bulge plus an exponential disk . in each case we assume the galaxies are comparable in half - light radius to sdss main sample galaxies . we use three different axis ratios @xmath83 ( as listed in figure [ test_nfit_plot ] ) by which we distort the galaxy image . we then project that image onto a grid of pixels of size @xmath84 , the size of sdss pixels , apply single gaussian seeing with a standard deviation of @xmath85 , and then extract radial profiles using a similar scheme as used by sdss photo . using the single gaussian model for the seeing , we fit the srsic models to the resulting galaxy profiles . we perfectly reconstruct the case of @xmath83 for the srsicprofiles ; for small axis ratios , we measure a higher concentration than we would for a circular image with the same profile as along the semi - major axis . for the bulge - plus - disk models , we find that the srsic index @xmath79 is monotonically related to the @xmath82 ratio of the bulge flux to the total flux . the bottom panels of figure [ test_nfit_plot ] show the dependence of the petrosian inverse concentration parameter @xmath81 for galaxies of the same shapes as , but larger radii than , the galaxies in the top panels . this situation is meant to be similar to the conditions under which galaxies were observed by @xcite and @xcite , who observed large , bright galaxies less affected by seeing than those in the sdss main sample . we find that @xmath81 is closely related to srsic index and @xmath82 . we define the half - light surface brightness @xmath86 to be the average surface brightness within the petrosian half - light radius for the srsic profile fit in the @xmath20-band , in magnitudes in one arcsec@xmath32 , @xmath16-corrected and corrected for cosmological surface brightness dimming : @xmath87 this measure of surface brightness is thus slightly different ( in general , higher ) than the raw estimate calculated from the petrosian half - light radius output by the sdss photometric pipeline , because it is corrected for seeing . note that we have implicitly assumed that the @xmath16-correction of the observation of surface brightness remains unaffected by color gradients in the galaxy ; because the @xmath16-correction from the observed @xmath1 band to the @xmath20 band are expected to vary little with redshift and color ( @xmath88 magnitudes ) this approximation is not bad . for each object in the sample we estimate the local galaxy density @xmath89 as follows . first , we distribute a large number ( @xmath90 ) of poisson random points within the sdss volume , distributed in redshift according to the selection function of the survey . the selection function is determined by an evolving @xmath19-band schechter function fit to the luminosity function of the following form ( following @xcite ) : @xmath91 where @xmath92 expressed in magnitudes , this implies @xmath93 . if @xmath94 , these formulae mean galaxies are brighter in the past . we indeed find that in order to explain the dependence of number counts on redshift in the survey , it is necessary to include the possibility that galaxy luminosities evolve . we perform this fit in the @xmath19 band and find the parameters : @xmath95 we do not give error bars for this fit , nor do we recommend this fit as our best estimate of the luminosity density or its evolution . we do not give error bars for this fit , nor do we recommend this fit as our best estimate of the luminosity density or its evolution . a more complete discussion of this topic is forthcoming in a separate paper . however , this fit is good enough to use to estimate local density . using this fit , the magnitude limits as a function of position on the sky ( @xmath96 and @xmath97 ) , and the spectroscopic sampling fraction as a function of position on the sky ( @xmath98 ) we can calculate the selection function at any point on the sky . given the galaxy catalog and the random catalog , we count the number of galaxies @xmath99 ( including the galaxy in question ) and the number of random points @xmath100 in a 8 @xmath0 mpc radius comoving sphere around each galaxy . we define @xmath101 where @xmath102 is the average expected number of objects in the sphere . we use this as an estimate of the local density relative to the mean . because we use a spherical cell only 800 km s@xmath40 in radius , this measure of density is affected somewhat by the velocity dispersions in the deep potential wells of clusters , in the sense that it will slightly underestimate the density in those regions . also , since it is centered on a galaxy , and galaxies are clustered on these scales , its mean is greater than unity . the local density is the only quantity considered here which is measured with a large amounts of noise for each object . the average `` expected '' number of objects for a poisson distribution in each cell is about 14 ; the average actual number is about 37 ( it is so much larger because the variance of the overdensity on this scales is around unity ) . this indicates that the average cell has errors of @xmath103 . in particular , distant galaxies ( which only have the brightest galaxies at that redshift within 8 @xmath0 mpc ) tend to have a measurement of @xmath89 noisier than do close galaxies . one can not really consider the joint distribution of the local density measured here and other galaxy properties , without accounting for the noise in the measure of the local density ( @xcite ) . because of these effects , the galaxies which have the highest density estimates are just the galaxies with the smallest expected number of galaxies around them ( and thus the highest expected noise in the measurement ) . furthermore , the galaxies with the smallest expected number of galaxies around them are those at the highest redshift , and are therefore the highest luminosity galaxies . this introduces a correlation between the luminosity of a galaxy and the noise in our estimate of the density . the galaxies with the highest density estimates are high luminosity galaxies simply because of this artificial correlation between luminosity and measurement noise . this effect means that one can not determine ( from our measurements ) the density dependence of luminosity ( or any property correlated with it that is , all galaxy properties ) . nevertheless , one can still meaningfully consider the median density within bins of other galaxy properties , because the increased noise for galaxies with higher mean distances does not affect the median density . as proof of this assertion , figure [ emg_density ] shows the density distribution as a function of redshift for several small ranges of absolute magnitude . the lines are the 10% , 50% , and 90% quantiles of the distribution . although at large distances the distribution widens ( the 10% and 90% quantiles decrease and increase , respectively ) , the median remains remarkably constant , indicating that it measures the density around galaxies consistently at high redshift and at low redshift . this consistency means that even when we divide galaxies into different groups which cover different redshift ranges ( such as dividing by luminosity ) we can make a fair comparison among the median local densities of the groups . this consistency happens to be the justification for the common practice of calculating the correlation function for galaxies of different luminosities , in different redshift ranges . to compute the global number and luminosity densities , it is necessary to compute the number - density contribution @xmath104 for each galaxy , where @xmath105 is the volume covered by the survey in which this galaxy could have been observed , accounting for the flux , surface brightness , and redshift limits as a function of angle ( schmidt 1968 ) . this volume is calculated as follows : @xmath106 where @xmath98 is described above and @xmath107 and @xmath108 are defined by : @xmath109 the flux limits @xmath110 ( equal to 14.5 mag across the survey ) and @xmath111 ( approximately 17.77 mag ) implicitly set @xmath112 by : @xmath113 the surface brightness limits @xmath114 ( in practice , too high surface brightness to matter for this sample ) and @xmath115 ( approximately 24.5 mag but variable in a known way across the survey ) implicitly set @xmath116 by : @xmath117 note that we have implicitly assumed that the @xmath16-correction of the observation of surface brightness remains unaffected by color gradients in the galaxy . in practice , the surface brightness limits only rarely affect the @xmath105 determination . the function @xmath98 is the sdss sampling fraction of galaxies as a function of position on the sky . the total sampling rate of galaxies is computed separately for each region covered by a unique set of spectroscopic survey tiles . we adopt the nomenclature of the 2df ( @xcite ) : each such region is a `` sector '' ( which corresponds identically to the `` overlap regions '' defined in @xcite ) . that is , in the case of two overlapping tiles , the sampling is calculated separately in three sectors : the sector covered only by the first tile , the sector covered only by the second tile , and the sector covered by both . each position on the sky is thereby assigned a sampling rate @xmath98 equal to the number of redshifts of galaxy targets obtained in the region divided by the number of galaxy targets in the region . in regions covered by a single tile , typically @xmath118 ; in multiple plate regions , typically @xmath119 . these completenesses average to 0.92 . with @xmath120 , we calculate the global number and luminosity density contributions of any group of galaxies as @xmath121 and @xmath122 . our primary results on the distributions of colors , surface brightnesses , profile shapes , luminosities and local densities of galaxies are shown in figures [ emg_num][emg_num_cond ] . each of these plots is a multi - panel figure showing some form of the bivariate distribution of every pair of galaxy properties in the off - diagonal plots . the plots on the diagonal show the univariate distribution of each quantity . figure [ emg_num ] shows the projected number density distribution for all pairs of seven of our parameters , excluding density ( for reasons described above ) . the @xmath123 values calculated above are used to calculate the contribution of each galaxy to the total number density . the panels above and below the diagonal are mirror images of each other . the grey scale is proportional to the square - root of the projected density in that plane . contours in the plot represent loci of constant projected density enclosing 68% and 95% of the density . figure [ emg_lum ] shows the bivariate @xmath20-band luminosity density distribution between all of our parameters . this figure is similar to figure [ emg_num ] , only now each object contributes @xmath124 rather than @xmath104 to the density . figure [ emg_num_cond ] shows the conditional number density distribution of all eight of our parameters relative to seven of them ( only excluding density as an independent variable ) . that is , the greyscale represents the probability of a particular value of the quantity on the vertical axis , given the value of the quantity on the horizontal axis . this is equivalent to taking figure [ emg_num ] , and for each vertical column of pixels renormalizing by the total density in that column . this plot is useful when the dependence of one quantity on another is strong in the regime where there are few galaxies ( such as at high density or high luminosity ) . the histograms along the diagonal are the number density distributions of each quantity , and are identical to the histograms in figure [ emg_num ] . some basic results are visible in these plots , which we review in the following subsections . many of these results have been noted in the literature in the past . we will concentrate mostly on figure [ emg_num_cond ] , in which many of the relationships are easiest to see . the first simple result to point out from figure [ emg_num_cond ] is the histogram in the right column and the second row from the top , which shows the luminosity function in the @xmath125-band . the function is close to a schechter function , although because the vertical scale is linear , it may not appear familiar . now consider the same panel in figure [ emg_lum ] , which shows the luminosity density as a function of absolute magnitude . it is clear from this plot that the luminosity density is almost entirely contained within the range of absolute magnitudes shown ( @xmath126 ) . we may further ask whether the surface brightness limits ( around @xmath127 ) greatly affect any estimate of the luminosity density from this sample . the histogram in figure [ emg_lum ] showing the luminosity density as a function of surface brightness demonstrates that the luminosity density falls off long before the surface brightness limit becomes important . an even clearer demonstration of this fact is the panel showing the luminosity density versus @xmath128 and @xmath129 . this panel shows that even at the lowest luminosities in the sample , the surface brightness limit excludes very little luminosity density . these results are in agreement with @xcite ( using early sdss data ) and @xcite ( using early 2dfgrs data ) . a more careful analysis of the luminosity function and its evolution from this sample is being carried out by blanton et al . ( in preparation ) . next , let us follow the right column of figure [ emg_num_cond ] down from the top . each panel below the top panel shows the conditional dependence of a galaxy parameter on luminosity . these panels show many of the familiar relationships of galaxy properties to luminosity from previous observations . in sequence from top to bottom local density is a strong function of luminosity . the most luminous galaxies exist preferentially in the densest regions of the universe . density increases with luminosity for all absolute magnitudes above @xmath130 . at lower luminosities , the local density starts to increase again , which is potentially related to the existence of dwarf spheroidal galaxies in clusters . this result agrees with results from cfa ( @xcite ) , the optical redshift survey ( hermit96a ) , as well as the recent results of @xcite in the sdss and @xcite in the 2dfgrs on the dependence of the correlation function amplitude on galaxy luminosity . 2 . highly luminous galaxies are more concentrated , and thus have higher srsic indices , than lower luminosity galaxies . this trend is consistent with high luminosity galaxies being preferentially `` early - type . '' however , even for the highest luminosity galaxies , the de vaucouleurs profile ( @xmath131 ) is always more concentrated than more than 75% of the galaxies ; the total number density of galaxies with @xmath132 in our absolute magnitude range is @xmath133 . thus , either the number density fraction of `` true ellipticals '' is small in this absolute magnitude range or many `` true ellipticals '' are not de vaucouleurs profile galaxies . that the srsic profile in most cases provides a better fit to the surface brightness profiles of morphologically - classified elliptical galaxies than the de vaucouleurs profile has been shown previously by several groups ( @xcite ) . we can try to compare to these results by dividing our sample by color . figure [ emg_nmgmr ] shows the dependence of @xmath79 on absolute magnitude for three ranges of @xmath64 . the srsic index @xmath79 is clearly independent of luminosity for very blue galaxies ( which are close to exponential ) . for very red galaxies the srsic index is a strong function of luminosity ( paralleling the dependence of srsic index on luminosity found by @xcite and @xcite for morphologically elliptical galaxies ) . luminosity and surface brightness are positively correlated at low luminosities , while at extremely high luminosity surface brightness declines with luminosity . as we will see below , this occurs because the surface brightnesses of highly concentrated galaxies , which tend to be highly luminous , decline with luminosity , while the surface brightnesses of exponential galaxies , which tend to be low luminosity , increase with luminosity . note that constant size as a function of surface brightness would be a 45 degree line on the luminosity - surface brightness plot . at all luminosities , the surface brightness increases more slowly than luminosity , indicating that the median size of galaxies increases strongly with luminosity , especially for the most luminous galaxies . the colors of galaxies depend strongly on luminosity , in the sense that more luminous galaxies are redder . the color - magnitude relation is most obvious in @xmath134 , where separate red and blue populations are evident . this bimodality has also been noted in the observed @xmath135 colors by @xcite and in the d4000 spectroscopic measurements of @xcite . in short , there are strong dependencies of all parameters on luminosity . the number density distribution of @xmath79 , the srsic index , shows a strong peak at @xmath75 in figures [ emg_num ] and [ emg_num_cond ] , indicating that exponential galaxies are extremely common in the universe . note that there is no peak in the number density at @xmath76 , which corresponds to a de vaucouleurs profile , indicating that de vaucouleurs galaxies lie along a continuum of galaxy concentration . naturally , this result is subject to the limitations of our fit which is based on an azimuthally averaged profile and uses a double gaussian approximation to the seeing . on the other hand , figure [ test_nfit_plot ] suggests that the effects of inclination are more likely to increase the observed srsic indices . to test the effects of inclination more empirically , we can restrict our sample to galaxies which are nearly face on ( axis ratio greater than 0.8 ) , according to the photo pipeline s ab_exp and ab_dev measurements of the best fit axis ratios assuming an exponential or de vaucouleurs model , respectively . using either ab_exp or ab_dev as a measure of axis ratio does not lead the srsic fits to be closer to @xmath76 or for any separate `` concentrated '' peak to appear in the distribution of srsic index @xmath79 . similarly , if we restrict our sample to galaxies which have @xmath80 less than @xmath136 ( where @xmath137 is the number of degrees of freedom in the fit ) , we also do nt find any peak at a de vaucouleurs profile . the only subset of galaxies which we have found in our parameter space whose srsic index even close to `` typically '' @xmath76 are very red ( @xmath138 ) and very luminous ( @xmath139 ) , which one can see in figure [ emg_nmgmr ] ; the actual number of such galaxies is very tiny . thus , if there is a single parameter one can measure about galaxies which indicates whether it would be classified morphologically as an elliptical , and the typical such elliptical really has @xmath76 , that parameter is not any of those considered here . the second column from the right in figure [ emg_num_cond ] shows the dependence of each quantity on @xmath79 , the srsic profile index . from top to bottom : 1 . local density does not depend on srsic index as much as it depends on luminosity , although exponential galaxies are in somewhat lower density regions than are concentrated galaxies ( about 25% less overdense ) . galaxies with @xmath140 have a luminosity function peaked at about @xmath141 . the peak luminosity declines as the profile shape becomes closer and closer to exponential ( @xmath75 ) , to the extent that it is clear that the peak of the luminosity function for pure exponential galaxies is well below our absolute magnitude limit of @xmath142 . due to these trends , the luminosity density distribution of @xmath79 in figure [ emg_lum ] is far flatter than the number density distribution , such that concentrated galaxies contribute a much larger fraction of the luminosity density than they do of the number density ( for a more detailed analysis of this phenomenon , see @xcite ) . the half - light surface brightnesses of exponential galaxies ( @xmath70 ) are distributed around @xmath143 while the surface brightnesses of concentrated galaxies are centered at much higher surface brightness ( @xmath144 ) . the galaxies which are nearly exponential tend to be very blue in all colors ( @xmath145 ) . for @xmath132 , galaxies are extremely homogeneous : constant and red colors , constant and high surface brightness , and constant average density . the third column from the right in figure [ emg_num_cond ] shows the dependence of each quantity on @xmath129 ; 1 . density is only a weak function of surface brightness on scales of 8 @xmath0 mpc . note that @xcite investigate the dependence of the correlation function on surface brightness ; they find little dependence on these scales , but a strong dependence on smaller scales . 2 . paralleling the results for the srsic index , the luminosity function of high surface brightness objects is peaked at high luminosity , while at low surface brightness the luminosity function continues to increase until it passes our low luminosity limit . this result is similar to that found in sdss commissioning data by @xcite and in the 2dfgrs by @xcite . the profiles of high surface brightness galaxies are concentrated , while the profiles of most galaxies at lower surface brightness than @xmath146 are consistent with exponential . the highest surface brightness objects are uniformly red , while fainter than @xmath146 , surface brightness and color are positively correlated the lower surface brightness a galaxy is , the bluer it tends to be . the four left columns in figures [ emg_num][emg_num_cond ] show the relation between the four optical colors of galaxies and the other parameters . in figure [ emg_num ] , two distinct peaks in the number density distribution are visible , most notably in the relation between @xmath63 and @xmath64 . this result corresponds closely to the work of @xcite , which found a double peaked distribution of ( observer - frame ) @xmath135 color . the four left columns in figure [ emg_num_cond ] show the dependence of all properties on the colors : 1 . density is a strong function of all colors . one expects this trend , based on the separate correlations of galaxy morphological classification with color ( @xcite and references therein ) and environment ( for example , @xcite ) . the luminosity function of galaxies which have @xmath147 and @xmath148 is peaked at high luminosity ( @xmath149 ) . the luminosity function of bluer galaxies increases right up to our lower luminosity limit . redder than @xmath147 or @xmath148 , the peak luminosity declines as well . this trend indicates that the reddest galaxies are a different population than those right around @xmath147 . blue galaxies have preferentially exponential profiles . the concentration of galaxy profiles is highest for galaxies with @xmath147 or @xmath148 ( although , again , the median never reaches @xmath131 ) and declines at redder colors again . again , this trend suggests that the reddest galaxies are a special population ; in particular , the fact that they tend to have profiles closer to exponential suggests that they are edge - on spiral galaxies reddened by dust . this speculation is supported by the fact that the axial ratios measured by photo for these objects are low ( @xmath150@xmath151 ) . the half - light surface brightness increases for redder galaxies , again until @xmath147 or @xmath152 , at which point the surface brightness declines again . this trend may also be partly due to the fact that the reddest galaxies are probably dusty spirals ; however , it is also due to the trend that at the higher luminosities , where galaxy colors are redder , galaxy half - light surface brightnesses tend to get lower with luminosity . furthermore , the colors are clearly highly correlated among themselves . in particular , measuring @xmath64 constrains the other colors to very small ranges . we have so far explored only two - dimensional projections of an eight - dimensional space of parameters . it is interesting to explore a third dimension of this space . to do so , we divide the galaxies into two groups according to their srsic index . we choose an exponential group ( @xmath153 ) and a concentrated group ( @xmath154 ) . ( the results for the intermediate group look like a combination of the contributions from each group separately ) . figures [ emg_n0_num_cond ] and [ emg_n2_num_cond ] show the conditional distribution of all our galaxy properties on all of the others for the exponential and concentrated groups , respectively . it is easy to identify which group of galaxies is considered from the panels involving the srsic index for example , there is clearly no data for @xmath155 in figure [ emg_n0_num_cond ] . the diagonal panels in each plot ( which can be inferred from figure [ emg_num ] ) demonstrate that the srsic index does indeed divide galaxies into red , high surface brightness , luminous galaxies and blue , low surface brightness , underluminous galaxies . in particular , the luminosity function of the exponential group clearly is rising steeply through our low luminosity limit , while the luminosity function of the concentrated group is peaked at high luminosity . constrasting the conditional distributions of the exponential group in figure [ emg_n0_num_cond ] to those of the concentrated group in figure [ emg_n2_num_cond ] reveals that the properties of these populations have very different interrelationships . most strikingly , the color - magnitude diagrams of these two classes separate extraordinarily well , into a tight red sequence ( probably corresponding closely to the color magnitude relation for morphologically elliptical galaxies @xcite ) and a less tight , but undeniable , blue sequence . similarly , the dependence of surface brightness on luminosity differs markedly for the two classes . the surface brightness of exponential galaxies increases with luminosity ( as does their size ) , while the surface brightness of concentrated galaxies decreases with their luminosity ( while their size increases even more strongly ) . another way of looking at the same issue is to consider the dependence of the physical half - light radius on absolute magnitude ; figure [ emg_r50mn ] shows this relationship for three ranges of srsicindex @xmath79 . again , this figure shows that galaxy size closely correlates with luminosity , with a stronger dependence for de vaucouleurs than for exponential galaxies . these results are comparable to those in figure 44 of @xcite ; if one simply inverts the slope of the relationship between @xmath22 and @xmath156 ( which yields a correct result if the relationship between the two is sufficiently linear and deterministic ) we find @xmath157 , close to their result . these results also hold for the dependence of surface brightness on color ; surface brightness correlates with color for exponential galaxies , but peaks at around @xmath158 for concentrated galaxies . the results for concentrated galaxies are in good agreement with the work of @xcite ( based on morphologically elliptical galaxies ) . some of the most interesting results from these figures are the dependence of local density on galaxy properties for each type of galaxy profile taken separately . for example , the local density of exponential galaxies does not depend on surface brightness ; for de vaucouleurs , local density _ decreases _ slightly with surface brightness , a fact probably related to the decrease of luminosity with surface brightness for concentrated galaxies . it is only that higher surface brightness galaxies are more concentrated , and that concentrated galaxies are on average in higher density regions , that keeps the dependence of density on surface brightness flat for all galaxies . the dependence of the local density on luminosity has a similar _ shape _ for concentrated galaxies and exponential galaxies , in the range of luminosities in which there is data for both types ( less luminous than about @xmath159 ) . however , concentrated galaxies are naturally in slightly denser regions at any given luminosity . meanwhile , for both concentrated and exponential galaxies , density depends strongly on color . figures [ emg_n0_num_cond ] and [ emg_n2_num_cond ] show that exponential and concentrated galaxies have colors , surface brightnesses , and local densities which are distinct from one another and which depend on galaxy luminosity . to quantify these relations , figure [ cmrplot ] shows the @xmath104-weighted median properties of both types as the boxes for exponential ( thin lines ) and concentrated ( thick lines ) galaxies ( these lines are just the same as the median lines in figures [ emg_n0_num_cond ] and [ emg_n2_num_cond ] ) . the statistical uncertainties in the median containing 68% of the probability are shown as vertical lines ( the width of the distribution is of course much larger ) . we have fit these median relations in the following manner . first , we add a minimum variance in quadrature to the uncertainties on the points , which is @xmath160 for the colors and surface brightness , and is @xmath161 for density . we do so in order that our @xmath80 minimization does not obsessively fit the few points with extremely tiny statistical uncertainties ( who could easily be off by a small amount due to systematic errors in the data and our processing ) . next , we fit a cubic polynomial to the resulting set of points and uncertainties . to be explicit , for property @xmath162 we fit the function @xmath163 where @xmath128 is expressed in absolute magnitudes , the surface brightness is in magnitudes in 1 arcsec@xmath32 , the colors are in magnitudes , and @xmath164 is unitless . we center the fit at @xmath165 because that is near the center of our luminosity range . the resulting curves are shown in figure [ cmrplot ] as continuous lines . they are all reasonable fits to the data over the range we have fit . the parameters @xmath166 are listed in table [ cmrtable ] for each property and galaxy profile type . we also list the standard deviations @xmath167 defined by the distribution of galaxies around the best fit curves , both in figure [ cmrplot ] and in table [ cmrtable ] . however , as can be seen in figures [ emg_n0_num_cond ] and [ emg_n2_num_cond ] , gaussians of constant width are by no means adequate descriptions of the distributions around the median . these relations may be useful for comparison to theory or to other observations . they allow one to predict the median properties of an exponential or concentrated galaxy given only the absolute magnitude . we caution that because there is significant scatter in all of these relations , one can not reverse _ these _ fits to predict absolute magnitude from other galaxy properties . the two - dimensional projections of the number and luminosity density data used to create figures [ emg_num ] , [ emg_lum ] , [ emg_nmgmr ] , [ emg_n0_num_cond ] , [ emg_n2_num_cond ] , and [ emg_r50mn ] are available in fits format from the electronic version of this article * [ note to editor : these electronic tables will be required ; for now these can be found at + http://wassup.physics.nyu.edu / manyd / paper/]*. the fits file each have a single hdu containing three columns and one row . the first column contains @xmath168 names ( in string format ) of the quantities listed in that file ( corresponding to the @xmath169 and @xmath170 axes of the figures ) . the second column contains the ranges over those @xmath168 quantities included in the figure ( the first element of each range indicates the position of the left edge of the left - most pixel , and the second element of each range indicates the position of the right edge of the right - most pixel ) . the third column is a four dimensional array of dimensions @xmath171 where @xmath168 is the number of quantities , and @xmath172 is the number of pixels on a side of each image . each @xmath173 subarray contains the data values used to create the greyscales in the corresponding figure of this paper . the values in the subarray are per unit `` quantity @xmath174 '' per unit `` quantity @xmath175 '' , where @xmath174 and @xmath175 represent the position in the @xmath176 sized dimensions in the four dimensional array . in this paper , we have presented the number and luminosity density distributions of photometric galaxy properties as measured by the sdss . galaxy properties are highly correlated . here are our main conclusions : 1 . the most luminous galaxies comprise a homogeneous red , highly concentrated , high surface brightness population , and reside in locally dense regions . underluminous galaxies are less homogeneous , but in general are bluer , less concentrated , lower surface brightness , and reside in less dense regions . these results are in qualitative agreement with previous work . 2 . the relations among galaxy luminosities , surface brightnesses , and colors separate neatly once one has separated them into concentrated and unconcentrated ( exponential ) groups . some of these relationships are well - known ( such as the color magnitude relation of concentrated galaxies ; @xcite ) while others are less commonly discussed ( such as the color magnitude relation for exponential galaxies ) . 3 . local density is a strong function of luminosity , at least for the most luminous galaxies , and on color for galaxies of all colors . this dependence of clustering on galaxy type has been quantified many times previously , early on in the work of @xcite , later in the cluster studies of @xcite , and also on larger scales by @xcite , and many others . 1 . optical galaxy colors all correlate very strongly with @xmath134 color . the distribution of @xmath134 is double - peaked . most of the luminosity density in the universe is contained in galaxies within our range of galaxy luminosities ( @xmath177 ) . 3 . the very reddest galaxies ( which are small in number ) are in optical colors exponential galaxies , not concentrated galaxies . 4 . for all types of galaxies , size increases with luminosity . for concentrated galaxies , surface brightness decreases with luminosity for the most luminous galaxies . for exponential galaxies , surface brightness increases with luminosity . simple quantitative expressions of some of these results are contained in the parametric fits to the conditional distributions of figure [ cmrplot ] and table [ cmrtable ] as well as the two - dimensional projections given in the associated electronic tables . however , in future work we may characterize the distribution using a full seven - dimensional model of the distribution of properties . a potentially important effect to account for in future analyses is the passive evolution of galaxy luminosities . measurements of this evolution in the sdss suggest that galaxy luminosities are brighter in the past ( a trend of @xmath178 , @xmath179 , @xmath180 , @xmath180 , and @xmath181 magnitudes per unit redshift in the @xmath182 , @xmath183 , @xmath184 , @xmath125 , and @xmath185 bands ) . while the difference in absolute magnitude over the whole redshift range is quite small ( 0.3 magnitudes ) compared to the dynamic range in @xmath20-band absolute magnitude , some of the colors may be affected significantly . because the survey is flux limited , this effect can , in principle , alter the slope of our measurement of the color - magnitude relationship . for the trends quoted here , accounting for evolution would make the measured slope steeper , because the high luminosity galaxies are observed primarily at high redshift , and are observed to be bluer than they would be at the lower redshifts of the lower luminosity galaxies to which we are comparing them . for example , @xcite have found such evolution of the @xmath186 color magnitude diagram in their early - type galaxy sdss sample ( their figure 24 ) . we note that one can detect the effect of not accounting for evolution by comparing the results of blanton et al . ( in preparation ) for the evolution - corrected luminosity function at @xmath56 to the @xmath187-based estimate of the @xmath20-band luminosity function presented here ( our luminosity function extends to slightly higher luminosities than that estimate ) . the scope of this paper has not allowed us to delve into a more detailed discussion of the individual relationships and their relation to galaxy formation theory . however , we could not do justice to these topics in the space and time available here , and leave this for the future , encouraging readers to use the data distributed here to perform similar work . we note that @xcite have performed a similar analysis using fits to stellar population models . they also see bimodality in the distribution of galaxy properties , particular in their measure of the 4000 break . in addition , they find that because their estimated mass - to - light ratio ( in their case in the @xmath21-band ) is a strong ( and increasing ) function of mass or luminosity , the dependence of surface mass density in a galaxy on total stellar mass is monotonic , even though the dependence of surface brightness on luminosity shows a maximum surface brightness at around an absolute magnitude of @xmath188 . a particularly interesting application along these lines would be to consider the dependence of local density on stellar mass . two important photometric parameters missing from this space are color gradients and axis ratios . as it turns out , the color gradients of galaxies in our sample are usually quite small , partly because the central bulges of our spiral galaxies are blurred with the disks by seeing . nevertheless , using a restricted sample of nearby galaxies , one can characterize this distribution . the distribution of axis ratios is related to the three - dimensional shape of the galaxy , so understanding the conditional distribution of this quantity on other parameters will yield a better understanding of galaxy geometry . preliminary tests have shown that the axis ratios of our red , concentrated objects tend to be close to unity , whereas the axis ratios of blue , unconcentrated objects tend to differ significantly from unity , agreeing with the common understanding of the nature of elliptical and spiral galaxies . including a characterization of the two - dimensional shape of each galaxy would be an interesting way to expand the results found here . we began this paper by noting that many galaxy properties have been known for decades to be correlated , and in particular to be correlated with position on the hubble sequence . many readers will ask how galaxies of different morpholopgical classifications are distributed in this space , a question we could address using a nearby subsample of the galaxies studied here . we have not done so here because we believe that morphological classification is not a sufficiently specified measurement to be straightforwardly interpreted . the position along the hubble sequence is determined by most galaxy classifiers from a consideration of a galaxy s surface brightness , smoothness , concentration , axis ratio , the prominence of dust lanes , and spiral arm pitch angle . however , astronomy has not standardized the weights to be accorded by the classifier to each of these qualities of a galaxy image when placing each galaxy along the one - dimensional hubble sequence . in addition , the decision about a galaxy s classification is highly dependent on the observing conditions , especially the distance of the galaxy from the observer , the dynamic range of the image , the passband of the observation , and the angle from which the galaxy is observed . even when considering a single image and allowing only the classifier to vary , the repeatability of classification appears to be low ( @xmath189 in units of revised hubble @xmath190 type ; @xcite ) . even this level of repeatability is not clearly due to the fact that classifiers all weight the various hubble type criteria similarly , since the properties under consideration all correlate . nevertheless , as reviewed by @xcite , morphological classification is clearly important , since it does correlate with many physical properties of galaxies . furthermore , we _ can _ specify certain aspects which determine morphological classification . for example , many investigators have quantified measures of surface brightness , concentration , smoothness / blobbiness ( e.g. @xcite ) , and lopsidedness ( e.g. @xcite ) , among others , and investigated the dependence of these measures on observing conditions . these efforts , of which this paper is a part , are forming a new approach to quantitative morphology that we hope will have a more direct and better specified connection to theoretical predictions . mb and dwh acknowledge nasa nag5 - 11669 for partial support . mb is grateful for the hospitality of the department of physics and astronomy at the state university of new york at stony brook , who kindly provided computing facilities on his frequent visits there . funding for the creation and distribution of the sdss has been provided by the alfred p. sloan foundation , the participating institutions , the national aeronautics and space administration , the national science foundation , the u.s . department of energy , the japanese monbukagakusho , and the max planck society . the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium ( arc ) for the participating institutions . the participating institutions are the university of chicago , fermilab , the institute for advanced study , the japan participation group , the johns hopkins university , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , university of pittsburgh , princeton university , the united states naval observatory , and the university of washington . hogg , d. w. , blanton , m. r. , strateva , i. , bahcall , n. a. , brinkmann , j. , csabai , i. , doi , m. , fukugita , m. , hennessy , g. , ivezi , z. , knapp , g. r. , lamb , d. q. , lupton , r. , munn , j. a. , nichol , r. , schlegel , d. j. , schneider , d. p. , & york , d. g. 2002 , , 124 , 646 kauffmann , g. , heckman , t. m. , white , s. d. m. , charlot , s. , tremonti , c. , peng , e. w. , seibert , m. , bernardi , m. , brinkmann , j. , castander , f. , csabai , i. , fukugita , m. , ivezi ' c , z. , munn , j. , nichol , r. , thakar , a. , & weinberg , d. 2002 , submitted to ( astro - ph/0204055 ) percival , w. j. , baugh , c. m. , bland - hawthorn , j. , bridges , t. , cannon , r. , cole , s. , colless , m. , collins , c. , couch , w. , dalton , g. , de propris , r. , driver , s. p. , efstathiou , g. , ellis , r. s. , frenk , c. s. , glazebrook , k. , jackson , c. , lahav , o. , lewis , i. , lumsden , s. , maddox , s. , moody , s. , norberg , p. , peacock , j. a. , peterson , b. a. , sutherland , w. , & taylor , k. 2001 , , 327 , 1297 , k. , fukugita , m. , doi , m. , hamabe , m. , ichikawa , t. , okamura , s. , sekiguchi , m. , yasuda , n. , brinkmann , j. , csabai , i. . , ichikawa , s. , ivezi ' c , z. , kunszt , p. z. , schneider , d. p. , szokoly , g. p. , watanabe , m. , & york , d. g. 2001 , , 122 , 1238 smith , j. a. , tucker , d. l. , kent , s. , richmond , m. w. , fukugita , m. , ichikawa , t. , ichikawa , s .- i . , jorgensen , a. m. , uomoto , a. , gunn , j. e. , hamabe , m. , watanabe , m. , tolea , a. , henden , a. , annis , j. , pier , j. r. , mckay , t. a. , brinkmann , j. , chen , b. , holtzman , j. , shimasaku , k. , & york , d. g. 2002 , , 123 , 2121 , i. , ivezi ' c , . , knapp , g. r. , narayanan , v. k. , strauss , m. a. , gunn , j. e. , lupton , r. h. , schlegel , d. , bahcall , n. a. , brinkmann , j. , brunner , r. j. , budav ' ari , t. . , csabai , i. . , castander , f. j. , doi , m. , fukugita , m. , gyry , z. , hamabe , m. , hennessy , g. , ichikawa , t. , kunszt , p. z. , lamb , d. q. , mckay , t. a. , okamura , s. , racusin , j. , sekiguchi , m. , schneider , d. p. , shimasaku , k. , & york , d. 2001 , , 122 , 1861 cccrrrrr concentrated ( @xmath191 ) & @xmath63 & mags & 0.321 & 1.697 & -0.0338 & 0.0023 & -0.01005 & @xmath64 & mags & 0.068 & 0.937 & -0.0238 & -0.0039 & -0.00400 & @xmath65 & mags & 0.036 & 0.410 & -0.0082 & -0.0007 & -0.00077 & @xmath66 & mags & 0.076 & 0.322 & 0.0132 & -0.0132 & -0.00053 & @xmath129 & mags in 1 arcsec@xmath192 & 0.542 & 19.131 & -0.1866 & 0.0759 & 0.00981 & @xmath164 & & 0.344 & 0.373 & -0.0413 & 0.0168 & -0.00269 & @xmath193 $ ] & & 0.109 & 0.305 & -0.2370 & 0.0154 & 0.00277 exponential ( @xmath153 ) & @xmath63 & mags & 0.262 & 1.142 & -0.0712 & 0.0095 & -0.00076 & @xmath64 & mags & 0.143 & 0.621 & -0.0606 & -0.0045 & 0.00175 & @xmath65 & mags & 0.063 & 0.352 & -0.0205 & -0.0006 & 0.00013 & @xmath66 & mags & 0.132 & 0.210 & -0.0139 & -0.0194 & 0.00243 & @xmath129 & mags in 1 arcsec@xmath192 & 0.600 & 20.266 & 0.2545 & 0.0348 & -0.00230 & @xmath164 & & 0.371 & 0.283 & 0.0080 & -0.0298 & 0.00782 & @xmath193 $ ] & & 0.123 & 0.533 & -0.1468 & 0.0061 & -0.00030
dwarf galaxies are the most common type of galaxy in the universe and are thought to be the progenitors of larger structures . therefore , a clear understanding of dwarf galaxy evolution is the first step toward a larger understanding of general galaxy evolution . fortunately , dwarf galaxies are also comparatively simple systems that can be effectively modeled and simulated . in particular , the satellite galaxies of the milky way , by virtue of being nearby , compact and situated in the sparser regions of the galactic halo , have proven to be exceptional laboratories for studying models of stellar and chemical evolution ( see , e.g. mateo 1998 ) . a useful tool for studying the dwarf galaxies is the comparison of their chemical properties to those of the well - studied milky way globular clusters . some globular clusters show evidence of multiple stellar populations ( omega centauri ( lee et al . 1999 ) ; ngc 1851 ( milone et al . 2008 ) , ngc 2808 ( piotto et al . others , particularly those associated with dsph galaxies or tidal streams , show peculiar chemical enrichment histories and/or young ages ( see , e.g. , letarte et al . however , most _ inner halo _ milky way globular clusters formed their stars a hubble time ago in a single burst ( marin - franch et al . since the early universe was dominated by massive stars , it is likely that only core collapse supernovae ( type ii sn ) were able to enrich the primordial interstellar medium ( ism ) before the globular clusters formed . these events would have enriched the ism with both fe and @xmath2-elements , resulting in relatively high @xmath2-element to iron ratios [ @xmath2/fe ] . by contrast , objects that formed later , after type ia sn had time to enrich the ism with more fe , would have lower relative @xmath2-element ratios . most dwarf galaxies have compound populations , having undergone slow sporadic star formation over their lifetimes ( see review in mateo 1998 ) . while their earliest populations were formed in the aftermath of the the first snii events , their later populations benefited from ongoing snia . the result is a steep decline in @xmath2-element abundance ratios with decreasing age / increasing metallicity . the most metal - poor stars have @xmath2-element ratios similar to globular clusters , while the more metal - rich populations are deficient in @xmath2-elements ( shetrone et al . 1998 , 2001 , 2003 ; tolstoy et al . 2003 ; venn et al . 2004 ; smith et al . 2006 ; helmi et al . 2006 ) . the chemistry of dwarf spheroidal ( dsph ) stars is of particular importance in untangling the hierarchical formation of the milky way . cdm cosmological models predict that galaxies like the milky way have grown through the merging of low - mass systems with perhaps some low - level merging continuing at the present time . among the current retinue of dsph satellite galaxies , the sagittarius dsph shows clear signs of having been torn apart by the milky way potential ( majewski et al . 2003 ) and ursa minor , leo i , coma berenices and ursa major ii also hint at a violent dynamical past ( palma et al . 2003 ; sohn et al . 2007 ; simon & geha 2007 ) . however , while the most metal - poor stars of the dsph galaxies are chemically similar to the most metal - poor stars of the galactic halo , the enrichment of snia kicks in at lower fe abundances in the dsph galaxies , resulting in consistent deficits in the @xmath2-element ratios of dsph stars compared to halo stars of similar [ fe / h ] ( venn et al . 2004 ) . this would indicate that the galactic halo formed from objects that had more rapid and therefore more snii - dominated chemical evolution and that objects like the current dsph galaxies have contributed little to the halo . this argument is , however , undermined by recent studies indicating that the stars the sagittarius dsph is contributing to the halo differ from those in the residual core ( bellazzini et al . 2006 ; chou et al . 2007 ; siegel et al . 2007 ) . if the dsph galaxies are the remnants of much larger objects that have been disrupted over a hubble time , the _ present _ stellar populations of the dsph galaxies may not reflect the _ initial _ stellar populations they would have preferentially contributed to the galactic halo . moreover , if the stellar populations of the dsph galaxies have not evolved in isolation , their chemical abundances may have been affected by continual dynamical harassment . these arguments over dsph stellar chemistry could be clarified by the examination of more distant dwarf galaxies that have not been perturbed by the milky way potential . if an isolated dsph were shown to have the same chemical abundance patterns as the relatively nearby dsphs , such a discovery would ( a ) strengthen the argument that the lack of halo stars with dsph - like " abundance patterns suggests that the vast majority of the halo formed very quickly and that the dsphs have contributed very few stars since this initial construction phase ; and ( b ) show that the @xmath2-element deficits have nothing to do with dynamical stirring and are entirely the product of dsph stellar and chemical evolution . in an effort to address these issues , we now expand the study of chemical abundances to the leo ii dsph galaxy . leo ii is the second most distant dsph galaxy assumed to be orbiting the milky way ( 218 kpc , bellazzini et al . 2005 ; siegel et al . 2008 , hereafter s08 ) . it is predominantly metal - poor ( hodge 1982 ; demers & harris 1983 ; aaronson et al . 1983 ; azzopardi et al . 1985 ; suntzeff et al . 1986 ; lee 1995 ; demers & irwin 1993 ; koch et al . 2007a , b ; bosler et al . 2007 , hereafter bss07 ) and dominated by intermediate age populations ( mighell & rich 1996 ; dolphin et al . 2005 ; bellazzini et al . 2005 ; gullieuszik et al . it has a handful of carbon stars ( azzopardi et al . 1985 ) , a rich population of rr lyrae variable stars ( swope 1967 , 1968 ; siegel & majewski 2000 ) , a high central velocity dispersion ( vogt et al . 1995 ; koch et al . 2007b ; bss07 ; s08 ) and an extended stellar distribution ( coleman et al . 2007 ; s08 ) . at present , however , the indications are that is had undergone little , if any , interaction with the galactic potential ( koch et . al 2007b ; coleman et al . 2007 ; walker et al . 2007 ; s08 ) and its modest radial velocity ( + 26.2 km s@xmath3 , s08 ) would be consistent with a circular orbit that does not bring it close to the milky way . the bss07 study cited above analyzed medium resolution nir spectra of a large number of leo ii rgb stars in an effort to measure radial velocities and metallicities . these spectra are dominated by the calcium ii infrared triplet ( cat ) features but also have numerous weak fe , ti , and mg lines that have yet to be exploited . in this paper , we show that these weak lines can be used to measure metallicity and abundance ratios for stars of sufficient s / n through the use of synthetic model spectra . this method has allowed us to study the chemical evolution of leo ii , to constrain the relationship between the equivalent width of the cat lines and the [ fe / h ] abundances as measured from fei lines and to make a preliminary examination of the age - metallicity relationship ( amr ) of the brightest leo ii rgb stars . 2 of this paper discusses the additional reduction and analysis we have applied to the bss07 data . 3 compares the abundances ratios measured in leo ii to those measured in the globular clusters to provide insight into the chemical evolution of leo ii . 4 examines the inferred ages of our bright leo ii stars while 5 summarizes our findings . a full description of the observations and data reduction can be found in bss07 but we give a brief summary of the observations here . low - dispersion spectra of red giants in the leo ii dsph were taken using the keck i 10-m telescope and lris ( oke et al . 1995 ) in 2002 and 2003 . the lris was configured with the 1200 l / mm grating blazed at 7500 which gives a dispersion of 0.62 per pixel and resolution of 1.55 (r @xmath4 10 000 ) . twenty to thirty stars were observed on each slit mask creating a total sample of 74 stars with snr between 10 and 44 . traditional equivalent width analysis of these spectra proved impossible due to the heavy blending of many lines and the difficulty in setting the continuum measure . however , these problems can be overcome by comparison of the observed spectra to synthetic spectra created by the 2007 version of the lte line - analysis code moog ( sneden 1973 ) . the initial line list was obtained from the r. l. kurucz cd - rom 23 . in order to further calibrate this line list we compared the list with the nearby red giant star , arcturus ( hinkle et al . we adopted atmospheric parameters for arcturus (= 4280 , = 1.55 , @xmath5 = 1.61 km s@xmath3 , [ fe / h ] = -0.50 , [ ti / fe ] = + 0.26 , [ mg / fe ] = + 0.39 and [ ca / fe ] = + 0.21 ) by combining the results of mcwilliam ( 1990 ) , fulbright et al . ( 2006,2007 ) and koch & mcwilliam ( 2008 ) . since we are performing a differential analysis , we chose to adjust the gf values in our synthetic spectrum only for lines that were significantly discrepant from the catalog arcturus spectrum , rather than adjust all 8345 lines . we smoothed the model spectra with a gaussian kernel of fwhm equivalent to the bss07 spectral resolution of 1.55 . initial metallicity estimates for globular cluster and leo ii stars were taken from bss07 . effective temperatures were calculated from interpolation of the temperature - color - metallicity relationship of ramirez & melendez ( 2005 ) and the @xmath6 photometry of s08 , the latter allowing the leo ii and cluster data to be analyzed using the same photometric system . surface gravities were calculated from the temperatures and bolometric - corrected magnitudes assuming a mass of 0.8 @xmath7 . we used distance moduli of 15.07 , 13.76 , 15.59 , 15.19 and 21.61 for m3 , m107 , ngc 1094 , ngc 4590 and leo ii , respectively . values by approximately 0.04 , an insignificant correction . ] the microturbulence was estimated as @xmath8 , in accordance with recent analyzes of globular cluster giants near the tip of the giant branch ( e.g. sneden et al . 2004 , lee , carney & habgood 2005 , ivans et al . 2001 , cohen & melendez 2005 ) . microturbulent velocities ranged from 1.4 km s@xmath3 to 2.0 km s@xmath3 in the globular cluster sample and from 1.6 to 2.0 km s@xmath3 in the leo ii sample . these parameters were fed into moog to produce preliminary synthetic spectra . we then adjusted the [ fe / h ] metallicity of the atmospheres until the fe lines matched the observed spectrum . a weighted mean of all the iron lines was used to determine the best [ fe / h ] value for each star . the model was then set to this metallicity and the abundance ratios were determined by holding all elements constant while varying the desired @xmath2-element . the best fit was determined by the output difference between the moog synthetic spectra and observed spectra . we found that we were able to effectively measure abundances for stars with snr @xmath0 per pixel . the setting of the continuum was critical for a proper analysis of the weak lines in these medium resolution spectra . to refine the continuum levels we created a synthetic spectrum for each star , divided it into the observed spectrum and fit the residual with a high order spline . this fitted spline was then divided into the observed spectrum to remove high order terms in the continuum placement . if our modeled spectrum changed significantly , e.g. when a new overall metallicity change was required , we repeated this procedure for the new model . we attempted to determine abundance ratios for all of the moderate lines in the spectra . these lines include fe , ca , na , mg , and ti . figure 1 compares one of our highest snr spectra to several synthetic spectra with varied fe and ti abundances . there are a large number of lines in this region with varying strength that can be used to determine abundances . for a more typical snr spectrum , most of these lines are not strong enough for abundance analysis . we identified sufficient ti and fe lines in all of our bright stars but the na lines and all but one of the mg lines were absent in most of the spectra . thus we limit our analysis to the stronger fe , ca , mg and ti lines listed in table 1 . the only ca lines in the observed spectral region are the extremely strong ca triplet lines . the cat lines are difficult to model with synthetic spectra ( hence the more common use of globular cluster - calibrated equivalent width analysis ) . at a constant ca abundance , the strength of the cat lines decreases as the electron pressure increases due to changes in the continuous opacities . for cool stellar atmospheres , the main sources of electrons are mg , fe , si , ca , na and al , but which element contributes the most depends in the effective temperature of the star and which layer of the atmosphere is surveyed . for example , for a = 4200 k star at 0.01 optical depth , the main sources of electrons are mg and fe ; while for a = 4000 k star at 0.01 optical depth , the main sources of electrons are ca , mg , na and al . thus , as the @xmath2-abundance declines , the continuous opacity changes to counteract the ca abundance decline . for example , a 0.5 dex change in the electron contributors will increase the cat line strength enough to mimic a ca abundance increase of 0.38 and 0.22 dex for stars of = 4000 , = 0.7 , [ m / h ] = -1.1 and = 4200 , = 0.7 , [ m / h ] = -1.9 , respectively . this means that as [ m / h ] rises , any decline in the [ ca / fe ] ratio would be muted in the measured cat lines . as a further complication , the cat lines become less sensitive to the ca abundance as the lines strengthen . for the globular cluster sample , the plez model atmospheres with [ @xmath2/fe]=+0.4 enhanced abundance ratios were appropriate and there was no need to change the model . however , our initial analysis showed that the ti and mg ratios of the leo ii stars decline with increasing metallicity ( 3 ) . this resulted in stars with low @xmath2-abundances having erroneously high ca abundances . to derive accurate ca abundance , we were therefore required to iterate . we first determined the fe abundance for each star and then adjusted the model appropriately . we next determined the ti and mg abundances , from which we calculated a preliminary @xmath2-abundance . finally , we adjusted the model abundance based on the average @xmath2-abundance at each star s [ fe / h ] . the result was that for the more metal - rich leo ii stars , we used model atmospheres that were more metal - poor than the fe abundance would imply , thus compensating for the lack of electrons owing to the lower @xmath2-abundances . measurement errors were determined for each fitted feature by adjusting the abundances up and down until the residuals of the fit were larger than the surrounding continuum regions , ie . larger than the snr . the measurement errors were then propagated to yield a single measurement error which was added in quadrature to the abundance errors from the modeling uncertainties k for , @xmath10 dex for , @xmath11 for v@xmath12 and @xmath10 dex for [ m / h ] . typically the measurement errors dominated over the modeling errors particularly for [ mg / fe ] which was derived from a single spectral feature . tables 2 and 3 list the globular cluster and leo ii stars from bss07 for which we have been able to determine abundances while tables 4 and 5 show the abundances derived from the moog software . we have recalculated the signal to noise ratio ( snr ) from the residual of the best fit synthetic spectra to the observed spectra and list that snr in tables 2 and 3 . figure 2 shows the photometrically derived effective temperatures and surface gravities for the leo ii sample . there is a clear rgb sequence , corresponding to the photometric sequence upon which the temperatures and gravities are based . the two outlier stars 195 and 166 are metal - poor radial velocity members . they are , however , according to s08 , well - removed from the dominant leo ii rgb locus in color - magnitude space although both have washington+ddo51 photometry consistent with giant stars . it is possible that these are interloping giants in the field , possibly from the metal - poor debris stream of the sagittarius dsph galaxy passing in front of leo ii ( s08 ) . alternatively , they could represent a young , metal - poor population in leoii ( 4 ) . given their outlier status , we have chosen to remove these two stars from the analysis of mean trends in the leo ii sample . figure 3 compares our derived [ fe / h ] metallicities against those derived from the cat lines by bss07 . on average , the cat - based metallicities are 0.17 dex higher than the our model [ fe / h ] metallicities with an rms of 0.18 dex . the abundances are correlated but there is a positive residual which shows increasing significance with decreasing metallicity . the offset is likely due to several factors . first , all of our fe lines are fei lines and it is possible that metal - poor giants overionize fe ( thevenin & idiart 1999 ; asplund & garcia - perez 2001 ) . in m3 , for example , sneden et al . ( 2004 ) find an underabundance factor of -0.13 in the fei lines and they adopt the feii abundances as the true metallicity . in m68 lee , carney & habgood ( 2005 ) find an underabundance factor of -0.37 in the fei lines and also adopt the feii abundances as the true metallicity . however , not all authors find these overionization factors . cohen & melendez ( 2005 ) find that the fei and feii abundances agree to within uncertainties . kraft & ivans ( 2003 ) find small and variable overionization in metal - poor globular clusters but of much smaller magnitude than thevenin & idiart ( 1999 ) predict . in light of the disagreement , we have chosen not to make any correction for overionization but will discuss the potential impact in later sections . a second potential source of the discrepancy between the [ m / h]@xmath13 and our derived [ fe / h ] comes from the calibration of the cat metallicity scale , which in bss07 is based on the carretta & gratton ( 1997 ) metallicity scale . the choice of calibration system can significantly impact the zero point of the metallicity . battaglia et al . ( 2008 ) , for example , find a 0.1 dex difference between the carretta & gratton ( 1997 ) metallicity scale and their high resolution analysis over the range of metallicity spanned by leo ii s rgb stars . we can explore the issue of the metallicity zero point by comparing the fei metallicities we derive for the bss07 globular clusters to the high resolution analyzes listed in pritzl , venn & irwin ( 2005 ) . the average abundances we derive for ngc 1904 , ngc 4590 and m3 are -0.14 ( @xmath14 ) dex more metal - poor than those listed in pritzl et al .. however , their ngc 1904 abundances come solely from the study of gratton & ortolani ( 1989 ) . pritzl et al . find that the gratton & ortolani abundances are typically 0.16 more metal - rich than more recent analyzes . correcting the ngc 1904 abundance down by 0.16 dex , we find our calculated fe i abundances for ngc 1904 , ngc 4590 and m3 are -0.08 ( 0.14 ) dex more metal - poor than the pritzl et al . values a zero point offset within the error of the weighted mean abundances listed in table 4 . figures 4 and 5 show the derived [ ti / fe ] and [ mg / fe ] abundance ratios , respectively , for both the leo ii and globular cluster sample . the leo ii and globular cluster star [ ti / fe ] abundances overlap at the poorest metallicities but depart dramatically at higher metallicities . the leo ii [ mg / fe ] ratios are offset from the globular cluster sample and show a pronounced decline with increasing metallicity . the large error - bars on the [ mg / fe ] limit the usefulness of the individual measurements but confirm the same trend seen in the [ ti / fe ] abundance ratios . we note that the [ ti / fe ] ratios in m107 are as high if not higher than those found in the more metal - poor globular clusters . such large [ ti / fe ] abundance ratios are not unprecedented . sneden et al . ( 1994 ) found that the relatively metal - rich globular cluster m71 had high ti abundance ratios ( [ ti / fe ] = + 0.5 ) , although ramirez & cohen ( 2002 ) found " normal ti ratios m71 . figure 6 shows the [ caii / fei ] abundance ratios for the globular cluster and leo ii samples . there is a slight decline in the [ caii / fei ] ratio with increasing metallicity among the globular cluster sample . the leo ii ca abundance ratios are offset from the globular cluster sample and the slope of the best fit line is slightly steeper than that of the cluster sample . this is very similar to the trends in mg abundances seen in figure 5 . for m3 , ngc 4590 and ngc 1904 , we compare our abundance ratios to the compilation of pritzl , venn & irwin ( 2005 ) . our [ ti / fe ] ratios are @xmath15 dex larger , our [ mg / fe ] ratios are @xmath16 dex larger , and our [ ca / fe ] ratios are @xmath17 dex larger . the only significant zero point is in our [ caii / fei ] abundances . this is not entirely unexpected . most analyzes compare ionized species to ionized species and neutral species to neutral species . we would obviously prefer to make a similar comparison , but the previously noted lack of feii lines prevents this . additionally , as noted above , the literature is divided on the over - ionization correction need to converted from fei to feii abundances . such corrections are likely to be metallicity dependent and could be the cause of the slope in the globular cluster sample . because we do not make a correction , the absolute value of our [ caii / fei ] abundances is uncertain . however , because our analysis is concerned with _ relative _ differences between the clusters and leo ii , such a zero point correction will have little impact on our conclusions . figure 7 show the [ ca / h ] ratios derived from our analysis compared to the [ ca / h ] ratios derived from the cat by bss07 . the overlap of the cluster and leo ii sample in this figure suggests that the corrections we made for the more metal - rich low-@xmath2 stars are consistent with ca abundances derived with the cat methodology . the slopes in these samples are less steep than those of figure 3 as would be expected since the cat lines should be sensitive to the overall ca abundance . however , there is a residual slope between the two [ ca / h ] scales , which remains unexplained . figure 8 shows the mean of the [ ca / fe ] , [ ti / fe ] and [ mg / fe ] ratios . the fit to the data with the lowest formal reduced chi square is a second order legendre polynomial which exhibits a steep decline in the @xmath2-abundance ratios of the leo ii stars with metallicities above [ fe / h]@xmath4 - 2.0 . this is consistent with what is seen in other dsph galaxies ( see references in 1 ) and is likely due to slower chemical evolution and concordant increased contribution of snia to leo ii . however , the data would also be consistent with a step function near [ fe / h]@xmath4 - 1.8 in which the metal - poor stars have globular cluster like @xmath2-abundances while the more metal - rich stars are underabundant in @xmath2 elements . the slow enrichment implied by our abundance analysis suggests that leo ii should have an age - metallicity relationship ( amr ) in which the more metal - rich stars are younger than the more metal - poor stars and have therefore been subject to more enrichment by snia . an amr would not be immediately obvious in color - magnitude diagrams due the photometric degeneracy of age and metallicity . however , our precise spectroscopic survey affords the ability to remove abundance from the equation . we can now combine precise broadband photometry with modern isochrones to approximate individual stellar ages . this method has had some success in untangling the stellar populations of the omega centauri globular cluster , for example ( hilker et al . 2004 ; sollima et al . 2005 ) and was recently applied to the leo ii dsph galaxy by koch et al . ( 2007a ) . the koch et al . study found that leo ii has undergone steady star formation with little enrichment , starting 9 gyr ago and continuing up until 2 gyr ago . more recently - formed stars appear to be somewhat enriched . our analysis uses similar data but with the advantage that we have measured @xmath2-abundances , rather than assumed abundances . because our stars are near the tip of the rgb , rather than the more age - sensitive turnoff , any analysis will necessarily be imprecise ( koch et al . estimate uncertainties of 40% using these techniques ) . the ideal stars for this analysis are those near the age - sensitive main sequence turnoff ( msto ) . however , given that leo ii s msto is well beyond the spectroscopic range of even the largest ground - based telescopes , the rgb is our best option for probing leo ii s age distribution . figure 9 demonstrates the analysis . photometry is taken from s08 , which has a precision of 0.01 mag in all passbands at the tip of the red giant branch ( trgb ) . @xmath18 colors were calculated by converting the washington measures in s08 using transformations in majewski et al . isochrones are taken from the dartmouth stellar evolution program ( dotter et al . 2007 ) and shifted to a distance modulus of @xmath19 and reddening of @xmath20=0.02 , in accordance with s08 s trgb distance derived from the same data . cursory examination of the individual panels , broken down by passband and metallicity bin , shows no obvious amr in leo ii . to make a more precise estimate , we interpolated the isochrones in 0.01 magnitude intervals and , with metallicity fixed at the spectroscopic value and [ @xmath21 fixed to the mean [ @xmath22 $ ] at each star s [ fe / h ] , we found the isochrone nearest to each star s photometric position . we estimated age uncertainties by offsetting the [ fe / h ] of each star by its uncertainty ( which dominates over s08 s small photometric uncertainties ) . the left panels of figure 10 show the resultant age - metallicity distribution . although the uncertainties are significant ( typically 2 - 3 gyr ) , some interesting trends can be seen . the mean age of the stars is @xmath23 gyr but has a spread significantly greater than the uncertainty , implying ongoing low - level star formation . the younger stars appear to be slightly more metal - rich and indeed , an amr that shows slight enrichment with age either gradually or step - wise would be more consistent with the data than a flat trend . however , we can not statistically rule out a flat amr , which would have a @xmath24 per degree of freedom of 0.9 . in fact , our derived amr is very similar to of koch et al . it should be noted that the exploration of the rgb stars has number of complications that could affect the inferred amr . in particular : @xmath25 leo ii is known to have prominent asymptotic giant branch ( agb ; mighell & rich 1996 ; gullieuszik et al . 2008 ) . stars on the agb will overlap young rgb isochrones and perhaps 10% of our stars could be agb stars . the most likely candidates would be the two apparently young metal - poor stars that are clear outliers in the bottom panels of figure 9 and both panels of figure 10 ( note that these are different stars than the two outliers rejected in figures 2 - 8 , which were excluded _ a priori _ from the age analysis ) . we attempted to measure ages for these stars using the padova isochrones ( marigo et al . 2008 ) , which include an agb . however , these stars are too blue for all but the youngest agb isochrones ( at these magnitudes , the rgb and agb are separated by only a few .01 mag ) . removing these outlier stars from the sample , however , increases the contrast between the younger and older stars of leo ii , making enrichment more tenable . but a flat amr would still be within the uncertainties ( @xmath26 ) . @xmath25 any age analysis will be sensitive to the distance modulus of leo ii . adjusting the distance of leo ii shifts the absolute ages , but not the relative ages . the exception is at short distance moduli , where the stars are too faint to correspond to anything but the maximum age in our isochrone set ( 15 gyr ) . if we assume that leo ii s stars must be younger than 15 gyr , then leo ii must be at a distance of at least @xmath27 . longer distances result in younger absolute ages but have little effect on the inferred amr . @xmath25 the overionization correction mentioned in 2 could alter the amr by shifting the sample to a more metal - rich distribution . we applied a 0.2 dex correction ( right panels of figure 10 ) and find that the amr tightens somewhat , producing a sharper peak in the age distribution at around 6 - 7 gyr . however , a flat amr would still be within the uncertainties ( @xmath26 ) , even if the two outlier stars are removed . @xmath25 our snr - limited sample is unable to probe the fainter more metal - rich stars in leo ii . improved spectra of these stars would extend the amr over a greater range of metallicity , providing better constraint of the enrichment and allowing a comparison to toy models of enrichment . in particular , it would determine if the apparent rise in abundance for the younger stars in figure 10 is merely statistical scatter or is the indicator of a larger trend . even with these caveats , the age - metallicity distribution of leo ii stars , while giving tantalizing hints of enrichment , remain statistically consistent with no enrichment . a flat enrichment would be unexpected for leo ii , since we _ do _ find that the more metal - rich stars have lower @xmath2-element abundances , suggesting gradual enrichment by snia over the course of several gyr . in the end , while this exercise demonstrates that precise photometry and spectroscopy can be used to probe the amr of objects out to 200 kpc , the enrichment of leo ii may be too subtle to be detected in our data . spectroscopic examination of the more numerous and age - sensitive turnoff stars may be the only way to tease out the amr of this distant and enigmatic galaxy . our analysis of the weaker spectral features in the bss07 spectra provides critical information about not only the leo ii dsph galaxy but other spectroscopic studies as well . @xmath25 we find the fe - line metallicities disagree with cat - based metallicities , although it is likely this is a result of the overionization of fe at low metallicity . we are unable to analyze the fainter more metal - rich stars to confirm the high - metallicity tail identified by bss07 . @xmath25 we confirm that leo ii , like the other low - luminosity dsph galaxies , has globular cluster - like @xmath2-abundances in its most metal - poor stars but shows a declining @xmath2-abundance at higher metallicities , likely due to the increasing influence of snia . @xmath25 leo ii s amr gives tantalizing hints of slow enrichment . however , a flat amr with no enrichment can not be ruled out with the present data , given the inherent uncertainties in analyzing only the brightest stars in the galaxy . nevertheless , this demonstrates that precise spectroscopic and photometric data can be used to get a general picture of the enrichment history of distant dsph galaxies . leo ii s particular influence on models of dwarf galaxy evolution lies in its difference from the other dsph galaxies . as noted in 1 , leo ii appears to have had little interaction with the milky way . this would indicate that the @xmath2-abundance patterns found in the dsph galaxies are a universal feature of dwarf galaxies and have no relation to any dynamical interaction with a larger parent galaxy . it also strengthens the contention that while objects like the present retinue of dsphs have contributed somewhat to the construction of the milky way given the similarities between their most metal - poor stars and some chemically peculiar stars within the halo the stark chemical dissimilarities between the bulk of the dsph and field halo stars indicates that this contribution has been small . further investigation is needed into the zero point issues intrinsic in our analysis so that we may more confidently translate measured fei abundances into intrinsic feii abundances . this could be accomplished by looking at existing high snr cat spectra of globular clusters , especially those known to be deficient in @xmath2-elements or those known or suspected to have been stripped from larger dsph galaxies . once the zero point issues are understood , the technique we have unveiled in this paper can be applied to the large cat spectral datasets in the literature to derive @xmath2-abundance measures for thousands of dsph and field halo stars . better and easier chemical fingerprinting of distant rgb stars has the potential to finally unravel the hierarchical history of the milky way . moreover , this technique expands the range of current and future investigations into dsph chemistry . it shows that current telescopes can probe dsph chemistry out to at least 200 kpc and indicates that next - generation telescopes will open up investigation into even more isolated dwarfs such as the phoenix dirr / dsph ( tobolewski et al . , in prep . ) this will expand the study of galactic chemical evolution into an even more diverse array of galactic environments . at this time , however , the existing models of dwarf galaxy chemical evolution appear to be valid out to at least 200 kpc from the galaxy . aaronson , m. , olszewski , e. w. & hodge , p. w. 1983 , , 267 , 271 asplund , m. , & garca prez , a. e. 2001 , , 372 , 601 azzopardi , m. , lequeux , j. & westerlund , b. e. 1985 , , 144 , 388 battaglia , g. , irwin , m. , tolstoy , e. , hill , v. , helmi , a. , letarte , b. , jablonka , p. 2008 , , 383 , 183 bellazzini , m. , gennari , n. & ferraro , f. r. , 2005 , , 360 , 185 bellazzini , m. , correnti , m. , ferraro , f. r. , monaco , l. , & montegriffo , p. 2006a , , 446 , 1 bosler , t.l . , smecker - hane , t.a . , stetson , p.b . 2007 , , 378 , 318 ( bss07 ) carretta , e. & gratton , r.g . 1997 , , 121 , 95 chou , m .- y . , et al . 2007 , , 670 , 346 cohen , j. g. , & melndez , j. 2005 , , 129 , 303 coleman , m. , jordi , k. , rix , h .- w . , grebel , e. k. & koch , a. , 2007 , , 134 , 1938 demers , s. & harris , w. e. 1983 , , 88 , 329 demers , s. & irwin , m. j. 1993 , , 261 , 657 dolphin , a. e. , weisz , d. r. , skillman , e. d. & holtzman , j. a. , 2005 , _ astro - ph/0506430 _ fulbright , j. p. , mcwilliam , a. , & rich , r. m. 2006 , , 636 , 821 fulbright , j. p. , mcwilliam , a. , & rich , r. m. 2007 , , 661 , 1152 gullieuszik , m. , held , e. v. , rizzi , l. , girardi , k. , marigo , p. & momany , y. , 2008 , , _ in press , astro - ph/0805.0735 _ gratton , r. g. , & ortolani , s. 1989 , , 211 , 41 helmi , a. , et al . 2006 , , 651 , l121 hilker , m. , kayser , a. , richtler , t. , & willemsen , p. 2004 , , 422 , l9 hinkle , k. , wallace , l. , valenti , j. , & harmer , dianne 2000 , visible and near infrared atlas of the arcturus spectrum 37279300 , ( san francisco : asp ) . hodge , p. w. , 1982 , , 87 , 1668 ivans , i. i. , kraft , r. p. , sneden , c. , smith , g. h. , rich , r. m. , & shetrone , m. 2001 , , 122 , 1438 koch , a. , grebel , e. k. , kleyna , j. t. , wilkinson , m. i. , harbeck , d. r. , gilmore , g. f. , wyse , r. f. g. , & evans , n. w. 2007a , , 133 , 270 koch , a. , grebel , e. k. , kleyna , j. t. , wilkinson , m. i. , harbeck , d. r. , gilmore , g. f. , wyse , r. f. g. & evans , n. w. 2007b , , 133 , 270 koch , a. , & mcwilliam , a. 2008 , 135 , 1551 kraft , r. p. , & ivans , i. i. 2003 , , 115 , 143 lee , m. g. , 1995 , , 110 , 1155 lee , y .- w . , joo , j .- m . , sohn , y .- j . , rey , s .- c . , lee , h .- c . , & walker , a. r. 1999 , , 402 , 55 lee , j .- w . , carney , b. w. , & habgood , m. j. 2005 , , 129 , 251 letarte , b. , hill , v. , jablonka , p. , tolstoy , e. , franois , p. , & meylan , g. 2006 , , 453 , 547 majewski , s. r. , ostheimer , j. c. , kunkel , w. e. , & patterson , r. j. , 2000 , , 120 , 2550 majewski , s. r. , skrutski , m. f. , weinberg , m. d. & ostheimer , j. c. , 2003 , 599 , 1082 marigo , p. , girardi , l. , bressan , a. , groenewegen , m. a. t. , silva , l. , & granato , g. l. 2008 , , 482 , 883 marin - franch , a. , aparicio , a. , piotto , g. , rosenberg , a. , chaboyer , b. , sarajedini , a. , siegel , m. , anderson , j. , bedin , l. r. , dotter , a. , hempel , m. , king , i. , majewski , s. , milone , a. , & reid , i. n. , , _ submitted _ mateo , m. l. 1998 , , 36 , 435 mighell , k. j. & rich , m. r. 1996 , , 111 , 777 milone , a. p. , et al . 2008 , , 673 , 241 oke , j. b. , et al . 1995 , , 107 , 375 palma , c. , majewski , s. r. , siegel , m. h. , patterson , r. j. , ostheimer , j. c. & link , r. , 2003 , 125 , 1352 piotto , g. , et al . 2007 , , 661 , l53 pritzl , b. j. , venn , k. a. , & irwin , m. 2005 , , 130 , 2140 ramrez , s. v. , & cohen , j. g. 2002 , , 123 , 3277 ramirez , i. & melendez , j. 2005 , , 626 , 465 shetrone , m. d. , bolte , m. , & stetson , p. b. 1998 , , 115 , 1888 shetrone , m. d. , ct , p. , & sargent , w. l. w. 2001 , , 548 , 592 shetrone , m. , venn , k. a. , tolstoy , e. , primas , f. , hill , v. , & kaufer , a. 2003 , , 125 , 684 siegel , m. h. & majewski , s. r. , 2000 , , 120 , 284 siegel , m. h. , et al . 2007 , , 667 , l57 siegel , m. h. , majewski , s. r. , sohn , s. t. , shetrone , m.d . , munoz , r. r. , & patterson , r. j. 2008 , , _ submitted _ [ s08 ] simon , j. d. & geha , m. , 2007 , , 670 , 313 smith , g. h. , siegel , m. h. , shetrone , m. d. , & winnick , r. 2006 , , 118 , 1361 sneden , c. , kraft , r.p . , langer , g.e . , prosser , c.f . , & shetrone , m.d . 1994 , , 107 , 1773 . sneden , c. , kraft , r. p. , guhathakurta , p. , peterson , r. c. , & fulbright , j. p. 2004 , , 127 , 2162 sohn , s. t. , majewski , s. r. , munoz , r. m. , kunkel , w. e. , johnston , k. v. , ostheimer , j. c. , guhathakurta , p. , patterson , r. j. , siegel , m. h. & cooper , m. , 2007 , , 663 , 960 sollima , a. , pancino , e. , ferraro , f. r. , bellazzini , m. , straniero , o. , & pasquini , l. 2005 , , 634 , 332 suntzeff , n. b. , aaronson , m. , olszewski , e. w. & cook , k.h . 1986 , , 91 , 1091 swope , h. h. 1967 , , 79 , 439 swope , h. h. 1968 , , 73 , s204 thvenin , f. , & idiart , t. p. 1999 , , 521 , 753 tolstoy , e. , venn , k. a. , shetrone , m. , primas , f. , hill , v. , kaufer , a. , & szeifert , t. 2003 , , 125 , 707 venn , k. a. , irwin , m. , shetrone , m. d. , tout , c. a. , hill , v. , & tolstoy , e. 2004 , , 128 , 1177 vogt , s. s. , mateo , m , olszewski , e. w. & keane , m. j. , 1995 , 109 , 151 walker , m. g. , mateo , m. , olszewski , e. w. , gnedin , o. y. , wang , x. , sen , b. & woodroofe , m. , 2007 , , _ in press , astro - ph/0708.0010 _ mcwilliam , a. 1990 , 74 , 1075 lrrrrrrr tii & 8378.25 & 3.72 & -2.30 + tii & 8382.53 & 0.82 & -1.63 + tii & 8435.65 & 0.84 & -1.30 + fei & 8220.38 & 4.32 & 0.20 + fei & 8327.06 & 2.20 & -1.64 + fei & 8387.77 & 2.18 & -1.60 + fei & 8468.41 & 2.22 & -2.17 + fei & 8514.07 & 2.20 & -2.25 + fei & 8611.80 & 2.85 & -2.15 + fei & 8674.75 & 2.83 & -2.00 + fei & 8688.63 & 2.18 & -1.21 + mgi & 8806.76 & 4.35 & -0.13 + ca ii & 8498.02 & 1.69 & -1.31 + ca ii & 8542.09 & 1.70 & -0.36 + ca ii & 8662.14 & 1.69 & -0.62 + lrrrrrr + 241 & 13.61 & 1.11 & 37.6 & 4419 & 0.94 & 1.77 + 131 & 13.02 & 1.20 & 34.9 & 4298 & 0.85 & 1.80 + 223 & 13.19 & 1.16 & 34.5 & 4350 & 0.72 & 1.85 + 160 & 14.06 & 1.52 & 31.2 & 3969 & 0.31 & 2.02 + 294 & 14.22 & 1.00 & 22.3 & 4586 & 1.29 & 1.62 + 181 & 13.95 & 1.25 & 21.3 & 4236 & 0.94 & 1.76 + + hi82 & 12.59 & 1.33 & 24.6 & 4188 & 0.34 & 1.98 + hi119&13.62 & 0.96 & 21.1 & 4633 & 1.09 & 1.70 + hi239&14.19 & 0.87 & 14.2 & 4788 & 0.89 & 1.79 + + 265 & 13.26 & 1.30 & 93.0 & 4155 & 0.83 & 1.81 + 250 & 14.11 & 0.95 & 82.1 & 4627 & 1.50 & 1.54 + 640 & 13.27 & 1.26 & 77.1 & 4199 & 0.87 & 1.79 + 589 & 12.87 & 1.33 & 76.6 & 4124 & 0.64 & 1.89 + 1217 & 14.00 & 1.05 & 76.4 & 4470 & 1.36 & 1.59 + 885 & 13.47 & 1.07 & 75.5 & 4441 & 1.13 & 1.69 + 334 & 13.24 & 1.20 & 55.3 & 4269 & 0.91 & 1.78 + 238 & 12.69 & 1.57 & 53.5 & 3918 & 0.35 & 2.01 + + sl & 14.04 & 1.47 & 56.1 & 4581 & 1.37 & 1.59 + sf & 13.39 & 1.70 & 49.0 & 4225 & 0.87 & 1.79 + sr & 14.66 & 1.29 & 42.2 & 4910 & 1.80 & 1.41 + sh & 13.84 & 1.61 & 38.9 & 4356 & 1.15 & 1.68 + ss & 14.79 & 1.40 & 37.9 & 4703 & 1.74 & 1.43 + su & 14.78 & 1.28 & 37.5 & 4930 & 1.86 & 1.39 + s62 & 13.97 & 1.62 & 34.1 & 4341 & 1.19 & 1.66 + lrrrcrrr 180 & 18.998 & 1.5117 & 43.7 & -1.60 & 3981 & 0.33 & 2.02 + 255 & 19.291 & 1.3114 & 40.0 & -1.66 & 4151 & 0.62 & 1.90 + 271 & 19.375 & 1.3561 & 39.5 & -1.48 & 4100 & 0.60 & 1.90 + 258 & 19.311 & 1.3507 & 36.9 & -1.57 & 4110 & 0.59 & 1.91 + 336 & 19.531 & 1.1669 & 36.8 & -2.03 & 4299 & 0.83 & 1.81 + 195 & 19.026 & 1.0915 & 36.2 & -1.60 & 4406 & 0.71 & 1.86 + 296 & 19.385 & 1.0488 & 35.7 & -2.02 & 4438 & 0.87 & 1.79 + 293 & 19.304 & 1.2481 & 35.5 & -1.84 & 4216 & 0.68 & 1.87 + 166 & 18.827 & 1.1629 & 35.3 & -1.52 & 4318 & 0.56 & 1.92 + 285 & 19.337 & 1.3657 & 35.0 & -1.50 & 4091 & 0.58 & 1.91 + 304 & 19.389 & 1.2711 & 34.7 & -1.81 & 4194 & 0.69 & 1.87 + 254 & 19.259 & 1.3981 & 33.6 & -1.43 & 4055 & 0.51 & 1.87 + 236 & 19.212 & 1.3330 & 32.6 & -1.90 & 4144 & 0.58 & 1.91 + 351 & 19.552 & 1.2835 & 32.5 & -1.40 & 4176 & 0.74 & 1.85 + 333 & 19.495 & 1.2278 & 32.4 & -1.45 & 4240 & 0.77 & 1.83 + 379 & 19.618 & 1.3171 & 32.1 & -1.52 & 4141 & 0.74 & 1.85 + 209 & 19.033 & 1.4749 & 31.7 & -1.59 & 4006 & 0.37 & 2.00 + 256 & 19.332 & 1.2784 & 30.4 & -1.54 & 4182 & 0.66 & 1.88 + 341 & 19.541 & 1.2780 & 29.2 & -1.43 & 4182 & 0.74 & 1.85 + 281 & 19.342 & 1.1127 & 28.6 & -2.05 & 4358 & 0.80 & 1.82 + 282 & 19.329 & 1.4611 & 27.6 & -1.52 & 4008 & 0.49 & 1.95 + 420 & 19.717 & 1.0824 & 27.1 & -1.46 & 4430 & 1.00 & 1.74 + 377 & 19.587 & 1.1853 & 25.8 & -1.73 & 4283 & 0.84 & 1.80 + 260 & 19.268 & 1.2811 & 25.7 & -1.57 & 4179 & 0.63 & 1.89 + 248 & 19.354 & 1.2399 & 25.1 & -1.51 & 4225 & 0.70 & 1.86 + 234 & 19.250 & 1.2238 & 24.5 & -1.80 & 4240 & 0.67 & 1.87 + 230 & 19.194 & 1.4458 & 24.1 & -1.56 & 4025 & 0.45 & 1.96 + lrrrr + 241 & -1.70 ( 0.18 ) & 0.11 ( 0.27 ) & 0.30 ( 0.38 ) & 0.70 ( 0.17 ) + 131 & -1.89 ( 0.21 ) & 0.70 ( 0.25 ) & 0.45 ( 0.37 ) & 0.70 ( 0.20 ) + 223 & -1.55 ( 0.18 ) & 0.57 ( 0.22 ) & 0.50 ( 0.39 ) & 0.70 ( 0.17 ) + 160 & -2.06 ( 0.26 ) & -0.03 ( 0.30 ) & 0.70 ( 0.59 ) & 0.35 ( 0.20 ) + 294 & -1.62 ( 0.22 ) & 0.65 ( 0.27 ) & 0.60 ( 0.42 ) & 0.45 ( 0.17 ) + 181 & -2.02 ( 0.22 ) & 0.46 ( 0.25 ) & 0.20 ( 0.43 ) & 0.85 ( 0.24 ) + average & -1.77 ( 0.08 ) & 0.44 ( 0.10 ) & 0.44 ( 0.17 ) & 0.61 ( 0.08 ) + + hi82 & -2.44 ( 0.16 ) & 0.13 ( 0.20 ) & 0.50 ( 0.30 ) & 0.50 ( 0.22 ) + hi119 & -2.25 ( 0.14 ) & 0.22 ( 0.20 ) & 0.70 ( 0.24 ) & 0.35 ( 0.27 ) + hi239 & -2.03 ( 0.18 ) & 0.28 ( 0.37 ) & -0.20 ( 0.35 ) & 0.30 ( 0.22 ) + average & -2.26 ( 0.09 ) & 0.19 ( 0.13 ) & 0.44 ( 0.17 ) & 0.39 ( 0.13 ) + + 265 & -1.62 ( 0.14 ) & -0.08 ( 0.16 ) & 0.20 ( 0.25 ) & 0.60 ( 0.17 ) + 238 & -2.10 ( 0.19 ) & 0.30 ( 0.18 ) & 0.10 ( 0.32 ) & 0.45 ( 0.19 ) + 250 & -1.40 ( 0.14 ) & 0.38 ( 0.16 ) & 0.50 ( 0.28 ) & 0.40 ( 0.15 ) + 334 & -1.71 ( 0.14 ) & 0.36 ( 0.16 ) & 0.30 ( 0.22 ) & 0.75 ( 0.15 ) + 589 & -1.77 ( 0.19 ) & 0.43 ( 0.16 ) & 0.30 ( 0.27 ) & 0.65 ( 0.15 ) + 640 & -1.51 ( 0.14 ) & -0.02 ( 0.16 ) & 0.30 ( 0.28 ) & 0.55 ( 0.15 ) + 885 & -1.61 ( 0.15 ) & 0.40 ( 0.17 ) & 0.05 ( 0.34 ) & 0.45 ( 0.16 ) + 1217 & -1.68 ( 0.16 ) & 0.60 ( 0.20 ) & 0.10 ( 0.23 ) & 0.50 ( 0.18 ) + average & -1.64 ( 0.05 ) & 0.28 ( 0.06 ) & 0.24 ( 0.09 ) & 0.55 ( 0.06 ) + + s62 & -1.03 ( 0.19 ) & 0.55 ( 0.23 ) & 0.50 ( 0.27 ) & 0.50 ( 0.20 ) + sf & -0.94 ( 0.18 ) & 0.60 ( 0.23 ) & 0.15 ( 0.26 ) & 0.40 ( 0.20 ) + sh & -1.18 ( 0.16 ) & 0.72 ( 0.20 ) & 0.35 ( 0.24 ) & 0.50 ( 0.17 ) + sl & -1.17 ( 0.21 ) & 0.68 ( 0.26 ) & 0.35 ( 0.33 ) & 0.50 ( 0.22 ) + sr & -1.01 ( 0.20 ) & 0.55 ( 0.24 ) & 0.25 ( 0.27 ) & 0.35 ( 0.21 ) + ss & -1.05 ( 0.19 ) & 0.89 ( 0.27 ) & 0.10 ( 0.29 ) & 0.40 ( 0.20 ) + su & -0.92 ( 0.23 ) & 0.13 ( 0.40 ) & -0.10 ( 0.33 ) & 0.15 ( 0.24 ) + average & -1.05 ( 0.07 ) & 0.63 ( 0.09 ) & 0.25 ( 0.11 ) & 0.41 ( 0.08 ) + lrrrr 180 & -1.94 ( 0.18 ) & -0.16 ( 0.16 ) & & 0.25 ( 0.15 ) + 209 & -1.97 ( 0.17 ) & -0.13 ( 0.16 ) & -0.20 ( 0.30 ) & 0.30 ( 0.15 ) + 271 & -1.77 ( 0.18 ) & -0.10 ( 0.18 ) & 0.20 ( 0.30 ) & 0.50 ( 0.17 ) + 351 & -1.73 ( 0.22 ) & -0.17 ( 0.21 ) & 0.25 ( 0.28 ) & 0.30 ( 0.18 ) + 166 & -1.85 ( 0.18 ) & 0.18 ( 0.19 ) & 0.10 ( 0.30 ) & 0.60 ( 0.16 ) + 293 & -1.93 ( 0.18 ) & -0.05 ( 0.17 ) & -0.20 ( 0.15 ) & 0.35 ( 0.15 ) + 336 & -1.98 ( 0.21 ) & & 0.10 ( 0.32 ) & 0.20 ( 0.17 ) + 236 & -2.26 ( 0.22 ) & 0.33 ( 0.21 ) & 0.50 ( 0.33 ) & 0.40 ( 0.20 ) + 195 & -1.70 ( 0.19 ) & 0.36 ( 0.23 ) & 0.70 ( 0.30 ) & 0.42 ( 0.18 ) + 258 & -1.64 ( 0.19 ) & -0.17 ( 0.19 ) & 0.05 ( 0.30 ) & 0.30 ( 0.17 ) + 254 & -1.80 ( 0.19 ) & -0.10 ( 0.18 ) & 0.40 ( 0.27 ) & 0.20 ( 0.17 ) + 255 & -2.05 ( 0.19 ) & 0.30 ( 0.17 ) & -0.35 ( 0.39 ) & 0.45 ( 0.16 ) + 296 & -2.13 ( 0.19 ) & 0.28 ( 0.22 ) & 0.30 ( 0.33 ) & 0.40 ( 0.17 ) + 304 & -2.09 ( 0.18 ) & -0.02 ( 0.18 ) & 0.45 ( 0.27 ) & 0.35 ( 0.16 ) + 281 & -2.07 ( 0.21 ) & 0.35 ( 0.26 ) & 0.15 ( 0.36 ) & 0.35 ( 0.20 ) + 282 & -1.68 ( 0.19 ) & -0.34 ( 0.19 ) & -0.30 ( 0.37 ) & 0.05 ( 0.17 ) + 333 & -1.95 ( 0.21 ) & 0.28 ( 0.21 ) & 0.35 ( 0.30 ) & 0.50 ( 0.19 ) + 285 & -1.73 ( 0.18 ) & -0.03 ( 0.20 ) & 0.05 ( 0.28 ) & 0.45 ( 0.16 ) + 379 & -1.71 ( 0.21 ) & -0.03 ( 0.21 ) & 0.15 ( 0.33 ) & 0.30 ( 0.20 ) + 260 & -1.75 ( 0.21 ) & -0.17 ( 0.20 ) & & 0.35 ( 0.18 ) + 341 & -1.42 ( 0.19 ) & -0.18 ( 0.19 ) & -0.00 ( 0.30 ) & 0.15 ( 0.17 ) + 256 & -1.73 ( 0.18 ) & 0.05 ( 0.18 ) & -0.20 ( 0.16 ) & 0.35 ( 0.16 ) + 248 & -1.72 ( 0.20 ) & 0.03 ( 0.21 ) & 0.05 ( 0.18 ) & 0.40 ( 0.19 ) + 420 & -1.80 ( 0.22 ) & 0.24 ( 0.25 ) & 0.45 ( 0.32 ) & 0.45 ( 0.21 ) + 230 & -1.88 ( 0.24 ) & 0.07 ( 0.25 ) & 0.75 ( 0.34 ) & 0.50 ( 0.23 ) + 377 & -1.98 ( 0.20 ) & 0.09 ( 0.24 ) & 0.15 ( 0.38 ) & 0.40 ( 0.19 ) + 234 & -1.76 ( 0.19 ) & -0.20 ( 0.21 ) & -0.10 ( 0.33 ) & 0.25 ( 0.18 ) +
gas - phase and gas - grain reactions are considered the two main paths by which molecules form in the interstellar medium ( ism ) @xcite . as of february 2016 , almost 200 molecular species have been observed in the ism and circumstellar shells ; a number that will only increase as observational techniques improve . it is well known that the observed abundance of molecular hydrogen ( h@xmath0 ) can not be accounted for by gas - phase reactions @xcite . thus , gas - grain reactions are invoked to account for the amount of h@xmath0 observed in the ism . the main regions of focus for h@xmath0 formation are molecular clouds ( mcs ) and the cold neutral medium ( cnm ) . these regions have temperatures that range from 10 k to 100 k , and make up a bit under half of the ism by mass @xcite . the interplay of dense mcs with the cnm allows for the formation of h@xmath0 and plays a crucial role in star formation . there are four main steps in the typical gas - grain reaction . the reactants must first adsorb to the interstellar dust surface ; they then must diffuse across the substrate . when reactants encounter each other , they must react to form a molecular species . finally , the newly formed molecule must desorb from the dust surface . in dense clouds , these interstellar grains / substrates consist of silicate or carbonaceous cores covered by a thin layer of ice about 0.1 @xmath1 m thick @xcite . the icy layers are believed to predominantly be composed of h@xmath0o ; although ices such as co , co@xmath0 , nh@xmath2 , and ch@xmath3 certainly occur as well @xcite . the h@xmath0o ice is generally regarded to be amorphous in structure @xcite . the formation of h@xmath0 on dust grain surfaces is thought to occur through three mechanisms . the first , dubbed the eley - rideal ( er ) mechanism , involves an impinging hydrogen atom directly hitting an already adsorbed hydrogen atom on the icy surface @xcite . in the langmuir - hinshelwood ( lh ) mechanism , two hydrogen atoms already adsorbed to the grain surface diffuse some distance until they meet one other and react to form h@xmath0 @xcite . the distinction between these two mechanisms lies in the thermalization of the atom with respect to the grain . in the er mechanism , the adsorbed atom has thermalized to the grain , and is directly hit by an impinging atom that has not yet thermalized . by contrast , in the lh mechanism , both atoms have already thermalized with the grain . the final mechanism , known as the hot - atom mechanism , involves a hydrogen atom with a high kinetic energy striking the substrate and thermally hopping along the substrate surface until it strikes another already adsorbed hydrogen atom and forms h@xmath0 @xcite . the common feature of these three mechanisms is the adsorption of at least one of the hydrogen atoms to the icy substrate . a key quantity characterizing the adsorption properties of a given gas species on a given substrate is the probability that the gas molecule sticks to the substrate when the two collide . in computational studies , the sticking probability can be deduced as the fraction of incident gas molecules that stick to the substrate , in the limit of a large number of trials . the sticking probability for a general species is @xmath4 where @xmath5 is the number of sticking events and @xmath6 is the number of trials . the sticking probability depends on the temperature of the dust or ice grain ( @xmath7 ) and the incident kinetic energy ( @xmath8 ) of the colliding species when the species and surface are very far apart . in many experimental and observational situations , one does nt have access to the incident kinetic energy of the gaseous species , but rather the temperature of the gas ( @xmath9 ) . the quantity characterizing the sticking in this case is called the sticking coefficient , @xmath10 , a function of both gas and grain temperatures . the sticking coefficient can be calculated by averaging the sticking probability weighted by a maxwellian velocity distribution of the desired gas temperature . hollenbach & salpeter ( 1970 ) derived a theoretical form for this sticking coefficient : @xmath11 where @xmath12 , @xmath13 , and @xmath14 . the variable @xmath15 is the well depth of the adsorption potential , @xmath16 is the energy transferred in a single collision , and @xmath17 is close to unity for lambert s law and equals two in the case of isotropic scattering @xcite . in astrophysical models , this equation is often approximated with @xmath18 where @xmath19 is dependent on the incident species and the type of substrate involved @xcite . some experiments ( e.g. manico et al . 2001 ) suggest that eqn . ( 3 ) underestimates the sticking coefficient . the sticking coefficient is of great importance to astrophysical models , as it is a crucial component for the rate equation of various grain - mediated processes ( including h@xmath0 formation ) in the ism . the goal of this work is therefore to obtain accurate sticking coefficients of atomic hydrogen sticking on amorphous water - ice for astrophysical modeling purposes . there have been many such studies , experimental and theoretical , with the same goal , though h@xmath0 is primarily formed on warm amorphous carbon and silicate grains @xcite . to our knowledge , the first simulational study was performed by buch & zhang ( 1991 ) , who studied the sticking of h and d atoms on an amorphous cluster of 115 water molecules using classical molecular dynamics ( md ) . they found sticking probabilities and coefficients on 10 k ice for both h and d atoms . a group of studies investigated the sticking and mobility of h atoms on amorphous water - ice slabs consisting of 1000 water molecules using classical md @xcite . they performed simulations on 10 k and 70 k ice and showed that the sticking probability can not be assumed to be equal to unity for the range of gas temperatures encountered in the ism . however , due to the computational restraints of that era , not many simulations were performed for each incident kinetic energy of the hydrogen atom . al - halabi et al . ( 2002 , 2007 ) also performed classical md simulations on both crystalline and amorphous water - ice substrates . their simulations confirmed that sticking depends on the incident kinetic energy of the h - atom and the temperature of the ice surface by carrying out a large number of trials . veeraghattam et al . ( 2014 ) performed large numbers of classical md simulations on both 10 k and 70 k amorphous water - ice slabs consisting of 1000 water molecules . they computed sticking probabilities and coefficients for their simulations that are reproduced and expanded upon in the current work . experimentally , manico et al . ( 2003 ) , hornekaer et al . ( 2003 ) , and watanabe et al . ( 2010 ) all observed h@xmath0 formation on lab - generated substrates and measured the recombination energy . in this study we perform an extensive analysis of atomic hydrogen sticking to amorphous water - ice surfaces for a wide range of grain temperatures ( 10 k - 70 k ) and for a wide range of atomic hydrogen kinetic energies ( 5 k - 400 k ) . therefore a key goal is to fill in the large gaps in predicted sticking coefficients from previous studies to provide richer , more in - depth information for astrophysical modelers . here we also pay particularly close attention to the definition of what constitutes a sticking event . the previous simulational studies mentioned above do not agree on a objective definition for a sticking event , and we have found that different criteria can lead to different results . the appendix provides additional detail on the subject . using classical molecular dynamics ( md ) , we study the interaction of a hydrogen atom with an amorphous water - ice slab @xcite . it is reasonable to question whether classical dynamics is appropriate for describing hydrogen . studies of hydrogen in other contexts have shown that quantum dynamics is preferable for hydrogen , but at a very high computational cost . in the meantime , classical trajectory calculations have been shown to prove useful in describing physical processes involving hydrogen on surfaces @xcite . but , as andersson et al . ( 2006 ) mention , quantum effects can be tricky to pin down and could be important as the kinetic energy of the hydrogen atom decreases . nevertheless , at this stage the vastly more complicated and expensive quantum calculations seem unwarranted . the present simulations require two pairwise potentials : ( 1 ) h@xmath0o - h@xmath0o and ( 2 ) h - h@xmath0o . these potentials are differentiated to define a force field , which is then used to numerically integrate the coupled , classical equations of motion for the many - particle system using the leapfrog algorithm at a fixed time - step of @xmath20 fs . in non - spherical systems containing rigid molecules , like h@xmath0o , the rotational equations of motion must also be solved @xcite . the h@xmath0o - h@xmath0o potential interaction is used to simulate the water - ice substrate while the h - h@xmath0o potential is used to model the hydrogen interaction with that substrate . these potentials are derived from quantum chemical calculations for a single pair of particles . our system is treated either as a canonical ensemble ( nve ) or as a isobaric isothermal ensemble ( nvt ) , depending on the stage of the simulation . we consider a system of 1000 water molecules to model the amorphous water - ice surface as a slab of finite thickness . the water molecules are treated as rigid and exist within a @xmath21 @xmath22 @xmath22 @xmath23 simulation box with periodic boundary conditions in the @xmath24 and @xmath25 directions . the interaction potential for these water molecules is a tip4p / ice configuration @xcite . this tip4p / ice model is a slight modification of the well - known transferable intermolecular potential model ( tip4p ) @xcite . the basic form consists of a lennard - jones ( lj ; lennard - jones & devonshire 1936 ) term added to a coulombic term : @xmath26.\ ] ] the lj term applies to a site that is on the oxygen atom but in the coulombic term , each h@xmath0o molecule is modeled by three charge sites : a positive charge on each hydrogen atom and a negative charge on the h - o - h molecular angle bisector , 0.1577 @xmath23 from the oxygen atom . the variable @xmath27 is the distance between two oxygen atoms and @xmath28 is the distance between charge sites on two different h@xmath0o molecules , where @xmath29 and @xmath30 each run over the three charge sites of the two molecules . the lj constants have values of @xmath31 @xmath23 and @xmath32 kcal mol@xmath33 . the charge on each hydrogen site is @xmath34 , where @xmath35 is the electron charge ; the charge on the h - o - h molecular angle bisector is @xmath36 . the h - o - h angle has a fixed value of 104.52@xmath37 , @xmath38 @xmath39 which is the fixed length of the o - h bond . we use the potential above to generate the amorphous ice slab from a quenched md simulation . within our simulation box , we restrict the initial molecular position in the slab to a @xmath21 @xmath22 @xmath40 @xmath23 region . the restriction on the @xmath25 coordinate ensures that we generate a slab rather than a cubic arrangement of water molecules . we give each molecule a random initial position and velocity within this region and the system is then numerically solved . a 10 @xmath23 cutoff was used for the coulombic interaction and ewald summation was used for long - range electrostatics . quenching the slab occurs as follows : initially , the system evolves for 150 ps in nvt at a temperature of 300 k using velocity rescaling to set the temperature . the system then evolves in nve for 50 ps , after which the system is cooled to 200 k and run again in nvt for 150 ps , followed by another 50 ps nve run . the temperature is then lowered to 100 k. the process is repeated at 100 k before we cool the slab to 70 k. the reason it is desirable to perform a simulation in nvt is to ensure that the temperature of the amorphous ice slab , which is only a small component of an actual interstellar ice mantle , stays approximately constant . as a result of the nvt simulation , the substrate will settle into an energetically favorable configuration . the system is then run in nve so that the substrate can be fixed in that energy configuration . to generate slabs of other temperatures we sequentially cool this 70 k slab in 10 k intervals , i.e. from 70 k to 60 k to ... to 10 k. the system runs in nve for 10 ps at each interval and velocity rescaling occurs every 5 ps . a restart file is output at the end of each interval , containing the positions and velocities of the water molecules to be used as the initial conditions of the next cooling interval . we tested this method by generating a 10 k slab from scratch and comparing calculated sticking probabilities on that slab to sticking probabilities on a 10 k slab made from the 70 k slab . the sticking probabilities we calculated were in very good agreement , and we determined that this method is sufficient for slab generation at a variety of temperatures . note that in this method of making the 10 k slab , we also make slabs of the other temperatures studied along the way . these slabs are then used for hydrogen atom trajectory calculations . a single hydrogen atom trajectory is initialized as follows . first , we assign the hydrogen atom random initial @xmath41,@xmath42 coordinates and a @xmath25 coordinate fixed at 18 , about 8 @xmath39 above the ice surface . because the ice is amorphous the surface of the slab is ill defined , and therefore we use an average surface for referencing h - atom / surface distances . random initial velocity components that correspond to a particular kinetic energy are assigned as well . the initial velocity in the @xmath25-direction must be negative so that the hydrogen atom is launched towards the slab . after the initialization , the trajectory is then integrated numerically using the leapfrog algorithm . the interaction between the hydrogen atom and the water molecules of the substrate is described by the potential derived by zhang et al . ( 1991 ) , which has a form given by @xmath43y_{l , m}(\theta,\phi).\ ] ] this potential was developed using open - shell hartee fock and fourth - order mller - plesset perturbation theory . the _ ab initio _ data were then fit to a lj form given by equation ( 5 ) . the variable @xmath27 represents the distance between the impinging hydrogen atom and the oxygen atom . the @xmath44 and @xmath45 coordinates are the spherical angles of the incident hydrogen atom with respect to the water molecule . the origin is at the oxygen atom and the polar axis coincides with the molecular symmetry axis . for a water molecule , the spherical coordinates of the two hydrogen atoms are given by @xmath46 , @xmath47 , and @xmath48 . the @xmath49 and @xmath50 parameters for this potential are given in table 1 . the simulation time for a single run is 10 ps and the time - step is 1 fs . the simulation time was selected in an attempt to avoid ambiguity in determining the sticking or scattering status of the hydrogen atom . periodic boundary conditions were used for the @xmath41 and @xmath42 coordinates ( parallel to the slab surface ) , but the simulation box was non - periodic in the @xmath25-coordinate ( normal to the surface ) . thus the hydrogen atom leaves the simulation box and the simulation terminates if the @xmath25-coordinate of the atom ever exceeds 20 . during the simulation , the system runs in nve , but the velocity of the substrate molecules are rescaled every 5 ps to ensure that the slab temperature remains ( approximately ) constant . for nine incident kinetic energies of hydrogen atoms ranging from @xmath515 k - 400 k , we performed 300 runs apiece for each of the 7 substrate temperatures . thus , simulation data for a single substrate temperature comprises of 2700 runs , for a total of 18,900 runs over the 7 substrate temperatures . we chose to perform 300 runs for each h - atom to obtain statistically significant values for the sticking probability . for each of the 18,900 runs , the sticking or scattering status was determined . we defined a sticking event to be a trajectory whose final @xmath25 coordinate ( @xmath52 ) was below 13 @xmath23 and whose final total energy ( @xmath53 ) was below -100 k. the physical motivation for this definition is discussed in the appendix . to get an estimate of error for each batch of runs we randomly selected 50 runs out of the 300 and calculated the sticking probability . we repeated this random sampling 25 times and computed the standard deviation of the probabilities to represent the error of our computations . mc and cnm regions of the ism contain dust grains covered with ice which have temperatures that range from 10 k to 100 k. we have performed simulations on 7 different ice substrate temperatures ( 10 k - 70 k ) for 9 different incident kinetic energies ( 5 k - 400 k ) of the h - atom . this wide selection of slab temperatures provides a thorough coverage of the temperature ranges of interest . all of our sticking probability and sticking coefficient results are shown in table 2 . for comparison with previous studies ( e.g. , veeraghattam et al . 2014 and al - halabi et al . 2007 ) , we plot the sticking probability ( @xmath54 ) vs. the incident kinetic energies on 10 k and 70 k ice in figures 1 and 3 , respectively . in addition , in figures 2 and 4 , are plots of the sticking coefficient ( @xmath55 ) vs. the gas temperature for 10 k and 70 k ice , respectively . because , to our knowledge , there have been no studies which calculated sticking probabilities and coefficients on substrate temperatures other than 10 k and 70 k , we focus the majority of our discussion on those two slab temperatures . figure 1 displays the sticking probability plotted vs. the incident kinetic energy of the hydrogen atom . virtually all the atoms stick " to the substrate at very low incident kinetic energies ( i.e. , 5 k and 10 k ) . then , as the incident kinetic energy of the h atom increases , the sticking probability decreases . the probability decreases slowly at first and then at around @xmath56 k , the probability begins to decrease more rapidly . our results ( represented by dots ) show good agreement with those results of veeraghattam et al . ( 2014 ) ( represented by triangles ) , in spite of the fact that that study used a tip4p ( rather than a tip4p / ice ) potential for the water - water interaction and used a different sticking criterion ( see appendix ) . in addition , our results show strong agreement with those of al - halabi et al . ( 2002 ) , who also utilized a tip4p potential . with that being said , the results of al - halabi et al . ( 2002 ) were on a crystalline water - ice slab , rather than an amorphous one . the results of buch & zhang ( 1991 ) and those of masuda et al . ( 1997 ) are not in good agreement with our study or those mentioned in the preceding paragraph . buch & zhang ( 1991 ) obtain sticking probabilities that are much lower than those obtained here , whereas the sticking probabilities of masuda et al . ( 1997 ) are in general higher . buch & zhang ( 1991 ) used a small cluster of ice comprising of 115 water molecules . the cluster is highly irregular and as a result it is more difficult for the h atom to stick to the cluster @xcite . masuda et al . ( 1997 ) performed simulations on a slab comprising of the same number of water molecules as our own ( 1000 ) but used a different potential to describe the water - water interaction , tip2s @xcite . we cite the same discrepancy that al - halabi et al . ( 2002 ) reported to explain the difference between their results and those of masuda et al . the study of masuda et al . ( 1997 ) applied an incorrect expression for @xmath57 that accounts for the greater chance of adsorption in their simulations @xcite . figure 2 shows the sticking coefficient as a function of gas temperature . the sticking coefficient is very close to unity at low gas temperatures ( 5 and 10 k ) and decreases as the temperature of the gas increases . again , our results show strong agreement with those of veeraghattam et al . ( 2014 ) , despite the technical differences in our simulations . the small differences between our calculated coefficients could possibly be attributed to the different number of runs that we performed , or the difference in time step . regardless , all of the results of veeraghattam et al . ( 2014 ) lie within the uncertainties of our own results . our results for the 70 k amorphous water - ice present a much more interesting case . these results largely disagree with those of previous studies . figure 3 , which displays the sticking probability as a function of incident kinetic energy , illustrates the discrepancies . our calculated sticking probabilities lie outside of the error bars of all three studies of comparison . overall the trends are similar : in all three studies , the sticking probability decreases as the incident kinetic energy of the hydrogen atom increases . in contrast with the 10 k slab , the sticking probability for atomic hydrogen does not go to unity at low incident kinetic energy . since the h@xmath0o molecules of the 70 k slab have greater average rotational and intermolecular vibrational energy than those of the 10 k slab , correspondingly larger energy transfers when the slab and h - atom collide yield a reduced likelihood of sticking . the same reasoning is invoked to explain the differences between our results and those of masuda et al . ( 1997 ) as for the 10 k amorphous water - ice . however , the discrepancies between our results and those of veeraghattam et al . ( 2014 ) are more subtle . the use of a new potential , tip4p / ice instead of tip4p , does not account for the discrepancy . we performed simulations using the tip4p potential to test this and found that our results were still much higher than those of veeraghattam et al . in fact , we found that the use of the new potential makes very little quantitative difference at all in these simulations . the only remaining difference between our results and those of veeraghattam et al . ( 2014 ) is the criterion used for assessing sticking events . we defined a sticking event to be a run in which the h - atom had a final @xmath25-coordinate that was less than 13 @xmath23 _ and _ had a final total energy less that -100 k. by contrast , veeraghattam et al . ( 2014 ) defined a sticking event as a run in which the h - atom had a final @xmath25-coordinate that was less than 11 @xmath23 and did not consider the final total energy of the h - atom . these details make little difference on the 10 k ice , because at the end of the simulation atoms are very rarely still hopping around with high amplitudes . however , on 70 k ice it takes longer for atoms to give up their kinetic energy to the slab , because the water molecules in the slab have higher rotational and translational kinetic energies . as a result , an atom may still be bouncing on the surface of the substrate at the end of the run . we have selected a higher cutoff @xmath25 coordinate to include these bouncing atoms . we have also opted for a longer simulation time : 10 ps , compared to the 5 ps of veeraghattam et al . ( 2014 ) , in order to allow for sufficient relaxation on the warmer substrate . we also performed a few trial simulations of 20 ps to confirm that 10 ps was an adequate selection . we found that the additional 10 ps had a negligible impact on the sticking or non - sticking status of the h - atom . in addition , we have applied an energy cutoff of -100 k in the h - atom total energy as part of the sticking criterion . this was done to align with the sticking criterion used by by al - halabi et al . ( 2002 ) . as a result of the contrasting definition with veeraghattam et al . ( 2014 ) , we observe large discrepancies between calculated probabilities and coefficients . however , we do see that the sticking probability tends to be in closer agreement with the results of veeraghattam et al . ( 2014 ) as the incident kinetic energy of the h - atom increases . this is because the bouncing atom " scenario play less of a role at higher incident kinetic energies , where the h atom generally either sticks , or leaves the simulation box within the simulation time . there are not many data points of al - halabi et al . ( 2002 ) to compare with and those that are present disagree with our predicted sticking probabilities . that fact that al - halabi et al . ( 2002 ) performed their simulations on crystalline ice is the most likely explanation for the discrepancy , since all other features of the calculations ( potentials , sticking criteria , etc . ) are either similar or the same . because the sticking coefficient is a function of the sticking probability , our calculated sticking coefficients are much larger than those of veeraghattam et al . ( 2014 ) , reflecting the discrepancy in our sticking probabilities . this is shown in figure 4 , where the sticking coefficient is plotted against the gas temperature . again there is better agreement at higher gas temperatures , where the particular criterion of a sticking event has less of an effect . having discussed the results on the low - end and high - end of the slab temperatures studied , it is reasonable to assume that the sticking probabilities and coefficients for the in - between slab temperatures lie somewhere in - between the above results . table 2 confirms that that is indeed the case . we provide these results for astrophysical modeling purposes . as the slab temperature increases , in general , the sticking probability and coefficient decrease for each respective h - atom incident kinetic energy . figure 5 shows the sticking coefficient for each gas temperature ( 5 k - 400 k ) as a function of substrate temperature . this plot demonstrates some key features of these simulations . for all but the highest gas temperatures , the sticking coefficient decreases with increasing substrate temperature , as one would expect . however , the overall drop in sticking coefficient from the value at substrates temperature of 10 k to 70 k decreases as gas temperature increases . by gas temperatures of 300 k and 400 k , the sticking coefficient is very nearly constant in substrate temperature , at least over the range of temperatures studied . this behavior can be understood by considering the average kinetic energy of the gas atoms . at high gas temperature the average kinetic energy is so high compared to that of the substrate h@xmath0o molecules that energy transfers to and from the substrate are insensitive to substrate temperature @xcite . as shown previously by @xcite , the sticking coefficients for 10 k ice obtained in our md simulations gives results significantly larger than those predicted by @xcite . the latter is adopted in most astrochemical models involving ice - surface chemistry . since the formation of h@xmath0 on the ice surface is linearly proportional to the sticking coefficient , the adoption of the current values would be expected to result in both an increased efficiency in the molecular hydrogen formed on the ice surface as well as faster removal of atomic hydrogen from the gas phase . further , since the current sticking coefficients on 70 k ice can be as much as 30% larger than those calculated earlier in @xcite , we can expect the h@xmath0 efficiency to be relatively enhanced even as the surface temperature increases ( see fig . 5 ) until water begins to evaporate from the surface above @xmath58100 k. such an effect was found for the sticking of h@xmath0 and other molecular species on amorphous ice in the time - dependent , hot core warm - up modeling of @xcite . we have reported the sticking probabilities and coefficients for nine gas temperatures of atomic hydrogen ( 10 - 400 k ) across seven different substrate temperatures ( 10 k - 70 k ) . we present our results so that they may be used in future astrophysical modeling simulations . the sticking of hydrogen atoms on amorphous water - ice is of significant importance in the ism , and its understanding is necessary before more complicated molecular species , which form through grain - mediated reactions , can be understood . this study shows in detail how grain temperature affects sticking probabilities and coefficients , thereby showing how grain temperature affects the rate of h@xmath0 formation . it is important however to point out some of the limitations of the study . first , the nature of the simulation ( i.e. , launching a single h atom at a surface and determining its adsorption ) is different from the actual adsorption process in the ism , in which a dilute cloud of gas comprising of h atoms passes through these dusty regions . in addition , the lack of a fully quantum treatment may be cause for some concern , because quantum effects tend to be more important for lighter atoms . this study has shown that the criterion for assessing a sticking event can greatly effect sticking probabilities . in the appendix of the paper we discuss the methods by which we have arrived at our sticking criterion . we have explored many different criteria ( e.g. velocity , standard deviation of position , etc . ) for this process and have determined that our sticking criterion is the most appropriate choice for wide application . further studies involving different species which have previous data ( e.g. h@xmath0 on h@xmath0o ice ) could confirm our criterion for sticking . the authors would like to thank vijay veeraghattam for his useful comments and guidance throughout this entire project . we also would like to thank the referee whose comments greatly improved the final version of the manuscript . abascal , j. l. f. , sanz , e. , fernandex , r. g. , & vega , c. 2005 jchph , 122 , 23451 al - halabi , a. , kleyn , a. w. , van dishoeck , e.f . , & kroes , g. j. 2002 , jpcb , 106 , 6515 al - halabi , a.,van dishoeck , e. f. , & kroes , g. j. 2004 , jchph , 120 , 3358 al - halabi , a. & van dishoeck , e. f. 2007 , mnras , 382 , 1648 allen , m. p. & tildesley , d.j . 1987 , _ computer simulation of liquids _ ( oxford : claredon press ) andersson , s. , al - halabi , a. , kroes , g. , & van dischoeck , e. f. 2006 , jchph , 124 , 064715 eley , d. d. 1941 , rspsa , 178 , 452 hagan , w. , tielens , a. g. g. m. , & greenberg , j.m . , 1981 , cp , 54 , 367 harris , j. & kasimo , b. 1981 , sursc , 105 , l281 he , j. , acharyya , k. , & vidali , g. 2016 , , 823 , 56 hinshelwood , c. n. 1930 , ann . rep . , 27 , 11 hollenbach , d. & salpeter , e. e. 1970 , jchph , 53 , 79 hollenbach , d. , kaufman , m. f. , bergin , e. a. , & melnick , g. j. 2009 , apj , 690 , 1497 hornekaer , l. , baurichter , l. , petrunin , v. v. , field , d. , & luntz , a. c. 2003 , sci , 302 , 1943 iqbal , w. , acharyya , k. , & herbst , e. 2012 , apj , 751 , 58 jorgenson , w. l. 1982 , jchph , 77 , 4156 langmuir , i. 1922 , trans . faraday soc . , 17 , 621 le bourlot , j. , le petit , f. , pinto , c. , roueff , e. , roy , f. 2012 , a&a , 541 , a76 lennard - jones , j. e. & devonshire , a. f. 1936 , natur , 137 , 1069 masuda , k. , takahashi , j. , & mukai , t. 1997 , a&a , 330 , 773 manico , g. , raguni , g. , pirronello , v. , roser , j. e. , & vidali , g. 2001 , apjl , 548 , l253 perets , h. b. , biham , o. , manico , g. , et al . 2005 , apj , 627 , 850 takahashi , j. , masuda , k. , & nagaoka , m. 1999 , mrnas , 306 , 22 tielens , a. g. g. m. 2005 , _ the physics and chemistry of the interstellar medium _ , cambrige university press veeraghattam , v. , manrodt , k. , lewis , s. , & stancil , p. c. 2014 , apj , 790 , 4 watanabe , n. , kimura , y. , kouchi , a. , et al . 2010 , apjl , 714 , l233 zhang , q. , sabelli , n. , & buch , v. 1991 , jchph , 95 , 1080 the purpose of this section is to provide physical reasoning for our definition of a sticking event . a sticking event occurs when a hydrogen atom has adsorbed to the water - ice substrate . there are three main outcomes of these simulations : back scattering events , trapping - desorption events , and adsorption events @xcite . a back scattering or elastic event is one in which the atom leaves the simulation box before the simulation ends . this event is a clear non - sticking event . an adsorption event is one in which the atom very clearly sticks to the surface within the timescale of the simulation . this is determined by examining a ) the final position of the hydrogen atom and b ) the final total energy of the atom . clear sticking events have a final position that is either embedded in the substrate or very close to the top of the substrate and low total energy , indicating that the atom has given up most of its kinetic energy and is bound to the surface . trapping - desorption events , however , are difficult to define within classical md . this is a case where an atom appears to be adsorbed to the substrate , but then gains kinetic energy from the substrate which causes the atom to desorb . the uncertainty with regards to the desorption is that it may occur after the timescale of the md simulation . trapping - desorption events have time scales on the scale of _ nanoseconds _ whereas our simulation has a timescale of 10 _ picoseconds _ @xcite . in the instance of trapping - desorption events , an atom may be thermally hopping with large amplitude at the end of the run , yet the atom will still be confined to the simulation box . in classical md simulations these trapping desorption runs must be designated as a sticking or non - sticking event . this designation is dependent on the particular definition of adsorption . it is worth briefly considering the definitions of adsorption in previous studies of this nature . buch & zhang ( 1991 ) , defines a sticking trajectory as one that has a final energy that is less than -100 k. they also state that no questionable runs occurred in their simulations . either the atom stuck , or it was scattered . k. masuda et al . 1997 studies the sticking probability of hydrogen atoms on the icy mantle of dust grains . they define a sticking event as one where an incident h atom resided on the surface of the amorphous water ice after 5000 fs of simulation time . " whereas a scattering / non - sticking event was one in which the atom returned to its initial height at the start of the simulation @xcite . this definition neglects the possibility of trapping - desorption trajectories and gives no quantitative value for what was considered to be the cutoff for a sticking event . al - halabi et al . ( 2004 ) who studied co adsorption on amorphous water ice , defines a sticking event as one that exhibits more than one turning point in the @xmath25 coordinate such that 1 ) the final @xmath25 coordinate of position @xmath52 @xmath59 @xmath39 and 2 ) the final energy of the adsorbed co molecule is below @xmath60 , where @xmath7 is the temperature of the slab , and @xmath61 is the boltzmann constant . the values for @xmath52 are converted from the original number to allow comparison to our simulation box , in which the top of the substrate has an average height of 10 . al - halabi et al . ( 2002 ) employed a similar definition , but does without the @xmath52 cutoff and changed the final total energy cutoff to be -100 k. finally , veeraghattam et al . ( 2014 ) says a sticking event occurs when the h atom stops bouncing on the surface and remains stuck for the rest of the simulation . " however , we found that the h atom does not stop bouncing on the surface within the simulation time . the h atom , even if embedded in the icy substrate continues to bounce with small amplitudes on or within the substrate . upon further investigation , veeraghattam et al . ( 2014 ) applied a restrictive height cutoff value ( @xmath62 ) . we initially took a sticking event to be any such run that leaves the simulation box before the timescale of the simulation . in doing so , we noticed a discrepancy between our results and those of veeraghattam et al . figures 6 and 7 , which illustrate the reason for this discrepancy , show the sticking probability as a function of cutoff value @xmath52 for 10 and 70 k substrates , respectively . we let @xmath52 equal 100 values between 0.0 and 1.0 in reduced units , where 1.0 corresponds to 20 @xmath23 and 0.5 corresponds to 10 . we then compute the sticking probability at each @xmath52 . from figure 6 , we can see that on a 10 k substrate the sticking probabilities plateau fairly quickly . the approximate top of the substrate is at 0.5 ( 10 ) in reduced units . probabilities plateau around 0.55 - 0.6 , which is about 11 - 12 . in addition , in the region of 0.55 - 0.6 the probabilities are not varying significantly . although uncertainties are not plotted here for aesthetic reasons , the probabilities in this region would be within uncertainties . compare this to figure 7 , in which probabilities plateau around 0.65 ( 13 ) , and the probabilities are very different from those calculated at 0.55 ( 11 ) . figure 7 further demonstrates the reason for our higher sticking probabilities when compared to those results of veeraghattam et al . ( 2014 ) on 70 k ice , while figure 6 demonstrates the reason that our results agree on 10 k ice . as a result of figure 7 , we decided to select @xmath63 as our cutoff value in all sticking calculations . however , based on the studies of al - halabi et al . ( 2002 , 2004 , and 2007 ) , we decided apply a second criterion for sticking . after exploring other parameters such as the final mean @xmath25 velocity of the h atom in the last 0.3 ps of a run and the standard deviation in @xmath25 position of the h atom in the final 0.3 ps of a run , we decided that final total energy ( @xmath53 ) was the appropriate choice . figures 8 and 9 plot the final total energy value against the final @xmath25 coordinate value for each of the 300 runs of a 150 k atom on 10 and 70 k ice , respectively . atoms in the dense region towards the upper right of each plot are clear non - sticking events . these are atoms that have left the simulation box within the timescale of the simulation . atoms whose location is near to the bottom left corner of the plots are definite sticking events . these are atoms that are embedded in the slab with virtually zero kinetic energy . comparing the middle regions of figure 8 to figure 9 , one is able to notice some significant differences . for instance , in figure 8 , the atoms on the 10 k substrate that have not left the simulation box ( i.e. those atoms not in the upper right hand corner ) are relatively densely packed when compared to those atoms on the 70 k substrate . this shows that the higher temperature of the substrate is affecting where the atoms will be and what energy the atoms will have at the end of the simulation . note also that on the 10 k substrate the atoms seem to be grouped together in clumps , whereas there is no grouping behavior on the 70 k substrate . the reasons for this clumping behavior are not completely clear at the present but they are likely due to regions of our slab configuration which are energetically favorable for the h - atoms to settle . there appear to be pockets " in our slab that the atoms have a preference to congregate towards . by contrast , in the 70 k substrate , there are no energetically favorable sites for the atoms to settle into so all regions of z@xmath64 and final total energy are equally likely . to summarize , the clumping behavior is likely due to the differences in the particular amorphous ice slab configuration between the 10 k and 70 k slabs . further analysis would be necessary to confirm this claim . based on figures 8 and 9 we can draw a box around the atoms which have stuck to the substrate using @xmath52 and @xmath53 . the atoms in the box have adsorbed to the substrate whereas those atoms which are outside the box have not adsorbed to the substrate . figure 10 displays how the selection of @xmath53 with a fixed @xmath65 affects sticking probability on 70 k ice . note that we only show the 70 k substrate here because it is the 70 k substrate that we are using to choose our sticking event parameters . we see that the sticking probability plateaus around -500 k , but in order to be conservative about our selection and to agree with the parameter of al - halabi et al . ( 2002 ) , we select @xmath66 k. c c c c c c 10 & 5 & 0.993 & 0.014 & 0.990 & 0.017 + & 10 & 0.993 & 0.017 & 0.980 & 0.023 + & 50 & 0.957 & 0.040 & 0.860 & 0.045 + & 100 & 0.917 & 0.049 & 0.647 & 0.046 + & 150 & 0.770 & 0.055 & 0.490 & 0.040 + & 200 & 0.567 & 0.068 & 0.381 & 0.033 + & 250 & 0.457 & 0.060 & 0.304 & 0.027 + & 300 & 0.373 & 0.042 & 0.248 & 0.023 + & 400 & 0.220 & 0.024 & 0.173 & 0.016 + 20 & 5 & 1.000 & 0.000 & 0.976 & 0.017 + & 10 & 0.977 & 0.030 & 0.948 & 0.022 + & 50 & 0.917 & 0.027 & 0.800 & 0.046 + & 100 & 0.843 & 0.064 & 0.594 & 0.047 + & 150 & 0.640 & 0.059 & 0.455 & 0.039 + & 200 & 0.553 & 0.066 & 0.361 & 0.032 + & 250 & 0.410 & 0.056 & 0.295 & 0.027 + & 300 & 0.297 & 0.037 & 0.247 & 0.022 + & 400 & 0.223 & 0.021 & 0.180 & 0.017 + 30 & 5 & 0.963 & 0.031 & 0.973 & 0.027 + & 10 & 0.967 & 0.027 & 0.979 & 0.021 + & 50 & 0.923 & 0.041 & 0.812 & 0.048 + & 100 & 0.800 & 0.062 & 0.613 & 0.050 + & 150 & 0.770 & 0.051 & 0.467 & 0.043 + & 200 & 0.523 & 0.059 & 0.365 & 0.035 + & 250 & 0.440 & 0.054 & 0.292 & 0.029 + & 300 & 0.377 & 0.064 & 0.239 & 0.024 + & 400 & 0.220 & 0.033 & 0.167 & 0.017 + 40 & 5 & 0.937 & 0.039 & 0.940 & 0.036 + & 10 & 0.937 & 0.034 & 0.941 & 0.037 + & 50 & 0.900 & 0.046 & 0.766 & 0.051 + & 100 & 0.743 & 0.057 & 0.572 & 0.050 + & 150 & 0.647 & 0.063 & 0.437 & 0.044 + & 200 & 0.547 & 0.057 & 0.345 & 0.037 + & 250 & 0.420 & 0.052 & 0.278 & 0.031 + & 300 & 0.310 & 0.047 & 0.230 & 0.027 + & 400 & 0.210 & 0.032 & 0.164 & 0.020 + 50 & 5 & 0.917 & 0.033 & 0.916 & 0.043 + & 10 & 0.910 & 0.047 & 0.914 & 0.052 + & 50 & 0.870 & 0.064 & 0.733 & 0.058 + & 100 & 0.707 & 0.057 & 0.541 & 0.050 + & 150 & 0.610 & 0.054 & 0.408 & 0.040 + & 200 & 0.510 & 0.064 & 0.317 & 0.032 + & 250 & 0.387 & 0.044 & 0.252 & 0.026 + & 300 & 0.307 & 0.040 & 0.206 & 0.022 + & 400 & 0.180 & 0.023 & 0.144 & 0.015 + 60 & 5 & 0.853 & 0.055 & 0.861 & 0.061 + & 10 & 0.873 & 0.062 & 0.844 & 0.066 + & 50 & 0.767 & 0.069 & 0.662 & 0.060 + & 100 & 0.647 & 0.058 & 0.501 & 0.048 + & 150 & 0.530 & 0.059 & 0.388 & 0.038 + & 200 & 0.457 & 0.047 & 0.306 & 0.031 + & 250 & 0.410 & 0.034 & 0.247 & 0.025 + & 300 & 0.340 & 0.033 & 0.203 & 0.021 + & 400 & 0.200 & 0.021 & 0.144 & 0.015 + 70 & 5 & 0.747 & 0.061 & 0.768 & 0.058 + & 10 & 0.780 & 0.063 & 0.771 & 0.052 + & 50 & 0.730 & 0.052 & 0.635 & 0.060 + & 100 & 0.633 & 0.078 & 0.481 & 0.053 + & 150 & 0.540 & 0.061 & 0.373 & 0.044 + & 200 & 0.433 & 0.062 & 0.297 & 0.036 + & 250 & 0.370 & 0.030 & 0.242 & 0.029 + & 300 & 0.290 & 0.046 & 0.201 & 0.024 + & 400 & 0.197 & 0.030 & 0.145 & 0.018
the possible existence of dark matter ( dm ) is one of the most intriguing aspects of modern particle physics , astrophysics and cosmology . in order to survey the nature of dm , a number of observations and experiments have been conducted or are underway around the world . the most recent cosmological results based on _ planck _ measurements of the cosmic microwave background ( cmb ) temperature and lensing - potential power spectra indicate that dm comprises about @xmath4 of the energy density of the universe which also includes about @xmath5 baryon matter and about @xmath6 dark energy @xcite . many theories beyond the standard model of particle physics predict natural candidates for dm , e.g. , the weakly interacting massive particles ( wimps ) which are a class of hypothetical stable neutral particles with a huge range in masses from @xmath7 gev to @xmath8 tev and interaction cross sections with normal matter ( proton ) from @xmath9 to @xmath10 @xmath11 @xcite . in terrestrial laboratory , dm might be directly detected through their elastic scattering off nuclei in particle detectors @xcite . a number of underground dm direct detection experiments have been performed , and among them an excess of events over the expected background has been observed by cogent @xcite , dama @xcite , cresstii @xcite as well as in the recent results presented by the cdms - ii(si ) collaboration @xcite . however , these results are in strong tension with the constraints set by some other experimental groups like xenon100 @xcite , lux @xcite and supercdms(ge ) @xcite , leaving a confusing situation for the community . this has led to a number of attempts trying to explain the discrepancy by considering atomic uncertainties @xcite or different mechanisms that deviate from standard assumptions about dm interactions or its astrophysical distributions @xcite . isospin - violating dark matter ( ivdm ) provides a very promising mechanism to reconcile the tension among different experiments @xcite . within the ivdm framework , dm is assumed to couple differently with protons and neutrons , and this assumption of the isospin violation has been supported by a number of theoretical works @xcite based on the particle physics point of view . many parameters need to be specified in the standard method of analyzing dm direct detection experiments @xcite and recently frandsen _ et al . _ @xcite presented a systematic discussion on the possible ways to ameliorate the tension among different experiments . they found that the tension between the cdms - ii(si ) results and the xenon100 bounds is independent of the astrophysical uncertainties concerning the dm halo and any momentum- and velocity - dependence of the cross section in particle physics , but it can be largely ameliorated or even resolved within the framework of ivdm . besides the uncertainties in astrophysics and particle physics concerning the interaction of a dm particle scattering off a single nucleon mentioned above , the uncertainties in nuclear physics describing how the struck nucleon is distributed inside the nucleus may also play an important role in interpreting dm signals . this is because in dm direct detection experiments , the nuclear form factors are generally applied to describe the dm - nucleus cross section and the bounds on dm - nucleon cross section are then obtained accordingly . in particular , the empirical helm form factor extracted from nuclear charge distributions @xcite has been commonly adopted in current direct detection experiments . however , the dm - nucleus interaction should be in principle described by using the form factors of both proton and neutron distributions in the nuclei rather than the charge distributions since the dm particles actually interact with the protons and neutrons in the nuclei . within the framework of relativistic mean field ( rmf ) model , chen _ et al . _ @xcite derived the nuclear form factor for the spin - independent scattering between the wimps and nucleus , and they found that the results can deviate from the empirical helm form factor by @xmath12 to @xmath13 in a large range of recoil spectrum of @xmath14 kev . a recent work by co _ et al . _ @xcite suggests that the use of different distributions for protons and neutrons instead of the commonly used empirical charge distributions for a target nucleus could be important in interpreting dm signals , especially if ivdm is considered . for stable nuclei , the proton distribution can be well determined from the charge distribution which can be accurately measured with electron scattering @xcite . in contrast , the neutron distribution is usually determined from hadron scattering experiments and the results are generally highly model dependent due to the unclear non - perturbative strong interaction @xcite . recently , the lead radius experiment ( prex ) collaboration at jefferson laboratory ( jlab ) published their results on the measurement of the parity - violating cross section asymmetry in the elastic scattering of polarized electrons from @xmath15 @xcite , which provides a model - independent determination of the neutron density distributions in @xmath15 . the prex measurement leads to a value of @xmath16 for the rms radius of the neutron distributions for @xmath15 , implying a large neutron skin thickness , i.e. , @xmath17 by assuming a point - proton rms radius of @xmath18 fm @xcite . one can see that the obtained value of @xmath19 has a large error , indicating a large uncertainty of the neutron distribution relative to the proton distribution in @xmath15 . theoretically , it has been established @xcite that the @xmath19 is intimately related to the nuclear matter symmetry energy which characterizes the isospin dependent part of the equation of state ( eos ) of asymmetric nuclear matter ( anm ) @xcite . in particular , it has been shown recently that the @xmath19 of heavy nuclei is uniquely determined by the density slope @xmath20 of the symmetry energy at a subsaturation cross density @xmath21 @xmath22 @xcite . these features imply that the uncertainties of @xmath19 , especially the neutron distributions , predicted by various nuclear models are essentially due to our poor knowledge about the symmetry energy . the symmetry energy is of critical importance for understanding not only the structure and reaction of radioactive nuclei , but also a number of interesting issues in astrophysics , such as the structure of neutron stars and the mechanism of supernova explosions , and has become a hot topic in current research frontiers of nuclear physics and astrophysics @xcite . the determination of the symmetry energy provides a strong motivation for studying isospin - dependent phenomena with radioactive nuclei at a number of new / planning rare isotope beam facilities around the world , such as csr / lanzhou and brif - ii / beijing in china , ribf / riken in japan , spiral2/ganil in france , fair / gsi in germany , spes / lnl in italy , raon in korea , and frib / nscl and t - rex / tamu in usa . in this work , using the proton and neutron form factors obtained from skyrme - hartree - fock calculations by varying the symmetry energy slope parameter @xmath23 within the uncertainty region set by the prex experiment , we investigate the form factor effects in the direct detection of dark matter . this article is organized as follows . we briefly introduce in sec . [ sec : model ] the theoretical models and methods used in the present work , and then present the results and discussions in sec . [ sec : sym ] . finally , a conclusion is given in sec . [ sec : conclusion ] . the eos of isospin asymmetric nuclear matter , defined by its binding energy per nucleon , can be well approximated by @xmath24 in terms of baryon density @xmath25 and isospin asymmetry @xmath26 with @xmath27 and @xmath28 denoting the proton and neutron densities , respectively . @xmath29 corresponds to the eos of symmetric nuclear matter , and the nuclear symmetry energy can be expressed as @xmath30 there are no odd - order @xmath31 terms in eq . ( [ eosanm ] ) due to the exchange symmetry between protons and neutrons ( isospin symmetry ) in nuclear matter . neglecting the contribution from higher - order terms in eq . ( [ eosanm ] ) leads to the well - known empirical parabolic law for eos of anm , which has been verified by all many - body theories to date , at least for densities up to moderate values @xcite . furthermore , around a reference density @xmath32 , the symmetry energy @xmath33 can be expanded as @xmath34 with @xmath35 the slope parameter @xmath36 characterizes the density dependence of symmetry energy around @xmath32 . it has been shown @xcite that the neutron skin thickness @xmath37 of heavy nuclei is uniquely fixed by the density slope @xmath23 at a subsaturation cross density @xmath38@xmath22 which roughly corresponds to the average density of the nuclei . for the calculations of finite nuclei , we use in the present work the standard skyrme - hartree - fock ( shf ) approach in which the nuclear effective interaction is taken to have a zero - range , density- and momentum - dependent form , i.e. @xcite , @xmath39 } , \label{v12sky}\end{aligned}\]]with @xmath40 and @xmath41 . in the above expression , the relative momentum operators @xmath42 and @xmath43 act on the wave function on the right and left , respectively . the quantities @xmath44 and @xmath45 denote , respectively , the spin exchange operator and pauli spin matrices . the skyrme interaction in eq . ( [ v12sky ] ) includes totally @xmath46 parameters , i.e. , the @xmath47 skyrme interaction parameters @xmath48 , @xmath49 , @xmath50 , and the spin - orbit coupling constant @xmath51 . this standard shf approach has been shown to be very successful in describing the structure of finite nuclei , especially the global properties such as binding energies and charge radii @xcite . instead of using directly the @xmath47 skyrme interaction parameters , we can express them explicitly in terms of @xmath47 macroscopic quantities , i.e. , @xmath52 , @xmath53 , the incompressibility @xmath54 , the isoscalar effective mass @xmath55 , the isovector effective mass @xmath56 , @xmath57 , @xmath58 , @xmath59 , and @xmath60 . the @xmath59 and @xmath60 are respectively the gradient and symmetry - gradient coefficients in the surface interaction part of the binding energies for finite nuclei which is defined as @xmath61 ^{2}/(2{\rho ) } .\ ] ] then , by varying individually these macroscopic quantities within their known ranges , one can examine more transparently the correlation of nuclear matter properties with each individual macroscopic quantity . recently , this correlation analysis method has been successfully applied to study the neutron skin @xcite and giant monopole resonance of finite nuclei @xcite , the higher - order bulk characteristic parameters of anm @xcite , and the relationship between the nuclear matter symmetry energy and the symmetry energy coefficient in nuclear mass formula @xcite . in the present work , we use the @xmath47 macroscopic quantities instead of using directly the @xmath47 skyrme interaction parameters . especially , we study the symmetry energy effects by varying the @xmath62 value while keeping the other macroscopic quantities unchanged . in this work , we use the standard method of analyzing dm direct detection experiments @xcite . the spin - independent differential event rate of nuclear recoils with a recoil energy of @xmath63 , occurring in a detector due to an elastic collision between a target nucleus of mass @xmath64 and a dm particle of mass @xmath65 , can be expressed as @xmath66 where a represents the mass number of the target nucleus and @xmath67 is the number of target nucleus per unit mass in the detector , @xmath68 is the local halo dm density , @xmath69 is the local dm velocity distribution evaluated in the galactic rest frame with @xmath70 , @xmath71 is earth velocity in the galactic rest frame , and @xmath72 is the galactic escape velocity @xcite . @xmath73 and @xmath74 denote , respectively , the spin - independent differential dm - proton scattering cross - section and the effective form factors , and they will be detailed below . the @xmath75 is the momentum transfer between the dm particle and the struck nucleus . the lower limit @xmath76 of the integration in eq . ( [ dre ] ) corresponds to the smallest dm velocity that can give a recoil energy of @xmath63 , i.e. , @xmath77 where @xmath78 is the dm - nucleus reduced mass . the predicted normalized recoil spectrum can thus be expressed as @xmath79 where @xmath80 is the natural abundance of the isotope with mass number a for the target element . the total event rate @xmath81 expected in an energy range @xmath82 $ ] in a detector consisting of compound targets with finite detector energy resolution , usually characterized by the unit of one event @xmath83 , can be written as @xcite @xmath84 where @xmath85 and @xmath86 are the detector acceptance and the detector response function , respectively . the @xmath63 integration region is determined by the cuts in the experiments . it is convenient , and usually adequate , to describe the matter distributions of a finite nucleus by a form factor , @xmath87 , in analyzing dm direct detection experiments @xcite . in the first - order born approximation , the form factors of spherical nuclei are the fourier transforms of the proton and neutron density distributions , i.e. , @xmath88 where the index p ( n ) denotes protons ( neutrons ) . the density distributions , @xmath89 , for spherical nuclei can be obtained by using a mean - field approach @xcite , e.g. , the shf approach as we discussed above . we further define the so - called effective form factors in eq . ( [ dre ] ) in terms of @xmath90 as @xmath91 where z ( n ) is the proton ( neutron ) number of the target nucleus , a = z+n , and @xmath92 is the isospin - violation factor with @xmath93 and @xmath94 denoting the effective coupling of dm to neutrons and protons , respectively . the spin - independent differential dm - proton scattering cross - section can then be expressed as @xcite @xmath95 where @xmath96 is the reduced dm - proton mass , and @xmath97 is the elastic dm - proton cross - section at zero momentum transfer ( @xmath98 ) . integrating the differential event rate in eq . ( [ dre ] ) with respect to @xmath63 , i.e. , eq . ( [ ter ] ) , one can relate the expected number of scattering events in a direct detection experiment with the dm - proton cross - section @xmath97 . furthermore , by adopting , e.g. , a profile likelihood analysis proposed by the xenon100 collaboration @xcite or the maximum gap method introduced by yellin @xcite _ et al . _ , one can constrain the elastic dm - proton cross - section @xmath97 in the @xmath65-@xmath99 plane with specific confidence level . however , in experiments usually reported is the dm - nucleon cross section @xmath100 , rather than @xmath97 . @xmath100 is an effective cross section by assuming that the dm particle couples with protons and neutrons equally ( i.e. , @xmath101 ) and the protons and neutrons have identical form factor which is in practice commonly taken as the empirical charge form factor parameterized by helm @xcite . obviously , the dm - proton cross section @xmath97 is generally different from the dm - nucleon cross section @xmath100 , but they can be related to each other by the so - called degradation factor @xmath102 defined as @xcite @xmath103 with @xmath104 in eq . ( [ degra ] ) , @xmath105 is the helm form factor which is obtained by using a charge density with a gaussian surface distribution and has the following simple analytical expression @xmath106 where @xmath107 is the first - order spherical bessel function , @xmath108 is an effective nuclear radius and @xmath109 is a measure of the nuclear skin thickness . for @xmath108 and @xmath109 , the following standard parametrization has been used @xcite @xmath110 in eq . ( [ degra ] ) , it has been assumed that the variations of @xmath111 and form factors do not change the shape of recoil spectra @xmath112 in eq . ( [ spec ] ) . in this case , varying @xmath111 and form factors does not change the event distributions but simply modifies the expected number of total events , @xmath81 in eq . ( [ ter ] ) , for a given cross section . in general cases with different @xmath111 and nuclear form factors , since @xmath113 , we can simply rescale the corresponding cross sections according to the degradation factor as @xmath114 to keep the likelihood function ( the maximum gap ) in a profile likelihood analysis ( the maximum gap method ) unchanged . however , if the recoil spectrum changes significantly , the degradation factor in eq . ( [ degra ] ) should be modified according to the following identity @xmath115 where @xmath116 denotes the degradation factor defined in eq . ( [ degra ] ) , @xmath117 and @xmath118 are the number of expected events with @xmath119 and @xmath120 being the upper limits of the dm - proton and dm - nucleon cross sections with certain confidence level . we estimate @xmath117 and @xmath118 by using the `` maximum gap method '' @xcite in this work . in this way , one can thus study the isospin - violating effects and the form factor effects simultaneously in analyzing the dm signals . as mentioned above , the neutron skin thickness of heavy nuclei is uniquely determined by the symmetry energy density slope parameter @xmath20 at a subsaturation cross density @xmath121 @xcite . therefore , the @xmath20 parameter controls the relative difference between the neutron and proton distributions ( and thus the neutron and proton form factors ) in the nuclei . in the present work , we investigate the form factor effects by varying the @xmath20 parameter in the shf calculations to fit the model - independent result of @xmath122 for @xmath0pb from the recent prex experiment at jlab . xe vs. that for @xmath0pb predicted by mean - field models with different interactions . constraints set by prex measurement @xcite are also shown.,width=321 ] shown in fig . [ nsts ] is the @xmath19 of @xmath123 as a function of that of @xmath15 in shf calculations with the skyrme interaction msl1 @xcite by varying @xmath20 while keeping the other 8 macroscopic quantities and the spin - orbit coupling constant @xmath51 fixed at their default values in the msl1 interaction , namely , @xmath124 @xmath22 , @xmath125 mev , the incompressibility @xmath126 mev , the isoscalar effective mass @xmath127 , the isovector effective mass @xmath128 , @xmath129 mev , @xmath130 mev@xmath131@xmath132 , @xmath133 mev@xmath131@xmath132 , and @xmath134 mev@xmath131@xmath132 . the msl1 interaction has been obtained by fitting a number of experimental data of finite nuclei , including the binding energy , the charge rms radius , the neutron @xmath135 energy level splitting in @xmath0pb , isotope binding energy difference , and neutron skin data of sn isotopes . for comparison , we also include in fig . [ nsts ] the results from shf calculations with @xmath136 other skyrme interactions ( bsk1 , bsk4 , bsk5 , bsk7 , bsk10 , bsk14 , bs15 , dutta , e - fit , esigma - fit , gsigma - fit , kde , kde0 , msl0 , ratp , rsigma - fit , sgi , sgii , sk255 , sk272 , ska , ski1 , ski2 , ski5 , skm , skmp , skm * , skt1 , skt4 , skt5 , skt6 , skz0 , skz1 , skz2 , skz3 , skz4 , skz-1 , sly4 , sly5 , sly9 , z - fit , zsigma - fit , and zsigmas - fit ) and rmf calculations with @xmath137 different interaction parameter sets ( fsu , iufsu , tm1 , pk1 , nl3 , and nl1 ) . the references of these skyrme and rmf interactions can be found in refs . @xcite . we have selected these interactions in such a way that their @xmath23 values scatter in large region and are not close to each other . one can see clearly a nice model - independent linear correlation between the neutron skin thicknesses of @xmath15 and @xmath123 within the non - relativistic and relativistic models with different interactions . in addition , although there are few interactions that predict a @xmath19 of @xmath15 larger than @xmath138 fm measured by the model - independent prex experiment , one can easily obtain a large @xmath19 of @xmath15 by increasing the @xmath20 value in the msl1 interaction . in particular , we find a value of @xmath139 mev in the msl1 interaction predicts @xmath140 fm for @xmath0pb and @xmath141 fm for @xmath123 . therefore , we use @xmath142 mev , @xmath143 mev and @xmath144 mev in the msl1 interaction , denoted as lc40 , lc92 and lc134 , respectively , to study the symmetry energy and form factor effects in the present work . we would like to point out that the large neutron skin thickness for @xmath0pb measured by the prex experiment can also be explained within the rmf model @xcite . .,width=510 ] to test the skyrme interaction parameter sets lc40 , lc92 , and lc134 , we calculate the binding energies and charge rms radii for a number of closed - shell or semi - closed - shell nuclei , i.e. , @xmath145o , @xmath146ca , @xmath147ca , @xmath148ni , @xmath149zr , @xmath150sn , @xmath151sn , and @xmath0pb , as well as the nuclei @xmath152si , @xmath153ge , and @xmath151xe . the elements si , ge and xe are widely used as targets in the dm direct detection experiments . for si and ge , their natural abundances are mostly dominated by their isotopes of @xmath152si and @xmath153ge , respectively . for xe , however , it has several isotopes with comparable natural abundances and here we just select one of them , i.e. @xmath151xe , as an example to show its properties . [ ebrc ] shows the relative deviation of the binding energies and charge rms radii of these nuclei from those measured in experiments @xcite . it is seen that the interactions lc40 and lc92 can describe the experimental data very well ( the deviations are within about @xmath154 ) except for the light nucleus @xmath145o for which the deviation of charge rms radius is about @xmath155 . it is interesting to see that , the interaction lc134 , which predicts a very strong density dependence of the symmetry energy at @xmath156 and gives a very large neutron skin thickness of @xmath157 fm for @xmath15 ( the upper limit of the prex measurement ) , still can give a reasonable description of the experimental data ( the deviations are within about @xmath158 as indicated by bands in fig . [ ebrc ] ) except the nucleus @xmath151sn for which the deviation of charge rms radius ( binding energy ) reaches about @xmath159 ( @xmath160 ) and for the light nucleus @xmath145o the deviation of charge rms radius is still about @xmath155 . these results are remarkable as lc40 , lc92 and lc134 are not obtained from fitting measured binding energies and charge rms radii of finite nuclei as in usual skyrme parametrization . it should be pointed out that our main motivation for introducing the interactions lc40 , lc92 and lc134 is not to construct new skyrme interaction parameter sets to describe data , but to use them as references to study the form factor effects in the following . for @xmath152si ( left panels ) and @xmath151xe ( right panels ) obtained from shf with lc40 , lc92 and lc134 . the momentum transfer @xmath161 , corresponding to a value of @xmath162 kev , is indicated by dash - dot - dotted lines.,width=510 ] to investigate the symmetry energy effects on the proton and neutron form factors in finite nuclei , we plot in fig . [ fnp ] the proton ( upper panels ) and neutron ( lower panels ) form factors as functions of momentum transfer @xmath163 for @xmath152si ( left panels ) and @xmath151xe ( right panels ) from shf calculations with lc40 , lc92 and lc134 . for comparison , the results are divided by the corresponding empirical helm form factors . the proton and neutron form factors of other isotopes or elements can be obtained from a generalized helm - like empirical parametrization given in the appendix where the proton and neutron form factors are distinguished and parameterized as functions of the neutron skin thickness of @xmath0pb by fitting the results obtained from shf calculations with lc40 , lc92 and lc134 . in addition , also indicated in fig . [ fnp ] is the momentum transfer @xmath161 corresponding to a value of @xmath162 kev which represents the possible maximum cutoff among various experiments . however , we would like to emphasize that the upper limits of the recoil energy @xmath63 usually vary significantly in different dm direct detection experiments , and here we just select a relatively large recoil energy of @xmath8 kev as an example to see the possible symmetry energy effects on the proton and neutron form factors in different nuclei . in calculations of the scattering cross sections in eq . ( [ degra ] ) for each experiment , one must use the recoil energy region given by each experiment . it is seen from fig . [ fnp ] that , for the nucleus @xmath152si with equal proton and neutron numbers , the symmetry energy effect is tiny and the shf calculations with lc40 , lc92 and lc134 predict almost the same form factors for neutrons and protons , which are further in good agreement with the empirical helm form factor for @xmath164 . on the other hand , for the neutron - rich nucleus @xmath151xe , one can see a clear symmetry energy effect on the form factors , namely , a larger @xmath23 shifts the neutron form factor to lower @xmath163 values while shifts slightly the proton form factor to higher @xmath163 values with the empirical helm form factor in between , leading to an isospin splitting between the proton and neutron form factors with respect to the @xmath163 value . these features actually reflect the symmetry energy effects on the neutron skin thickness , namely , increasing the value of @xmath23 increases the rms radius of neutrons and reduces slightly the rms radius of protons , and thus leads to a larger neutron skin thickness as shown in fig . . for si ( left panels ) and xe ( right panels ) with @xmath165 ( upper panels ) and @xmath166 ( lower panels ) obtained from shf with lc40 , lc92 and lc134 . the empirical helm form factor is also included for comparison.,width=510 ] since the effective form factor ( eq . ( [ eff ] ) ) should be used to analyze the dm signals for general cases , it is expected that the isospin splitting between the proton and neutron form factors would cause significant effects on the effective form factor , especially for a negative @xmath111 which could cause strongly destructive interference of dm scattering with protons and neutrons . on the other hand , the presence of different isotopes with comparable abundances in some detector target elements ( e.g. , xenon ) suggests that it is better to study the averaged effective form factors defined as @xmath167 this is illustrated in fig . [ ffs ] where the averaged effective form factor is plotted as a function of @xmath63 for si ( left panels ) and xe ( right panels ) from shf calculations with lc40 , lc92 and lc134 . the results from the empirical helm form factor with @xmath168 are also included for comparison . in fig . [ ffs ] , two cases for the dm are considered : one is for the standard isospin - invariant dm with @xmath165 ( upper panels ) and the other is for the ivdm with @xmath166 ( lower panels ) . it should be pointed out that @xmath169 , firstly suggested by feng _ et al . _ @xcite , leads to nearly complete destructive interference of the scattering amplitudes for dm - proton and dm - neutron collisions in xe target and gives maximum suppression of the sensitivity for xe - based detectors if the empirical helm form factor is adopted . for such kind of ivdm , the confidence region of cdms - ii(si ) and the xenon100 exclusion contours are consistent with each other @xcite . as will be shown later , for ivdm with @xmath169 , both the results from recent lux @xcite and the most recent supercdms(ge ) @xcite are also consistent with the confidence region of cdms - ii(si ) . theoretically , such negative values of @xmath111 can arise , e.g. , in models with a new light neutral gauge boson @xmath170 @xcite . one can see from fig . [ ffs ] that the negative @xmath111 ( i.e. , @xmath171 ) indeed causes strongly destructive interference of dm scattering with protons and neutrons for both si and xe targets and thus the averaged effective form factor with @xmath169 is strongly suppressed compared to its value with @xmath172 . furthermore , it is interesting to see from fig . [ ffs ] that for @xmath172 , the averaged effective form factors of both si and xe targets from shf calculations with lc40 , lc92 and lc134 are in a very good agreement with the corresponding results from the empirical helm form factors , indicating that there are essentially no symmetry energy effects . however , for @xmath169 , one can see that the averaged effective form factors of xe target display very different behaviors for different distributions from shf calculations with lc40 , lc92 and lc134 as well as that from the empirical charge distributions , although the averaged effective form factors of si target are essentially the same for these different distributions . these features indicate that the symmetry energy effect is tiny for the averaged effective form factor of si target but is very strong for that of xe target in the case of @xmath169 . this is due to the fact that for a negative @xmath111 , the scattering amplitudes for dm - proton and dm - neutron collisions may interfere destructively , and the dm particle is almost completely decoupled from the xe isotopes , i.e. @xmath173 ( see panel ( d ) of fig . [ ffs ] ) , especially for @xmath174 @xcite . in this case , a small difference between @xmath175 and @xmath176 may lead to a significant change for the effective form factor . as a result , the form factor of xe exhibits a very strong symmetry energy effect for @xmath169 , namely , a larger @xmath20 value leads to a larger form factor as shown in panel ( d ) of fig . in particular , one can see that the averaged effective form factors of xe target from shf calculations with lc40 , lc92 and lc134 can be significantly larger than that from the empirical helm form factor in a large range of recoil energy . this implies that the symmetry energy with a larger @xmath20 ( and thus a larger neutron skin thickness ) can give a more significant enhancement for the effective dm - nucleus form factors in eq . ( [ dre ] ) when the ivdm with @xmath169 is considered . therefore , for @xmath169 , one can expect a significant enhancement for the sensitivity of the xe - based detectors to the dm - proton cross sections by taking into account the form factor effects due to different symmetry energies compared with the corresponding result using the empirical helm form factor in analyzing the experimental data . confidence level ( c.l . ) limits from xenon100 , lux and supercdms(ge ) together with @xmath177 c.l . favored regions from cdms - ii(si ) in @xmath65-@xmath97 plane for @xmath172 ( left panel ) and @xmath169 ( right panel ) with form factors from the empirical charge distributions ( helm ) and shf calculations with msl1 by varying @xmath23 from @xmath178 mev to @xmath144 mev.,width=510 ] to see the effects of the form factor and the isospin - violation factor @xmath111 on the extraction of elastic spin - independent dm - proton cross sections @xmath97 in dm direct detection experiments , we show in fig . [ css ] the results of the @xmath177 confidence level ( c.l . ) limits from xenon100 @xcite , lux @xcite and supercdms(ge ) @xcite along with the @xmath177 c.l . favored regions from cdms - ii(si ) @xcite in @xmath65-@xmath97 plane for @xmath172 ( left panel ) and @xmath169 ( right panel ) with form factors from the empirical charge distributions and shf calculations with msl1 by varying @xmath23 from @xmath178 mev to @xmath144 mev . for the results from supercdms(ge ) , only the mass region less than @xmath179 gev is shown since no data are available for @xmath180 gev @xcite . in these analyses , we use the standard astrophysical parameters in the standard halo model , namely , a maxwell - boltzmann distribution for @xmath181 with @xmath182 and galactic escape velocity @xmath183 , and a dm density of @xmath184 @xcite . in fig . [ css ] , the form factor effects due to the symmetry energy and the effects from @xmath111 are considered through dividing the experimental results by the degradation factor @xmath102 as discussed earlier . in particular , for xenon100 , we assume that the energy resolution is dominated by poisson fluctuations in the number of photoelectrons ( pe ) as in ref . we use s1 efficiency obtained from fig . 1 of ref . @xcite and the scintillation efficiency @xmath185 from fig . 1 of ref . @xcite , respectively , and integrate the differential rate over s1 from @xmath2 to @xmath179 pe to get the total rate . for the lux analysis which is similar with xenon100 s , the signal region is limited within @xmath186 to @xmath179 pe and the efficiencies are taken from fig . 1 of ref . @xcite . for cdms - ii(si ) , we use the acceptance from fig . 1 of ref . @xcite , and choose an energy interval between @xmath187 and @xmath188 in the analysis . for supercdms(ge ) , with the efficiency from fig . 1 of ref . @xcite , we use the signal region in the range @xmath189 in the calculations as in ref . @xcite . furthermore , since xenon100 reported @xmath186 candidate events in the low recoil energy region , we calculate the degradation factor according to eq . ( [ degra2 ] ) instead of eq . ( [ degra ] ) to take into account the effects due to shape changing of the recoil spectrum as we will show later . similarly , for lux , we follow the analysis in ref . @xcite and calculate the degradation factor according to eq . ( [ degra2 ] ) by using one observed event ( with @xmath190 pe ) . for cdms - ii(si ) and supercdms(ge ) that focus on the light dm , we have checked that the shape changing of their recoil spectra can be negligible and eq . ( [ degra ] ) is thus applied to evaluate the degradation factor . for the isospin - invariant dm with @xmath172 , one can see from fig . [ css ] that the cdms - ii(si ) results are in some tension with the upper limit placed by the xenon100 experiment , and they become in even strong disagreement with the upper limits set by the recent lux @xcite experiment and the more recent supercdms(ge ) @xcite experiment . however , for the ivdm with @xmath169 , one can see that the tension between cdms - ii(si ) and xenon100 is essentially disappeared as found by frandsen _ et al . _ furthermore , it is remarkable to see from the right panel of fig . [ css ] that the disagreement of the cdms - ii(si ) results with lux and supercdms(ge ) can also be largely ameliorated within the framework of ivdm with @xmath169 ( we note that a similar conclusion is also obtained in ref . therefore , these results indicate that ivdm indeed provides a very promising mechanism to reconcile the tension among various experiments . it should be mentioned here that although @xmath169 leads to a maximum suppression for the sensitivity of xe - based detector for the empirical helm form factor , it does not for the more realistic shf form factors as will be shown later . furthermore , it is interesting to see that , although there are essentially no form factor effects on the extracted spin - independent dm - proton cross sections @xmath97 in different experiments in the case of isospin - invariant dm with @xmath172 , the form factor effects can significantly affect the extraction of @xmath97 for the xe - based experiments ( xenon100 and lux ) in the case of ivdm with @xmath169 . therefore , our results indicate that , for isospin - invariant dm , the widely used empirical helm form factor is well grounded and this is consistent with our previous discussions about the form factor effects . on the other hand , for ivdm with @xmath169 , using the form factors obtained from shf calculations in msl1 with varied @xmath23 values generally increases the sensitivity of the xe - based detectors ( xenon100 and lux ) compared with using the empirical helm form factor , and a larger @xmath23 value , which leads to a larger neutron skin thickness in xe isotopes , generally leads to a stronger sensitivity of the xe - based detectors . this form factor effect becomes more pronounced with the increment of the dm mass @xmath65 . while the form factor effect is about @xmath191 for the dm mass of @xmath192 gev , it can become very significant for the dm with mass above a few tens gev . for @xmath193 gev , which roughly corresponds to the dm mass upper limit of cdms - ii(si ) bounds , the sensitivities of xenon100 and lux can be enhanced by a factor of @xmath2 using the form factor with @xmath194 mev compared with that using the empirical helm form factor . in particular , the sensitivities can have a factor of more than @xmath46 improvement for @xmath195 gev where many constraints for supersymmetric wimps have been put by the data from lhc @xcite . on the other hand , the relative variation between the results extracted from the helm form factor and the shf calculations is small ( @xmath196 ) for supercdms(ge ) and it can be negligible ( @xmath197 ) for cdms - ii(si ) for which the results from the shf calculations are not shown in fig . [ css ] . with @xmath169 for @xmath198 ( panel ( a ) ) , @xmath199 ( panel ( b ) ) and @xmath200 ( panel ( c ) ) , respectively . cases for different form factors from the empirical helm charge distribution ( black solid lines ) and shf calculations with msl1 by varying @xmath23 from @xmath178 mev to @xmath144 mev are shown in the figure . for comparison , spectra calculated with helm form factors for isospin - invariant cases ( @xmath201 ) are also shown.,width=529 ] the dm mass dependence of the form factor effects observed in xenon100 and lux for @xmath169 can be understood as a result of the shape changing of the expected recoil spectra for varied @xmath65 due to the enhancement effects of the form factors when we take the symmetry energy effects into account as discussed in the previous subsection . shown in fig . [ spe ] is the normalized dm - xenon recoil spectrum @xmath112 for @xmath202 , @xmath203 and @xmath204 with @xmath169 and various form factors from shf calculations with lc40 , lc92 and lc134 as well as the empirical helm form factor . we note that for @xmath172 , the @xmath112 essentially does not depend on the variation of the form factors and the results are almost the same as that from the empirical helm form factor with @xmath169 . it is very interesting to see from fig . [ spe ] that for the case with @xmath169 , although the form factor effect essentially does not change or only slightly modifies the predicted normalized dm - xenon recoil spectra for light dm ( see , e.g. , figs . [ spe ] ( a ) and ( b ) ) , it can drastically change the shape of the recoil spectra for heavy dm ( see , e.g. , fig . [ spe ] ( c ) ) . for both xenon100 and lux , we have considered this spectrum shape changing effect in eq . ( [ degra2 ] ) by multiplying eq . ( [ degra ] ) by another term @xmath205 for the degradation factor , which will result in , e.g. , @xmath206 ( @xmath207 ) enhancement of the upper limit of @xmath97 from xenon100 ( lux ) relative to the results obtained from eq . ( [ degra ] ) for the lc134 case with @xmath208 . more specifically , one can see from fig . [ spe ] that for heavier dm , the expected dm - xenon recoil spectra become harder and thus the effective form factors are typically probed at higher momentum transfer , and this will significantly enhance the effects caused by different form factors ( see , e.g. , fig . [ ffs ] ( d ) ) adopted for the @xmath63 integration in the numerator of r.h.s of eq . ( [ degra ] ) , leading to the observed dm mass dependence of the form factor effects in xenon100 and lux for @xmath169 . it is also very interesting to see that for the case with @xmath169 , the bounds of xenon100 and lux overlap with each other in the mass region of @xmath209 gev although the lux gives much more stringent limit on the spin - independent elastic dm - proton scattering cross section than xenon100 in the lower mass region as shown in the right panel of fig . this can also be explained as a result of the shape changing of the expected recoil spectrum for varied @xmath65 as shown in fig . [ spe ] . particularly , fig . [ spe ] suggests that a significant fraction of the events is expected at larger recoil energies for heavier dm , and this is especially the case for the ivdm with @xmath169 and shf form factors with lc92 and lc134 where one can see the non - standard recoil spectra peaked around @xmath210 kev . since lux adopts a relatively small energy range , i.e. @xmath211 kev @xcite , compared with xenon100 where @xmath212 kev is adopted @xcite , it thus exhibits relatively weaker sensitivity than xenon100 for heavier dm . as a matter of fact , even for the isospin - invariant case where the form factor effects are negligible as shown in fig . [ css ] ( a ) , the sensitivity of lux becomes relatively weaker with the increment of @xmath65 compared with that of xenon100 . it should be mentioned that the small form factor effect observed for supercdms(ge ) in fig . [ css ] ( b ) is also partially due to the smaller threshold energy cut , i.e. , @xmath213 kev @xcite . our results imply that the non - standard recoil spectra shown in fig . [ spe ] ( c ) may provide an extremely important experimental evidence for non - standard dark matter interactions and they also give valuable information on the direct detection of ivdm . in above calculations for the case of ivdm , we have fixed the isospin - violation factor at @xmath169 . as mentioned earlier , for the specific ivdm with @xmath169 , the sensitivity of xe - based detectors is maximumly suppressed when the empirical helm form factor is used and the tension among various experiments using different target elements has been shown to be largely ameliorated . for more realistic shf form factors , it is interesting to see how the form factor effects change with more general values of the isospin - violation factor @xmath111 . following the method by feng _ et al . _ @xcite , we study in the following how the form factor effects change for various isospin - violation factor @xmath111 . shown in fig . [ xepho ] is the degradation factor @xmath102 as a function of the isospin - violation factor @xmath111 obtained by using the empirical helm form factors as well as the form factors from shf calculations with lc40 , lc92 and lc134 for @xmath214 gev ( panel ( a ) ) and @xmath215 gev ( panel ( b ) ) . since both xenon100 and lux are xe - based experiments and their results have similar dependence on the @xmath111 and the symmetry energy , we thus only consider cdms - ii(si ) , supercdms(ge ) and xenon100 in fig . [ xepho ] ( a ) for @xmath214 gev . in addition , for @xmath215 gev , the form factor effect is negligible for cdms - ii(si ) and relatively small for supercdms(ge ) which adopts a small threshold energy cut and has been designed especially for light dm searching , we thus only show the results of xenon100 ( panel ( b ) ) . as a function of the isospin - violation factor @xmath111 obtained by using the empirical helm form factors as well as the form factors from shf calculations with lc40 , lc92 and lc134 for @xmath214 gev ( panel ( a ) ) and @xmath215 gev ( panel ( b ) ) . for @xmath214 gev , the results of cdms - ii(si ) , supercdms(ge ) and xenon100 are shown while only the results of xenon100 are shown for @xmath215 gev.,width=510 ] it is seen from fig . [ xepho ] that generally the sensitivity of each detector is significantly suppressed for a negative @xmath111 within the range of @xmath216 due to destructive interference between dm - neutron and dm - proton scattering amplitudes as pointed out in ref . however , the @xmath102 reaches its minimum value ( where the detector is most insensitive ) at different values of @xmath111 for different detectors due to the different compositions and distributions of neutrons and protons in each target element of the detector . in particular , one can see that the corresponding value of @xmath111 which leads to a minimum value of @xmath102 , denoted as @xmath217 , is about @xmath218 , @xmath219 and @xmath220 for cdms - ii(si ) , supercdms(ge ) and xenon100 , respectively , for the case of @xmath214 gev when the empirical helm form factor is used . for @xmath214 gev , the form factor effect is very small for cdms - ii(si ) and the largest relative variation with respective to the results from the empirical helm form factor when @xmath23 varies from @xmath221 mev to @xmath144 mev is @xmath222 at @xmath223 . the largest relative variation due to the form factor effect is @xmath224 at @xmath225 for supercdms(ge ) , and it becomes @xmath226 at @xmath227 for xenon100 . as a matter of fact , a clear shift of curves for the xe - based detectors calculated with different form factors has been already seen even for such a small dm mass in fig . [ xepho ] ( a ) . these features indicate that the form factor effect can be already pronounced for xenon100 even for a small mass dm with @xmath214 gev . for xenon100 , one can see from fig . [ xepho](b ) that the form factor effect becomes very strong for @xmath215 gev . in particular , a larger @xmath23 value gives significantly stronger sensitivity of the detectors and the largest relative variation due to the form factor effect is @xmath228 ( almost one order ) at @xmath229 . furthermore , one can see from fig . [ xepho ] that the value of @xmath217 also depends significantly on the form factor , and it is @xmath230 , @xmath231 and @xmath232 for the shf form factor with @xmath233 , @xmath143 and @xmath144 mev , respectively , smaller than @xmath169 for the empirical helm form factor . in addition , it is very interesting to see that , while using the shf form factors enhances the sensitivity of the xe - based detector relative to the case with helm form factor for @xmath234 , it suppresses the sensitivity for @xmath235 . these features indicate that the form factor effects can either enhance or suppress the sensitivity of the detector relative to the helm s case depending on the specific value adopted for @xmath111 . it is also very interesting to see that one can not find a @xmath111 value such that @xmath236 , leading to zero sensitivity for scattering off the elements , and this is mainly due to the fact that the elements have multiple isotopes and completely destructive interference can not be simultaneously achieved for all isotopes @xcite . we would like to point out that even for elements with only one naturally abundant isotope , one still can not find a @xmath111 value such that @xmath236 since the neutron form factor is generally different from the proton form factor in a nucleus . the above results indicate that , the form factor effect can not only significantly change the sensitivities of the detectors , but also affect the interaction behaviors of the target element with the isospin - violating dm in various isospin - violating regions we are interested in for dm direct detection experiments . in future when different direct detection experiments using various target elements had specify the dm signals precisely , one may determine simultaneously the isospin - violation factor @xmath111 and the form factor ( and thus the neutron skin thickness as well as the symmetry energy ) , especially if the ivdm particle has a larger mass of above tens of gev . conversely , our results indicate that precise knowledge on the symmetry energy or the neutron skin thickness ( and thus the neutron and proton form factors ) of @xmath0pb is of critical importance for the direct detection of ivdm using xe - based detectors . in particular , the future experiment prex - ii @xcite at jlab is expected to improve significantly the measurement accuracy of the neutron skin thickness for @xmath0pb and thus could make important contribution to this issue . in the present work , we have shown that isospin - violating dark matter ( ivdm ) indeed provides a possible mechanism to ameliorate the tension among recent direct detection experiments , including cdms - ii(si ) , xenon100 , lux , and supercdms(ge ) . for ivdm , we have demonstrated that the results of the dm direct detection experiments based on neutron - rich target nuclei , e.g. , xe - based detector , may strongly depend on the density slope @xmath23 of the symmetry energy at a subsaturation cross density @xmath38@xmath22 , which is presently largely unknown and uniquely determines the neutron skin thickness and thus the relative difference of neutron and proton form factors of the target nuclei . in particular , using the proton and neutron form factors obtained from skyrme - hartree - fock calculations by varying the @xmath23 within the uncertainty region set by the latest model - independent measurement of the neutron skin thickness from prex experiment at jlab , we have found that although the form factor effects on the extracted bounds on dm - proton cross sections are negligible in the direct detection for isospin - invariant dm , they could become critically important in the detection for ivdm . especially , for ivdm with neutron - to - proton coupling ratio fixed to @xmath237 in the mass region constrained by cmds - ii(si ) , the form factor effect may enhance the sensitivity of xe - based detectors ( e.g. , xenon100 and lux ) to the dm - proton cross section by a factor of @xmath2 , compared with the results using the empirical helm nuclear form factors extracted from charge distributions . this form factor effect can even enhance the sensitivity by more than a factor of @xmath46 for such kind of ivdm with mass larger than @xmath3 gev . furthermore , we have found that the form factor effect can significantly modify the recoil spectrum of xe - based detectors for heavy ivdm with @xmath237 . we have also studied how the form factor effects change with the variation of @xmath238 and found that the @xmath238 value maximumly suppressing the sensitivity of the detector may depend on the form factor , and the form factor effects can either enhance or suppress the sensitivity of the detector relative to the helm s case depending on the specific value adopted for @xmath238 . our results imply that the precise determination of the symmetry energy or the neutron skin thickness ( and thus the neutron and proton form factors ) of @xmath0pb is extremely useful for the direct detection of ivdm based on detectors with neutron - rich targets ( e.g. , xenon ) . the authors would like to thank fei gao for useful discussions . this work was supported in part by the nnsf of china under grant nos . 11275125 and 11135011 , the shanghai rising - star program under grant no . 11qh1401100 , the `` shu guang '' project supported by shanghai municipal education commission and shanghai education development foundation , the program for professor of special appointment ( eastern scholar ) at shanghai institutions of higher learning , and the science and technology commission of shanghai municipality ( 11dz2260700 ) . for si ( upper panels ) , ge ( middle panels ) and xe ( lower panels ) with @xmath165 ( left panels ) and @xmath166 ( right panels ) obtained from shf with lc40 , lc92 and lc134 ( thick lines ) and from the empirical parametrization for proton and neutron form factors in eq . ( [ fitting ] ) ( thin lines ) with @xmath239 , @xmath240 and @xmath241 . the empirical helm form factor is also included for comparison.,width=12 ] shown in fig . [ fnp ] are the proton and neutron form factors as functions of momentum transfer for @xmath152si and @xmath151xe from shf calculations with lc40 , lc92 and lc134 . the results for other isotopes of si and xe as well as ge isotopes can also be calculated in the same approach . thanks to the approximately linear relationship between the symmetry energy density slope parameter @xmath20 and the neutron skin thickness @xmath242 of @xmath0pb , it is possible to approximate the proton and neutron form factors obtained from shf calculations with varied @xmath20 by analytical parameterizations with explicit dependence on the @xmath242 of @xmath0pb . in particular , here we construct the analytical expressions for the proton and neutron form factors separately according to helm - like form factor with an additional term to consider isospin dependence and the symmetry energy effects ( and thus @xmath242 of @xmath0pb ) , i.e. , @xmath243 where the effective proton ( neutron ) distribution radius is parameterized as @xmath244 with @xmath245 compared with the parametrization of helm form factor ( i.e. , eq . ( [ helm ] ) ) , the parametrization here includes an additional term , namely , the last term in eq . ( [ cp ] ) and eq . ( [ cn ] ) . in addition , our parametrization here distinguishes protons and neutrons , and depends on explicitly the proton and neutron numbers of the nucleus and the @xmath242 of @xmath0pb . by fitting a number of proton and neutron form factors calculated from shf approach with varied symmetry energies , including the 9 stable xe isotopes , 5 stable ge isotopes and 3 stable si isotopes with the symmetry energy density slope parameter @xmath20 varying from @xmath246 to @xmath247 , we obtain all parameters in the above expressions and they are listed in table [ table ] . in order to test the quality of the above parametrization , we compare in fig . [ ffs02 ] the averaged effective form factors as functions of @xmath63 for si , ge and xe using the above parameterized proton and neutron form factors with the results obtained from the shf approach . it is very interesting to see that one can indeed simply use just one set of parameters given in table [ table ] to well describe the proton and neutron form factors in different nuclei in a large mass region with very different values of the symmetry energy density slope parameter @xmath20 ( i.e. , @xmath242 of @xmath0pb ) . * cogent * collaboration , c. e. aalseth et al . , _ results from a search for light - mass dark matter with a p - type point contact germanium detector _ , _ phys * 106 * ( 2011 ) 131301 [ arxiv:1002.4703 ] . c. savage , g. gelmini , p. gondolo , and k. freese , _ xenon10/100 dark matter constraints in comparison with cogent and dama : examining the @xmath248 dependence _ , _ phys . _ * d 83 * ( 2011 ) 055002 [ arxiv:1006.0972 ] . g. duda , a. kemper , and p. gondolo , _ model independent form factors for spin independent neutralino - nucleon scattering from elastic electron scattering data _ , _ jcap _ * 0704 * ( 2007 ) 012 [ hep - ph/0608035 ] . chen , y.z . luo , l. li , h. shen , and x.q . li , _ determining nuclear form factor for detection of dark matter in relativistic mean field theory _ , _ commun . * 55 * ( 2011 ) 1059 [ arxiv:1101.3049 ] . a. ong , j. c. berengut , and v. v. flambaum , _ the effect of spin - orbit nuclear charge density corrections due to the anomalous magnetic moment on halonuclei _ , _ phys . _ * c 82 * ( 2010 ) 014320 [ arxiv:1006.5508 ] . brown , _ neutron radii in nuclei and the neutron equation of state _ , _ phys * 85 * ( 2000 ) 5296 ; s. typel and b.a . brown , _ neutron radii and the neutron equation of state in relativistic models _ , _ phys . rev . _ * c 64 * ( 2001 ) 027302 . z. zhang and l.w . chen , _ constraining the symmetry energy at subsaturation densities using isotope binding energy difference and neutron skin thickness _ , _ phys . * b 726 * ( 2012 ) 234 [ arxiv:1302.5327 ] . p. klpfel , p .- reinhard , t.j . brvenich , and j.a . maruhn , _ variations on a theme by skyrme : a systematic study of adjustments of model parameters _ , _ phys . _ * c 79 * ( 2009 ) 034310 [ arxiv:0804.3385 ] . f. le blanc , l. cabaret , e. cottereau , j.e . crawford , s. essabaa , j. genevey , r. horn , g. huber , j. lassen , j.k.p . lee , g. le scornet , j. lettry , j. obert , j. oms , a. ouchrif , j. pinard , h. ravn , b. roussiere , j. sauvage , and d. verney , _ charge - radius change and nuclear moments in the heavy tin isotopes from laser spectroscopy : charge radius of * * 132sn _ , _ phys . rev . _ * c 72 * ( 2005 ) 034305 . c. strege et al . updated global fits of the cmssm including the latest lhc susy and higgs searches and xenon100 data _ , _ jcap _ * 1203 * ( 2012 ) 030 [ arxiv:1112.4192 ] ; a. fowlie , m. kazana , k. kowalska , s. munir , l. roszkowski , e. sessolo , s. trojanowski , and y .- l . tsai , _ the cmssm favoring new territories : the impact of new lhc limits and a 125 gev higgs _ , _ phys . rev . _ * d 86 * ( 2012 ) 075010 [ arxiv:1206.0264 ] ; o. buchmueller et al . , _ higgs and supersymmetry _ , _ eur . j. _ * c 72 * ( 2012 ) 2020 [ arxiv:1112.3564 ] .
blind surveys for optical transients have revealed a class of super - luminous supernovae ( sl ) sne that may in some cases be powered by a collision between the sn ejecta and a massive shell or wind of circumstellar material ( csm ) ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? examples include sn 2003ma @xcite , 2006gy ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and 2008es @xcite , among others . as a consequence of the collision with the csm , a significant fraction of the kinetic energy is converted into radiation via shock dissipation , which is responsible for the observed emission ( see figure 1 ) . the rate of slsne with absolute magnitude @xmath1 is order of @xmath2 , @xmath3% of the normal core - collapse sn rate @xcite , but some normal - luminosity sne such as sn 2005ip @xcite , 2006jc @xcite , 2008iy @xcite and ptf 09uj @xcite , which may also be powered by ejecta - csm interactions , are more common @xcite . finally , recent observations of sn 2009ip and 2010mc suggest that the csm eruption is timed to occur months - to - years before the core collapse @xcite . interaction - powered sne may be efficient cosmic - ray ( cr ) accelerators , where one can expect that the diffusive shock acceleration mechanism operates at the forward and reverse shocks by a collision between the sn ejecta and csm . for a range of csm parameters ( mass and shock dissipation radius ) , the shock is radiation - mediated the thomson optical depth is larger than @xmath4 @xcite , where @xmath5 is the shock velocity and efficient cr acceleration is not expected because the cr collisionless mean free path is much shorter than the deceleration length . however , as the shock propagates in the csm , photons can stream out ahead of the shock , and photon energy can no longer support the shock ( i.e. , shock breakout ) . after the breakout , for wind - like csm profiles , the shock will become collisonless and cr acceleration can be efficient @xcite . recently , murase et al . ( 2011 ) considered a collision between the sn ejecta with a csm shell and found that cr protons may be accelerated , and furthermore that the protons may experience strong pionic losses via inelastic @xmath0 collisions , producing gamma - rays and neutrinos . thus , interaction - powered sne may be interesting cr accelerators and high - energy / multi - messenger emitters . in this work , we continue our study of the possibility of the non - thermal emission from the shock interaction of a sn embedded in a dense csm . in particular , we focus on the secondary electrons and positrons expected from the same @xmath0 collisions that give rise to neutrinos and gamma rays . importantly , we show that these secondaries can emit detectable synchrotron radiation at high - frequency radio wavelengths including mm / submm and fir bands . in section 2 , we review the shock physics and the potential for cr acceleration in interaction - powered sne , providing a much more detailed discussion than murase et al . section 3 gives a brief discussion of the high - energy emission expected , and recipes that connect the observed optical emission to the non - thermal signatures are provided in appendix a. in section 4 we discuss high - frequency radio diagnostics . for a range of csm parameters , we show that secondary leptons from @xmath0 interactions should radiate synchrotron at @xmath6 , and with fluxes of @xmath7 mjy at distances of hundreds of mpc . in section 5 , we summarize our results . throughout this work , we use the notation @xmath8 in cgs unit unless we give notice . in this preparatory section , before we discuss non - thermal signatures , we explain the picture of interaction - powered sne and describe the basic physical setup . let us consider sn ejecta with the kinetic energy @xmath9 and the velocity @xmath10 . noting @xmath11 for the ejecta mass @xmath12 , the momentum and energy conservation laws give @xmath13 where @xmath14 is the total csm mass and @xmath15 is the csm velocity . the total dissipated energy @xmath16 is written as @xmath17 where @xmath18 is used in the last equality . the above equation suggests that a significant fraction of @xmath9 can be dissipated if the csm mass is large ( see also , e.g. , * ? ? ? * ; * ? ? ? density profiles of both the ejecta and csm are important for detailed predictions . for example , when the density profile of the ejecta is steep enough and most of its energy is carried by lower - velocity ejecta material , the explosion has driven waves that can be described by chevalier - nadezhin self - similar solutions @xcite . when the shock wave sweeps up ambient mass comparable to @xmath12 and it is non - radiative , we expect blast waves that can be described by sedov - taylor - like self - similar solutions ( see * ? ? ? * and references therein ) . in this work , to push the basic idea and avoid uncertainty in the ejecta profile and many other complications due to radiation processes , we discuss non - thermal properties without relying on such details . our treatment still provides an order of magnitude estimate of expected non - thermal signals , and a more detailed study will be presented in an accompanying paper @xcite . hereafter , we assume that the csm has a wind - like power - law density profile and extends to the edge radius of the wind , @xmath19 . we expect that this is reasonable ( see , e.g. , * ? ? ? * ) , although details are uncertain due to poor understandings of the csm eruption mechanism . then , the csm density is written as @xmath20 where @xmath21 should be expressed in cm , @xmath22 cm , and @xmath23 is defined that is different from ours . ] for the mass - loss rate of @xmath24 and the wind velocity of @xmath25 . this can also be expressed by @xmath26 the csm mass within @xmath21 is estimated to be @xmath27 where @xmath28 is the csm inner edge radius . in particular , in the wind case ( @xmath29 ) , we have @xmath30 where we have used @xmath31 and @xmath32 . note that , in the one - zone model where the calculation is performed for a csm density @xmath33 at a given radius @xmath21 , qualitative pictures for different density profiles are simply obtained by using @xmath14 instead of @xmath34 ( although the dynamics and temporal evolution depend on density profiles ) . the deceleration is significant after the ejecta accumulates the csm mass equivalent to its own mass , whose radius is characterized by @xmath35 if @xmath36 , most of the ejecta energy is dissipated by the ejecta - csm collision . one of the important quantities is the thomson optical depth . using the csm electron density , @xmath37 the thomson optical depth is estimated to be @xmath38 for @xmath39 , where @xmath40 is the thomson cross section . the thomson optical depth in the downstream is also @xmath41 although the density in the thin , interacting shell is compressed by the shock compression ratio . as seen below , the emission is mostly observed when @xmath42 after photons can leave the system . however , while the interaction with a dense csm shell happens at @xmath43 , hard x rays and soft gamma rays produced at the shock can not avoid compton down - scattering and a significant fraction of the emission would be thermalized @xcite . equal lines of various optical depths , csm density and luminosities in the ( @xmath21 , @xmath44 ) plane and ( @xmath21 , @xmath14 ) plane are shown in figures 2 and 3 , respectively . the equal csm density line , @xmath45 , is also overlaid . explanations for the optical depths other than the thomson optical depth are given below . as argued by murase et al . ( 2011 ) , efficient cr acceleration is possible at @xmath42 , and for @xmath46 we expect almost all the accelerated cr ions to produce neutrinos , hadronic gamma rays , and secondary electrons and positrons . in the system , hadronic gamma rays can interact with photons via the two - photon annihilation process and/or matter via the bethe - heitler ( bh ) pair - production process , respectively ( see section 3 ) . the attenuation of gev gamma rays due to the bh process is insignificant at @xmath47 , which is not far from @xmath48 . in this work , we consider the forward shock , so the shock velocity @xmath5 is regarded as the forward shock velocity @xmath49 . the reverse shock power is smaller when the ejecta profile is steep ( e.g. , * ? ? ? but this might not be the case if the profile is changed , e.g. , possibly by experiencing many interactions with many csm shells . , @xmath44 ) plane . see the text for meanings of @xmath50 , @xmath51 , @xmath52 and @xmath53 . the shaded region suggests the range where we do not expect either production or escape of gamma rays . the thick curves represent @xmath54 while the thin curves do @xmath55 . with quadruplicate - dotted cures , lines of the constant csm density ( @xmath33=const . ) , constant post - breakout radiation luminosity ( @xmath56=const . ) , constant optically - thin free - free luminosity ( @xmath57=const . ) , and constant csm mass ( @xmath14=const . ) are also shown for comparison . ] , @xmath14 ) plane , where @xmath14 has the solar - mass unit . lines of the constant csm density ( @xmath33=const . ) , constant post - breakout radiation luminosity ( @xmath56=const . ) , and constant optically - thin free - free luminosity ( @xmath57=const . ) are also shown for comparison . for explanations of specific sn examples , see sub - section 2.6 . ] when a collision with csm occurs at @xmath58 , photons should experience many compton scatterings , and it takes time for them to leave the system . the photon diffusion time is roughly @xmath59 , which can further be approximated to be @xmath60 for the wind profile if @xmath61 ( c.f . * ; * ? ? ? photons can not essentially diffuse out from the system , when @xmath62 is longer than the dynamical timescale is @xmath63 . hence , when the collision begins at @xmath64 , we start to observe a significant fraction of the emissions in the rise time @xmath65 such that @xmath66 ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? we define the breakout radius @xmath67 , where photons can essentially leave the system . when the effective diffusion radius @xmath68 is sufficiently smaller than @xmath19 ( for @xmath69 ) , we have @xmath70 , and the shock breakout radius can be written as @xmath71 . when the csm is so dense that @xmath68 is larger than @xmath19 , the rise time can be smaller than @xmath72 because @xmath73 at the breakout @xcite . we expect that thermal radiation carries a significant fraction ( @xmath74 ) of the energy carried by the interacting shell ( @xmath75 ) , where the radiated energy , @xmath76 is obtained in the mini - shell model @xcite . see also ofek et al . ( 2014 ) . ] is @xmath77 and @xmath75 after the collision is roughly comparable to @xmath16 ( see equation 3 ) . noting @xmath78 , the ( bolometric ) radiation luminosity just after the breakout is @xmath79 the constant radiation luminosity line , @xmath80 ( with @xmath81 ) , is depicted in figures 2 and 3 . a more sophisticated model is given by chevalier & irwin ( 2011 ) , which is summarized in appendix a and @xmath82 @xcite ] . the collision between the sn ejecta and the csm may start from the optically - thick regime . then , after the shock breakout , the interaction eventually enters the optically - thin regime . this regime typically comes after the time of @xmath83 when @xmath19 is large enough . hence , for optically slsne such as sn 2006gy , it usually happens only after the shock crosses @xmath84 . in order to expect optically - thin ejecta - csm interactions within @xmath19 , relatively large @xmath85 and/or low @xmath14 are needed . alternately , the collision may occur at @xmath42 if csm effectively has an inner edge and can be regarded as a shell . especially for optically - thin interactions at @xmath86 , hard x rays easily leave the system although ultraviolet photons and soft x rays may be attenuated due to bound - free absorption . indeed , such x - ray and radio emissions have been observed in some sne like sn 1988z @xcite and 2006jd @xcite . let us consider a csm extending to @xmath19 . for @xmath87 , the characteristic duration of emission is expected to be @xmath88 ( or @xmath89 ) . when the dominant loss process for thermal electrons is the free - free emission , the ( bolometric ) radiation luminosity mainly comes from bremsstrahlung emission . in the non - radiative case , we have @xcite @xmath90 where @xmath91 is the gaunt factor , @xmath92 for hydrogen is used , and absorption is not considered yet . here the compression factor @xmath93 is taken to be 4 but can be larger . the constant radiation luminosity line , @xmath94 , is depicted in figures 2 and 3 . when the shock is radiative , @xmath95 should be limited by the kinetic luminosity as in equation ( 11 ) . it roughly occurs when the cooling time @xmath96 is shorter than time @xmath97 . note that @xmath57 is proportional to @xmath98 and @xmath99 k is assumed , otherwise cooling by line emissions becomes relevant ( e.g. , * ? ? ? * ) . whether efficient cr acceleration occurs or not depends on the shock properties . when an ejecta - csm collision occurs at sufficiently small radii , the shock is initially radiation - mediated in the sense that the upstream shock structure is modified by radiation from the downstream . for non - relativistic shocks where effects of pairs are irrelevant , the shock is radiation - mediated when @xmath64 @xcite . when coupling with radiation is strong enough , electrons transfer their energy to photons . in the thermal equilibrium limit , at the far downstream , photons , electrons and protons have the temperature of @xmath100 where @xmath101 is used and @xmath102 is the gray - body factor . note that this emission does not have to be the observed emission since photons start to escape only after @xmath103 . thermal equilibrium may not be realized if sufficient photons are not produced by the bremsstrahlung process @xcite . at the nearer downstream , protons and electrons have much higher temperatures . in the absence of collisionless - plasma processes , the electron temperature is determined by the balance between coulomb heating and cooling processes . when the relevant cooling process is the compton cooling , we have @xcite @xmath104 over the length scale of @xmath105 ( where @xmath106 is the ion - electron collision frequency ) , where is defined for the downstream energy density . on the other hand , as in murase et al . ( 2011 ) , @xmath107 is defined for the total energy while @xmath108 is defined for the ram pressure of the upstream flow . ] @xmath109 , where @xmath110 is the energy density of photons . note that the above equation is valid for sufficiently high - velocity shocks , since @xmath111 is limited by the proton temperature @xmath112 @xcite . in reality , collisionless - plasma processes can be faster than coulomb collisional processes , where @xmath111 can be higher than in equation ( 14 ) , but should be lower than the equipartition temperature . after @xmath42 , the shock is no longer radiation - mediated , and we expect collisionless ( or collisional ) shocks . for strong , non - relativistic shocks , the proton temperature at the immediate downstream is @xmath113 where @xmath114 is the adiabatic index . when the adiabatic index is @xmath115 , we have a well - known result , @xmath116 . the electron temperature is affected by energy transfer from protons , which may be coulomb heating or faster collisionless plasma processes . if @xmath115 and when electrons and protons achieve the equipartition , we have ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) @xmath117 therefore , in the interaction - powered sn scenario , we can naturally expect x rays via bremsstrahlung or inverse - compton ( ic ) processes since electrons should be heated by the shock ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? also , we may assume the material is highly ionized at least in the immediate upstream . detections of x rays allow us to probe the existence of strong shocks , supporting the scenario @xcite . however , there are several complications . first , free - free emission may be suppressed if thermal electrons mainly cool via the ic process . secondly , when the ejecta - csm interaction occurs at @xmath118 , hard x rays lose their energies in both the emission zone ( downstream ) and screen zone ( upstream ) , and softer x rays are down - graded via bound - free absorption . if the amount of non - ionized atoms similarly exists in the far upstream , the bound - free optical depth for soft x rays is roughly estimated to be ( e.g. , * ? ? * ; * ? ? ? * ) @xmath119 at @xmath120 kev . naively , the x - ray luminosity is then ( e.g. , * ? ? ? * ; * ? ? ? * ) @xmath121 where @xmath122 is the suppression by other losses , @xmath123 is the optical depth in the emission zone and @xmath124 is the optical depth in the screen zone . predictions for both the thermal and non - thermal x rays depend on details including the frequency - dependent opacity and the ionization in the upstream ( see * ? ? ? * ; * ? ? ? * and references therein ) . this work does not study x - ray emissions in detail , since non - thermal x rays can be largely contaminated or masked by thermal x rays . now , we consider particle acceleration . cr acceleration may become efficient when the shock is no longer radiation - mediated and becomes collisionless , which can be realized when @xmath42 @xcite . strong small - scale magnetic fields are expected as a result from plasma instabilities , and mhd mechanisms such as the turbulent dynamo can also play crucial roles especially in the downstream ( e.g. , * ? ? ? * ; * ? ? ? in addition , crs themselves excite turbulence via cr stream instabilities , which can be important in the upstream ( see * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) . since details of these processes are uncertain , for simplicity , we parameterize the magnetic field with the ratio of @xmath125 to @xmath126 as @xmath127 . then , we obtain @xmath128 in the bohm limit , the proton acceleration time scale is estimated to be ( see a review * ? ? ? * ) @xmath129 and the proton maximum energy is estimated by comparing it to various cooling time scales ( see below ) . in the fully - ionized plasma ( that is the case in the vicinity of the shock ) , the coulomb cooling time of thermal protons @xcite , @xmath130 where @xmath131 is assumed but the velocity of injected protons may be higher , depending on details of injection processes . the ratio between the two is @xmath132 when the upstream is cold enough . hence , the coulomb energy loss timescale is longer than their acceleration time , so cr proton acceleration is possible . note that the coulomb cooling is much less relevant in the immediate downstream , where the temperature is much higher . as protons are accelerated above the pion production threshold , @xmath133 , inelastic @xmath0 interactions occur , leading to production of electrons and positrons via @xmath134 . their minimum injection lorentz factor is @xcite @xmath135 the important point here is that the minimum injection lorentz factor is unique for hadronic injections , which is different from the case of primary electron acceleration . primary electrons can also be shock accelerated via the fermi acceleration mechanism . however , the larmor radius of thermal electrons is smaller than that of thermal protons . thus , electrons need to be energized via some plasma processes to cross the shock length and to get injected to the fermi acceleration process . in other words , the larmor radius of relativistic electrons , @xmath136 , should be larger than that of thermal protons , @xmath137 . the lorentz factor of elections that can be accelerated by the conventional shock acceleration mechanism satisfies @xmath138 keeping this in mind and introducing the energy fraction ( @xmath139 ) and number fraction ( @xmath140 ) of relativistic electrons distributed with a power law , the injection lorentz factor of primary shock - accelerated electrons ( @xmath141 ) is expressed as @xmath142 where @xmath143 for @xmath144 and @xmath145 for @xmath146 . here , @xmath147 is the injection spectral index of accelerated electrons and @xmath148 is the maximum lorentz factor of accelerated electrons . note that equation ( 24 ) is obtained from @xmath149 and @xmath150 . the values of @xmath139 and @xmath140 are uncertain . the cr spectra observed at the earth imply @xmath151 , and such values are inferred from modeling of radio emission from type iib sne ( e.g. , * ? ? ? smaller values of @xmath152 are obtained in the leptonic scenario for young sn remnants ( e.g. , * ? ? ? sufficiently large values of @xmath140 imply @xmath153 , where other electron acceleration processes in the shock transition layer should be relevant , and the spectrum for @xmath154 may be steeper than that for @xmath155 . the energy carried by accelerated crs can be summarized as follows . the cr acceleration is efficient only after the radiation escapes from the system and the strong shock jump is formed by collisionless shocks . hence , if @xmath156<r_{\rm bo}$ ] , we do not expect many crs . normal sne correspond to @xmath157 , so only a fraction of the sn explosion energy @xmath9 is converted to crs at the time the shock reaches @xmath19 . if the csm is massive and @xmath158 ( thus @xmath159 ) , we expect @xmath160 , so a significant fraction of the sn explosion energy can be converted into the energy of crs . the energy of accelerated cr protons in interaction - powered sne ( @xmath161 ) is roughly estimated to be @xmath162<r_{\rm bo}$ ) } \end{array } \right.\ ] ] here equation ( 3 ) is used . note that , for @xmath163 ( i.e. , @xmath164 ) , only the fraction of the sn explosion energy can be dissipated by one collision . also , @xmath165 is the energy fraction carried by crs above @xmath166 , and @xmath167 is typically used in the context of sn remnants . when particles are accelerated up to very high energies , high - energy gamma rays and neutrinos should be accompanied via hadronuclear interactions like the @xmath0 reaction . in particular , cr protons interact with nucleons while they are advected to the downstream , so neutrinos and pionic gamma rays are expected in the gev - pev range @xcite . by comparing the @xmath0 interaction time scale @xmath168 and @xmath169 , we get the @xmath0 optical depth as @xmath170 where @xmath171 is used . from figures 2 and 3 , one sees large parameter space satisfying @xmath42 and @xmath46 , where neutrinos and gamma rays provide promising signals of cr proton acceleration at collisionless shocks . even if @xmath172 , we can say that a significant fraction of the cr energy is converted into hadronic emission via @xmath0 interactions . from equation ( 26 ) , one sees @xmath173 or @xmath174 . thus , as seen from figure 3 , @xmath0 processes are typically efficient for ejecta - csm interactions at radii of @xmath175 cm , for @xmath176 . note that @xmath0 interactions are relevant even in the optically - thin regime of @xmath86 . we here discuss several examples of sne to see if ejecta - csm interactions satisfying @xmath42 and @xmath177 are indeed indicated by recent observations . first , we consider the optically - thick regime around the shock breakout . observationally , the csm nucleon density @xmath33 ( or @xmath34 ) can be estimated from the radiation luminosity @xmath95 ( or the radiated energy @xmath178 ) at the time of the shock breakout , the rise time @xmath65 , and the shock velocity @xmath5 ( or @xmath67 that is the radius at the shock breakout ) . note that the approximation @xmath61 is valid when @xmath179 . 1 . sn 2006gy : sn 2006gy is one of the slsne ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? the radiated energy is @xmath180 and the breakout time is @xmath181 d @xcite . the breakout radius is estimated to be @xmath182 cm , corresponding to @xmath183 . these observational parameters imply @xmath184 ( see appendix a ) , leading to @xmath185 and @xmath186 within @xmath67 . these are roughly consistent with numerical modeling of optical light curves @xcite . 2 . sn 2009ip : sn 2009ip is one of the luminous type iin sne , which showed re - brightening in 2012 . the radiated energy is @xmath187 and the rising time given here , are regarded as an upper limit . ] is @xmath188 d. the breakout radius is estimated to be @xmath189 cm , corresponding to @xmath190 . these observational parameters imply @xmath191 , @xmath192 and @xmath193 within @xmath67 @xcite , which are consistent with observational constraints @xcite . in figure 3 , both of the examples roughly lie on the @xmath50 lines . the condition @xmath194 is satisfied , so neutrinos and gamma rays should be produced in the presence of cr protons . sn 2006gy almost lies on the constant luminosity line of @xmath195 . note that @xmath160 can be expected for sn 2006gy but not for sn 2009ip . next , we consider post - breakout emission in the later phase . for @xmath42 , the shock crossing time @xmath196 or the deceleration time @xmath197 can also be used instead of @xmath65 . then , one can observationally estimate @xmath33 ( or @xmath34 ) from @xmath95 , @xmath198 ( or @xmath199 ) , and @xmath5 . alternatively , if we assume that the csm is not completely ionized , then the x - ray measurements of the bound - free absorption can provide an estimate ( or at least a lower limit ) on @xmath200 . two examples are given below . 1 . sn 2006jd : sn 2006jd was a type iin sn , showing bright x - ray emission with the unabsorbed x - ray luminosity of @xmath201 in the @xmath202 kev range @xcite . radio emission was detected after @xmath203 d. with @xmath54 , we expect @xmath204 at @xmath205 d. then , with the temperature @xmath206 kev , @xmath207 implies @xmath208 and @xmath209 within @xmath21 . this estimated csm density is comparable to @xmath210 obtained with @xmath211 , @xmath212 ( for @xmath22 cm ) , and @xmath213 cm @xcite . on the other hand , x - ray absorption allows us to estimate the column density of non - ionized hydrogen atoms ( @xmath214 ) that may exist in the far upstream . it is also suggested that this density is different from values based on x - ray observations , which may imply that the csm is highly ionized even in the far upstream so there are not many non - ionized atoms in the line of sight of x rays . however , detections of radio emission may imply the absence of too strong free - free absorption in the upstream . 2 . sn 2010jl : sn 2010jl was also a type iin sne , showing bright x - ray emission without radio detections . the unabsorbed x - ray luminosity of @xmath215 is obtained in the @xmath202 kev range @xcite . however , ofek et al . ( 2014 ) suggested a higher intrinsic luminosity and obtained @xmath216 from modeling of the optical emission in the early phase around @xmath217 , which is consistent with a value indicated by x - ray observations . with @xmath218 , we expect @xmath219 at @xmath220 d , implying @xmath221 and @xmath222 at that time . although there is uncertainty , the csm with several solar masses is likely @xcite , and this energetic sn seems one of the promising targets of dedicated searches for neutrinos and gamma rays . based on these parameters , sn 2010jl has @xmath194 , indicating efficient neutrino and gamma - ray production . on the other hand , @xmath223 is suggested for sn 2006jd , but the @xmath0 efficiency is still significant . sn 2006jd also lies on the constant luminosity line of @xmath224 . note that another estimate is possible based on detailed modeling of radio sne ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , although this work does not focus on such more model - dependent studies . most of type iin sne have not been seen by low - frequency radio observations , but we show that some of them may be detectable at high - frequency radio wavelengths including mm / submm and fir bands ( see below ) . in the figure 3 , we indicate the above examples by stars for the purpose of demonstration . note that their parameters have large uncertainty so such plots do not have to be very precise . also , @xmath21 increases as the observation time @xmath97 , so one can ideally draw evolution curves in the ( @xmath21 , @xmath44 ) plane and ( @xmath21 , @xmath14 ) plane . in addition to the four sne , we indicate sn 2006jc and 2008iy , which are also likely to be interaction - powered sne . for sn 2006jc , we use @xmath225 and @xmath226 cm at the x - ray peak time of @xmath227 d based on immler et al . ( 2008 ) . for sn 2008iy , we adopt @xmath228 and @xmath229 cm at the peak time of @xmath230 d from miller at al . ( 2010 ) . sn remnants have been largely believed to be the origin of galactic crs up to the knee of @xmath231 gev ( see a review * ? ? ? * and references therein ) . it would also be interesting to consider interaction - powered sne as potential accelerators of high - energy crs . the maximum energy of accelerated protons , @xmath232 , is determined by comparing @xmath233 to the cooling time @xmath234 and dynamical time @xmath169 . if the time scales of energy losses ( including adiabatic losses ) are long enough , the maximum energy is limited by the dynamical time , @xmath235 then , the maximum energy is @xcite @xmath236 note that the larmor radius ( @xmath237 ) of protons is smaller than the system size ( @xmath21 ) , where the protons are confined . at small @xmath21 and/or large @xmath44 , the maximum energy is limited by energy losses . the @xmath0 cooling time of protons is expressed as @xmath238 where @xmath239 is the proton inelasticity . equating @xmath240 and @xmath241 gives @xcite @xmath242 before crs propagate in a galaxy , they need to escape from the system without significant losses ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? while the ejecta interacts with a csm , cr escape may be limited by the free escape boundary @xmath243 , which could be determined , e.g. , by magnetic field amplification processes or wave damping via ion - neutral collisions @xcite . by comparing the diffusion length @xmath244 ( in the bohm limit ) to @xmath243 , we obtain @xmath245 where @xmath246 is the maximum energy of escaping protons . if @xmath232 is too low for crs to escape within @xmath169 , crs are confined and their escape is non - trivial . if magnetic fields rapidly decay after the shock crossing time ( as often supposed in gamma - ray bursts ) , the condition can be @xmath247 , otherwise it depends on diffusion and adiabatic losses . as a reasonable condition for crs not to be depleted , we here assume @xmath248 and @xmath223 . in figures 4 and 5 , we show the parameter range allowing @xmath249 ev protons . parameter space allowing for higher @xmath246 is narrower . if an ejecta - csm interaction occurs at @xmath46 , crs are largely depleted . we need @xmath250 to expect @xmath251 , and @xmath252 cm is typically favored for escaping crs to avoid significant @xmath0 cooling . higher - energy protons can generate pairs via the bh process , which occurs at @xmath253 , i.e. , @xmath254 . sufficiently high - energy protons may dominantly lose their energies via the bh process , especially in the optically - thick regime . for @xmath118 , the number density of thermalized photons is @xmath255 , where @xmath256 is the luminosity of thermalized ( re - processed ) photons and @xmath257 is their energy fraction . assuming @xmath258 , the bh cooling time @xmath259 becomes @xmath260 where @xmath261 is used at @xmath262 in the proton rest frame @xcite . this is typically longer than the @xmath0 cooling time , so we may mainly consider the @xmath0 reaction . as indicated in figures 4 and 5 , on the other hand , the optically - thin regime of @xmath263 is more likely in cases where protons with @xmath264 gev survive . x - ray photons interact with tev protons and the number density of optically - thin bremsstrahlung photons is @xmath265 , so we have @xmath266 which is negligible compared to @xmath241 . in addition , photomeson production may also occur at @xmath267 . although it seems that the proton energy has to be quite high , some interactions with x - ray photons are possible in the attenuation scale of x rays . finally , we briefly discuss a possible contribution of interaction - powered sne to the observed galactic crs . contributions from various types of sne including `` hypernovae '' erg , which are usually broadline type ibc sne . note that only a fraction of them are trans - relativistic sne that show a mildly relativistic component . ] have been considered ( e.g. , * ? ? ? it is still under debade which astrophysical accelerator is responsible for crs around the knee , although normal sne have been commonly believed as a leading candidate . an issue is how cr protons are accelerated up to the knee , and interaction - powered sne could have some advantages because of higher densities and possible stronger fields . the observed cr energy flux at @xmath268 gev is smaller than that at @xmath269 gev by @xmath270 . the contribution of interaction - powered sne at @xmath271 compared to the contribution of normal sne at @xmath272 is ( e.g. , * ? ? ? * ) @xmath273 where @xmath274 is the conversion factor from the total energy to the energy spectrum ( see appendix a ) , @xmath275 is the rate of interaction - powered sne , @xmath276 is the sn rate , and @xmath277 is the confinement time of galactic crs . although the confinement time is highly uncertain , if @xmath278 and @xmath279 , interaction - powered sne could contribute to the observed cr flux around the knee if the rate of interaction - powered sne is as high as @xmath280 . ev protons in the ( @xmath21 , @xmath44 ) plane . cr acceleration is possible at @xmath42 and @xmath281 and @xmath282 are required to achieve @xmath283 ev in the acceleration region . the shaded region suggests the range where we do not expect either production or escape of @xmath284 ev protons . ] , @xmath14 ) plane , where @xmath14 has the solar - mass unit . ] by taking into account the inelasticity @xmath239 , we obtain the effective @xmath0 optical depth as @xmath285 which gives the meson production efficiency . using @xmath286 , the neutrino energy fluence ( per flavor ) is estimated to be @xmath287e_p^2\frac{dn_p}{de_p}\nonumber\\ & \simeq&1.4\times{10}^{-4}~{\rm erg}~{\rm cm}^{-2}~{\rm min}[1,f_{pp}]{\left(\frac{e_\nu}{0.05m_pc^2}\right)}^{2-q_p}\nonumber\\ & \times&\epsilon_{p,-1}{\mathcal e}_{51}{(d/10~{\rm mpc})}^{-2}{\mathcal r}_{p0,1}^{-1},\end{aligned}\ ] ] where @xmath288 is the power - law index of the proton spectrum and @xmath289 is @xmath290 at @xmath166 ( see appendix a ) . when the csm mass is sufficiently large , assuming @xmath291 at the collision , we expect the typical energy fluence of @xmath292 due to @xmath293 . on the other hand , when the csm mass is not large enough , assuming @xmath291 at the collision , the typical energy fluence is expected to be @xmath294 due to @xmath295 . the typical neutrino energy is @xmath296 , so we expect gev - pev neutrino emission given that @xmath232 reaches @xmath297 pev . the icecube effective area for muon neutrinos is order of @xmath298 in the 100 tev range while @xmath299 in the 1 tev range , implying the fluence sensitivity of @xmath300 @xcite . hence , high - energy neutrinos are detectable by icecube for nearby sne that occur at @xmath301 mpc , and stacking analyses for aggregated signals from a number of interaction - powered sne will also be useful @xcite . in appendix a , we provide recipes that connect the observed optical emission to the neutrino signal . the pionic gamma - ray energy fluence is 2 times larger than the neutrino fluence per flavor ( after neutrino mixing ) , so we expect @xmath302e_p^2\frac{dn_p}{de_p}\nonumber\\ & \simeq&2.8\times{10}^{-4}~{\rm erg}~{\rm cm}^{-2}~{\rm min}[1,f_{pp}]{\left(\frac{e_\gamma}{0.1m_pc^2}\right)}^{2-q_p}\nonumber\\ & \times&\epsilon_{p,-1}{\mathcal e}_{51}{(d/10~{\rm mpc})}^{-2}{\mathcal r}_{p0,1}^{-1},\end{aligned}\ ] ] or using @xmath303 , the typical gamma - ray energy flux is @xmath304e_p^2\frac{dn_p}{de_pdt}\nonumber\\ & \sim&1.1\times{10}^{-11}~{\rm erg}~{\rm cm}^{-2}~{\rm s}^{-1}~{\rm min}[1,f_{pp}]\nonumber\\ & \times&{\left(\frac{e_\gamma}{0.1m_pc^2}\right)}^{2-q_p}\epsilon_{p,-1}d_\nonumber\\ & \times&{(v_s/5000~{\rm km}~{\rm s}^{-1})}^{3}{(d/10~{\rm mpc})}^{-2}{\mathcal r}_{p0,1}^{-1}.\end{aligned}\ ] ] the _ fermi _ 1 yr sensitivity at gev energies is @xmath305 @xcite , so interaction - powered sne may be detectable with _ fermi _ for powerful explosions at @xmath306 mpc @xcite . stacking analyses are useful for gev gamma rays as well as neutrinos ( see appendix a ) . in addition , future ground cherenkov detectors will be more sensitive . for example , the cta 100 hr sensitivity at tev is @xmath307 @xcite , whereas the hawc 3 yr sensitivity at 2 tev is @xmath305 @xcite . if the cr spectrum is as hard as @xmath308 , we expect that detections are possible up to @xmath309 mpc via followup observations by cta within days - to - years . for extragalactic gamma rays , one has to keep in mind attenuation processes . there are two effects : attenuation by the extragalactic background light ( ebl ) and attenuation by target photons in the source . the attenuation by ebl is relevant above @xmath310 tev @xcite , so we can neglect it to discuss the detestability in the gev - tev range . the attenuation in the source can be relevant especially for emission from the reverse shock , but we show below that gamma rays from the shocked csm can typically escape from the source after the shock breakout , @xmath42 . gamma rays around mev energies are downgraded via compton scattering with electrons in the matter . the bh pair - creation process occurs at @xmath311 and becomes dominant over the compton scattering above the gev range , where the compton scattering is reduced by the klein - nishina suppression . at sufficiently high energies , the approximate cross section of the bh process is @xcite @xmath312.\ ] ] then , the bh opacity for gamma rays is expressed as @xmath313 where @xmath314 is evaluated at @xmath315 . the parameter space such that @xmath316 is shown in figures 2 and 3 . interestingly , it is comparable to the parameter space such that @xmath317 for typical shock velocities , since gamma - ray attenuation due to the bh process is irrelevant when @xmath5 is higher than @xmath318 . note that the bh opacity may be even smaller . several observations have suggested that csm is clumpy rather than uniform . the bh opacity becomes irrelevant when the csm is anisotropic or so clumpy that most of the csm mass is concentrated in dense clumps . for example , if the density enhancement in the clumps is @xmath319 , the bh opacity is changed by @xmath320 , where the attenuation by the bh process becomes even less relevant . gamma rays interact with photons via @xmath321 . the interaction typically happens at @xmath322 , and the @xmath323 opacity for sufficiently high - energy gamma rays is @xmath324 where @xmath325 is the photon number density at @xmath326 . when the collision with csm occurs at @xmath43 , most of the emissions are thermalized . using @xmath327 , we obtain @xmath328 at @xmath329 . above this energy , interactions mainly happen in the klein - nishina regime , so the @xmath330 opacity decreases as @xmath331e_\gamma^{-1}$ ] . hard x rays , which can be produced by bremsstrahlung emission in the downstream , could potentially prohibit lower - energy gamma rays from leaving the emission region because the pair - creation threshold energy is lower for target photons with higher energies . using @xmath322 and @xmath332 , the typical energy of gamma rays interacting with x - ray photons is estimated to be @xmath333 as explained below we expect that gev gamma rays can leave the system without significant attenuation . first , we consider an optically - thin collision at @xmath86 with @xmath334 . the black - body approximation is invalid , and hard x - ray emission becomes largely visible . the number density of optically - thin bremsstrahlung photons is @xmath265 , so we obtain @xmath335 so gamma rays will be able to leave the emission region for ejecta - csm interactions at @xmath86 . the radiation luminosity may be limited by the kinetic luminosity , and interactions with x rays can be relevant in the length scale of @xmath336 even in the optically - thick regime of @xmath337 . then , we may roughly expect @xmath338 and @xmath339 at @xmath67 . therefore , given @xmath74 , gev gamma rays would be able to escape from the system . note that the bh attenuation is also avoidable for sufficiently high shock velocities . electrons and positrons ( for which we simply say electrons ) can generate synchrotron emission . the synchrotron cooling time of relativistic electrons is @xmath340 on the other hand , the ic cooling timescale is @xmath341 where @xmath342 is the compton @xmath343 parameter , @xmath344 is the synchrotron self - compton ( ssc ) @xmath343 parameter , and @xmath345 is the external inverse - compton ( eic ) @xmath343 parameter . the ic emission is bolometrically more important than the synchrotron emission if @xmath346 . we expect that the ssc emission is typically weak . this is because the ssc @xmath343 parameter in the thomson regime ( e.g. , ? ? ? * ) , @xmath347 is less than unity . here , @xmath348 $ ] and @xmath349 ( or @xmath350 ) , where @xmath351 ( or @xmath352 ) is the injection lorentz factor and @xmath353 is the cooling lorentz factor defined below . external radiation fields are mainly supplied by sn emission , which is more relevant in our cases . if the system is optically thin , the energy density of thermal photons is @xmath354 whereas the magnetic field energy density is @xmath355 because of @xmath356 , we see that the synchrotron cooling would be typically stronger than the eic cooling , although the situation can be altered depending on parameters such as @xmath357 . as a result , non - thermal x rays can be expected mainly due to eic emission . but they will be weaker than thermal x rays ( except at hard x rays ) , and this work focuses on radio signals . in the dense csm environment , one also has to care about other losses such as bremsstrahlung and coulomb losses . the relativistic bremsstrahlung cooling time scale is @xcite @xmath358 which is longer than @xmath169 , @xmath359 and @xmath360 for our typical parameters . the coulomb loss time scale of relativistic electrons is @xcite @xmath361)}\nonumber\\ & \simeq&2.2\times{10}^8~{\rm s}~\gamma_{e,2}\mu_ed_{,-1}^{-1}r_{16}^{2 } , \end{aligned}\ ] ] which suggests that sufficiently high - energy electrons typically cool via the synchrotron emission . at lower energies , however , the coulomb loss can be the shortest time scale , and the resulting synchrotron spectrum becomes complicated . the cooling lorentz factor is given by equating @xmath362 as @xmath363 if @xmath364 , the system is in the fast cooling regime , so the energy flux has a peak at @xmath365 . if @xmath366 , the system is in the slow cooling regime , so the energy flux has a peak at @xmath367 . primary protons with @xmath133 lead to inelastic @xmath0 reactions , providing secondary electrons . the characteristic frequency of electrons with @xmath352 is @xmath368 for our typical parameters , @xmath369 and @xmath370 cm , we expect @xmath371 ghz . we here point that these secondary electrons can be more relevant in the interaction - powered sn scenario . very naively , the secondary electronic emission dominates over the primary electronic emission when @xmath372\epsilon_p>\epsilon_e,\ ] ] which is likely to be realized if @xmath286 is as high as @xmath373 . one should keep in mind that this is the crude argument applied to the bolometric emission , and the relative importance changes with frequency , depending on spectral energy distributions of cr protons and electrons . assuming the fast cooling regime , the resulting synchrotron flux from hadronically - injected electrons ( at @xmath374 ) is @xmath375e_p^2\frac{dn_p}{de_pdt}\nonumber\\ & \sim&2.7\times{10}^{-14}~{\rm erg}~{\rm cm}^{-2}~{\rm s}^{-1}\nonumber\\ & \times&{\rm min}[10,f_{pp,-1}]{\left(\frac{\nu}{\nu_h}\right)}^{2-q_p}\epsilon_{p,-1}\nonumber\\ & \times&d_{,-1}{(v_s/5000~{\rm km}~{\rm s}^{-1})}^{3}{\mathcal r}_{p0,1}^{-1}{(d/10~{\rm mpc})}^{-2},\end{aligned}\ ] ] which corresponds to @xmath376 for @xmath308 . hence , the synchrotron signal is detectable with high - frequency radio telescopes when several absorption processes we discuss below are irrelevant . in particular , the alma sensitivity at 100 ghz is @xmath377 , allowing detections up to @xmath378 gpc if followup observations are successful . primary electrons can also emit synchrotron photons , and the corresponding characteristic frequency of electrons with @xmath141 is @xmath379 it suggests that studying radio sne at relatively low frequencies can probe electron acceleration in the shock transition layer , whereas investigations at high frequencies allow us to see the conventional shock acceleration of electrons and/or hadronic injections . the synchrotron energy flux ( at @xmath380 ) in the fast cooling case is @xmath381 where @xmath147 is the electron spectral index and we expect @xmath382 for @xmath383 . note that @xmath147 may naturally be different from values expected in the conventional diffuse shock acceleration theory at @xmath384 , corresponding to electrons with @xmath385 , and steep indices of @xmath386 are indeed indicated in non - relativistic radio sne . the synchrotron signal from primary electrons is also detectable by current radio telescopes and mm / submm facilities including alma if secondary electrons are sub - dominant . for a given @xmath326 , the secondary electronic emission dominates over the primary electronic emission when @xmath387\frac{\epsilon_{p,-1}}{\epsilon_{e,-3}}{\left(\frac{\nu_l}{\nu_h}\right)}^{2-q}\gtrsim0.06,\ ] ] where the fast cooling regime is assumed for @xmath388 . hence , we expect that secondaries are typically more important for high - frequency radio emission from interaction - powered sne . however , detecting radio signals may suffer from scattering and various absorption processes . first , if @xmath43 , the synchrotron emission can be modified by comptonization due to thermal electrons . in particular , thermal electrons in the hot downstream may up - scatter low - energy photons the condition that the comptonization does not change the synchrotron spectrum is roughly given by @xmath389\lesssim1 $ ] , so we focus on ejecta - csm collisions satisfying @xmath390 . in addition , there are three important absorption processes that can hinder observations at the radio band , razin - tsytovich ( rt ) suppression , synchrotron self - absorption ( ssa ) and free - free absorption . we discuss these suppression and absorption processes below . , @xmath44 ) plane , at @xmath391 ghz , where @xmath54 , @xmath392 and @xmath393 and @xmath394 are used . the suppression or absorption of radio emission is insignificant at @xmath395 , @xmath396 , @xmath397 and @xmath398 . the light shaded region indicates the forbidden region for the radio emission in the pessimistic case , where the upstream is assumed to be ionized with @xmath399 k. the dark shaded area indicates the forbidden region in the optimistic case , which may be realized for different upstream properties . the downstream temperature is set to @xmath400 kev . from this figure , one sees that radio emission from interaction - powered sne satisfying @xmath401 is suppressed at this band . ] , @xmath14 ) plane , where @xmath14 has the solar - mass unit . ] ghz . from this figure , one sees that observations at 100 ghz are useful for probing interaction - powered sne satisfying @xmath401 . ] , @xmath14 ) plane , where @xmath14 has the solar - mass unit . ] ghz . ] , @xmath14 ) plane , where @xmath14 has the solar - mass unit . ] synchrotron emission in a plasma is different from that in a vacuum . when a cold plasma plays a role , it is suppressed at low frequencies due to collective effects . the suppression occurs at the rt frequency @xcite , @xmath402 the line of @xmath403 is shown in figures 6 - 11 . instead , given a frequency , we can obtain the upper limit on @xmath34 as @xmath404 the ssa opacity is estimated to be ( see appendix b ) @xmath405 where @xmath406 is the spectral index of electrons ( that is generally different from @xmath407 for injected particles ) , @xmath408 , and @xmath409 here , when the coulomb cooling is irrelevant , @xmath410 $ ] and @xmath411 is evaluated as a function of @xmath406 , which is order of @xmath412 ( see appendix b ) . when only synchrotron and ic ( in the thomson regime ) losses are relevant , we have @xmath413 at @xmath414 in the fast cooling case or @xmath415 when @xmath416 ( i.e. , the slow cooling case ) . here , @xmath417 is related to the number of electrons swept by the shock as @xmath418 where @xmath419 is the number fraction of electrons defined against @xmath420 . note that , for primary electron acceleration , the different parameter @xmath140 satisfies @xmath421 assuming a power - law injection spectrum , we have @xmath422 for @xmath423 and @xmath424 for @xmath425 , and we typically expect @xmath426 . for secondary electrons , we have @xmath427\epsilon_p}{6\gamma_h\ln(\gamma_p^m)}\frac{m_p}{m_e},\ ] ] for @xmath423 and @xmath427\epsilon_p(q-2)}{6\gamma_h(q-1)}\frac{m_p}{m_e},\ ] ] for @xmath425 . the proton spectrum is assumed to be a power law above @xmath428 , where we typically obtain @xmath429 $ ] . introducing @xmath419 allows us to discuss primary and secondary electrons in parallel . setting @xmath430 , the ssa frequency is estimated to be @xmath431 for @xmath413 ( leading to @xmath432 ) , and @xmath433 for @xmath434 ( leading to @xmath435 ) . instead , given a frequency , we can obtain the upper limit on @xmath34 as @xmath436 for @xmath413 , and @xmath437 for @xmath434 . note that we should use @xmath438 and @xmath413 for the fast cooling case . results for @xmath423 and @xmath439 are shown in figures 6 - 11 . in this work , we assume the case of @xmath440 $ ] , although ssa heating is relevant when @xmath441 $ ] @xcite . the free - free absorption is important especially when photons propagate in the ionized plasma . for simplicity , here we assume that ions are protons . in the hot downstream , the free - free opacity for photons with @xmath442 is @xcite @xmath443 where @xmath444 is the free - free absorption coefficient and @xmath91 is the gaunt factor . high temperatures of @xmath445 k are expected in the immediate downstream , while the temperature is lower at the far downstream due to bremsstrahlung cooling especially if the shock is radiative . the free - free absorption frequency is given by @xmath446 , and we have @xmath447 instead , given a frequency , we can obtain the upper limit on @xmath34 as @xmath448 if the free - free absorption in the emission region is dominant , we expect the suppression factor of @xmath449 at @xmath450 . however , one typically expects that upstream material would be more crucial for absorbing low - frequency emission ( but see * ? ? ? the upstream temperature is lower than the immediate downstream temperature , so the free - free optical depth can be much larger . before the shock reaches @xmath19 , assuming ionized material , the free - free optical depth in the upstream is @xmath451 the free - free absorption frequency is given by @xmath452 , and we have @xmath453 instead , given a frequency , we can obtain the upper limit on @xmath34 as @xmath454 note that , if the free - free absorption in the screen region is dominant , we expect the suppression of @xmath455 , which can be in principle distinguished from the other possibilities . as shown in figures 8 and 9 , at @xmath456 ghz , the free - free absorption is typically the most important attenuation process in sne with dense csm . even at @xmath457 ghz , it is the dominant attenuation process for ejecta - csm interactions at @xmath458 cm . this process is sensitive to the electron temperature , and @xmath459 k is used in figures 6 - 11 . the temperature may indeed be high enough , as suggested by successful radio detections of sn 2006jd @xcite . in figure 7 , sn 2006jd lies near the lines of @xmath452 and @xmath430 . on the other hand , non - detections of radio emission from many other type iin sne like sn 2010jl seem consistent with the large absorption that is easily realized with more conservative values of @xmath460 k @xcite . at @xmath457 ghz , as shown in figures 6 and 7 , the free - free absorption and ssa processes suppress radio signals , and the parameter space around @xmath401 is located in the dark shaded area . so it is difficult to see hadronic signatures with synchrotron emission at radio wavelengths . but the situation drastically changes at higher frequencies . at @xmath456 ghz , the free - free absorption is still an obstacle for ejecta - csm interactions at @xmath461 cm , but not for @xmath458 cm . importantly , as shown in figures 8 and 9 , a large parameter space of @xmath401 is free of absorption and scattering processes . so observations at high - frequency radio wavelengths including mm / submm and fir bands are indeed powerful to test the hadronic model and probe cosmic - ray proton acceleration . at higher frequencies such as 1000 ghz , all the absorption processes discussed here are negligible compared to the comptonization due to thermal electrons . not all interaction - powered sne allow us to expect high - frequency radio signals from secondaries . as pointed out by murase et al . ( 2011 ) , except at sufficiently late phases , optically - bright slsne are difficult to detect with synchrotron emission at radio bands . on the other hand , as shown in figures 9 and 11 , some type iin sne such as sn 2006jd and 2008iy seem very promising . there are some possibilities that we have effectively lower values of @xmath462 . first , the csm may be anisotropic or clumpy as suggested in several type iin sne like sn 2005ip @xcite and 2009ip ( e.g. , * ? ? ? * ) , where the emission more easily escape from partial regions where the csm density is much lower . secondly , @xmath462 is smaller when the csm is little ionized , which may be realized especially in the far upstream . this is different from soft x rays that are more strongly absorbed in the non - ionized csm . such more transparent cases correspond to the light shaded area in figures 6 - 11 . note that the free - free optical depth declines after the shock reaches the outer edge of csm , @xmath19 . after @xmath463 , it becomes @xmath464 where @xmath465 ( at @xmath466 ) is assumed . for example , a possible value of @xmath467 is suggested in the late phase of the re - brightening of sn 2009ip @xcite . in this work , we provide a broad discussion of multi - messenger diagnoses of interaction - powered sne including type iin sne and some slsne , focusing on non - thermal signals . the shock would be radiation - mediated at very early times , and thus cr acceleration is inefficient . however , as photons escape the system , a collisionless shock can form and cr acceleration becomes possible . while shock heating leads to x rays , crs are expected to produce broadband non - thermal emission , including gamma rays , x rays , radio waves and neutrinos . photon emission in general may be largely thermalized depending the optical depth , which in turn depends on details of the csm . neutrinos are the most direct probe in the sense that they do not suffer from attenuation in the source . in addition , by advancing the idea proposed by murase et al . ( 2011 ) , we have shown that gev gamma rays can typically escape after the shock breakout , although tev gamma rays are attenuated due to the two - photon annihilation process . along with neutrinos , gev gamma rays can provide unique opportunities to probe the formation of collisionless shocks and the onset of cr acceleration . interestingly , the physical parameters suggested by observed interaction - powered sne imply densities similar to those inferred from gamma - ray novae @xcite , allowing us to expect analogous high - energy emission and to probe the physics of cr acceleration in the dense environment . detecting signals from one sn requires a nearby event , but stacking analyses are still useful . gamma rays and neutrinos are especially powerful for optically - bright slsne , for which the recipes provided in appendix a can be used . for normal luminosity interaction - powered sne , broad - band non - thermal emissions from radio to tev gamma - ray bands are possible . in particular , high - frequency radio observations in the mm / submm and fir bands can probe cr proton acceleration and test the hadronic model . we pointed out that secondary electrons produced via @xmath0 reactions play an important pole role in the synchrotron emission from some interaction - powered sne such as sn 2006jd . comprehensive observations from ghz to 1000 ghz may also be relevant to study acceleration of primary electrons that may not be accelerated by the conventional shock acceleration . our work demonstrates the importance of multi - messenger approaches in revealing the mechanism of type iin sne and cr acceleration in real time . the interaction - powered sn scenario has been commonly used to interpret slsne , but slsne are diverse and other scenarios also possible . for example , some slsne such as sn 2007bi may be rather pair - instability sne originating from progenitors with @xmath468 , where the stellar collapse is caused by the pressure decrease due to electron - positron pair - production @xcite . or luminous sne including slsne ic ( e.g. , * ? ? ? * ) may be driven by newborn pulsars ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? some slsne seem difficult to explain using these scenarios @xcite . high - energy emissions including gamma rays and neutrinos have been predicted in both the interaction - powered sn scenario @xcite and the pulsar - driven sn scenario @xcite . detecting thermal and non - thermal signals from shocks in the dense csm and studying time - dependent spectra are crucial in order to discriminate among the scenarios . the most important point of this work is that secondary electrons and positrons from inelastic @xmath0 collisions will radiate detectable synchrotron photons efficiently at high - frequency radio wavelengths including the mm / submm and fir bands . although details depend on the mass of csm , its physical location relative to the progenitor star at the time of explosion , and the velocity of the ejecta , for typical parameters , we expect the synchrotron spectrum to peak at @xmath469 ghz and with flux of @xmath7 mjy for 2006jd - like interaction - powered sne at distances of hundreds of mpc . in particular , high - frequency radio signals using instruments like the high - frequency channels of the vla and alma can be very powerful to probe physics of collisionless shocks . for this reason , we encourage followup observations especially at the mm / submm band within months - to - years and at the gev - tev gamma - ray band within days - to - years . k. m. thanks john beacom , roger chevalier , boaz katz , brian lacki , keiichi maeda , jose prieto , masaomi tanaka and ryo yamazaki for useful discussions . k. m. is supported by nasa through hubble fellowship , grant no . 51310.01 awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , for nasa , under contract no . nas 5 - 26555 . k. m. also thanks supports by ccapp since this project was started in the beginning of 2011 when k. m. was a member of ccapp . the preliminary results of this work was presented in march 2013 , hubble fellows symposium . here , we provide recipes to test the hadronic model for gamma - ray and neutrino emissions from interaction - powered sne . basically , we need to know two quantities , the cr energy @xmath470 and the @xmath0 efficiency @xmath286 . the latter depends on @xmath33 , @xmath21 and @xmath5 , which can be determined from the breakout emission . the more sophisticated approach is possible based on self - similar solutions in the engine - driven case @xcite . we here overview the prescription by chevalier & irwin ( 2011 ) . let us consider the ejecta whose outer density profile is @xmath471 at @xmath472 . the shock is initially radiation - mediated and the flows are radiation - dominated since @xmath473 . the contact discontinuity is located at @xmath474 , where @xmath475 . the forward shock radius is estimated to be @xmath476 and the forward shock velocity is @xmath477 . the energy carried by the interacting shell is estimated to be @xmath478 , where @xmath479 . for @xmath480 , the radiation energy is @xmath481 , so we have @xmath482 in the wind case , the photon diffusion timescale is comparable to the breakout time , and we have @xmath483 as long as the breakout happens at @xmath484 . the breakout radius can also be estimated from the evolution of the radiation luminosity and temperature if the black - body approximation is valid , or if @xmath49 is known . as a result , if we can observe @xmath178 , @xmath65 and @xmath49 ( or @xmath67 ) , we can evaluate @xmath44 , @xmath485 and @xmath486 as @xcite @xmath487 @xmath488 @xmath489 one should keep in mind that the engine - driven self - similar solution is valid as long as @xmath490 . then , we can estimate @xmath33 , allowing us to calculate @xmath286 and resulting neutrino and gamma - ray spectra . there are two important free parameters , the cr spectral index ( @xmath288 ) and cr energy ( @xmath470 ) . for @xmath491 , the cr proton spectrum is given by @xmath492 and @xmath493 for our typical parameters . for @xmath494 , we have @xmath495 here , @xmath274 is the conversion factor from the total energy to the differential energy spectrum and @xmath496 is the minimum proton energy . these equations can be rewritten as @xmath497 where @xmath498 . the cr energy is parametrized as @xmath499 where @xmath500 is the cr loading parameter that is commonly introduced in the literature of hadronic emissions from gamma - ray bursts @xcite . since both @xmath165 and @xmath501 are order of 0.1 , we expect @xmath502 and we can make predictions for breakout high - energy emissions , based on observational quantities . such an application was done in margutti et al . note that high - energy emissions continue after the breakout . given sufficient time - dependent data , later contributions can easily be taken into account by more detailed modeling . for example , one can directly use self - similar solutions for @xmath480 or @xmath115 at @xmath42 @xcite . most naively , instead , the overall contribution can be incorporated in the cr loading parameter . here we provide formulas to calculate ssa . for a power - law electron distribution , the ssa coefficient is @xcite @xmath503 where @xmath504 is the angle between the electron velocity and magnetic field . the normalization is determined by @xmath505 assuming @xmath506\gg\gamma_n={\rm min}[\gamma_i,\gamma_c]$ ] , we have @xmath507 , where @xmath508 for @xmath509 and @xmath510 for @xmath511 . when only the synchrotron emission is relevant , we have @xmath413 at @xmath414 in the fast cooling case or @xmath415 at @xmath416 in the slow cooling case . averaged over @xmath504 , the ssa coefficient is written as @xmath512 where @xmath513 for example , we obtain @xmath514 for @xmath413 and @xmath515 for @xmath434 , respectively . note that the spectrum in the optically - thick limit is obtained from @xmath516 , where @xmath517 is the synchrotron emissivity @xcite . the coefficient agrees with katz ( 2012 ) . one should keep in mind that the @xmath518 is obtained only if @xmath519 , whereas we expect @xmath520 if @xmath521 . balberg , s. , loeb , a. , 2012 , mnras , 414 , 1715 bell a. r. , 2004 , mnras , 353 , 550 bell , a. r. , 2013 , astropart . phys . , 43 , 56 bykov , a. m. , brandenburg , a. , malkov , m. a. , osipov , s. m. , 2013 , space sci . , 178 , 201 budnik , r. , katz , b. , macfadyen , a. , waxman , e. , 2008 , apj , 673 , 928 caprioli , d. , blasi , p. , amato , e. , 2009 , mnras , 396 , 2065 the cta consortium , 2011 , experimental astronomy , 32 , 193 chandra , p. , chevalier , r. a. , chugai , n. , fransson , c. , irwin , c. m. , soderberg , a. m. , chakraborti , s. , immler , s. , 2012a , apj , 755 , 110 chandra , p. , chevalier , r. a. , irwin , c. m. , chugai , n. , fransson , c. , soderberg , a. m. , 2012b , apj , 750 , l2 chevalier , r. a. , 1982a , apj , 258 , 790 chevalier , r. a. , 1982b , apj , 259 , 302 chevalier , r. a. , 1984 , apj , 285 , l63 chevalier , r. a. , 1998 , apj , 499 , 810 chevalier r. a. , fransson c. , 2003 , in weiler k. , ed . , lecture notes in physics vol . 598 , supernovae and gamma - ray bursters . springer - verlag , berlin , p. 171 chevalier , r. a. , irwin , c. m. , 2011 , apj , 729 , l6 chevalier , r. a. , irwin , c. m. , 2012 , apj , 747 , l17 chodorowski , m. j. , zdziarski , a. a. , sikora , m. , 1992 , apj , 400 , 181 chornock , r. et al . , 2013 , apj , 767 , 162 chugai , n. n. , danziger , i. j. , 1994 , mnras , 268 , 173 falk , s. w. and arnett , w. d. , 1973 , apj , 180 , l65 fransson , c. , lundqvist , p. , chevalier , r. a. , 1996 , apj , 461 , 993 fransson , c. et al . , 2013 , arxiv:1312.6617 funk , s. , hinton , j. a. , for the cta consortium , 2013 , astropart . , 43 , 348 immler , s. et al . , 2008 , apj , 674 , l85 inoue , t. , yamazaki , r. , inutsuka , s. , 2009 , apj , 695 , 825 inoue , t. , yamazaki , r. , inutsuka , s. , fukui , y. , 2012 , apj , 744 , 71 inserra , c. et al . , 2013 , apj , 770 , 128 kasen d. , bildsten l. , 2010 , apj , 717 , 245 kashiyama , k. , murase , k. , horiuchi , s. , gao , s. , mszros , p. , 2013 , apj , 769 , l6 katz , b. , 2012 , mnras , 420 , l6 katz , b. , budnik , r. , waxman , e. , 2010 , apj , 716 , 781 katz , b. , sapir , n. , waxman , e. , 2011 , arxiv:1106.1898 katz , b. , waxman , e. , 2008 , jcap , 01 , 018 kotera , k. , phinney , e. s. , olinto , a. v. , 2013 , mnras , 432 , 3228 kulsrud , r. m. , cesarsky , c. j. , 1971 , al , 8 , 189 maeda , k. , 2012 , apj 758 , 81 mauerhan , j. c. et al . , 2013 , mnras , 430 , 1801 metzger , b. d. , giannios , d. , thompson , t. a. , bucciantini , n. , quataert , e. , 2011 , mnras , 413 , 2031 miller , a. a. et al . , 2009 , apj , 690 , 1303 miller , a. a. et al . , 2010 , mnras , 404 , 305 margutti , r. et al . , 2014 , apj , 780 , 21 moriya , t. j. , blinnikov , s. i. , tominaga , n. , yoshida , n. , tanaka , m. , maeda , k. , nomoto , k. , 2013a , mnras , 428 , 1020 moriya , t. j. , maeda , k. , taddia , f. , sollerman , j. , blinnikov , s. i. , sorokina , e. i. , 2013b , mnras , 435 , 1520 murase , k. , mszros , p. , zhang , b. , 2009 , phys . d , 79 , 103001 murase , k. , nagataki , s. , 2006 , phys . d , 73 , 063002 murase , k. , thompson , t. a. , lacki , b. c. , beacom , j. f. , 2011 , phys . d , 84 , 043003 murase , k. et al . , 2014 , in preparation ofek , e. o. , lin , l. , kouveliotou , c. , younes , g. , ggs , e. , kasliwal , m. m. , cao , y. t. , 2013a , apj , 768 , 47 ofek , e. o. et al . , 2007 , apj , 659 , l13 ofek , e. o. et al . , 2013b , apj , 763 , 42 ofek , e. o. et al . , 2010 , apj , 724 , 1396 ofek , e. o. et al . , 2013c , nature ( london ) , 494 , 65 ofek , e. o. et al . , 2014 , apj , 781 , 42 ohira , y. , murase , k. , yamazaki , r. , 2010 , a&a , 513 , a17 quimby , r. m. et al . , 2011 , nature , 474 , 487 quimby , r. m. , yuan , f. , akerlof , c. , wheeler , j. c. , 2013 , mnras , 431 , 912 rest , a. et al . , 2011 , apj , 729 , 88 rybicki g. b. , lightman a. p. , 1986 , radiative processes in astrophysics . wiley , new york sari , e. , & esin , a. a. , 2001 , apj , 548 , 787 schlickeiser , r. , 2002 , cosmic ray astrophysics , ( new york : springer ) smith , n. , 2013 , mnras , 429 , 2366 smith , n. , chornock , r. , silverman , j. m. , filippenko , a. v. , foley , r. j. , 2010 , apj , 709 , 856 smith , n. , foley , r. j. , filippenko , a. v. , 2008 , apj , 680 , 568 smith , n. et al . , 2007 , apj , 666 , 1116 smith , n. , mccray , r. , 2007 , apj , 671 , l17 smith , n. et al . , 2009 , apj , 695 , 1334 sveshnikova , l. g. , 2003 , a&a , 409 , 799 svirski , g. , nakar , e. , sari , r. , 2012 , apj , 759 , 108
neutrino physics in the past two decades has seen an astounding transformation from a collection of anomalies to precision science , which most recently resulted in the measurement of @xmath1 at accelerator @xcite and reactor @xcite experiments . the next goals in neutrino oscillation physics are the determination of the mass hierarchy ( _ i.e. _ , the ordering of the neutrino mass eigenstates ) and of the leptonic cp phase , which will require control of systematic errors at an unprecedented level of accuracy in neutrino physics . in particular the measurement of the cp phase will put very stringent demands on the determination of neutrino versus antineutrino interactions in the gev region , see _ e.g. _ refs . it is widely recognized that our current understanding of nuclear effects in neutrino - nucleus interactions is insufficient to guarantee the required control of systematical errors . a number of studies have been performed focusing on nuclear effects in this context , see for instance refs . given the complexity of the full problem with respect to neutrino and antineutrino cross - sections , we investigate here the simpler case of @xmath0 disappearance . in particular , we would like to extend the results of ref . @xcite and perform a comparison of _ different _ nuclear models and their impact on oscillation parameters . it is a common approach to estimate errors arising from theories using several available theoretical calculations and take the spread in their results as a measure of the associated uncertainty ; we follow the same logic and specifically , will compare the results obtained using genie @xcite and gibuu @xcite . clearly , this type of inter - comparison can be extended to a larger number of event generators as well . our choice is largely guided by the fact that genie is widely used , and gibuu represents an unique and complementary theoretical approach based on transport theory . moreover , both codes are open sources which seem a necessary condition for meaningful comparisons . we evaluate the impact of several aspects of nuclear models and their differing implementation with respect to the ability to measure the so - called atmospheric parameters , @xmath2 and @xmath3 in an experimental setup which is similar to t2k . our main result is that effects from changing the target nucleus from carbon to oxygen induces a bias in @xmath2 of about 1@xmath4 . fitting data obtained with one generator and the other results in dramatic shifts due to different modeling and implementation of final state interactions and nuclear models . also , the absence of multi - nucleon correlations in the fit can induce a bias between 1@xmath4 and 3@xmath4 on the results for both of the oscillation parameters . as previously found in ref . @xcite , a near detector does not resolve these issues . the paper is organized as follows . in sec . [ sec : events ] , we outline the principle of energy reconstruction for charged - current quasi - elastic - like events , and in sec . [ sec : gen ] we perform a detailed comparison of the various physics models implemented in both generators for a number of relevant interaction models . this is followed by a detailed description of the simulations in sec . [ sec : sim ] which leads to our results in sec . [ sec : results ] . finally , in sec . [ sec : concl ] we present our conclusions . in a water erenkov detector , the event sample is restricted to charged - current ( cc ) `` quasi - elastic - like '' ( qe - like ) events , a definition which was first introduced in refs . these are selected by requiring that there be only one charged particle above erenkov threshold in the final state , the so - called `` single ring '' events . for a qe event considering the neutron at rest , the neutrino energy can be reconstructed from the kinematic variables of the charged lepton @xmath5 in the final state as : @xmath6 where @xmath7 is the free neutron rest - mass , @xmath8 , and @xmath9 is the binding energy . both genie and gibuu use @xmath9=30 mev , a value which is obtained from electron scattering data @xcite . [ eq : recengwoproton ] is exact only for the qe interaction with a neutron at rest . however , for any neutrino experiment observing qe - like events , the final sample will also contain events that are not qe . for example , in a single pion neutrino interaction , _ i.e. _ , an event with a charged lepton and a pion in the final state , the pion can be absorbed by the nucleus during final state interactions ( fsi ) . such event will therefore be classified as qe - like . in this case , if eq . [ eq : recengwoproton ] is used to reconstruct the neutrino energy , it will unavoidably lead to a value of the reconstructed neutrino energy _ lower _ than the true incident value for the reason that part of the neutrino energy is carried away by the pion and eventually absorbed by the nucleus . the actual value of the reconstructed neutrino energy will depend on the energy of unobserved particles in the final state @xcite . therefore , while the reconstructed energy will mostly coincide with the true energy of the incident neutrino for a true qe event , there exists a certain probability to have a non - qe event end up being reconstructed with a significantly different neutrino energy . this defines a migration matrix between true and reconstructed neutrino energies , @xmath10,@xmath11 , where each element represents the probability that an event for each given true neutrino energy @xmath12 ends up being reconstructed with a different energy @xmath13 . the final qe - like event sample will comprise qe as well as non - qe events , where pions are not present in the final state : @xmath14 where @xmath15 is the true neutrino energy , @xmath16 stands for the oscillation probability of @xmath17 , @xmath18 is the initial flavor neutrino flux , the matrices @xmath19 account for the probability that an event with a true neutrino energy in the bin @xmath20 ends up being reconstructed in the energy bin @xmath21 . @xmath22 stands for the cross - section for a given non - qe process in which there are no pions in the final state . in this work , we will study a muon neutrino disappearance experiment , so @xmath23 . finally , it should be kept in mind the classification of event types in neutrino experiments is not well - defined , since the incident neutrino energy is not known . in this work , we focus on the migration of non - qe events into the qe - like sample . however , it should be kept in mind that there is a second source of mis - identifying events which takes place in the opposite direction . for instance , an initially purely qe interaction , where the outgoing proton re - interacts inside the nucleus , producing a delta resonance ( @xmath24 ) . the @xmath24 will then decay and produce a pion in the final state . as a result , this event will be classified as non - qe due to the resonance production . in summary , both the qe and resonance interactions are entangled , and sometimes it is hard to distinguish one from the other . neutrino oscillation experiments rely on different event generators to help them estimate the portion of mis - identified qe and resonance production events . however , there are considerable theoretical uncertainties in predicting both event classes . this results in different event generators having different theoretical implementations of the same event types . this is a major source of systematic uncertainty which , like all theory - related errors , is difficult to quantify . in this work , both migration matrices and cross - sections in eq . [ eq : events - qelike ] are computed by the event generator . we have considered two different event generators in this work : genie ( generates events for neutrino interaction experiments ) 2.8.0 @xcite and gibuu ( giessen boltzmann - uehling - uhlenbeck ) 2.6 @xcite . genie is used by the major neutrino accelerator experiments in the us , such as miner@xmath25a @xcite , minos @xcite , microboone @xcite , no@xmath25a@xcite and lbne @xcite , and is also used by the t2k @xcite experiment . gibuu , on the other hand , is based on a semiclassical transport model @xcite , and therefore constitutes a complimentary and independent theoretical approach . figure [ fig : tot_xsec_genie_gibuu ] shows the total cross - sections per neutron on @xmath26o with no pions in the final state for all qe - like interactions using genie and gibuu . in this work , we consider charged - current quasi - elastic ( qe ) , charged - current single pion production ( res ) from @xmath24 resonant decay , charged - current non - resonant pion production ( non - res ) , and neutrino interactions involving the two - nucleon currents arising from meson exchange processes and nucleon - nucleon correlations ( mec/2p2h ) . we analyze simulated events with the additional requirement of no pion in the final - state . it should be noted that we do not consider the contributions of deep inelastic scattering ( dis ) and the production of higher resonances , even though they were included in the analysis performed in ref . the reason for this is the following . the cross - section for the production of higher resonances with no pion in the final state is found to be very small , see fig . 1 in ref . @xcite , and so is its contribution to the total number of events . in addition , for the particular setup in this simulation work , the neutrino flux decreases very rapidly above 1 - 1.5 gev , where the dis cross - section is still negligible . therefore , since there is no sizable contribution from dis events either , they have been removed from our simulations . it should be kept in mind , though , that the removal of events coming from the higher resonances and dis may not be the case for neutrino experiments in which the flux peaks at higher energies , such as lbne @xcite . the migration matrices are computed as follows . each neutrino interaction sample consists of 50,000 events , generated in bins of true neutrino energy from 0.2 gev to 4.8 gev , for a total of 46 bins . each matrix is built by counting the number of entries in each bin of the reconstruction energy as defined in eq . [ eq : recengwoproton ] and then dividing it by the total amount of entries . this is done separately for each type of interaction as defined in fig . [ fig : tot_xsec_genie_gibuu ] . for each given value of the initial true energy , a probability distribution in reconstructed energies @xmath27 can be computed . in fact , @xmath10,@xmath28 can be regarded as filling one two - dimensional probability density plane where the y - axis indicates the density in reconstructed energies , @xmath27 , and the x - axis represents the true energy value . this effectively implements an energy smearing due to nuclear effects . we require the sum of densities for a given true energy to be normalized to the unity , so that the number of events before and after migration remains the same . both genie and gibuu event generators differ on the nuclear models used as well as on how the different types of interactions are computed . in the rest of this section , we list some of the main differences between them . when generating neutrino - nucleus qe interactions , both genie and gibuu use variants of the relativistic fermi gas model ( rfg ) @xcite to describe the nuclear structure and the dynamics of neutrino - nucleus interactions under the hypothesis of plane wave impulse approximation ( pwia ) , see for instance ref . @xcite for a review . in the pwia , the struck nucleon in every single neutrino - nucleus interaction is treated as a quasi - free particle due to the high momentum transfer q@xmath29 , while the rest of the nucleus , the so - called spectator system , is left unperturbed . in the rfg , the double differential cross - section can be written down as a function of scattering angle and outgoing lepton s total energy , that is approximated as the lepton s kinematic energy @xcite : @xmath30 with @xmath31 where @xmath32 is a normalization factor , and @xmath33 is the fermi momentum ( @xmath34 , or @xmath35 ) , which is the same for all nuclear targets in genie and gibuu . @xmath36 , where @xmath9 is the average binding energies ( _ i.e. _ , the binding energy of the nucleon ) . @xmath37 is the elementary cross - section used to describe the probability of interactions between the incident neutrino and the nucleon . the integration limits in eq . [ eq : tot_xsec_formula ] are determined by the boundaries of the kinematically allowed region @xcite . the nuclear model used in genie to simulate qe interactions is a modified rfg that includes short - range nucleon - nucleon correlations according to the model developed by a. bodek and j. l. ritchie @xcite . gibuu also uses the rfg , but in this case , the rfg is modified by adding corrections from the nucleon s momentum and density dependent mean - field potential , where all nucleons are considered to be bound . the phase - space density function in gibuu also includes the real part of the self - energy for the knock - out nucleon @xcite . both generators use the same value for the axial mass , @xmath381 gev / c@xmath39 . the vector form factor in genie is bbba05 @xcite , while in gibuu it is bbba07 @xcite . the qe cross - sections for the two generators are practically the same but with small discrepancies which result from different electromagnetic form factor shapes and separate corrections added to the rfg . the qe event distribution for genie and gibuu on @xmath26o is shown in fig . [ fig : events_qe ] . we have noted some differences in the event distribution as function of reconstructed neutrino energy between genie and gibuu as shown in fig . [ fig : events_qe ] . the nuclear model used to describe the qe neutrino interactions is essentially the same between genie and gibuu . for example , we find no differences in the qe cross - section as shown in fig . [ fig : tot_xsec_genie_gibuu ] between genie and gibuu . we do not find any shift between the event distributions as function of reconstructed neutrino energy if we repeat the same simulation but removing fsi effects for both genie and gibuu . nevertheless , we find a shift of 10% between the event distribution as a function of reconstructed neutrino energy ( defined as in eq . [ eq : recengwoproton ] ) between genie and gibuu , as it is shown in fig . [ fig : events_qe ] . this is not surprising due to the fact that genie and gibuu follow a completely different approach to describe fsi . for example gibuu , the fsi is modeled by solving the semi - classical boltzmann - uehling - uhlenbeck equation . for more details on genie and gibuu , see for instance refs . @xcite . the migration matrices for qe events are shown in fig . [ fig : genie_matrix_qe ] and fig . [ fig : gibuu_matrix_qe ] in app . [ app : matrices ] for genie and gibuu , respectively . these matrices include fsi effects . at the neutrino energies relevant for this work ( @xmath401 gev ) , the second dominant neutrino interaction is the single pion production via @xmath24 resonances ( res ) . the res includes : ( 1 ) the baryon resonance decay ; ( 2 ) the charge - exchange of a neutral pion inside the nucleus ; ( 3 ) the absorption of multiple pions . gibuu contains 13 kinds of resonance modes . the vector form factors for each resonance mode are obtained from the maid @xcite analysis of the electron scattering data . the axial form factors for gibbu are derived from a fit to the data as described in ref . @xcite . in the case of gibuu , nuclear effects include collisions within nucleons and the nucleon s momentum and density dependent mean - field potential . the fsi is determined by modeling a coupled - channel transport method . this model is tested against electron- , photon- , pion- and proton - scattering data . genie , instead , applies the rein - sehgal @xcite model with 16 resonance models in the region of transferred energy @xmath41 . fermi motion and pauli blocking are the only nuclear effects included in this case . the fsi is modeled using an intra - nuclear model as described in detail in ref . genie res is validated with electron scattering data from @xmath42fe and @xmath43c targets . the res event distributions for genie and gibuu using @xmath26o are shown in fig . [ fig : events_res ] in app . [ app : matrices ] . the corresponding migration matrices are shown in figs . [ fig : genie_matrix_res ] and [ fig : gibuu_matrix_res ] for genie and gibuu , respectively . in fig . [ fig : tot_xsec_genie_gibuu ] , we only show the contribution of @xmath24-resonant production to the cross - section without regarding the heavier resonance modes . as already explained at the beginning of this section , the higher resonance effects become more significant and dominant at higher neutrino beam energies and would therefore have little impact on our results . aside from qe and res , at neutrino energies below 1 gev an additional contribution to the total neutrino cross - section comes from processes involving two particle - two hole ( mec/2p2h ) excitations , as shown in fig . [ fig : tot_xsec_genie_gibuu ] . several models have been proposed to compute these contributions in neutrino experiments , and the field has been object of a very intense research in the past few years . for an incomplete list of references on this topic , see @xcite . sources of mec/2p2h excitation include : ( 1 ) nucleon - nucleon correlations in the initial state , ( 2 ) neutrino coupling to 2p2h and ( 3 ) fsi . in processes in which two - nucleons are knocked out from the target nucleus , the nucleon s momentum distribution in the spectator system is influenced . therefore , an excess of qe - inclusive events is produced . furthermore , it is obvious that if eq . [ eq : recengwoproton ] is applied to obtain the reconstructed neutrino energy , this will most likely differ from the true incident neutrino energy , due to the non - qe nature of the interaction . the effect of 2p-2h has been recently revealed by the theoretical interpretation @xcite of the results obtained in the miniboone experiment @xcite . a detailed description of the implementation of mec/2p2h in genie is available in ref . @xcite , and for gibuu is available in ref . @xcite . the mec/2p2h event distribution for genie and gibuu on @xmath26o is shown in fig . [ fig : events_2p2h ] of app . [ app : matrices ] . the associated migration matrices are shown in figs . [ fig : genie_matrix_2p2h ] and [ fig : gibuu_matrix_2p2h ] for genie and gibuu , respectively . these matrices seem to be rather different . to simulate this particular interaction , both gibuu and genie have been tuned based on the measurements on @xmath43c done by the miniboone experiment and they do not yet include full theoretical model implementations like the ones presented in refs . @xcite . in addition , it should be noted that the miniboone experiment measures the sum of the qe and mec/2p2h contributions , but the experiment is not capable of discriminating between them . in other words , the tuning of the neutrino interaction generators is performed in such a way that the sum of the qe and the mec/2p2h contributions to the double differential cross section agrees with the data . therefore , if a difference exists between genie and gibuu in the qe interactions , then the mec/2p2h by construction will be different as well . finally , it is worth mentioning that no difference in the total cross - section between genie and gibuu is found for both @xmath43c and @xmath26o as a result of the heavy tuning of the model on miniboone @xmath43c results . the non - resonant pion production ( non - res ) includes the contributions of coherent and non - coherent pion productions . the coherent pion production results in one muon and one single pion in the final state , in which the pion is not produced by a @xmath24-resonance decay . the residual nucleus system remains in its ground state in the coherent production while in the incoherent pion production this is not the case . genie includes only the coherent ( exclusive ) pion production @xcite , whereas gibuu considers only the incoherent pion production process . a full theoretical calculation for this interaction mode can be found in ref . @xcite . genie uses pcac @xcite with the rein - sehgal @xcite model to simulate the single pion interaction with the nucleon inside the ground - state target nucleus . the neutrino - nucleon cross - section is computed at the initial time ( t=0 ) with the elastic nuclear form factors and an absorption factor to simulate the fsi of the outgoing pion . in contrast , gibuu takes into account only the incoherent part of the initial - state pion , and does not make any local approximation to the @xmath24 propagator in the medium @xcite . the non - res event distribution for genie and gibuu on @xmath26o is shown in fig . [ fig : events_coh ] of app . [ app : matrices ] . the associated migration matrices are shown in figs . [ fig : genie_matrix_coh ] and [ fig : gibuu_matrix_coh ] for genie and gibuu , respectively . in this work we consider a setup very similar to the t2k experiment , simulated following ref . the main details of the setup are summarized in tab . [ tab : setup ] . we consider two detectors : a far detector , placed at 295 km from the source , with a fiducial mass of 22.5 kton ; and a near detector of 1 kton fiducial mass , placed at 1 km from the decay pipe . the size and location of the near detector have been chosen so as to guarantee that it observes enough events to be able to constrain the systematic errors included in the analysis ( for details on the systematic errors and the @xmath44 implementation , see app . [ app : chi2 ] ) . in this work , we assume that the two detectors observe the same flux , and that they are identical in terms of their composition and detection properties . it should be kept in mind , however , that these assumptions would most likely _ not _ be realized in an actual neutrino beam experiment . this may lead to a _ larger _ impact of nuclear effects on the extraction of the oscillation parameters than what is found in this work . l@c@c@c@c@c@ & baseline & fiducial mass & flux peak & beam power & running time + far & 295 km & 22.5 kt & & & near & 1.0 km & 1.0 kt & & & + the only oscillation channel considered is @xmath45 disappearance . for this channel , the only relevant background would come from neutral current ( nc ) interactions . we do include such a background in our analysis , which produces a total of @xmath46 events at the far detector . nevertheless , since our aim is to explore the impact of nuclear effects on the cc signal , we keep the background event rates the same for all the configurations under consideration in this work , and no study is done on the variation of these rates with different nuclear models and/or energy reconstruction effects . background rates are smeared using a gaussian with @xmath47 mev , following ref . @xcite . for the signal , on the other hand , since the migration matrices due to nuclear effects already introduce a rather coarse energy smearing , we do not consider additional effects due to the finite resolution from the detector . we believe the effect due to this will be certainly minor , considering the large smearing that we observe already for the qe event sample , see figs . [ fig : genie_matrix_qe ] and [ fig : gibuu_matrix_qe ] in app . [ app : matrices ] . finally , energy dependent detection efficiencies are implemented for the signal , following ref . these are applied after the events are migrated to reconstructed neutrino energies . the expected number of events at the far detector for the different contributions to the qe - like event sample are shown in tab . [ tab : events ] . these are computed using the cross - sections produced for the two event generators under consideration , using oxygen as the target nucleus . [ fig : rates ] shows the expected event rates for the qe - like event sample , binned as a function of the neutrino energy . results are shown for the near and far detectors in the right and left panel , respectively . the gray shaded areas show the event rates for the events before migration to reconstructed energies , while the solid blue ( dashed red ) lines show the results after migration to reconstructed neutrino energies , when the matrices and cross - sections are computed using gibuu ( genie ) , as explained in sec . [ sec : gen ] . in all cases , the oscillation parameters have been set to the values in eq . [ eq : oscparams ] in app . [ app : chi2 ] , and detector efficiencies are accounted for . we find that the event distributions using the matrices generated with gibuu are in agreement with those shown in fig . 10 of ref . similar results were also shown in fig . 3 of ref . @xcite for a different nuclear model , the event distribution before migration has been smeared with a gaussian energy resolution function with a constant width , to account for the detectors finite energy resolution . as a result , the lowest two energy bins have slightly more events before migration than after migration . we have checked that for a detector of perfect energy resolution this is not the case and therefore our results are in agreement with those in refs . @xcite . ] . [ cols="<,^,^,^,^,^",options="header " , ] it is also worth mentioning the large contribution to the qe - like sample coming from mec/2p2h contributions . it has recently been argued @xcite that this may be the source of the large discrepancy between the value of @xmath48 reported in ref . @xcite and the world average value of @xmath49 gev , see for instance ref . @xcite . as it can be seen from tab . [ tab : events ] ( see also fig . [ fig : events_2p2h ] in app . [ app : matrices ] ) , the expected contribution to the qe - like sample from mec/2p2h interactions is rather large for the setup considered in this work as well . table [ tab : events ] shows that around a @xmath50 of the final qe - like sample in our simulated setup would come from mec/2p2h interactions . it is also noticeable the difference between the number of events ( @xmath51 ) obtained when the cross - section is computed using gibuu or genie . this difference is not coming from the size of the cross - section , since in both cases the mec/2p2h cross - section is tuned to the miniboone data and the size of the peak is roughly the same , see fig . [ fig : tot_xsec_genie_gibuu ] . the reason for the difference in the number of events is a shift in neutrino energy in the cross section obtained from the two generators when they are compared against each other : while the genie cross - section reaches its maximum at around 1 gev , the gibuu cross - section peaks at slightly higher energies , around 1.2 gev . since the peak in the neutrino flux considered in this work lies below 1 gev or so , the number of events from mec/2p2h interactions when the genie cross - section is used will be larger . finally , small differences can also be appreciated in the number of events coming from resonant and non - resonant pion production . the first one comes from a different number of resonances included in the computation , see also sec . [ subsec : gen - nonres ] . in the second case , a different result is obtained since gibuu computes incoherent pion production while genie computes only the coherent contribution to this process . in this section , we explore the impact on the extraction of the oscillation parameters in three different scenarios : a. when the target nucleus is changed in the fit ; b. when the nuclear model is changed in the fit ; c. when multi - nucleon contributions ( mec/2p2h ) are completely removed from the fit . it is common practice to `` tune '' event generators according to a given target nucleus in a certain experiment . however , the event generator may be used later on for an oscillation experiment using a different target . in this section we evaluate the effect on the oscillation analysis if an event generator is tuned according to data obtained for a certain target but the experiment is performed using a different target . results are shown in fig . [ fig : targets ] . in the left panel , the binned expected event rates are shown as a function of the reconstructed neutrino energy , when oxygen ( solid black lines ) and carbon ( dashed gray lines ) are used to compute the cross - sections and migration matrices in eq . [ eq : events - qelike ] . in both cases , the genie event generator is used to compute the matrices for all contributions . as it can be seen from the figure , there is a slight shift towards lower energies in the event rates for carbon with respect to those obtained for oxygen . we have checked that this shift is already present for the true qe event sample , and that it remains if final state interactions are removed from the simulation . this is automatically translated into a shift for the best fit value of the mass - squared splitting . the effect is shown in the right panel , where the gray shaded areas show the results when the true and the fitted event rates are computed using the same set of migration matrices , while the solid lines show the results when a different target is used to compute the matrices used in the true and the fitted event rates . in our particular example , oxygen is used to generate the matrices and cross - sections used to compute the true rates , and these are then fitted using carbon migration matrices and cross - sections . as it can be seen from the figure , there is a shift in the best fit for the atmospheric mass splitting from @xmath52 to @xmath53 . there is also a shift in the best fit for the mixing angle , although in this case the effect is minor . the value of the @xmath44 found at the best fit is also shown , together with the number of degrees of freedom in the analysis , @xmath54 , where @xmath55 is the number of energy bins and @xmath56 is the number of parameters that are being determined from the fit . finally , it is worth mentioning that we find the size of this effect to be practically negligible if the gibuu event generator is used to produce the migration matrices and cross sections instead of genie . let us now address the impact of using a different nuclear model . for this purpose , we take the event rates computed using migration matrices produced with one event generator and we try to fit them using the matrices obtained with a different generator . the difference in the event rates when the matrices are computed with different generators is significant , as it can be seen from the comparison of the solid blue and dashed red lines in fig . [ fig : rates ] . in fact , we find that the main source of the discrepancy appears in the qe sample already , as discussed in sec . [ sec : gen ] . therefore , a large effect in the fit to the oscillation parameters should be expected in this case . the results are shown in fig . [ fig : generators ] . again in this case , the shaded regions show the confidence regions when the same set of matrices is used to compute the true and the fitted event rates . for the solid lines , on the other hand , we compute the true event rates using matrices produced with gibuu , and try to fit them with the event rates computed using the matrices from genie . in all cases , oxygen is chosen as the target nucleus . we find that the difference between the two models is so large that the best fit for the atmospheric mass splitting takes place around @xmath57 , as shown in fig . [ fig : gen1 ] . such large value would be in strong tension with the presently allowed region at @xmath58 from global fits , see for instance ref . the fit would disfavor the true input value for the mass splitting at much more than @xmath58 and for the atmospheric mixing angle at roughly @xmath59 . in a real experiment , this would most likely be attributed to a systematic error that has not been correctly evaluated , or to a signal of new physics . we focus on the first possibility here . in particular , since we observe that the main difference in the event rates is that the events suffer an energy shift ( see fig . [ fig : rates ] , or fig . [ fig : events_qe ] ) , we introduce a calibration error as an additional nuisance parameter . in order to do so , the event rates are effectively recomputed as : @xmath60 \rightarrow n[(1+a)\,e ] , \nonumber\ ] ] where @xmath61 is the calibration error . an additional pull term @xmath62 is added to the @xmath44 in eq . [ eq : chi2 ] , where @xmath63 is the prior uncertainty for this parameter . after a calibration error of around a 5% is added , the resulting best fit is in much better agreement with the true input value , as it can be seen from fig . [ fig : gen2 ] . nevertheless , there is still a significant shift in the best fit ( black triangle ) with respect to the true value ( red dot ) , which brings it from @xmath64 to @xmath65 . the best fit for the mixing angle is also shifted a couple of degrees into the second octant . the value of the @xmath44 at the minimum over the number of degrees of freedom is also computed , and found to be @xmath66 . we find that the largest contribution to the minimum of the @xmath44 comes from the tension in the energy bins below @xmath67 gev between the two event distributions . this can be understood from the comparison of the event histograms in fig . [ fig : rates ] , where the differences between the solid and dashed lines are observed to be largest in this energy region . in order to accommodate the large differences in these bins , the associated nuisance parameters during the @xmath44 minimization tend to take large values . as a consequence , their respective pull - terms will significantly contribute to the final @xmath44 . details on the systematics implementation can be found in app . a. recently , a lot of attention has been drawn to the miniboone experiment and the extraction of the value of the axial mass from the qe data . it seems that the @xmath68 disagreement between the value of the axial mass extracted from the results in ref . @xcite and the one obtained from previous experiments may be explained ( at least partially ) if multi - nucleon contributions are included in the analysis , see for instance refs . @xcite . in this section , we evaluate the impact that neglecting mec/2p2h contributions may have on the extraction of neutrino oscillation parameters . the results are shown in fig . [ fig : mec ] when the migration matrices are computed using gibuu ( left panel ) and genie ( right panel ) . in both cases , oxygen is used as the target nucleus . the shaded areas show the confidence regions at 1,2 and @xmath69 cl when all contributions to the qe - like event sample are considered as in eq . [ eq : events - qelike ] , and the same set of migration matrices is used to compute the true and fitted event rates . for the solid lines , on the other hand , the fit is done when the mec/2p2h contribution is completely removed from the fitted event rates . as expected , the impact on the confidence regions is rather relevant , and induces large deviation for the best fit values of the oscillation parameters , between 1 and 3@xmath4 depending on the event generator that is used to produce the migration matrices . a final comment is in order here . as it was shown previously in the literature ( see for instance refs . @xcite ) and confirmed here , the energy dependence of the qe - like event sample is very different when mec/2p2h contributions are included . we have shown in this section that the impact of multi - nucleon contributions on the extraction of oscillation parameters can be rather large in the disappearance channels . similar effects are in principle expected in the appearance channels as well . this could have a significant impact on the sensitivity to cp violation at future oscillation experiments ( as mentioned , for instance , in refs . @xcite ) , since the sensitivity to cp violation in neutrino oscillations comes from the analysis of _ both _ the energy dependence of the signal _ and _ the comparison between neutrino and antineutrino rates . both genie and gibuu have tuned their mec/2p2h interactions to the miniboone data , and their results for mec/2p2h contributions with oxygen and carbon therefore give exactly the same results . however , _ a priori _ there is no reason to think that these effects should be the same for different nuclei . as for antineutrinos , there are currently very few measurements available . the miniboone collaboration has recently reported some measurements in the antineutrino channel , where again it seems that mec/2p2h may play a leading role @xcite . this result has also been confirmed by the minerva collaboration @xcite . nevertheless , we would like to stress out the fact that the current proposals for the next generation of neutrino experiments would use either water ( t2hk @xcite or ess@xmath25sb @xcite , for instance ) or liquid argon ( lbne @xcite and lbno @xcite ) detectors , for which there are practically no measurements available at the relevant neutrino energies . again in this case , theoretical calculations show that in principle one should not expect these effects to be similar for neutrinos and antineutrinos . @xcite and may be even larger for the latter , see for instance refs . @xcite . nuclear effects in neutrino interactions will be one of the leading sources of systematical errors in future neutrino - beam oscillation experiments . already in the current t2k appearance result they are among the largest contributors to the overall systematic error budget @xcite . in this paper we try to estimate the size of the systematic error associated with theoretical models of nuclear effects as embodied by event generators , specifically genie and gibuu . apart from providing a quantitative estimate we also developed a methodological framework which lends itself to be extended to a larger class of event generators and in principle also to cp violation studies . given that lbne has chosen argon as the detector material , one question is whether changing the nuclear target will have a profound impact on the ability to extract oscillation physics . to get a first glimpse of an answer we study the @xmath0 disappearance channel and determine the bias resulting from simulating data with oxygen as target and fitting those data with a carbon interaction model . the results of this experiment are shown in fig . [ fig : targets2 ] and the quantitative findings are summarized in tab . [ tab : summary ] , which correspond to a 1@xmath70 bias in @xmath2 . these results are only an indication , but it is noteworthy that most nuclear models have been tuned on carbon data and thus , the generators can be expected to be most accurate for nuclei around carbon . nevertheless , it is somewhat surprising that a small step in atomic mass from @xmath71 to @xmath72 leads already to a sizable bias . what would happen if one tried to extrapolate to argon with @xmath73 remains pure speculation . l@l@\|@c@c@c@c@c@ true & fitted & @xmath74 & @xmath75 $ ] & @xmath76 & @xmath63 & fig . no . + genie ( @xmath26o ) & genie ( @xmath43c ) & @xmath77 & 2.49@xmath78 & 2.28 & & [ fig : targets ] + & & @xmath79 & 2.69@xmath78 & 47.64 & & [ fig : gen1 ] & & @xmath80 & 2.55@xmath78 & 20.95 & 5% & [ fig : gen2 ] + gibuu ( @xmath26o ) & gibuu ( @xmath26o ) w / o mec & @xmath81 & 2.44@xmath78 & 22.38 & & [ fig : mec1 ] + genie ( @xmath26o ) & genie ( @xmath26o ) w / o mec & @xmath82 & 2.36@xmath78 & 19.54 & & [ fig : mec2 ] + interestingly , we find even for carbon very large differences in the shape of the qe event rate spectrum between genie and gibuu which we trace back to differences in the implementation of final state interactions , see sec . [ sec : gen ] . these differences are large enough to introduce a bias in the mass splitting of many standard deviations . the resulting minimum of the @xmath44 would also be very large as shown in fig . [ fig : gen1 ] . introducing an uncertainty on the energy scale of 5% reduces the resulting tension and brings also the @xmath44 back to acceptable levels , but still leaves a 1@xmath4 bias in both the mass splitting and mixing angle , see fig . [ fig : gen2 ] . besides , it is also not clear whether an oscillation experiment would have any freedom left to include such a large calibration error in the fit . finally , the recent miniboone results seem to imply that multi - nucleon effects play an important role for neutrino energies @xmath83 . there is a large variety of models in the literature trying to describe these effects . therefore , we test the effect of removing the multi - nucleon correlation in the fit of data which has been generated including those . this is again a case where the two generators produce different effects in gibuu the mixing angle is most affected whereas in genie it is the mass splitting seeing the bulk of the effect , which is obvious from fig . [ fig : mec ] . in summary , we find that changing nuclear targets , the used event generator or the implementation of multi - nucleon effects , each leave a bias comparable to the statistical errors in the determination of the mixing parameters , as illustrated in tab . [ tab : summary ] . any experiment aiming at high precision measurements of oscillation parameters like the leptonic cp phase will have to develop a strategy to deal with these uncertainties in a transparent fashion . one important step in this direction would be to make the event generators accessible to the community . we have only considered light targets like carbon and oxygen and it is unclear how to extrapolate to heavier targets like argon without additional data . the methods presented here are well suited to be extended to experiments aiming to determine the cp phase and the neutrino mass hierarchy . the authors would like to thank s. dytman , u. mosel and o. lalakulich for their support , which lead to successful simulation results . we would also like to thank d. meloni , j. nieves , n. rocco and m. vicente vacas for useful discussions , and o. benhar for the precious discussions and for careful reading of this manuscript . this work has been supported by the u.s . department of energy under award number . in this appendix we briefly describe the details on the implementation of the @xmath44 and the inclusion of systematic uncertainties . all fits to the oscillation parameters presented in sec . [ sec : results ] are performed using globes @xcite . unless otherwise stated , the true values of the oscillation parameters are set to the following : @xmath84 a @xmath44 analysis is done to extract the best fit values for the oscillation parameters as well as the allowed confidence regions at 1 , 2 and 3@xmath4 cl . for each energy bin @xmath21 and detector @xmath85 , a contribution to the @xmath44 is computed as : @xmath86 with @xmath87 here , @xmath88 ( @xmath89 ) refer to the true ( fitted ) event rates observed at a detector @xmath85 in an energy bin @xmath21 , @xmath90 indicates the dependence on the test values for the oscillation parameters , and @xmath91 and @xmath92 stand for the nuisance parameters associated to flux and normalization uncertainties , respectively . it should be noted that @xmath88 depends only on the true values assumed for the oscillation parameters , while @xmath89 depends on the pair of values we are testing as well as on the nuisance parameters . in addition , the nuisance parameter associated to the normalization error is bin - to - bin correlated , while the one associated to the flux uncertainty is bin - to - bin uncorrelated . these will help to accommodate the normalization and shape differences in the event rates due to different nuclear models . the final @xmath44 reads : @xmath93 where the two last terms are the pull - terms associated to the nuisance parameters @xmath94 , and @xmath95 is the prior uncertainty assumed for each systematic error ( which have been set to 20% in all cases ) . all oscillation parameters are kept fixed during the fit . since the analysis is done for the disappearance @xmath96 channel only , we believe that marginalization will not have a relevant impact on our results . finally , the sum in eq . [ eq : chi2 ] is done over 100 mev bins between 0.2 and 2.0 gev . nevertheless , the migration matrices and cross - sections are computed up to @xmath97 gev in true neutrino energy in order to get the full contribution from the high - energy tail of the flux . in this appendix we show the complete set of migration matrices used in the oscillation analysis for genie ( fig . [ fig : genie_matrix ] ) and gibuu ( fig . [ fig : gibuu_matrix ] ) using oxygen as the target nucleus . the number of events as function of reconstructed neutrino energy for the various neutrino interaction modes , as described in sec . [ sec : gen ] , is also shown in fig . [ fig : event_distribution_carbon ] for oxygen . the results for carbon are similar and therefore will not be shown here .
several years ago , observations of supernovae of type ia ( sne ia ) demonstrated that the expansion of the universe is accelerating @xcite . associated with this acceleration is dark energy , a component of the universe with negative pressure , that makes up a significant fraction of the total energy density . recent years have seen the supernova evidence for dark energy continue to mount @xcite . meanwhile , cosmic microwave background ( cmb ) data from the wmap project @xcite , in combination with information about either the hubble constant @xcite or the galaxy power spectrum @xcite , also requires the existence of dark energy . cross - correlations between cmb anisotropies and matter power spectrum inhomogeneities provide evidence for a late - time integrated sachs - wolfe ( isw ) effect @xcite . this isw effect indicates a recent change in the inhomogeneous gravitational potential , providing further evidence for dark energy . one way to characterize the dark energy is to measure its equation of state , the ratio @xmath8 of its pressure @xmath9 and energy density @xmath10 . the simplest model of the dark energy is a cosmological constant @xmath11 , which has a constant @xmath12 . quintessence models describe the dark energy as a dynamical scalar field with an equation of state @xmath13 , and in general , @xmath0 will not be constant in time @xcite . furthermore , models of extended quintessence or `` phantom energy '' even allow @xmath14 @xcite . models of the universe incorporating a cosmological constant or quintessence , in addition to cold dark matter , are known as @xmath11cdm or qcdm models , respectively . discovering whether or not the dark energy is a cosmological constant is the primary goal of the study of dark energy . if dark energy is shown not to be a cosmological constant , then the next important issues are whether or not @xmath14 , which can be theoretically problematic @xcite ( however , see @xcite for an alternative viewpoint ) , and whether or not @xmath0 is changing with time . while a quintessence with constant @xmath15 is very difficult to distinguish quantitatively from a cosmological constant , the qualitative difference for fundamental physics is enormous , so it is critical that the maximum effort be made to reduce the uncertainty in @xmath16 and its time derivative . ultimately , the precise quantitative values of @xmath0 and its time derivative are important for model building , but this is less crucial , at present , than the qualitative issue of whether the dark energy is dynamical or not . in order to address such questions about the nature of the dark energy , it is crucial that the equation of state and its time variation be determined observationally . it has been known for some time that the cosmological probes used for stuyding the dark energy are plagued by numerous parameter degeneracies @xcite . because of these degeneracies , a large uncertainty in a cosmological parameter not directly related to the dark energy sector ( _ e.g. _ , the matter density ) can lead to a large uncertainty in the dark energy equation of state . this sensitivity to other parameters means that one must incorporate the uncertainties in all parameters to obtain a realistic estimate of the uncertainties in the dark energy equation of state . in particular , one should avoid the use of strong priors on the form of @xmath17 , the values of other parameters , and independent information coming from other experiments . these can lead to underestimates of the uncertainty in @xmath0 by a very large factor . we will also discuss how standard likelihood marginalization can give a misleading impression of the uncertainty in @xmath0 . the analysis presented here employs a @xmath18 minimization procedure in order to avoid possible problems with marginalization , and assumes only weak prior constraints . we compare our @xmath18 minimization with the standard approach , the marginalization of a probability function computed using monte carlo markov chains . it is argued that our procedure gives a more conservative assessment of constraints in certain degenerate parameter spaces . in this sense , our procedure and the markov chain are complementary , as we will discuss . we apply this @xmath18 minimization analysis to the currently available data . the simultaneous determination of the dark energy equation of state and its red shift derivative is difficult without the combination of data from several cosmological probes . thus we analyze the latest data from wmap , sn ia searches , and the sloan digital sky survey . strong priors are avoided to prevent the underestimation of dark energy constraints . although the data favor @xmath19 and @xmath20 , we are unable to rule out the cosmological constant conclusively . furthermore , we find the constraints on @xmath17 to depend strongly on the parameterization chosen for @xmath17 , as well as on the location of the best - fit model in parameter space , despite the fact that we have restricted our study to two - parameter equations of state . we also list possible biases in the cosmological probes , and discuss their effects on our results . our conclusion that the cosmological constant is consistent with current data agrees with previous analyses , as we discuss in sec . [ sec : discussion ] . this establishes the validity of our analytic methods , which we can then apply with confidence to projections of future measurements . since the cosmological constant is consistent with current data , it is useful to ask precisely how well dark energy may be constrained in the future using known cosmological probes . we address this question by applying our @xmath18 minimization procedure to simulations of the data from these probes . besides the cmb and sne ia , we simulate data from a probe of weak gravitational lensing ( wl ) . in the future , wl is expected to be useful for constraining dark energy @xcite . most of these studies have demonstrated that wl breaks parameter degeneracies in the cmb analysis , and have shown the effectiveness of the combination of cmb and wl data in constraining dark energy . our study is the first to use a joint analysis of simulated cmb , sn ia , and wl data , rather than imposing prior constraints as a `` replacement '' for one of these data sets . this is especially important for parameters such as @xmath21 and @xmath22 , which have strong degeneracies with other parameters . since our goal is the study of @xmath17 constraints , rather than general parameter constraints , we assume the cosmological model outside of the dark energy sector to be as simple as possible . within this model , we are careful to avoid strong priors on the cosmological parameters . it is well understood that each of these cosmological probes has systematic uncertainties associated with it , and that significant progress needs to be made in order for each one to reach the level of precision necessary for constraining dark energy . therefore , we attempt at every opportunity to strike a reasonable balance between optimism and realism . the paper is organized as follows . [ subsec : demodel ] discusses several parameterizations of the dark energy equation of state . our choice of cosmological parameters , and the prior constraints imposed on them , are listed in sec . [ subsec : cosmoparamdegenpriors ] . @xmath18 minimization is compared with marginalization in sec . [ subsec : gridmin ] . sec . [ subsec : currentchi2 ] summarizes our analysis of current data , and secs . [ subsec : snmc]-[subsec : wlmc ] describe the simulation and analysis of future data . finally , sec . [ sec : discussion ] lists and interprets our findings , for our analysis of current data as well as our forecasts of future constraints , and sec . [ sec : conclusions ] discusses our conclusions . in the spirit of reasonable optimism , we have assumed that the dark energy may be parameterized using an equation of state @xmath23 , where @xmath9 and @xmath10 are the pressure and dark energy density , respectively , and @xmath17 is an unknown function of red shift . we have also assumed that the dark energy sound speed is @xmath24 . without any theoretical guidance about the form of @xmath17 , the space of all possibilities is equivalent to an uncountably infinite set of cosmological parameters . the data depend on @xmath17 only through a multiple integral relation @xcite , so we lack the large number of measurements of @xmath17 that would be necessary for non - parametric inference @xcite . thus we must describe @xmath17 using a small number of parameters in order to keep the analysis tractable . the danger is that there is no `` natural '' way to parameterize @xmath17 , so the choice of parameterizations is essentially arbitrary . ideally , one would like to verify that any constraints obtained on @xmath17 are parameterization - independent . however , the space of all possible parameterizations is infinite - dimensional , so we re back to square one . the only feasable option is to test for parameterization - independence by comparing constraints on a small number of `` reasonable - looking '' parameterizations . to begin with , we assume that @xmath17 is an analytic function , and that it can be approximated at @xmath25 by the first few terms in its taylor series about @xmath5 . all known probes are sensitive to an integral of some function of @xmath17 , rather than to @xmath17 itself , so any `` bumps and wiggles '' in @xmath17 are effectively smoothed out . it is reasonable to assume that this smoothed @xmath17 can be approximated by a low - degree polynomial in @xmath1 out to @xmath26 . from now on we proceed based on this assumption , with the caveat that the `` true '' equation of state can have added `` bumps and wiggles '' that disappear when smoothed . ( note , however , that any large - scale kinks or sharp changes in the actual equation of state can be missed by such simple parameterizations . for an alternative approach , see , _ e.g. _ , @xcite . ) the simplest parameterizations useful for describing dynamical dark energy have at least two parameters , one describing the equation of state @xmath21 at @xmath5 and the other parameterizing its red shift derivative @xmath27 . a few parameterizations in the literature are @xmath28(note that @xmath29 in ( [ eqn : eosuros ] ) differs by a factor of @xmath30 from the definition introduced in @xcite . we chose ( [ eqn : eosuros ] ) in order to facilitate comparison with ( [ eqn : eoslinder ] ) and ( [ eqn : eosparam ] ) . ) we are interested in the simplest two - parameter equations of state , and as we will see , ( [ eqn : eossimple ] ) is of limited utility for studying high red shift data . this leaves only eq . ( [ eqn : eoslinder ] ) out of the three parameterizations above . since we would like to compare two `` reasonable '' two - parameter equations of state , we introduce a fourth parameterization , @xmath31 dark energy constraints based on ( [ eqn : eosparam ] ) and ( [ eqn : eoslinder ] ) may be compared in order to estimate the parameterization - dependence of our results . table [ tab : eosparam ] compares all four parameterizations using their present - day derivatives @xmath22 , their mean low red shift ( @xmath32 ) derivatives @xmath33 , and their high red shift limits @xmath34 . .summary of the dark energy equation of state parameterizations used in the literature . [ tab : eosparam ] [ cols="^,^,^,^ " , ] contours for cmb and sn ia `` data '' simulated using the fiducial models ( @xmath35 , @xmath36 ) ( top ) and ( @xmath37 , @xmath38 ) ( bottom ) . the @xmath39 , @xmath40 , and @xmath41 contours are shown . the location of the fiducial model is marked by a `` + '' sign.,width=316 ] @xmath42 & @xmath43 + sn ia & prior @xmath44 & @xmath45 & @xmath46 + cmb + & standard & @xmath47 & @xmath48 + sn ia & @xmath49 in ( [ eqn : dctt]-[eqn : covteee ] ) & @xmath50 & @xmath51 + & cmb @xmath52 & @xmath53 & @xmath54 + & 1 yr . wmap & @xmath55 & @xmath56 + & sn @xmath57 & @xmath58 & @xmath59 + & sn @xmath60 & @xmath61 & @xmath62 + & sn @xmath63 & @xmath64 & @xmath65 + cmb + & standard & @xmath66 & @xmath67 + sn ia + & wl on linear & @xmath68 & @xmath69 + wl & scales only & & + & cmb @xmath52 & @xmath70 & @xmath71 + & sn @xmath57 & @xmath72 & @xmath73 + & sn @xmath63 & @xmath74 & @xmath75 + next , the analysis was repeated with two different fiducial models , chosen to lie approximately along the line of degeneracy in the ( @xmath76,@xmath77 ) plane . as shown in fig . [ fig : cmbsn.varfid ] , the model ( @xmath35 , @xmath36 ) had @xmath2 uncertainties of 0.73 and 2.58 on @xmath76 and @xmath77 , respectively , which are significantly larger than those reported above . meanwhile the model ( @xmath37 , @xmath38 ) had uncertainties of 0.48 and 1.58 . these are consistent with our constraints obtained using the @xmath11cdm fiducial model . returning to our @xmath11cdm fiducial model , we repeated the analysis with different assumptions about the data in order to test the robustness of our forecast constraints . table [ tab : checkfore ] lists the new constraints obtained . the standard deviations in our original constraints are about @xmath78 , so we did not consider a modification to be significant unless it changed constraints by at least @xmath79 . first we checked to what extent a cmb simulation without the `` extra '' factor of @xmath80 in the denominator of the covariance matrix ( [ eqn : dctt ] , [ eqn : dcee ] , [ eqn : dcte ] , [ eqn : covxx ] , [ eqn : covttte ] , [ eqn : covttee ] , [ eqn : covteee ] ) would improve our constraints . the difference turns out to be negligible . in fact , the final constraints are relatively independent of the details of the cmb simulation and analysis . cutting off the cmb power spectrum at a maximum multipole of @xmath81 , or replacing the wmap 8 year data set with a simulated 1 year data set , lead to insignificant changes in the dark energy constraints . on the other hand , the constraints are very sensitive to changes in the quality of the supernova data . changing the sn ia systematic uncertainty , or its red shift dependence , leads to significant changes in the dark energy constraints . meanwhile , the final @xmath2 uncertainties provided by the combination of cmb and sn ia data are nearly identical to those found by fixing @xmath82 in the sn analysis alone . thus , in some sense , adding the cmb data set is equivalent to fixing @xmath82 as a function of @xmath76 and @xmath77 in the supernova analysis . we began sec . [ subsec : forecasts ] by identifying three milestones for comparing constraints on @xmath76 and @xmath77 . recall that the first milestone is to distinguish between a cosmological constant and a dark energy with @xmath83 . from our forecasts we find @xmath84 , which is more than twice as high as the uncertainty of @xmath85 needed to distinguish between the two dark energy models . the second milestone , to distinguish between @xmath11cdm and tracker models of dark energy , is also not yet reached by the combination of cmb and sn ia data sets . we find @xmath86 at @xmath87 , twice the uncertainty of @xmath88 needed to reach this milestone . we can see from fig . [ fig : fore - cmbsn ] that , if we shifted the contours to be centered on ( @xmath89 , @xmath90 ) , then the cosmological constant model as well as some of the tracker models ( @xmath91 , @xmath92 ) would be within the @xmath2 contours . thus we can not confidently claim that the combination of cmb and sn ia data can rule out either the @xmath11cdm model or the trackers . finally , this combination is far from reaching the third milestone , which would require that the uncertainties in @xmath76 and @xmath77 be reduced by over an order of magnitude . the addition of weak lensing to the analysis resulted in the contours shown in fig . [ fig : fore - cmbsnwl ] , with grid spacings @xmath93 and @xmath94 . weak lensing tightened the @xmath2 constraints to @xmath95 and @xmath96 . the standard deviations in the @xmath2 uncertainties , found from five separate simulations , were @xmath97 and @xmath98 in @xmath42 and @xmath43 , respectively . as above , the standard deviations in each of the @xmath76 and @xmath77 uncertainties were about @xmath78 . we checked our results using a fisher matrix calculation and found the @xmath2 constraints @xmath99 and @xmath100 in @xmath76 and @xmath77 , respetively , consistent with our minimization results . meanwhile , the minimum @xmath2 uncertaity in @xmath17 was found to be @xmath101 at @xmath102 . our constraints are about the same as those forecast for snap sne ia with planck priors , but worse by a factor of two than snap sne ia with weak lensing , as listed in table [ tab : forecast ] . meanwhile , galaxy cluster measurements based on the more ambitious lsst survey claim an improvement by a factor of three over our forecast @xmath76 constraints @xcite . , @xmath40 , and @xmath41 contours are shown . the location of the fiducial model is marked by a `` + '' sign.,width=316 ] contours for cmb , sn ia , and wl `` data '' simulated using the fiducial models ( @xmath35 , @xmath36 ) ( top ) and ( @xmath37 , @xmath38 ) ( bottom ) . the @xmath39 , @xmath40 , and @xmath41 contours are shown . the location of the fiducial model is marked by a `` + '' sign.,width=316 ] the analysis was repeated with two different fiducial models in order to assess the dependence of dark energy constraints on the fiducial model . moving from the @xmath11cdm fiducial to the fiducial model ( @xmath35 , @xmath36 ) increases uncertainties in @xmath76 and @xmath77 by about @xmath103 each , to @xmath104 and @xmath105 , respectively . moving in the other direction along the degeneracy curve , to the fiducial model ( @xmath37 , @xmath38 ) , leads to a modest decrease in @xmath76 and @xmath77 uncertainties to @xmath106 and @xmath107 , respectively . qualitatively , this is the same behavior as was seen with the cmb and sn ia data combination . further tests of the robustness of our constraints revealed that , once again , cmb data at @xmath108 do not contribute much to the dark energy constraints ( see table [ tab : checkfore ] ) . meanwhile , if the weak lensing analysis is restricted to linear scales , then constraints on @xmath76 and @xmath77 weaken considerably , as shown in table [ tab : checkfore ] . this is consistent with the findings of @xcite . as a function of @xmath1 , for the data combinations cmb + sn ia ( upper , solid line ) and cmb + sn ia + wl ( lower , dashed line).,width=220 ] comparing @xmath2 constraints with and without weak lensing ( including nonlinear scales ) , we see that the uncertainties in @xmath76 shrink by a factor of @xmath109 , and the uncertainties on @xmath77 shrink by a factor of @xmath110 . in order to understand the contribution of weak lensing to the overall analysis , we compared @xmath17 uncertainties as functions of redshift for the combinations cmb + sn ia and cmb + sn ia + wl , as shown in fig . [ fig:2_sigma_w_vs_z ] . without weak lensing , @xmath17 is well constrained only around @xmath111 . evidently , weak lensing adds information on @xmath0 at some higher red shift , complementing the constraints from cmb and sn ia . the result is a @xmath17 uncertainty that is not only lower , but much more uniform across the red shift range . this improvement allows the combination of cmb , sn ia , and wl data to reach two of the three milestones identified at the beginning of this section . at @xmath6 we find @xmath112 , which will easily allow us to distinguish between a cosmological constant and dark energy models with @xmath113 . thus , weak lensing will either confirm or conclusively rule out the dark energies with @xmath114 favored by our analysis of current data . secondly , the constraint @xmath115 at @xmath102 is tight enough that this combination of data can distinguish between the cosmological constant and tracker models of dark energy . thus , weak lensing will allow us to rule out a portion of the most interesting region of parameter space . on the other hand , our forecast dark energy constraints do not reach the third milestone , which calls for these @xmath2 uncertainties to be reduced to @xmath116 . in particular , the uncertainty in @xmath77 is higher than this by over an order of magnitude . we have not included in our analysis either weak lensing tomography or galaxy cluster measurements , which may lower statistical uncertainties even further @xcite . these have been analyzed by others and look promising . they suggest that we can approach the third milestone , but there is no method suggested so far for pushing substantially beyond their forecast constraints . moreover , we remind the reader that the constraints discussed above , as well as the improvement due to weak lensing , are based on several optimistic assumptions listed at the beginning of sec . [ subsec : forecasts ] . systematic effects , such as dust - related dimming of the supernovae , can increase uncertainties in @xmath76 and @xmath77 . also , if the best - fit dark energy model is near the boundary of the ( @xmath76 , @xmath77 ) parameter space , then @xmath17 parameterization dependence will further weaken our constraints on the dark energy parameters . in the worst - case scenario @xmath117 , the best fit model will be near the boundary of parameter space ; parameterization - related uncertainties in @xmath76 and @xmath22 could completely swamp any improvements . the purpose of this investigation has been to determine how well @xmath17 can be resolved using currently planned astronomical observations . the well - known challenge is that individual measurements do not constrain @xmath17 directly , but rather some functional that depends on @xmath17 , integrals of @xmath17 , and a large set of additional parameters . we have shown that the uncertainties in measuring @xmath0 and @xmath4 at @xmath5 can be reduced dramatically by the beginning of the next decade , using a combination of the highest - quality cmb , sn , and wl data . however , the remaining uncertainties , @xmath3 and @xmath7 , will not be enough to determine definitively whether dark energy is inert ( a cosmological constant ) or dynamical ( quintessence ) , unless the true value of @xmath0 differs from @xmath30 by significantly more than @xmath88 . unfortunately , many quintessence models have @xmath118 and @xmath119 . our numerical studies illustrate how the measurements combine to produce this constraint . even the best supernova measurements are degenerate under certain combinations of variations in @xmath21 , @xmath22 , and @xmath82 @xcite . cmb constraints are degenerate along a surface in the space spanned by @xmath21 , @xmath22 , @xmath120 , and @xmath82 . however , we have found ( fig . [ fig : fore - cmbsn ] ) that combining cmb and sn measurements effectively collapses the degeneracy in the @xmath82 direction , leaving only the degeneracy in the @xmath21 , @xmath22 plane . we have shown that the cmb contribution to this degeneracy breaking comes entirely from the @xmath52 region of the power spectrum . since even the first year of wmap data has reduced uncertainties in this range to near the cosmic variance level , we do not expect a significant improvement in dark energy constraints from further wmap or planck cmb data . of course , this relies on our choice and number of cosmological parameters , and in particular , our assumption that @xmath121 . if curvature were considered , the angular diameter distance degeneracy would degrade the information on @xmath82 provided by the first cmb acoustic peak , and better information on subsequent peaks could prove valuable . also , we have not considered nonlinear effects such as the gravitational lensing of the cmb power spectra @xcite , which could provide valuable information on structure formation . meanwhile , the final constraints are sensitive to the supernova systematic uncertainties . a @xmath122 increase in @xmath123 , from @xmath124 to @xmath125 , changes the @xmath2 constraints on @xmath21 by about @xmath126 . after the cmb and sn data sets are combined , the remaining degeneracy runs along curves of @xmath17 which intersect one another near @xmath127 . the uncertainty in @xmath128 is roughly @xmath129 at the @xmath2 level , depending on what functional forms for @xmath17 are considered . to break the degeneracy between @xmath21 and @xmath22 , more data must be co - added that can constrain @xmath17 at a greater red shift . we have studied the weak lensing power spectrum as a means of breaking the @xmath21-@xmath22 degeneracy . our previous conclusion about the highest cmb multipoles remains valid ; cmb data at @xmath108 do not contribute to dark energy constraints even when weak lensing is considered . furthermore , any improvements due to weak lensing depend crucially on our ability to use shear measurements on nonlinear scales . this will require a better understanding of systematics , such as intrinsic alignments of galaxies , as well as more accurate computations of the matter power spectrum on nonlinear scales in dynamical dark energy cosmologies . when cmb , sn , and wl are combined , the supernovae constrain @xmath17 at low red shifts @xmath130 , while weak lensing constrains @xmath17 at higher red shifts ( see fig . [ fig:2_sigma_w_vs_z ] ) , leading to improvements of a factor of @xmath109 in @xmath76 and a factor of @xmath110 in @xmath77 . * distinguish between @xmath131 and @xmath113 . cmb and sn alone are unable to reach this milestone , while the combination of cmb , sn , and wl data can distinguish between these equations of state at the @xmath132 level . * distinguish between @xmath12 and tracker models with @xmath91 , @xmath133 . once again , cmb and sn alone are unable to reach this milestone . with wl added , @xmath11 and trackers can be distinguished at the @xmath2 level . * reduce @xmath2 uncertainties in @xmath0 and @xmath4 to @xmath134 . even with cmb , sn , and wl data , this milestone remains unreached . thus the combination of cmb , sn ia , and weak lensing is a promising tool for improving dark energy constraints . however , in the worst - case scenario that experiments find @xmath135 , these three probes can not decisively rule out quintessence models in which @xmath0 differs from @xmath30 by a few percent . other works have considered highly ambitious surveys of the cluster abundance evolution @xcite , assuming large numbers of observed clusters , or of weak lensing tomography @xcite , based on multiple red shift bins and observations of large fractions of the sky . both probes measure the structure growth rate as a function of red shift . these have the potential to improve statistical uncertainties significantly , by factors of 3 - 5 compared with our results , but concerns remain about systematic uncertainties @xcite . even taking the current estimated errors , uncertainties in @xmath76 and @xmath77 are several percent , which still allows a range of plausible quintessence models . thus , unless we are lucky enough to find a dark energy that is very different from the cosmological constant , new kinds of measurements or an experiment more sophisticated than those yet conceived will be needed in order to settle the dark energy issue . we thank n. bahcall , r. brustein , j. gunn , c. hirata , s. ho , e. linder , a. makarov , i. maor , j. ostriker , a. riess , u. seljak , d. spergel , c. stubbs , y. wang , and d. wesley for useful conversations . our analysis was run on a beowulf cluster at princeton university , supported in part by nsf grant ast-0216105 . this material is based upon work supported under a national science foundation graduate research fellowship ( au ) . m.i . acknowledges the support of the natural sciences and engineering research council of canada . this work was supported in part by us department of energy grant de - fg02 - 91er40671 ( pjs ) . a. g. riess , _ et al . . j. * 116 * , 1009 - 1038 ( 1998 ) . s. perlmutter , _ et al . _ , astrophys . j. * 517 * , 565 - 586 ( 1999 ) . a. g. riess , _ et al . _ , astrophys . j. * 536 * , 62 ( 2000 ) . a. g. riess , _ et al . _ , astrophys . j. * 560 * , 49 - 71 ( 2001 ) . j. l. tonry , _ et al . _ , astrophys . j. * 594 * , 1 - 24 ( 2003 ) . r. a. knop , _ et al . _ , astrophys . j. * 598 * , 102 - 137 ( 2003 ) . b. j. barris , _ et al . _ , astrophys . j. * 602 * , 571 - 594 ( 2004 ) . a. g. riess , _ et al . _ , astrophys . j. * 607 * , 665 - 687 ( 2004 ) . c. l. bennett , _ et al . _ , astrophys . . ser . * 148 * , 1 ( 2003 ) . d. n. spergel , _ et al . _ , astrophys . ser . * 148 * , 175 ( 2003 ) . w. l. freedman , _ et al . _ , astrophys . j. * 553 * , 47 - 72 ( 2001 ) . m. tegmark , _ et al . _ , astrophys . j. * 606 * , 702 - 740 ( 2004 ) . m. tegmark , _ et al . d * 69 * , 103501 ( 2004 ) . p. fosalba , e. gaztanaga , and f. castander , astrophys . j. * 597 * , l89 - 92 ( 2003 ) . p. fosalba and e gaztanaga , mon . not . * 350 * , l37-l41 ( 2004 ) . r. scranton , _ et al . _ , ( submitted ) [ astro - ph/0307335 ] ( 2003 ) . n. afshordi , y .- s . loh , and m. a. strauss , phys . d 69 * 8 * , 083524 ( 2004 ) . s. boughn and r. crittenden , nature , * 427 * , 45 - 47 ( 2004 ) . n. padmanabhan , _ et al . d ( submitted ) [ astro - ph/0410360 ] ( 2004 ) . p. j. e. peebles and b. ratra , astrophys . * 325 * , l17-l20 ( 1988 ) . b. ratra and p. j. e. peebles , phys . rev . d 37 * 12 * , 3406 - 3427 ( 1988 ) . i. zlatev , l wang , and p. j. steinhardt , phys . 82 * , 896 - 899 ( 1999 ) . p. j. steinhardt , l. wang , and i. zlatev , phys . d * 59 * , 123504 ( 1999 ) . r. r. caldwell , phys . b * 545 * , 23 - 29 ( 2002 ) . s. m. carroll , m. hoffman , and m. trodden , phys.rev . d * 68 * , 023509 ( 2003 ) . v. k. onemli and r. p. woodard , class . * 19 * , 4607 ( 2002 ) . v. k. onemli and r. p. woodard , phys . d * 70 * , 107301 ( 2004 ) . t. brunier , v. k. onemli , and r. p. woodard , class . * 22 * , 59 ( 2005 ) . m. doran and j. jckel , phys . d * 66 * , 043519 ( 2002 ) . s. dedeo , [ astro - ph/0411283 ] ( 2004 ) . j. r. bond , _ et al . * 72 * , 13 - 16 ( 1994 ) . j. r. bond , g. efstathiou , and m. tegmark , mon . not . * 291 * , l33-l41 ( 1997 ) . m. zaldarriaga , d. spergel , and u. seljak , astrophys . j. * 488 * , 1 - 13 ( 1997 ) . m. white , astrophys.j . * 506 * , 495 ( 1998 ) . g. huey , _ et al . _ , phys.rev . d59 , 063005 ( 1999 ) . g. efstathiou and j. r. bond , mon . not . r. astron . soc . * 304 * 75 - 97 ( 1999 ) . i. maor , r. brustein , and p. j. steinhardt , phys.rev.lett . * 86 * , 6 - 9 ( 2001 ) . w. hu , phys . d * 65 * , 023003 ( 2002 ) . d. huterer , phys . d * 65 * , 063001 ( 2002 ) . g. aldering , _ ( submitted ) [ astro - ph/0405232 ] ( 2004 ) . [ see also http://snap.lbl.gov ] . y .- s . song and l. knox , phys . d * 70 * , 063510 ( 2004 ) . l. knox , a. albrecht , and y.s . song , [ astro - ph/0408141 ] ( 2004 ) . s. wang , _ et . d * 70 * , 123008 ( 2004 ) . d. huterer and m. s. turner , phys . d * 64 * , 123527 ( 2001 ) . k. benabed and l. van waerbeke , phys.rev . d * 70 * , 123515 ( 2004 ) . k. abazajian and s. dodelson , phys . * 91 * 041301 ( 2003 ) . a. refregier _ j. * 127 * , 3102 ( 2004 ) . a. heavens , mon . not . * 343 * 1327 ( 2003 ) . p. simon , l. j. king and p. schneider , accepted by a&a , [ astro - ph/0309032 ] ( 2003 ) . b. jain and a. taylor , phys.rev.lett . * 91 * , 141302 2003 . g. m. bernstein , b. jain , astrophys.j . * 600 * , 17 ( 2004 ) . w. jang , _ et al . _ , third statistical challenges in modern astronomy conference , 221 - 241 ( 2003 ) . b. a. bassett , p. s. corasaniti , and m. kunz , astrophys . j. lett . * 617 * , 1 ( 2004 ) . m. chevallier , d. polarski , and a. starobinsky , int . j. mod d * 10 * , 213 ( 2001 ) . e. v. linder , phys . lett . * 90 * , 091301 ( 2003 ) . u. seljak , _ et al . _ , [ astro - ph/0407372 ] ( 2004 ) . r. r. caldwell , _ et al . _ , astrophys.j . * 591 * , l75-l78 ( 2003 ) . r. r. caldwell and m. doran , phys . d * 69 * , 103517 ( 2004 ) . m. ishak , _ et al . d * 69 * , 083514 ( 2004 ) . i. tereno , _ et al . _ , astron . ( submitted ) [ astro - ph/0404317 ] ( 2004 ) . m. ishak , c. hirata , phys . d phys.rev . d * 71 * , 023002 ( 2005 ) . w. h. press , s. a. teukolsky , w. t. vetterling , and b. p. flannery , _ numerical recipes in c : the art of scientific computing _ , ( cambridge university press , cambridge , 1992 ) . m. m. phillips , astrophys . j. lett . * 413 * , l105-l108 ( 1993 ) . m. hamuy , _ et al . * 26 * , 1362 ( 1994 ) . a. g. riess , w. h. press , and r. p. kirshner , astrophys . j. * 438 * , l17 - 20 ( 1995 ) . a. g. riess , w. h. press , and r. p. kirshner , astrophys . j. * 473 * , 88 ( 1996 ) . s. perlmutter , _ et . _ , astrophys.j . * 483 * , 565 ( 1997 ) . u. seljak and m. zaldarriaga , astrophys . j. * 469 * , 437 - 444 ( 1996 ) . r. caldwell ( private communication ) . l. verde , _ et al . _ , astrophys . ser . * 148 * , 195 ( 2003 ) . g. hinshaw , _ et al . _ , . ser . * 148 * , 63 ( 2003 ) . a. kogut , _ et al . _ , astrophys 148 * , 161 ( 2003 ) . http://www.hep.upenn.edu/~max/sdsspars.html ( 2003 ) . w. m. wood - vasey , _ et al . _ , new astron . rev . * 48 * , 637 - 640 ( 2004 ) . [ see also http://home.fnal.gov/~annis/astrophys/decam/ ] . a. g. kim , _ et al . _ , mon . not . * 347 * , 909 - 920 ( 2004 ) . [ see also http://snls.in2p3.fr/conf/posters/aas203/aas04snls.v4.ppt for a recent poster ] . c. stubbs ( private communication ) . w. l. freedman , [ astro - ph/0411176 ] ( 2004 ) . r. c. smith , _ et al . . soc . * 34 * , 1232 ( 2002 ) . p. m. garnavich , _ et al . 34 * , 1233 ( 2002 ) . r. p. kirshner , _ et al . _ , am . astron . soc . meeting 202 ( 2003 ) . [ see also http://www.ctio.noao.edu/~wsne ] . g . strolger , am . meeting 205 ( 2005 ) . d. s. madgwick , _ et al . _ , astrophys . j. * 599 * , l33-l36 ( 2003 ) . r. pain , _ et al . . soc . * 34 * , 1169 ( 2002 ) . pritchet [ astro - ph/0406242 ] ( 2004 ) . m. sullivan , [ astro - ph/0410594 ] ( 2004 ) . j. r. bond , a. h. jaffe , and l. knox , astrophys . j. * 533 * , 19 ( 2000 ) . ya . b. zeldovich and r. a. sunyaev , astrophys . & space sci . , * 4 * , 301 ( 1969 ) . r. a. sunyaev and ya . b. zeldovich , comments astrophys . space phys . , * 4 * , 173 ( 1972 ) . a. blanchard and j. schneider , astron . astrophys . * 184 * , 1 ( 1987 ) . a. kashlinsky , astrophys . * 331 * , l1 ( 1988 ) . k. m. huffenberger , u. seljak , and a. makarov , phys . d * 70 * , 063002 ( 2004 ) . u. seljak and c. m. hirata , phys . d * 69 * , 043005 ( 2004 ) . c. m. hirata , d * 70 * , 103501 ( 2004 ) . lawrence and a.e . lange , bull . soc . , * 29 * , 1273 ( 1997 ) . r. j. laureijs , international astronomical union . symposium no . 216 n. kaiser , astrophys . j. * 388 * , 272 , ( 1992 ) . b. jain and u. seljak , astrophys . j. * 484 * , 560 , ( 1997 ) . n. kaiser , astrophys . j. * 498 * , 26 , ( 1998 ) . j. m. bardeen , j. r. bond , n. kaiser , and a. s. szalay , astrophys . j. , * 304 * , 15 ( 1986 ) . chung - pei ma , r.r , caldwell , paul bode and limin wang , astrophys . letter * 521 * , 1 , ( 1999 ) . r. e. smith , _ et al . _ mon . not . r. astron . soc.341 , 1311 ( 2003 ) . s. hannestad and e. mrtsell , jcap * 0409 * , 001 ( 2004 ) . y. wang and m. tegmark , phys . * 92 * , 241302 ( 2004 ) . u. alam , v. sahni , and a. a. starobinsky , jcap * 0406 * , 008 ( 2004 ) . t. r. choudhury and t. padmanabhan , astron . astrophys . * 429 * , 807 ( 2005 ) . d. rapetti , s. w. allen , and j. weller , [ astro - ph/0409574 ] ( 2004 ) . p. s. corasaniti , _ et . al . d * 70 * , 083006 ( 2004 ) . a. makarov ( private communication ) . a. slosar , u. seljak , and a. makarov , phys . d , * 69 * , 123003 ( 2004 ) . u. seljak , _ et al . _ , [ astro - ph/0406594 ] ( 2004 ) . y. wang , astrophys . j. , * 531 * , 676 ( 2000 ) . y. wang , [ astro - ph/0406635 ] ( 2004 ) . l. stman and e. mrtsell , [ astro - ph/0410501 ] ( 2004 ) . i. zlatev , l. wang , and p. j. steinhardt , phys .rev . lett . * 82 * , 896 - 899 ( 1999 ) . p. brax and j. martin , phys . b * 468 * , 40 - 45 ( 1999 ) . a. refregier , ann.rev.astron.astrophys . * 41 * , 645 ( 2003 ) . i. maor , _ et al . _ , phys.rev . d * 65 * , 123003 ( 2002 ) . virey , _ et al . d * 70 * , 121301 ( 2004 ) . e. v. linder and r. miquel , phys . d * 70 * , 123516 ( 2004 ) . m. ishak ( submitted to mon . not . r. astron . soc . ) [ astro - ph/0501594 ] ( 2005 ) . c. m. hirata and u. seljak , phys . d * 68 * , 083002 ( 2003 ) .
just a couple of years after the introduction of the concept of ( non - abelian ) gauge fields , in his classical paper utiyama @xcite already intended to interpret general relativity as a gauge theory , in which the lorentz group emerges as the gauge group of transformations taking locally defined orthonormal bases to orthonormal ones . then mller @xcite formulated general relativity as a dynamical theory of frame fields , and in this new form einstein s theory showed indeed remarkable resemblance to yang mills theories . this similarity of the two theories made it possible to import techniques developed in one field to the other ( see e.g. @xcite ) , yielded a deeper understanding of both ( see e.g. @xcite ) , or motivated the search for a proper gauge theoretic foundation of general relativity ( see e.g. @xcite ) . in particular , the concept of charge density of non - abelian gauge fields appeared to be similar to the ( ill - defined ) concept of the energy - momentum density of the gravitational field : the vanishing of the gauge - covariant divergence of the ( gauge - covariant ) non - abelian charge 4-current ( e.g. of the fermionic matter ) in itself does _ not _ yield any gauge - covariant notion of conserved charges . to overcome this difficulty , tod @xcite introduced the notion of quasi - local charges in non - abelian yang mills theories , following the analogous twistorial definition of quasi - local energy - momentum and angular momentum of penrose in general relativity . his construction is based on the use of the holomorphic _ and _ anti - holomorphic cross sections of a hermitian vector bundle . hence the original construction could give a well defined notion of charge in theories _ with ( pseudo-)unitary gauge groups _ ( or subgroups of them ) . however , the lorentz group is not such a group , thus , to be able to see the closest analogy between general relativity and proper non - abelian gauge theories as it could be possible , it would have to be considered the latter with the lorentz ( or its covering group , the @xmath0 ) gauge group . therefore , we need to generalize the notion of charge of tod to arbitrary gauge groups , or at least to the groups above . in general relativity the energy - momentum and ( relativistic ) angular momentum take the form of 2-surface integrals , like the charge in maxwell theory . however , in einstein s theory there are ten additional quantities ( and in the coupled einstein maxwell system 16 quantities ) , taking the form of 2-surface integrals on spherical cuts of future null infinity , which are _ absolutely conserved_. these are the newman penrose ( or np ) quantities @xcite . though their meaning is still not quite well understood , in the linear approximation they appear to characterize the radiation profile of the _ incoming radiation _ , while in stationary configurations they are connected with the multipole structure of the source . thus the analogy between einstein s theory and the non - abelian gauge theories raises the question if the np quantities can be introduced in yang mills theories , and whether or not they are conserved . though there is an argumentation that the analogous np quantities can not be conserved @xcite , as far as we are aware no such mathematical analysis of the question has been given . although the charges and the np type quantities are conceptually different physical quantities , technically they are similar . thus it seems natural to consider both . therefore , the aim of the present paper is twofold : first , to generalize tod s definition of quasi - local charges to arbitrary gauge group and investigate the resulting expression in the standard small sphere and large sphere limits . the second is to form the analog of the np quantities and clarify the conditions of their conservation . mostly to fix the notation , in section 2 , we review the theoretical background and recall some basic facts that we use . then , in section 3 , we introduce the general notion of the self - dual / anti - self - dual charge integrals . we will see that the notion of electric and magnetic charges can be formulated and studied in terms of these more naturally and easier than the original charge integrals of the field strength . in subsection 3.2 we summarize the key points of tod s construction for hermitian vector bundles , and its generalization for arbitrary vector bundles and connections is suggested in subsection 3.3 , using _ either _ anti - holomorphic or holomorphic cross sections of the vector bundle . here it is also shown that for globally trivializable bundles the modified construction also gives vanishing magnetic charges . in subsection 3.4 . the limit of the quasi - local charge integrals are calculated in three limits . first , the charges are associated with small spheres near a point . we show that _ in the first two non - trivial orders the electric charge integral is independent of the actual construction _ , i.e. has some universal nature . the solution of the field equations near a point with the necessary accuracy is given in appendix [ sub - a.1 ] . then we calculate the spatial infinity limit . we show that under the standard boundary conditions for the gauge fields this limit is finite and constant in time , i.e. these define _ conserved _ total charges . in the case of the vanishing total magnetic charges the total electric charges in the holomorphic and in the anti - holomorphic constructions coincide . we also calculate the null infinity limit of the general charge integrals . we show that there is _ no _ reasonable definition for the total charges at _ null infinity _ that , for semisimple gauge groups , could be conserved in general radiative configurations . there can be charge conservation only in special ( e.g. not radiative ) field configurations or for gauge groups whose lie algebra contains a non - trivial solvable ideal , provided the charges are introduced in the anti - holomorphic construction . the holomorphic construction appears naturally at _ past _ null infinity . section 4 . is devoted to the newman penrose type quantities . first we show that for semisimple gauge groups _ no _ reasonable definition yields conserved np type quantities in the generic , radiative configurations , and their conservation can be expected only in special configurations or for special gauge groups . we also calculate them in special stationary field configurations and find that they are built from the expansion coefficients of the field strength with dipole and higher moments . to be able to do these calculations we had to increase the accuracy of the asymptotic solutions of the yang mills equations of newman @xcite by two orders . these solutions are given in appendix [ sub - a.2 ] . the signature of the spacetime metric is @xmath1 , and in connection with the spin weighted spherical harmonics we use the conventions of @xcite . let @xmath2 be a real or complex vector bundle over the spacetime manifold @xmath3 with an @xmath4 dimensional real or complex vector space @xmath5 as the typical fiber , let @xmath6 , @xmath7 , be a locally defined frame field on some open @xmath8 , i.e. at each point @xmath9 the vectors @xmath10 , ... , @xmath11 span the fibre @xmath12 at @xmath13 . ( thus the capital latin indices @xmath14 are abstract indices referring to the internal fiber space , while the capital boldface latin indices are concrete name indices taking numerical values . small latin indices are abstract spacetime indices . ) the dual frame field will be denoted by @xmath15 , and hence @xmath16 and @xmath17 hold . clearly , such a frame field is not unique , and we allow the vector basis to be transformed as @xmath18 , where @xmath19 is a function on @xmath20 taking its values in some subgroup @xmath21 of @xmath22 or @xmath23 . this yields the transformation @xmath24 of the dual frame field on @xmath20 , where @xmath25 . thus we think of the set of the frames above as a principal fibre bundle @xmath26 with structure group @xmath21 , and the matrices @xmath27 provide the free action of @xmath21 on @xmath28 . let a connection @xmath29 be given on @xmath2 . this can be specified by its action on the vectors @xmath30 of the frame field , yielding the connection 1-form @xmath31 on @xmath20 . the connection is a @xmath21connection if @xmath32 , as a locally defined matrix valued 1-form , takes its values in the lie algebra @xmath33 of the lie group @xmath21 . the curvature ( or field strength ) , also in the local frame field on @xmath20 , is the @xmath33 valued 2-form @xmath34 , which satisfies the bianchi identity @xmath35 . @xmath36 is already gauge - covariant , and globally well defined . mills equations with gauge group @xmath21 and gauge - covariant source current @xmath37 are @xmath38 these equations imply , in particular , that the spacetime _ and _ gauge - covariant divergence of @xmath39 is vanishing , and hence its spacetime divergence in itself is not : @xmath40 . therefore , by taking the flux integral on some spacelike hypersurface @xmath41 , it is the @xmath42-divergence free and _ not _ gauge - covariant current @xmath43 that should be used to define charge . thus it is more convenient to use the non - gauge - covariant form @xmath44 of the field equations . in the rest of the paper we frequently use the matrix notation instead of writing out the boldface ( matrix lie algebra ) or abstract fiber indices explicitly . for example , equation ( [ eq:2.1 ] ) takes the form @xmath45)=-4 \pi j_a$ ] . if we fix a basis @xmath46 in the lie algebra @xmath47 , @xmath48 , then we can write @xmath49 and @xmath50 . we define the structure constant @xmath51 of @xmath33 in this basis in the usual way by @xmath52 . then the definitions yield @xmath53 , and that the gauge - covariant derivative of any field e.g. with components @xmath54 can also be written in the form @xmath55 . the expression between the parentheses is just the gauge - covariant derivative on the vector bundle defined via the adjoint representation of @xmath21 as an abstract lie group , i.e. in which the typical fiber is @xmath33 as its abstract lie algebra . then it is straightforward to rewrite the yang mills equations in terms of these notions . also for later use , we introduce the symmetric metric @xmath56 on the lie algebra @xmath33 . it is a direct calculation to check that @xmath57 ( which are constant functions on the domain @xmath20 of the frame field @xmath58 ) as a field on @xmath20 is constant with respect to the gauge - covariant derivative @xmath29 , too . therefore , the connection is compatible with this metric : @xmath59 . in the adjoint representation , in which the basis @xmath46 is chosen to be the structure constants @xmath60 , this metric reduces to the killing cartan metric @xmath61 on @xmath33 . @xmath62 is well known to be non - singular precisely for semisimple @xmath33 , and this is negative definite precisely for lie algebras of compact semisimple ( and hence real ) lie groups . let @xmath63 be any smooth , orientable , closed spacelike 2-surface . let us fix a newman - penrose complex null tetrad @xmath64 on @xmath63 ( i.e. @xmath65 and @xmath66 are future pointing real null vectors orthogonal to @xmath63 , @xmath67 and @xmath68 are complex null tangents of @xmath63 , and they satisfy the normalization conditions @xmath69 and @xmath70 ) , and recall that the area 2-form on @xmath63 can also be given as @xmath71}$ ] , and the area element is @xmath72 ( see . e.g. @xcite ) . note that while @xmath65 and @xmath66 are globally defined on @xmath63 , @xmath67 and @xmath68 are defined only on the local trivialization domains of the tangent bundle of @xmath63 . then we search for the electric and magnetic charges in the form @xmath73\!\!\!\!&:=\!\!\!\!&\frac{\rm i}{4\pi } \oint_{\cal s}\varepsilon^a{}_b\ast f^b{}_{aab}\bar m^am^b{\rm d } { \cal s}=\frac{1}{4\pi}\oint_{\cal s}{\rm tr}\bigl(\varepsilon(\chi_1 + \tilde\chi_1)\bigr){\rm d}{\cal s } , \label{eq:3.1.a } \\ { \cal m}\bigl[\mu\bigr]\!\!\!\!&:=\!\!\!\!&\frac{\rm i}{4\pi}\oint _ { \cal s}\mu^a{}_b f^b{}_{aab}\bar m^am^b{\rm d}{\cal s}=\frac{\rm i } { 4\pi}\oint_{\cal s}{\rm tr}\bigl(\mu(\chi_1-\tilde\chi_1)\bigr){\rm d}{\cal s},\label{eq:3.1.b } \end{aligned}\ ] ] where @xmath74 and @xmath75 are two appropriately chosen globally defined cross sections of @xmath76 , where @xmath77 is the dual bundle , @xmath78 is the dual of the field strength , where @xmath79 is the spacetime volume 4-form . in the expressions on the right we adopted the matrix notation , in which @xmath80 is just the @xmath81 component of the anti - self - dual part of the curvature , and @xmath82 is the @xmath81 component of the self - dual part of the curvature ( see e.g. @xcite ) . : @xmath83 . moreover , though @xmath84 is defined as the @xmath81 component of @xmath85 and the vectors @xmath67 and @xmath68 are only locally defined , @xmath84 is globally well defined . in fact , it can be rewritten into the form @xmath86 , where @xmath87 is the area 2-form on the timelike 2-planes orthogonal to @xmath63 . in terms of the electric and magnetic field strengths defined with respect to some timelike unit vector field @xmath88 orthogonal to @xmath63 , @xmath89 and @xmath90 , respectively , @xmath91 holds , where @xmath92 is the outward pointing unit spacelike normal of @xmath63 such that @xmath93 . ) for real bundle and connection @xmath84 and @xmath82 are complex conjugate of each other , but in the complex case these are independent . note that for @xmath94 the first integral can be rewritten into the flux integral of the trace ( in the lie algebra indices ) of the current @xmath95 on any spacelike hypersurface @xmath41 whose boundary is @xmath63 , while the second is precisely the integral of the first chern class and @xmath96 is @xmath97 ( rather than only @xmath98 ) , the proper normalization factor of the @xmath99th chern class is @xmath100 ( instead of @xmath101 ) . ] of @xmath102 , the pull back of @xmath2 to @xmath63 . thus @xmath103 $ ] and @xmath104 $ ] are interpreted as the electric and magnetic charges , respectively , surrounded by @xmath63 and defined _ with respect to the fields _ @xmath74 and @xmath75 . if we define @xmath105 , then @xmath106:={\cal e}\bigl[q\bigr]-{\rm i}{\cal m}\bigl[q \bigr]=\frac{1}{2\pi}\oint_{\cal s}{\rm tr}\bigl(q\chi_1\bigr){\rm d } { \cal s},\label{eq:3.2}\ ] ] in which only the _ anti - self - dual _ curvature component @xmath84 appears . @xmath82 appears in @xmath107:={\cal e } [ \tilde q]+{\rm i}{\cal m}[\tilde q]$ ] with @xmath108 . if the bundle @xmath2 and the connection @xmath29 are real and @xmath74 and @xmath75 are chosen to be real , then @xmath107 $ ] is simply the complex conjugate of @xmath109 $ ] . for complex bundles and connections @xmath110 $ ] and @xmath107 $ ] are independent . as we will see , it is more convenient to work with these anti - self - dual and self - dual charges than the original @xmath103 $ ] and @xmath104 $ ] . since @xmath74 and @xmath75 are still arbitrary , the number of the ` charges ' @xmath103 $ ] and @xmath104 $ ] is infinite . however , we expect that the number of the independent charges be the number of the gauge potentials , i.e. @xmath111 . thus @xmath74 and @xmath75 should be restricted so that the number of independent charges be the correct ones , and the resulting integrals define reasonable electric and magnetic charge multiplets . tod s suggestion @xcite for the quasi - local charges has the form ( [ eq:3.1.a])-([eq:3.1.b ] ) with @xmath112 for appropriately chosen cross sections @xmath113 of @xmath102 and @xmath114 of @xmath77 : suppose that @xmath2 is a _ complex _ vector bundle endowed with a non - degenerate _ fiber metric @xmath115 , and assume that the connection @xmath29 on @xmath2 is compatible with this fiber metric , @xmath116 . thus the gauge group is assumed to be some ( pseudo ) unitary group ( or a subgroup of such ) . then by means of @xmath115 the complex conjugate bundle @xmath117 can be identified with @xmath118 via the fiber map @xmath119 . finally , the section @xmath120 was considered to be anti - holomorphic and @xmath121 to be holomorphic , i.e. on the local trivialization domains to be satisfying @xmath122 here @xmath123 and @xmath124 are the standard edth operators on @xmath125 and @xmath126 and @xmath127 are of zero - spin - weight functions on the domain of the frame field ( see e.g. @xcite ) . if @xmath128 is the space of solutions @xmath120 of ( [ eq:3.3.a ] ) and @xmath129 is the space of solutions @xmath121 of ( [ eq:3.3.b ] ) , then tod defines the quasi - local charges to be the eigenvalues of the bilinear mappings @xmath130 @xmath131 $ ] and @xmath132 @xmath133 $ ] . in the generic case on @xmath134 this construction gives @xmath4 ( not necessarily real ) charges , independently of the dimension of the gauge group . the magnetic charges are identically vanishing for globally trivializable bundle @xmath102 . ( for the details see @xcite . ) our aim is to modify tod s construction in such a way that ( 1 ) the gauge group can be arbitrary , ( 2 ) the number of the electric and the magnetic charges ( at least in the generic case ) be just @xmath111 , and ( 3 ) the charges be real for real bundles and connections ; but at the same time keeping the advantageous properties of the original construction ( especially the vanishing of the magnetic charges for globally trivializable bundles ) . in principle there are several ways to ensure the correct number of charges . the first is to consider those cross sections whose components @xmath135 and @xmath136 in a _ local _ frame field @xmath6 ( as real or complex @xmath137 matrix valued functions on the domain of the frame field ) take their values in the matrix lie algebra @xmath33 . in this case , we could always write them as @xmath138 and @xmath139 for some , still arbitrary ( real or complex ) valued functions @xmath140 and @xmath141 . then it would be the ( all together @xmath142 ) functions @xmath140 and @xmath143 that would have to be restricted appropriately to depend only on @xmath142 parameters in the generic case . though conceptually this appears to be the most natural approach , it turns out that this framework is less flexible as it is more difficult to prove statements ( e.g. the vanishing of the magnetic charges for globally trivializable bundles ) than e.g. in the third one below . another strategy is to build the charge integrals from the curvature and the fields @xmath74 and @xmath75 _ in the adjoint representation _ , i.e. when the typical fiber @xmath5 of the bundle @xmath2 is @xmath33 as an abstract lie algebra , and it is the structure constants that provide the matrix representation of a basis of @xmath33 . however , the kernel of the action of the lie algebra on itself via the adjoint map is not trivial , it is the center of the lie algebra . thus this construction may work e.g. for semisimple lie algebras , but it would not yield any charge e.g. for abelian lie algebras . the third possibility is to construct the charge integrals in a vector bundle based on a representation in which the dimension @xmath4 of the fiber space is just @xmath111 ( though the representation is _ not a priori _ the adjoint one ) . since typically the dimension of the classical lie algebras grows with the _ square _ of the dimension of their defining representation , the representation space with the dimension of the lie algebra appears to provide enough room for a faithful representation . restricting the original construction of tod in this way we get the correct number of charges . for the sake of simplicity we follow this strategy , but , instead of @xmath144 and @xmath75 , we formulate our conditions in terms of @xmath145 and @xmath146 . we search for @xmath145 and @xmath146 in the form @xmath147 and @xmath148 , respectively , where @xmath120 and @xmath149 are anti - holomorphic while @xmath150 and @xmath151 are holomorphic . if the bundle and the connection are real , then @xmath120 and @xmath149 are anti - holomorphic cross sections of the _ complexified _ vector bundles @xmath152 and @xmath153 , respectively , and the tilde denotes complex conjugation . in terms of their locally defined components these conditions are @xmath154 then clearly @xmath155 is anti - holomorphic and @xmath156 is holomorphic . next let us calculate the integrand of @xmath104 $ ] . since @xmath157 , by forming total derivatives and using ( [ eq:3.4.a ] ) and ( [ eq:3.4.b ] ) , the integrand of @xmath158 $ ] is @xmath159 where @xmath160 and @xmath161 , and , for the sake of brevity , we used the matrix notation . for general anti - holomorphic @xmath145 and holomorphic @xmath146 we could not prove more . if , however , we use the special form of @xmath162 and @xmath146 given in terms of @xmath120 , @xmath149 , @xmath150 and @xmath151 and using ( [ eq:3.4.a ] ) and ( [ eq:3.4.b ] ) again , we obtain @xmath163 where @xmath164 is the intrinsic levi - civita derivative operator on @xmath63 . thus if @xmath102 is globally trivializable and hence the connection 1-form @xmath32 can be chosen to be globally defined on @xmath63 , then the right hand side of ( [ eq:3.5 ] ) is also globally defined , yielding identically vanishing magnetic charges . therefore , the modified construction also satisfies one of the minimal requirement of reasonableness . let @xmath165 denote the space of the anti - holomorphic cross sections @xmath120 of @xmath102 , and @xmath166 the space of the anti - holomorphic cross sections @xmath167 of the dual bundle @xmath168 . similarly , let @xmath169 be the space of the holomorphic cross sections @xmath150 and @xmath170 the space of the holomorphic cross sections @xmath171 . then the anti - self - dual and the self - dual charges are defined to be the eigenvalues of the @xmath172-bilinear maps @xmath173 , \label{eq:3.6.a}\\ & { } & \tilde{\cal q}:\tilde{\cal h}\times\tilde{\cal h}^\rightarrow \mathbb{c}:(\tilde\phi^a,\tilde\psi_b)\mapsto\tilde{\cal q}\bigl [ \tilde\phi^a\tilde\psi_b\bigr ] , \label{eq:3.6.b}\end{aligned}\ ] ] respectively . if @xmath2 and @xmath29 are real , then the complex conjugate of an anti - holomorphic cross section is holomorphic , and hence @xmath174 is the complex conjugate of @xmath165 . similarly , @xmath175 is the complex conjugate of @xmath166 . therefore , @xmath176 is the complex conjugate of @xmath177 , yielding real electric and magnetic charges . repeating the argumentation of tod in @xcite one can show that for globally trivializable bundles over @xmath134 in the generic case the number of the electric charges is @xmath111 . on the other hand , there are exceptional 2-surfaces for which the construction still works , but the number of charges is different . though in general @xmath178 is not an element of @xmath33 , it can be considered as the generator of _ the central extension of the matrix lie algebra _ @xmath33 . since it is both holomorphic and anti - holomorphic , it is worth considering this extension . @xmath179 $ ] and @xmath180 $ ] can be thought of as the ` mean ' electric and magnetic charge , respectively , and , as we already noted , the latter is just the first chern class of @xmath181 . though the quasi - local charge integrals @xmath109 $ ] are well defined even in curved spacetime , for the sake of simplicity in the rest of the paper we assume that the spacetime is the minkowski spacetime . in particular , we calculate their small and large sphere limits in the flat spacetime . let @xmath182 be an arbitrary point , @xmath88 a future pointing timelike unit vector at @xmath13 , and let us consider the future pointing null geodesics starting from @xmath13 with affine parameter @xmath183 and tangents @xmath65 satisfying @xmath184 at @xmath13 . then for sufficiently small @xmath183 the set @xmath185 of the points on these null geodesics whose affine distance from @xmath13 is @xmath183 is a smooth spacelike 2-surface , which is homeomorphic to @xmath186 . such a surface is called a _ small sphere about @xmath13_. our aim is to calculate the charge integral @xmath109 $ ] for small spheres of radius @xmath183 , and to determine the coefficients @xmath187 in its expansion @xmath110={\cal q}^{(0)}+\cdots+r^4{\cal q}^{(4)}+o(r^5)$ ] . ( since @xmath188 is contractible to @xmath13 , any bundle over @xmath185 is globally trivializable , and hence @xmath104 $ ] is identically vanishing . ) the strategy is to expand the curvature component @xmath189 and the solutions @xmath190 and @xmath191 of ( [ eq:3.4.a ] ) ( or of ( [ eq:3.4.b ] ) ) by powers of @xmath183 . since @xmath192 , clearly @xmath193 and @xmath194 , and to calculate @xmath109 $ ] with @xmath195 accuracy we need to know @xmath189 , @xmath190 and @xmath191 with @xmath196 accuracy . similarly , expanding the component @xmath197 of the connection 1-form as a series of @xmath183 , substituting this to ( [ eq:3.4.a ] ) and taking into account that @xmath198 , where @xmath199 is the edth operator on the unit sphere ( and is given explicitly in the appendix ) , we see that we need to know the expansion coefficients @xmath200 for @xmath201 . however , as the analysis of appendix [ sub - a.1 ] shows , to determine the solution of the yang mills equations on the light cone ( more precisely , on the null boundary of the chronological future ) of @xmath13 for @xmath84 with @xmath196 accuracy , formally @xmath202 , @xmath203 and @xmath204 must be expanded with accuracy @xmath205 . the solution of the hypersurface equations ( [ eq : a.1a])-([eq : a.1c ] ) is given in appendix [ sub - a.1 ] , by means of which the equations ( [ eq:3.4.a ] ) themselves are @xmath206 since @xmath190 and @xmath191 are multiplets of type @xmath207 scalars , @xmath208 , @xmath209 , @xmath210 and @xmath211 are _ constant _ on @xmath63 . to solve the third of ( [ eq:3.4.1a ] ) and ( [ eq:3.4.1b ] ) first let us recall that in terms of the spinor form of the field strength , @xmath212 , and the ghp spinor dyad @xmath213 the curvature component @xmath214 is given by @xmath215 . ( in the present subsection we use two - component spinors , and hence , in particular , the capital latin indices @xmath216 , @xmath217 , ... _ in this subsection _ are abstract spacetime spinor indices . these should not be confused with the indices referring to the internal fiber space or the concrete name indices @xmath218 , @xmath219 , ... etc . ) however , in terms of the cartesian spinor dyad @xmath220 at @xmath13 the ghp dyad is @xmath221 and hence it is straightforward to integrate ( [ eq:3.4.1a ] ) and ( [ eq:3.4.1b ] ) in the coordinates @xmath222 : @xmath223 here @xmath224 , @xmath225 , @xmath226 ; and @xmath227 and @xmath228 are arbitrary constant sections , being the general solution of the homogeneous equation . then , using the pattern of ( [ eq:3.4.2a])-([eq:3.4.2b ] ) , it is easy to write down the holomorphic solutions , too . though for small spheres the solution spaces @xmath165 and @xmath229 are @xmath111 dimensional , the space of the @xmath230 accurate solutions is @xmath231 dimensional . the presence of the ` extra ' , spurious solutions is a consequence of the lack of a natural isomorphism between the solution spaces on _ different _ two - surfaces , and hence , in particular , with different radii . these yield some ambiguity in higher accurate approximations . ( for a more detailed discussion of this issue in general relativity , see e.g. @xcite . also , in the calculation of certain integrals , we will use equation ( 2.5 ) of @xcite . ) since @xmath232 and @xmath210 are constant , by ( [ eq : a.1.3a ] ) the @xmath233 order term @xmath234 is vanishing . for the @xmath235 order term we obtain @xmath236 where the integration is taken on the unit sphere . thus @xmath237 is the volume of the 3-ball of radius @xmath183 times the timelike component of the gauge covariant current at @xmath13 . this result coincides with that of tod , too . in fact , let us observe that in @xmath238 only the zeroth order terms of the cross sections @xmath190 and @xmath191 matter ( the others integrate to zero ) , which are constant both for the holomorphic and anti - holomorphic sections . in the fourth order we obtain @xmath239 though formally @xmath240 is built from the @xmath233 accurate cross sections , their @xmath233 order correction terms are integrate to zero . hence we obtain the same result in the holomorphic and the anti - holomorphic constructions . on the other hand , in general @xmath241 _ does _ depend on the @xmath242 order ` gauge solutions ' @xmath209 and @xmath211 : @xmath240 is well defined only if @xmath243 , i.e. when @xmath244 at @xmath13 . the calculation of @xmath245 is much longer and technically more difficult , so we summarize only the main message of that . first , ambiguities similar to that in @xmath240 , parameterized by the ` extra ' solutions @xmath209 , @xmath211 , ... etc , appear , unless the third and fourth order terms , @xmath238 and @xmath246 , are vanishing . the difference of the holomorphic and anti - holomorphic constructions can appear in this order . thus @xmath247 and @xmath240 have some ` universal ' nature , not being sensitive to the details of the defining equations for the cross sections @xmath190 and @xmath191 , provided @xmath190 and @xmath191 are constant in the zeroth and first orders . let @xmath248 be the standard spherical , complex stereographic coordinate system ( at least in a neighbourhood of spatial infinity ) in minkowski spacetime , and consider the charge integral @xmath109 $ ] on the metric 2-spheres @xmath185 of radius @xmath183 in some @xmath249 spacelike hyperplane . let @xmath88 and @xmath92 be their future pointing unit timelike and outward pointing unit spacelike normals , respectively , such that @xmath88 is orthogonal to the @xmath249 hyperplanes and @xmath93 . let @xmath67 and @xmath68 be the complex null tangents of the 2-spheres . then we impose the following a priori fall - off conditions for the scalar potential and the radial and tangential parts of the spatial vector potential : @xmath250 , @xmath251 , @xmath252 , @xmath253 , respectively , for some @xmath254 . in addition , we assume that the time derivative of the last three satisfy @xmath255 , @xmath256 and @xmath257 also for some @xmath258 . by technical reasons we also assume that the differentiation with respect to @xmath183 reduces these orders by one , but the differentiation with respect to @xmath259 and @xmath260 does not change these orders . then , expressing @xmath261 by the potentials , we find that @xmath262 if @xmath263 and @xmath264 . therefore , these boundary conditions ensure the existence of the @xmath265 limit of the charge integrals @xmath109 $ ] on 2-spheres @xmath185 of radius @xmath183 provided @xmath266 . ( if @xmath267 were not involved in the 2-surface integral of the curvature , i.e. if that were only the 2-surface integral of the _ components of the field strength _ in some globally defined frame @xmath6 , then one could rewrite that as the 3-volume integral of its divergence on some spacelike hypersurface @xmath41 with boundary @xmath63 . then , using the constraint equation , one could show that even weaker boundary conditions could ensure the finiteness of the original 2-surface integral @xcite . since , however , @xmath267 is defined _ only _ on @xmath63 , here this strategy can not be followed , and hence we have the above ` conservative ' fall - off conditions . ) next let us determine the conditions that ensure the vanishing of the @xmath265 limit of the time derivative of @xmath109 $ ] , @xmath268=\frac{1}{2\pi}\oint{\rm tr}\bigl(\dot q\chi_1+q\frac{1}{2}\bigl(\dot e_a+{\rm i}\dot b_a\bigr ) v^a\bigr){\rm d}{\cal s}. \label{eq:3.4.5}\ ] ] to determine the order of the second term of the integrand we use the evolution equations for the field strengths in the present geometrical situation ( i.e. flat spacetime foliated by spacelike 3-planes with vanishing shift vector ) . these are @xmath269 , \hskip 20pt \dot b_a=-\bigl(\mathbb{d}_ce_d\bigr)\varepsilon^{cd}{}_{ab}t^b- \bigl[\phi , b_a\bigr],\ ] ] where @xmath270 denotes the space- and _ spatial _ gauge - covariant derivative operator ( i.e. for example its action on any cross section @xmath271 of @xmath2 near the spatial infinity is @xmath272 and @xmath273 is the orthogonal projection to the @xmath249 hyperplanes ) . by these equations @xmath274 if @xmath275 . thus if @xmath276 and @xmath277 , then the @xmath265 limit of the time derivative of @xmath109 $ ] is zero . note that by @xmath275 the asymptotic form of @xmath278 ( and of @xmath279 , too ) is @xmath280 i.e. its leading order term does _ not _ depend on @xmath281 . finally , let us write @xmath282 and substitute into the equation of the anti - holomorphicity of the cross section @xmath267 . we find @xmath283= \\ ^{(0)},q^{(0)}\bigr]\bigr)+\frac{1}{r^{1+\epsilon}}\bigl({}_0 { { \mbox{\symbol{'360}}}}q^{(1)}+\bigl[\gamma^{(0)},q^{(1)}\bigr]+\bigl[\gamma^{(1 ) } , q^{(0)}\bigr]\bigr)+o\bigl(r^{-1-\epsilon}\bigr).\end{aligned}\ ] ] thus its solution is such that the leading term @xmath284 does not depend on time , and hence @xmath285 . therefore , under the fall - off conditions for the vector potential above , the spatial infinity limit of @xmath109 $ ] exists and is conserved in time . it might be interesting to note that here we used only the anti - holomorphicity of @xmath145 , but did not assume that it has the dyadic product structure @xmath286 . note that in general @xmath287 can not be transformed by gauge transformations to zero _ globally _ unless the leading term of the @xmath92 component of the magnetic field strength is vanishing . thus @xmath284 is _ not _ constant on the unit sphere , and hence the total charge depends on whether @xmath267 is anti - holomorphic or holomorphic . if the connection coefficients @xmath287 and @xmath288 can be made vanishing globally , then the total magnetic charges are vanishing and the total electric charges introduced in the holomorphic and anti - holomorphic constructions coincide . let us consider the charge integral @xmath109 $ ] on any @xmath289 cut of @xmath290 for still _ arbitrary _ globally defined @xmath145 . ( for a summary of the relevant results on the asymptotic structure of the yang mills fields near @xmath290 , see appendix [ sub - a.2 ] . ) because of the peeling this integral can be written in terms of the expansion coefficient @xmath291 as @xmath292 , where the integration is taken on the unit sphere . hence this integral is still a function of the retarded time coordinate . in this subsection we show that for semisimple gauge groups _ no _ reasonable definition yields conserved total charge in general , radiative configurations . in particular , neither tod s original definition ( as pointed out in @xcite ) nor the modified one yield _ conserved _ total charges . by integration by parts and using ( [ eq : a.2.1.b ] ) , for the @xmath293-derivative of @xmath109 $ ] we obtain @xmath294=\frac{1}{2\pi}\oint{\rm tr}\bigl(\dot q \chi^0_1-\bigl({}_0{{\mbox{\symbol{'360}}}}q+\bigl[\gamma^0,q\bigr]\bigr)\chi^0_2 \bigr){\rm d}{\cal s}_1 . \label{eq:3.4.7}\ ] ] since in any given retarded time instant @xmath295 in a generic ( radiative ) configuration @xmath291 and @xmath296 are independent data ( because @xmath297 and @xmath298 can be specified independently , see appendix [ sub - a.2.1 ] ) , the integrand can be zero precisely when the coefficient of @xmath299 and of @xmath296 are vanishing . in particular , @xmath145 must be anti - holomorphic , but in addition it would have to be constant in time even though the anti - holomorphic structure might be changing from cut to cut on @xmath290 : @xmath300=0 . \label{eq:3.4.8}\ ] ] the condition of the integrability of the system of these equations is @xmath301=0 $ ] . thus the charge @xmath109 $ ] with anti - holomorphic @xmath145 is conserved in the absence of outgoing radiation , @xmath302 , or when @xmath267 is anti - holomorphic and belongs to the center of the lie algebra @xmath33 , too . therefore , for semisimple gauge groups in the generic radiative configurations there is _ no _ total conserved charge . charge conservation requires the non - triviality of the radical ( i.e. the maximal solvable subalgebra ) of @xmath33 in its levi malcev decomposition . remarkably enough , though in the presence of outgoing radiation there is no charge conservation , the structure of the yang mills fields at _ future _ null infinity singles out just the anti - holomorphic cross sections @xmath145 in a natural way . thus our suggestion for the quasi - local charges is justified by the conservation properties of the total charges at future null infinity . at past null infinity the holomorphic cross sections would appear . even for semisimple lie algebras and non - stationary , but special configurations , in which @xmath303 and @xmath297 are not quite independent , the total charge can be conserved . such are the linard wiechert type solutions . in @xcite tod defined them geometrically as the algebraically special solutions of the yang mills equations for which the field strength components @xmath304 and @xmath305 are vanishing , and , in contrast to our present framework , the origin of the coordinate system @xmath306 ( i.e. the ` source s world line ' ) is allowed to be an arbitrary timelike curve . he showed that there is an anti - holomorphic frame in which @xmath291 is constant both in time and on the cut of @xmath290 , and hence that the total charge is , in fact , conserved . ( for another approach to the linard wiechert type configurations , based on certain assumptions on the explicit form of the 4-potential in terms of a pointlike source , the so - called colour , see @xcite , and for a recent review of them see e.g. @xcite ) . in the special stationary case discussed in appendix [ sub - a.2.2 ] @xmath307 and @xmath308 are vanishing and @xmath309 for some real or complex _ constants _ @xmath310 . thus , in particular , the anti - holomorphicity of @xmath145 yields that it is constant , i.e. @xmath311 for some constants @xmath312 . therefore , @xmath109=c^\mu e^a_{\mu b}q^b{}_a = c^\alpha g _ { \alpha\beta}q^\beta$ ] , yielding the meaning of the constants @xmath310 of the solution : they should represent the total charge of the system . in fact , if we search for @xmath145 in the form @xmath286 for anti - holomorphic @xmath120 and @xmath149 , then their anti - holomorphicity yields that they are constant , and hence that @xmath109=c^\mu e^a _ { \mu b}\phi^b\psi_a$ ] . thus , indeed , the charges introduced in subsection [ sub-3.3 ] are the eigenvalues of @xmath313 . the newman penrose conserved quantities in general relativity and in maxwell theory ( and , in fact , in any theory based on a linear zero - rest - mass field equation with any spin or on the conformally invariant wave equation ) are built from the _ second _ nontrivial expansion coefficient of the tetrad component of the curvature with the _ highest spin - weight _ , weighted with appropriate spin - weighted spherical harmonics @xcite . thus it is natural to define @xmath314:=\oint{\rm tr}\bigl(\omega\chi^1_0\bigr ) { \rm d}{\cal s}_1 , \label{eq:4.1}\ ] ] where @xmath315 is a cross section of ( in the case of real bundles , the complexified ) @xmath316 over the @xmath295 cut of @xmath290 with @xmath317 spin weight so that the integrand of ( [ eq:4.1 ] ) is globally well defined . here we show that for semisimple gauge group and _ any _ non - trivial choice for @xmath315 the quantity @xmath318 $ ] is _ not _ conserved in generic , radiative configurations . using ( [ eq : a.2.1.d ] ) a straightforward calculation ( by integration by parts ) yields @xmath319:=\oint{\rm tr}\bigl\{\dot\omega\chi^1_0 -\bigl({}_0{{\mbox{\symbol{'360}}}}\bigl({}_0{{\mbox{\symbol{'360}}}}'\omega+\bigl[\tilde\gamma^0,\omega \bigr]\bigr)+\bigl[\gamma^0,\bigl({}_0{{\mbox{\symbol{'360}}}}'\omega+\bigl[\tilde \gamma^0,\omega\bigr]\bigr)\bigr]-3\bigl[\omega,\chi^0_1\bigr]\bigr ) \chi^0_0\bigr\}{\rm d}{\cal s}_1 . \label{eq:4.2}\ ] ] since @xmath320 and @xmath321 are freely and independently specifiable functions at @xmath295 ( see appendix [ sub - a.2.1 ] ) , @xmath318 $ ] is constant in time precisely when their coefficients in ( [ eq:4.2 ] ) are vanishing : @xmath322\bigr)+\bigl[\gamma^0,\bigl({}_0{{\mbox{\symbol{'360}}}}'\omega+\bigl[\tilde \gamma^0,\omega\bigr]\bigr)\bigr]=3\bigl[\omega,\chi^0_1\bigr ] . \label{eq:4.3}\end{aligned}\ ] ] using ( [ eq : a.2.1.a])-([eq : a.2.1.b ] ) of the appendix , the integrability condition of this system of equations is @xmath323 \bigr)\bigr]+\bigl[\tilde\chi^0_2,\bigl({}_0{{\mbox{\symbol{'360}}}}'\omega+\bigl [ \tilde\gamma^0,\omega\bigr]\bigr)\bigr]=2\bigl[\bigl({}_0{{\mbox{\symbol{'360}}}}\chi ^0_2+\bigl[\gamma^0,\chi^0_2\bigr]\bigr),\omega\bigr].\ ] ] thus , in particular , @xmath318 $ ] is conserved if @xmath315 satisfies ( [ eq:4.3 ] ) and there is no outgoing radiation , i.e. @xmath324 ; or if @xmath315 belongs to the center of the lie algebra @xmath33 and solves @xmath325=0 $ ] . therefore , the np type quantities could be conserved in special situations ( e.g. for a collection of independent maxwell fields ) , but not conserved in general . however , this non - conservation of the np quantities in non - abelian yang mills theories appears to be known for a while , as bazanski pointed out to newman in an unpublished private communication @xcite . since for the linard wiechert type solutions of the yang mills equations , considered in @xcite , the curvature components @xmath304 and @xmath305 are identically vanishing , the np quantities for these solutions are vanishing for any gauge group . it is known that in stationary configurations the np quantities in the theories based on the _ linear _ zero - rest - mass field equations ( e.g. in particular in maxwell theory or the linearized einstein theory ) are vanishing . on the other hand , in stationary spacetime the gravitational np quantities of the full non - linear theory reduce to a nontrivial combination of the mass and the dipole and quadrupole moments of the source @xcite . thus their non - vanishing is due to the nonlinearity of the theory . this result raises the question whether the analogous quantities @xmath318 $ ] in non - abelian yang mills theories could be non - vanishing in the stationary configurations , and if they could , then what would be their meaning ? in the special stationary case discussed in subsection [ sub - a.2.2 ] equation ( [ eq : a.2.1.e ] ) reduces to @xmath326-\bigl[{}_0{{\mbox{\symbol{'360}}}}\chi^0_0,\tilde\chi^0_0\bigr]-2\bigl[\chi^0_0 , { } _ 0{{\mbox{\symbol{'360}}}}\tilde\chi^0_0\bigr]-3\bigl[\chi^0_0,{}_0{{\mbox{\symbol{'360}}}}'\chi^0_0 \bigr],\ ] ] while the defining equation ( [ eq:4.3 ] ) of @xmath315 to @xmath327 . \label{eq:4.4}\ ] ] writing @xmath328 for some spin weight @xmath317 functions @xmath329 and expanding the latter as @xmath330 , equation ( [ eq:4.4 ] ) yields @xmath331 thus @xmath332 can always be non - zero , and if @xmath333 , the ` charge matrix ' in the adjoint representation , has only one eigenvector with zero eigenvalue then @xmath334 for arbitrary constants @xmath335 . if , in addition to @xmath310 , there are other eigenvectors with zero eigenvalue , say @xmath336 , ... etc , then @xmath332 will be a linear combination of all them with arbitrary coefficients @xmath335 , @xmath337 , ... etc . however , @xmath338 for @xmath339 can be nonzero only in the exceptional case when @xmath340 is an eigenvalue of @xmath333 . ( for a more detailed discussion of @xmath333 see subsection [ sub - a.2.2 ] . ) substituting these into the definition of @xmath341 $ ] , by integration by parts we obtain @xmath314=-\frac{1}{2}\oint\omega^\mu g_{\mu\nu}c^\nu _ { \alpha\beta}\bigl(4c^\alpha k^\beta+\bigl({}_0{{\mbox{\symbol{'360}}}}h^\alpha\bigr ) \tilde h^\beta+2h^\alpha\bigl({}_0{{\mbox{\symbol{'360}}}}\tilde h^\beta\bigr)+3h ^\alpha\bigl({}_0{{\mbox{\symbol{'360}}}}'h^\beta\bigr)\bigr){\rm d}{\cal s}_1 , \label{eq:4.6}\ ] ] where the functions @xmath342 , @xmath343 and @xmath344 have been introduced via @xmath345 , @xmath346 and @xmath347 , respectively . thus , in particular , if the lie algebra @xmath33 is abelian , then by ( [ eq:4.6 ] ) @xmath318=0 $ ] , in accordance with the known fact that the np quantities are vanishing for ( an abelian multiplet of ) stationary maxwell fields . thus in the rest of the paper we assume that @xmath33 is not abelian . first suppose that @xmath333 is generic ( see appendix [ sub - a.2.3 ] ) . then the only non - zero expansion coefficients of @xmath329 in terms of spin weighted spherical harmonics is @xmath348 , while the expansion coefficients of @xmath342 and @xmath344 ( and in the complex case of @xmath343 , too ) are determined in appendix [ sub - a.2.3 ] . substituting those solutions into equation ( [ eq:4.6 ] ) , using the notation of the appendix and elementary properties of the edth operators and the spin weighted spherical harmonics ( e.g. the orthogonality of these harmonics with different indices ) , we obtain @xmath314=-\frac{1}{2}\oint\omega^\mu g_{\mu\nu}\bigl ( \sum_{m=-1}^1\bigl(c^\nu{}_\alpha k^\alpha_{1m}+h^\nu_{1m}\bigr)\,{}_1 y_{1m}\bigr){\rm d}{\cal s}_1.\ ] ] in the special case when the only eigenvector of @xmath333 with zero eigenvalue is @xmath310 and @xmath349 is not an eigenvalue of @xmath333 this reduces to zero . thus the non - abelian nature of the lie algebra @xmath33 in itself is _ not _ enough to have non - trivial np quantities . if , however , @xmath349 is an eigenvalue and @xmath350 are the corresponding eigenvectors , then @xmath318 $ ] can be non - zero . in the other extreme case when @xmath351 ( e.g. when the total charge is vanishing ) it gives @xmath352 , which is a _ homogeneous quadratic expression of the coefficients @xmath353 and @xmath354 characterizing the dipole structure of the yang mills field_. we saw in appendix [ sub - a.2.3 ] that for the lie algebras @xmath355 , @xmath356 , and @xmath357 the charge matrix has only one eigenvector with zero eigenvalue while for @xmath358 there are two , moreover there could be exceptional configurations for the last three lie algebras , but not for @xmath355 . therefore , for the lie algebra @xmath355 in general and for @xmath356 , @xmath359 and @xmath358 in the generic case the np quantities can be non - vanishing e.g. when the total charge is zero . finally we consider the exceptional configurations of appendix [ sub - a.2.3 ] . in this case the solution of ( [ eq:4.5 ] ) is @xmath360 . substituting this and the solution of ( [ eq : a.2.11 ] ) into ( [ eq:4.6 ] ) , and using the orthonormality of the spin weighted spherical harmonics on the unit sphere , we obtain that @xmath314=\frac{1}{4}g_{\mu\nu}\bigl(\sum_{m=-1}^1 ( -)^m\omega^\mu_{1-m}\bigl(2c^\mu{}_\alpha k^\alpha_{1m}+k^\nu_{1 m } \bigr)+\sum_{m =- j}^j(-)^m\omega^\mu_{j - m}\bigl(2c^\mu{}_\alpha k ^\alpha_{jm}+k^\nu_{jm}\bigr)\bigr),\ ] ] which can also be non - zero in general , even if the charge matrix has a single eigenvector with zero eigenvalue . in this case , however , @xmath318 $ ] is built from _ higher _ , actually the @xmath361th order multipole moments @xmath362 and @xmath363 of the curvature . in this appendix we summarize the structure of the yang mills fields near a point ( subsection [ sub - a.1 ] ) and near the future null infinity in minkowski spacetime ( subsection [ sub - a.2 ] ) . for the generalization to the coupled einstein mills equations in the newman penrose formalism , see @xcite . following @xcite we use the coordinates @xmath306 , where @xmath364 is the standard retarded null coordinate and @xmath365 are the complex stereographic coordinates on the unit sphere . the complex null newman penrose tetrad @xmath64 is adapted to the @xmath366 , @xmath367 2-spheres , the edth operators on the unit metric sphere will be denoted by @xmath368 and @xmath369 and their action on a scalar @xmath370 of ghp type @xmath371 is @xmath372 and @xmath373 , respectively . the components of the connection 1-form @xmath374 in the np basis are denoted by @xmath375 , @xmath204 , @xmath202 and @xmath203 , respectively . the components of the curvature are @xmath376 , @xmath377 , @xmath378 , @xmath379 , @xmath380 and @xmath381 . note that for real connections @xmath375 and @xmath204 are real and @xmath382 is the complex conjugate of @xmath202 . similarly , also for real connections , @xmath82 , @xmath383 and @xmath384 are complex conjugate of @xmath385 , @xmath84 and @xmath386 , respectively , but for general complex connections they are independent . in the gauge @xmath387 of newman @xcite the expression of the components of the field strength by the components of the connection 1-forms and the field equations ( [ eq:2.1 ] ) together with the bianchi identities take the form @xmath388 , \label{eq : a.1b } \\ \frac{\partial\chi_1}{\partial r}\!\!\!\!&+\!\!\!\!&\frac{2}{r}\chi_1 -\frac{1}{r}{}_0{{\mbox{\symbol{'360}}}}'\chi_0=-\bigl[\chi_0,\gamma_{10'}\bigr]+ 2\pi j_{00 ' } , \label{eq : a.1c } \\ \frac{\partial\chi_2}{\partial r}\!\!\!\!&+\!\!\!\!&\frac{1}{r}\chi_2 -\frac{1}{r}{}_0{{\mbox{\symbol{'360}}}}'\chi_1=-\bigl[\chi_1,\gamma_{10'}\bigr]+ 2\pi j_{10 ' } , \label{eq : a.1d } \\ \frac{\partial\gamma_{10'}}{\partial u}\!\!\!\!&=\!\!\!\!&-\chi_2 + \frac{1}{2}\frac{\partial\gamma_{10'}}{\partial r}+\frac{1}{2r } \gamma_{10'}+\frac{1}{r}{}_0{{\mbox{\symbol{'360}}}}'\gamma_{11'}+\bigl[\gamma_{10 ' } , \gamma_{11'}\bigr ] , \label{eq : a.1e } \\ \frac{\partial\chi_0}{\partial u}\!\!\!\!&=\!\!\!\!&\frac{1}{2 } \frac{\partial\chi_0}{\partial r}+\frac{1}{2r}\chi_0+\frac{1}{r } { } _ 0{{\mbox{\symbol{'360}}}}\chi_1+\bigl[\chi_0,\gamma_{11'}\bigr]-\bigl[\chi_1 , \gamma_{01'}\bigr]-2\pi j_{01 ' } , \label{eq : a.1f } \\ \frac{\partial\chi_1}{\partial u}\!\!\!\!&=\!\!\!\!&\frac{1}{2 } \frac{\partial\chi_1}{\partial r}+\frac{1}{r}\chi_1+\frac{1}{r } { } _ 0{{\mbox{\symbol{'360}}}}\chi_2+\bigl[\chi_1,\gamma_{11'}\bigr]-\bigl[\chi_2 , \gamma_{01'}\bigr]-2\pi j_{11'}. \label{eq : a.1 g } \end{aligned}\ ] ] the first four are only the constraint ( or ` hypersurface ' ) equations , while the remaining three are the evolution equations . in the complex case there are analogous equations for the independent curvature components @xmath305 , @xmath82 and @xmath383 , too . in the small sphere limit calculations of subsection [ sub-3.4.1 ] we need to know only @xmath84 and @xmath202 on the light cone of the point @xmath182 . thus we need to solve only ( [ eq : a.1a])-([eq : a.1c ] ) with the accuracy required by the small sphere calculations . hence we expand the components of the curvature and the connection 1-form as @xmath389 and @xmath390 , respectively , as well as @xmath391 . then the solution of ( [ eq : a.1a ] ) and its tilded version is equations ( [ eq : a.1.1a])-([eq : a.1.2 ] ) give all the components of the connection 1-form in terms of the curvature , except @xmath394 , which remains undetermined . this , however , can be made zero by an appropriate gauge transformation . the various components of the curvature are not independent either , as they are restricted by ( [ eq : a.1c ] ) and ( [ eq : a.1d ] ) . since we need only @xmath84 , we evaluate only the former . using ( [ eq : a.1.1a ] ) it gives the difference of these and their tilded version reproduces the expressions for @xmath396 , ... etc obtained from ( [ eq : a.1b ] ) and its tilded version . the expressions for the expansion coefficients of @xmath397 can be derived from ( [ eq : a.1d ] ) in an analogous way . in @xcite newman solved the source free yang mills equations near the future null infinity , and determined the freely specifiable functions of such solutions . in subsection [ sub - a.2.1 ] we present the results of the analogous analysis with the higher accuracy of the asymptotic expansion required by the analysis of section [ sec-4 ] . in subsection [ sub - a.2.2 ] the asymptotic field equations are solved in certain special stationary configurations . in subsection [ sub - a.2.3 ] we discuss some examples with specific gauge groups . @xmath400\bigr ) , \label{eq : a.2.1.b } \\ & { } & \dot\chi^0_0={}_0{{\mbox{\symbol{'360}}}}\chi^0_1+\bigl[\gamma^0,\chi^0_1\bigr ] , \label{eq : a.2.1.c } \\ & { } & \dot\chi^1_0=-{}_0{{\mbox{\symbol{'360}}}}'\bigl({}_0{{\mbox{\symbol{'360}}}}\chi^0_0+\bigl[\gamma ^0,\chi^0_0\bigr]\bigr)-\bigl[\tilde\gamma^0,{}_0{{\mbox{\symbol{'360}}}}\chi^0_0 + \bigl[\gamma^0,\chi^0_0\bigr]\bigr]+3\bigl[\chi^0_1,\chi^0_0\bigr ] , \label{eq : a.2.1.d } \\ & { } & 2\dot\chi^2_0=-{}_0{{\mbox{\symbol{'360}}}}\bigl({}_0{{\mbox{\symbol{'360}}}}'\chi^1_0+\bigl[\tilde \gamma^0,\chi^1_0\bigr]+\bigl[\chi^0_0,\tilde\chi^0_0\bigr]\bigr)- \bigl[\gamma^0,{}_0{{\mbox{\symbol{'360}}}}'\chi^1_0+\bigl[\tilde\gamma^0,\chi^1_0 \bigr]+\bigl[\chi^0_0,\tilde\chi^0_0\bigr]\bigr]- \nonumber \\ & { } & \quad -3\chi^1_0+\bigl[\chi^1_0,3\chi^0_1 + 2\tilde\chi^0_1\bigr]- \bigl[\chi^0_0,3\bigl({}_0{{\mbox{\symbol{'360}}}}'\chi^0_0+\bigl[\tilde\gamma^0,\chi ^0_0\bigr]\bigr)+\bigl({}_0{{\mbox{\symbol{'360}}}}\tilde\chi^0_0+\bigl[\gamma^0 , \tilde\chi^0_0\bigr]\bigr)\bigr ] ; \label{eq : a.2.1.e}\end{aligned}\ ] ] and there are the analogous ones for @xmath401 , @xmath402 , ... etc obtained from ( [ eq : a.2.1.a])-([eq : a.2.1.e ] ) by formally taking their complex conjugate . ( the dot denotes derivative with respect to the retarded time coordinate @xmath293 . ) here @xmath403 is an arbitrary function ( and , for complex bundle and connection , @xmath404 is also ) , but the functions @xmath405 , @xmath406 , ( and , in the complex case , @xmath407 , too ) can be specified freely only at a given retarded time instant @xmath295 . their time evolution is governed by ( [ eq : a.2.1.c])-([eq : a.2.1.e ] ) ( and the analogous tilded equations ) . similarly , the sum @xmath297 can be freely specified at @xmath295 , but the difference @xmath408 is already determined by @xmath307 and @xmath308 via it is a simple calculation to check that the evolution equation ( [ eq : a.2.1.b ] ) and its tilded version preserve ( [ eq : a.2.2 ] ) . geometrically @xmath307 and @xmath308 are the leading ( @xmath410 order ) term in the asymptotic expansion of @xmath202 and @xmath203 , respectively , and they are analogous to the asymptotic shear in general relativity , while @xmath296 and @xmath401 represent the outgoing radiation . in terms of these the expansion coefficients @xmath411 , @xmath412 , @xmath413 and @xmath414 are given by @xmath415\bigr ) , \label{eq : a.2.3.a } \\ & { } & \chi^2_1=-\frac{1}{2}\bigl({}_0{{\mbox{\symbol{'360}}}}'\chi^1_0+\bigl[\tilde \gamma^0,\chi^1_0\bigr]+\bigl[\chi^0_0,\tilde\chi^0_0\bigr]\bigr ) , \label{eq : a.2.3.b } \\ & { } & \chi^1_2=-\bigl({}_0{{\mbox{\symbol{'360}}}}'\chi^0_1+\bigl[\tilde\gamma^0,\chi^0 _ 1\bigr]\bigr ) , \label{eq : a.2.3.c } \\ & { } & \chi^2_2=\frac{1}{2}\bigl({}_0{{\mbox{\symbol{'360}}}}'\bigl({}_0{{\mbox{\symbol{'360}}}}'\chi^0_0 + \bigl[\tilde\gamma^0,\chi^0_0\bigr]\bigr)+\bigl[\tilde\gamma^0,{}_0 { { \mbox{\symbol{'360}}}}'\chi^0_0+\bigl[\tilde\gamma^0,\chi^0_0\bigr]\bigr]-\bigl [ \chi^0_1,\tilde\chi^0_0\bigr]\bigr ) , \label{eq : a.2.3.d}\end{aligned}\ ] ] let the yang mills field be stationary with respect to @xmath293 , i.e. all the dot - derivatives are vanishing . newman and penrose showed @xcite that in the analogous gravitational case the asymptotic shear can be made zero by an appropriate supertranslation of the origin cut . however , now the integrability condition of the _ local _ existence of a gauge transformation @xmath421 taking @xmath307 and @xmath308 to zero is the vanishing of the corresponding curvature , which is proportional to the right hand side of ( [ eq : a.2.2 ] ) . hence , for example , in the presence of magnetic charges ( see equation ( [ eq:3.1.b ] ) ) there is no such a gauge transformation . thus in the rest of this subsection we consider those special stationary solutions of the asymptotic field equations in which @xmath422 and @xmath423 can be ensured even _ globally _ , and hence @xmath424 holds . thus the total magnetic charges are vanishing . ( if the connection is real , then @xmath291 is also real . ) then equation ( [ eq : a.2.1.c ] ) immediately gives that @xmath291 is _ constant _ on @xmath63 , and hence has the form @xmath425 for some constants @xmath310 . these constants are real for real bundles and connections , while in the complex case they are complex but @xmath426 . we saw in subsection [ sub-3.4.3 ] that the constants @xmath310 represent the total charges . thus the expansion coefficients @xmath432 must be eigenvectors of @xmath333 with the _ real non - positive _ eigenvalues @xmath437 . by the definition of the charge matrix @xmath438 holds , i.e. @xmath439 is always an eigenvalue with eigenvector @xmath310 . thus , in particular , @xmath440 can always be non - zero , and if @xmath310 is the only eigenvector of @xmath333 with zero eigenvalue then @xmath441 for some constants @xmath442 . if the dimension of the kernel space of @xmath333 is @xmath443 , then @xmath440 will be the linear combination of the vectors of a basis of this kernel with arbitrary constants @xmath442 , @xmath444 , ... etc . however , since @xmath333 is a finite matrix , it may have only finitely many eigenvalues , and hence only finitely many coefficients @xmath432 can be non - zero . nevertheless , since the total ( electric ) charges @xmath310 can be specified freely , in general no eigenvalues of @xmath333 has the form @xmath445 for some @xmath339 . therefore , apart from the exceptional cases in which the charges have this structure , the expansion coefficient functions @xmath303 and @xmath446 are combinations only of the spherical harmonics @xmath447 and @xmath448 , respectively . thus the general solution of the homogeneous equation contains non - trivial @xmath452 for @xmath454 only if the matrix @xmath455 is chosen to be very special . this solution depends only on finitely many parameters . a particular solution of the inhomogeneous equation ( [ eq : a.2.6 ] ) can also be found by using the spin weighted spherical harmonics and the expansion of their product , @xmath456 , by the appropriate clebsch gordan coefficients and the spherical harmonics @xmath457 , @xmath458 . thus , to summarize , in the present special stationary case the asymptotic field equations can be integrated completely and these solutions ( with the second order accuracy , i.e. keeping only @xmath459 , @xmath460 and @xmath291 ) can be indexed by the freely specifiable total ( electric ) charges @xmath310 and the expansion coefficients @xmath440 and @xmath461 ( and , in the complex case , by @xmath462 and @xmath463 , too ) . other parameters may occur in the solutions ( viz . in the expansion coefficients @xmath464 , @xmath460 , etc ) _ only _ when the charges are chosen in very specific ways . first consider the generic case , i.e. when no non - vanishing eigenvalue of @xmath333 has the form @xmath465 for some @xmath466 . then by ( [ eq : a.2.5 ] ) we have that @xmath467 and @xmath468 , and then , using the expressions how the edth operators act on the spin weighted spherical harmonics , ( [ eq : a.2.6 ] ) reduces to its second line can be written as @xmath470 , where @xmath471 , @xmath472 , are explicitly given , homogeneous quadratic expressions of @xmath440 and @xmath462 , and are built also from the structure constants and the clebsch gordan coefficients . with this substitution we obtain that moreover , if @xmath333 has an eigenvalue @xmath474 for some @xmath475 then @xmath476 are its eigenvectors while @xmath477 for all @xmath478 ; otherwise the coefficients @xmath452 are vanishing for all @xmath479 . the first of ( [ eq : a.2.9 ] ) can be inverted to find the unique solution @xmath350 precisely when @xmath349 is not an eigenvalue of @xmath333 , otherwise the corresponding eigenvector can be freely added to the particular solution . similarly , @xmath480 is determined only up to the general solution of the homogeneous equation @xmath481 . if the only eigenvector of @xmath333 with zero eigenvalue is the charge vector @xmath310 , then @xmath441 and @xmath482 , which imply that @xmath483 . ( for the sake of simplicity we assume that the structure constants are real . ) hence @xmath350 is either an arbitrary combination of the eigenvectors of @xmath333 with eigenvalue @xmath349 or zero , while @xmath480 is a solution of @xmath481 . if , in addition to @xmath310 , @xmath336 is another eigenvector of @xmath333 with zero eigenvalue , then @xmath484 , implying that @xmath471 are proportional to @xmath485 , i.e. @xmath483 again . in the other extreme special case when the charge matrix is vanishing ( e.g. if @xmath486 ) @xmath487 holds , @xmath480 are arbitrary , but now @xmath488 must hold . this yields restrictions on the coefficients @xmath440 and @xmath462 . the evaluation of the explicit formulae show that then @xmath489 , but @xmath490 are still non - zero . to illustrate the above possibilities , let us specify the lie algebra @xmath33 . first it is chosen to be the abstract lie algebra @xmath491 , @xmath355 ( i.e. @xmath492 ) or @xmath357 ; and let us choose the basis @xmath493 such that @xmath494=\mp e_1 $ ] , @xmath495=e_2 $ ] and @xmath496=e_3 $ ] , where the upper sign corresponds to @xmath356 and the lower to @xmath355 and @xmath357 . in this basis the components of the killing cartan metric are @xmath497 . the eigenvalues of @xmath498 are @xmath439 , in which case the corresponding eigenvector is proportional to the total charge vector @xmath310 , and @xmath499 . the non - zero eigenvalues can be real for @xmath500 if @xmath501 and for @xmath502 also in special cases ( e.g. for purely imaginary @xmath310 ) , but never for @xmath503 . thus the configurations for @xmath503 are always generic . from the point of view of general relativity the most interesting choice for @xmath33 is the abstract lorentz lie algebra @xmath358 . we choose its basis @xmath504 to be the standard one , i.e. in which @xmath493 is the basis above for @xmath355 while @xmath505 , @xmath506 and @xmath507 are the ` boost generators ' such that @xmath508=0 $ ] , @xmath509=e_6 $ ] , ... etc . in this basis the killing cartan metric is @xmath510 . one eigenvalue of the charge matrix is zero with eigenvectors @xmath310 and @xmath511 ( as a column vector ) . clearly , @xmath512 . the ( square of the ) non - zero eigenvalues are @xmath513 . thus @xmath514 is real precisely when @xmath515 and @xmath516 , and , in contrast to @xmath355 , for special values of @xmath310 the non - zero eigenvalues could have the form @xmath517 or @xmath518 for some @xmath466 . where @xmath440 and @xmath362 are arbitrary combinations of the eigenvectors of @xmath333 with zero and @xmath522 eigenvalue , respectively . if we substitute this and the analogous expression for @xmath343 into ( [ eq : a.2.6 ] ) , use elementary properties of the edth operators and the spin weighted spherical harmonics and write the products of such harmonics as combinations of spherical harmonics with clebsch gordan coefficients , we obtain here @xmath524 can be given explicitly in terms of @xmath525 , @xmath526 , @xmath362 , @xmath363 , the structure constants and clebsch gordan coefficients . since we assumed that @xmath333 has non - zero eigenvalues , it can not be identically zero , and , for the sake of simplicity , let us assume that @xmath455 does not have more than two eigenvectors with zero eigenvalue . then @xmath527 ( see the paragraph following equation ( [ eq : a.2.9 ] ) ) . then , evaluating ( [ eq : a.2.11 ] ) , we find that @xmath477 for @xmath528 , @xmath452 are uniquely determined for @xmath529 , and @xmath461 are determined up to the addition of the general solution of the homogeneous equation @xmath530 . the authors are grateful to ted newman and paul tod for the helpful discussions on various aspects of the problem , and to stanislaw bazanski and piotr chrusciel for their remarks on the np quantities and the spatial infinity limit of the charges , respectively , in yang mills theories . this work was partially supported by the hungarian scientific research fund ( otka ) grant k67790 . f. w. hehl , j. d. mccrea , e. w. mielke , y. neeman , metric - affine gauge theory of gravity : field equations , noether identities , world spinors , and breaking of dilation invariance , phys . rep . 285 * 1 - 171 ( 1995 )
a topological insulating ( ti ) phase is a new state of quantum matter featuring massless dirac - like boundary states whose robustness is guaranteed by the time - reversal symmetry ( trs)@xcite . in this rapidly growing field , search for new topological phases has recently turned to extending the consideration of symmetry - protected states to include non - spatial symmetries@xcite such as , the particle - hole and chiral symmetries as well as the spatial symmetries@xcite . in particular , the nontrivial band topology can be shown to be protected by certain crystal symmetries , leading to the new class of tis called topological crystalline insulators ( tcis)@xcite . like the tis , a three dimensional ( 3d ) tci is also predicted to host metallic surface states on surfaces which preserve appropriate crystal symmetries@xcite . to date , the semiconducting pb@xmath0sn@xmath1(te , se ) alloys are the only experimentally realized tci materials , whose topological nature has been verified through direct observation of dirac - like surface states via angle - resolved photoemission spectroscopy ( arpes ) experiments@xcite . these gapless surface states are protected by the mirror symmetry with respect to the ( 110 ) or ( @xmath510 ) lattice plane . the characteristic features of the associated dirac states such as linear dispersion , lifshitz transition , spin / orbital texture , among others , have been examined via scanning tunneling microscopy / spectroscopy ( stm / sts ) and transport measurements@xcite . moreover , it has been found that , below a critical temperature , which depends on chemical composition , the cubic lattice structure can be distorted by strains , resulting in an orthorhombic or a rhombohedral structure@xcite . remarkably , in a recent stm experiment by okada _ et al._@xcite , two unexpected landau levels have been observed for the ( 001 ) surface of pb@xmath0sn@xmath1se under a perpendicular magnetic field . the presence of these two extra levels is believed to be associated with the gap opening of the two surface dirac cones induced by a ferroelectric - like lattice distortion , which breaks the corresponding mirror symmetry . it is clear that it is important to understand effects of strains on the electronic structure of tcis in order to gain a deeper handle on the nature and origin of their topological states , and how these states could be manipulated for practical applications of the tcis . despite some strain - related studies in the literature@xcite , we are not aware of a systematic investigation of how spin - textures and other properties of dirac states in the tcis evolve under various symmetry breaking strains . so motivated , here we examine strain effects on the ( 001 ) surface of the pb@xmath0sn@xmath1(te , se ) tcis . our analysis is based on an effective four - band model , which is shown to capture all essential features of the topological surface states . the form of strain related perturbations on the hamiltonian is clarified through general symmetry considerations , allowing us to delineate how characteristic features of the topological states in tcis evolve under strains . significantly , our analysis shows that certain mirror - symmetry - breaking strains induce hedgehog spin texture with out - of - the - plane spin - polarization at the dirac point . such a spin texture of dirac states has been reported previously in a 3d ti , but only in the presence of an exchange field@xcite . our findings thus identify a possible new pathway for realizing tci - based spintronics devices without requiring a trs breaking field@xcite . moreover , we show that the interplay of strains and applied zeeman field can be used to tune the chern number of the surface ground state , which is a topological invariant characterizing a quantum anomalous hall ( qah ) or a quantum hall insulator . electron - electron interactions effects are ubiquitous in condensed matter systems , especially when charge screening is relatively poor as is the case generally for dirac - like states in graphene@xcite or tis and tcis.@xcite in this connection , we discuss possible symmetry breaking orders generated under short - range repulsion @xmath6 as a first step toward understanding correlation driven effects on dirac states in the presence of applied strains . we numerically obtain the zero - temperature phase diagram as a function of @xmath6 and the particle density @xmath7 and delineate its evolution with increasing strength of strain . our results not only give insight into the nature of competing orders , but also build the foundation for what may be called `` straintronics '' applications driven by electron correlations . the paper is organized as follows . in sec . ii , we briefly discuss the effective four - band model for the ( 001 ) surface states in the low - energy regime , and its extensions for addressing strain effects in tcis . iii turns to consider strain effects on properties of the dirac states , including the interplay between applied strains and perpendicular zeeman fields . these results then allow us to investigate in section iv the possible interaction - driven competing orders in dirac states in the absence as well as presence of strains . finally , section v comments on potential applications and implications of our study , and concludes with a summary of our results . we start by reviewing the four - band @xmath2 model for the ( 001 ) surface states in a tci developed previously in ref . , and discuss its generalization to account for effects of strains . on the ( 001 ) surface , the low - energy surface states can be viewed as two sets of interacting coaxial dirac cones , originating from the interface between the inverted bands of the tci and the vacuum : one set is centered at @xmath3 and the other at @xmath4 in the surface brillouin zone ( sbz ) , see fig . [ fig : lattice ] ( b ) . defining [ 110 ] , [ @xmath510 ] , and [ 001 ] as @xmath8 , @xmath9 , and @xmath10 directions [ see fig . [ fig : lattice ] ( a ) ] , respectively , the effective model @xmath11 around @xmath3 must obey the following three essential symmetries which leave @xmath3 invariant : the mirror reflection about the @xmath12-plane , the mirror reflection about the @xmath13-plane , and time - reversal symmetry with the corresponding symmetry operations , i.e. m_xzh_\|x(k_x , k_y)m_xz^-1=h_\|x(k_x ,- k_y ) , + m_yzh_\|x(k_x , k_y)m_yz^-1=h_\|x(-k_x , k_y ) , + th_\|x(k_x , k_y)t^-1=h_\|x(-k_x ,- k_y ) , where @xmath14 is measured from @xmath3 . informed by the orbital characters of surface states as revealed by first - principles calculations,@xcite one may choose , for instance in the case of snte , @xmath15 , @xmath16 , latexmath:[$ @xmath11 because one of the interacting dirac cones is @xmath18-orbital ( sn ) rich , while the other is @xmath19-orbital ( te ) rich . as a result , the symmetry operations can be represented by the matrices , @xmath20 , @xmath21 , and @xmath22 , where @xmath23 with the pauli matrices @xmath24 and @xmath25 acting on orbital and spin spaces , respectively , and @xmath26 denotes complex conjugation . these considerations restrict @xmath11 to the form h_\|x&= & m_30+m^_22+k_x(v_1x_02+v_2x _ 20+v_3x_32 ) + & + & k_y(-v_1y_01+v_2y_13-v_3y_31),[h ] [ eq : surfaceh ] up to first order in @xmath27 . ( @xmath9 ) denotes the direction along [ 110 ] ( [ @xmath510 ] ) . ( b ) surface brillouin zone of snte . without strain , there are four gapless dirac cones , protected by the mirror symmetry either with respect to @xmath28 or @xmath29.,width=302 ] for @xmath30=0 and @xmath31=@xmath32=@xmath33=@xmath34=0 , @xmath11 becomes block diagonal , reflecting the structure of the two underlying parent dirac cones with two associated dirac points , @xmath35 ( @xmath18-rich ) and @xmath36 ( @xmath19-rich ) at @xmath3 [ see fig . [ fig : b1 ] ( a ) ] . a non - vanishing value of any of the parameters @xmath37 , @xmath31 , @xmath32 , @xmath33 and @xmath34 induces interaction between the two cones , leading to two immediate consequences : ( 1 ) the original dirac points are shifted to @xmath38 ; and ( 2 ) all degeneracies along the intersection of the two cones are lifted except for the two points of a time - reversal symmetric pair on the @xmath28 line , indicating the emergence of two child dirac cones protected by the mirror reflection about the @xmath12-plane [ see fig . [ fig : lattice ] ( b ) ] ; hereafter , referred to as the low - energy dirac points associated with @xmath39=0 . our effective four - band @xmath2 model correctly capture two key features of surface states of tcis . firstly , for @xmath40 topology of the constant energy contours changes from two separate fermi circles at low energies to two concentric ellipses at high energies . this lifshitz transition@xcite indicates the presence of van hove singularities ( vhss ) in the underlying electronic spectrum , which are located along the @xmath41 line at e_vhs+ = \|m^\|,= ( 0 , ) , if @xmath31 , @xmath32 , @xmath33 , and @xmath34 are neglected . a similar situation arises for @xmath42 with @xmath43 . in all cases , we have a logarithmically diverging density of states proportional to @xmath44 with @xmath45 and @xmath46 representing an energy scale away from the vhs and a cutoff energy for surface states , respectively . secondly , there is no out - of - the - plane spin component [ fig . [ fig : b1 ] ( b)-(c ) ] by using symmetry arguments in that the net out - of - the - plane spin polarization , @xmath47 , yielding @xmath48 . we turn next to generalize our four - band effective model to include low - energy states throughout the sbz in the presence of inter - cone interactions . this can be done by noticing that the surface states in the vicinity of @xmath3 and @xmath4 are related by @xmath49 symmetry , so that the effective model around @xmath4 can be obtained explicitly by a @xmath49 rotation , h_\|y(k_x , k_y)=_4 h_\|x(k_y ,- k_x)_4 ^ -1 , where @xmath50 and the @xmath19 orbital for te atoms in the basis states is now replaced by @xmath51 orbital in @xmath52 . the total surface hamiltonian then is h_(001)=h_\|x(k_x , k_y)h_\|y(k_x , k_y ) . although @xmath53 describes an even number of dirac cones at low energies like the case of a weak ti@xcite , note that the four dirac points in a tci do not locate at time - reversal invariant momenta . this key difference can lead to a rather different phase diagram in the presence of the electron - electron interactions@xcite in a tci , as discussed in sec . we now discuss effects on the surface states due to strains , which could be either intrinsic or extrinsic . since the gapless surface states in the pb@xmath0sn@xmath1(te , se ) tcis are mainly protected by mirror symmetries , a perturbation , which breaks one of these symmetries can be expected to at least partly gap the surface spectrum , and modify the spin textures around the associated ( massive ) dirac points . although this is generally true , we will see below that this is not always the case due to other symmetry considerations . for this purpose , we will carry out a systematic analysis of strain induced effects along the lines of sec . ii above . a general strain can be described in terms of a symmetric strain tensor @xmath54 , where @xmath55 denotes the displacement field and @xmath56 are chosen to coincide with the principal crystal axes as defined in sec . any effective , strain - induced perturbation term corresponding to @xmath57 will transform in the same way as the strain tensor under time - reversal and mirror symmetries . by examining the transformation properties of @xmath58 up to first order in @xmath27 , the possible resulting perturbations around @xmath3 are listed in table [ t1 ] . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] as expected , the strain induced perturbations in the first four columns of table i can open gaps at both low - energy dirac points along @xmath28 due to @xmath59 broken symmetry . more significantly , the resulting spin texture becomes hedgehog - like at low energies , similar to the spin reorientation phenomenon found on the surfaces of manganese - doped bi@xmath60se@xmath61 thin films@xcite . as a concrete example , let us add the perturbation @xmath62 in eq.([eq : surfaceh ] ) , where @xmath63 denotes the electron - phonon coupling strength . clearly , in fig . [ fig : f1](a ) , a gap opens at the dirac point with its magnitude proportional to @xmath64 . the spin - texture of the conduction band of the massive dirac cone is also depicted in fig . [ fig : f1](b ) . as the absolute value of the eigenenergy @xmath65 becomes smaller ( _ i.e. _ , closer to the bottom of the upper cone ) , the induced out - of - the - plane spin component , _ 03= , [ eq : sz ] becomes larger , where we have set @xmath66 for simplicity . several points should be noted here as follows . firstly , since the strain perturbation does not break trs , the spin texture for the two massive cones around @xmath3 must form time - reversed partners . this can be inferred from eq . ( [ eq : sz ] ) by changing the sign of @xmath67 . secondly , the out - of - the - plane spin component of the lower cone is basically opposite to that of the upper cone at the same @xmath27 . finally , the whole spin texture resulting from the strain differs from the case where the hedgehog - like texture is induced by the perpendicular zeeman field when both massive dirac cones around @xmath3 are considered , see fig . [ fig : f1](c ) . other strain perturbations in table [ t1 ] , which break either the @xmath13-mirror symmetry or both the mirror symmetries , neither open a gap at the dirac point nor induce an out - of - the - plane spin component . the robustness of this gapless dirac point originates from an underlying symmetry : the former type of perturbation is due to the presence of @xmath59 symmetry , while the latter type ( shear deformation ) is protected by a `` space - time '' symmetry , @xmath68 , with a rotation @xmath69 followed by a time - reversal operation . the space - time protection now allows persistence of gapless dirac points sitting at generic @xmath70 points away from the mirror line @xmath28 . this special feature has been noted previously in refs . , and it leads to a new type of tci@xcite . vanishing out - of - the - plane spin polarization can be proven via symmetry arguments . for instance , consider the surface states around @xmath3 under the strain perturbation @xmath71 . the corresponding hamiltonian now reads : @xmath72 . assuming @xmath73 is an eigenstate of @xmath74 , the @xmath12-mirror symmetry guarantees that 0=\|[_02,_\|x]\|=2iv_1yk_y \|_03\| . thus , the out - of - the - plane spin component @xmath75 vanishes everywhere around @xmath3 . other perturbations can be analyzed in a similar manner . there other cases deserve some comment . the first two cases involve a uniform expansion and a uniaxial stretch ( @xmath49 breaking ) , which correspond to @xmath76 and @xmath77 , respectively@xcite . because both these cases respect trs as well as the mirror symmetries , their net effect is only to renormalize the parameters in the original @xmath11 . consequently , these perturbations do not open a gap at dirac points , although positions of the dirac points could shift in opposite directions along the @xmath28 line . the third case is an experimentally observed ferroelectric - like distortion@xcite , in which two kinds of atoms are displaced along a certain direction in an opposite manner . denoting the displacement vector @xmath78 , the non - vanishing component of this distortion , @xmath79 ( @xmath80 ) , preserves the mirror symmetry with respect to the principal axis @xmath8 ( @xmath9 ) , but breaks the rotation and the mirror symmetries along a perpendicular direction . to the zeroth order in @xmath70 around @xmath3 , the perturbation due to such distortion can be straightforwardly shown to have the form : h_f^= _ fxd_x_10+_fyd_y_23 , where @xmath81 and @xmath82 denote the coupling strengths . in fact , the former term is similar to the effect of strain @xmath83 , preserving the gapless dirac points , while the latter term is similar to the effect of strain @xmath84 , resulting in opening a gap instead . with a fixed zeeman field , where @xmath85 . the strength of the strain perturbation is given by @xmath86 ( a ) 0 , ( b ) 0.03 , ( c ) 0.05 , and ( d ) 0.08 . from ( b ) to ( d ) , one of the dirac cones around @xmath4 closes and reopens again due to a band inversion . the model parameters are the same as those used in fig . [ fig : b1].,width=264 ] like a perpendicular zeeman field , the strain field on the surface of a tci may not only give rise to spin reorientation , but it may also induce a charge gap in the dirac cones . however , there are key differences in the effects of strain and zeeman fields . an applied zeeman field , which respects @xmath69 rotational symmetry around the @xmath10-axis , opens gaps with the _ same _ sign in the pair of dirac cones around @xmath3 , while an applied strain field , which breaks @xmath69 rotational symmetry , would induce gaps with _ opposite _ sign . this observation provides the foundation for tenability of the chern number via the interplay of applied strain and zeeman fields , suggested in ref . using a simplified two - band model for each dirac point . as a concrete example based on our four - band model , consider an applied zeeman field along the @xmath10 direction along with a strain field which breaks @xmath87 but preserves @xmath59 around @xmath4 on the ( 001 ) surface . from symmetry considerations , the strain and zeeman fields around @xmath3 and @xmath4 become coupled as h_\|x & + & h_z_03 , + h_\|y & + & h_z_03+v_s_23 , where @xmath88 ( @xmath89 ) is the field strength of zeeman ( strain ) field . note that for a given @xmath88 , all four low - energy dirac points open gaps in the absence of an applied strain [ see fig . [ fig : biv](a ) ] . when a non - vanishing @xmath90 is introduced , the gaps of the two massive dirac cones around @xmath4 evolve in an opposite manner in the sense that one cone increases while the other decreases . this is consistent with a picture in which the strain field induces out - of - the - plane spin components around the two massive dirac cones which form time - reversed partners with respect to their spin textures [ see fig . [ fig : f1](b ) ] , and therefore , respond oppositely to the existing @xmath88 field . with increasing strength of @xmath89 , the gap in the dirac cone with a decreasing gap continues to decrease further , becomes gapless at @xmath91 , and reopens inverted again , as shown in figs . [ fig : biv](b)-(d ) . this band inversion indicates that the system undergoes a topological phase transition . the topological nature of a ground state with broken time - reversal symmetry can be characterized by the chern number . a non - zero value of the chern number , @xmath92 , given by @xmath93}{(e_n - e_m)^2},\end{aligned}\ ] ] where the summation over momentum space is implicit , indicates a finite hall conductance @xmath94 , which can be obtained by integrating the berry curvature of the wave functions of the occupied bands over the momentum space@xcite . [ fig : chern ] presents the chern number of the system as a function of the zeeman field strength @xmath88 and the strain strength @xmath89 within the framework of our four - band model . in the presence of a zeeman field , by varying the strength of strain from small to large values is seen to drive the system from @xmath95 to @xmath96 , demonstrating the tunability of the system to the strain field . note that in our consideration here we have implicitly assumed that the sample is thick enough so that the hybridization between the top and bottom surfaces of the sample can be neglected ; with @xmath97 the bottom surface will contribute another chern number @xmath95 , which is not shown in fig . [ fig : chern ] for simplicity . and model parameters are fixed as : @xmath98=-2 , @xmath37=-1 , @xmath99=-1 , @xmath100=-1.5 , with other parameters taken to be zero . the phase boundary ( blue line ) is given by @xmath101 . ] in discussing electron - electron interactions , we consider short - range repulsive interactions between electrons on the ( 001 ) surface of a tci , and focus on delineating how the phase diagram evolves in the presence of a strain , which breaks one of the mirror symmetries , say , @xmath87 , based on the the interacting hamiltonian : @xmath102 where ^()&=&[c^_\|x , p_z(),c^_\|x , p_z ( ) , c^_\|x , p_x(),c^_\|x , p_x ( ) , + & & c^_\|y , p_z(),c^_\|y , p_z ( ) , c^_\|y , p_y(),c^_\|y , p_y ( ) ] , [ eq : basis ] with the subscript @xmath103 ( @xmath104 ) denoting the momentum point involved in the @xmath70 expansion . for @xmath53 , we have added the strain perturbation and hence , @xmath105 . the interaction term , @xmath106 ; @xmath107 and @xmath108 denote repulsive contact interactions between like and unlike orbitals , respectively , with _ u & = & ud^2 r_=p_z , p_x , p_y _ ( ) _ ( ) , + _ v & = & d^2 r_^ _ ( ) _ ^ ( ) , [ hint ] where @xmath109 ( @xmath110 , @xmath111 ) [ @xmath112 are defined in eq . ( [ eq : ftc ] ) below ] . since we are treating only the low - energy , long - wavelength physics , it is reasonable to consider field operators @xmath113 , which vary slowly on the scale of the lattice constant . this can be done by fourier transforming the operators : _ , s ( ) = _ e^i c_,s ( ) = e^i c_,s ( ) , [ eq : ftc ] where @xmath114 , expanding from the origin ( 0,0 ) . in terms of these field and density operators , and the identities given in the appendix , the full hamiltonian @xmath115 can be rewritten as @xmath116 where @xmath117 has the same form as eq . ( [ eq : basis ] ) except that the @xmath70-dependence is now replaced by @xmath118 . in order to determine the ground state of the system as a function of the interaction strength and fermion density , we employ the self - consistent mean - field ( mf ) approach , which is expected to be reliable as long as the interaction strength is much smaller than the bulk band gap . taking @xmath119 as a representative case , we decouple the @xmath120 into bilinear fermion terms . after some straightforward but tedious algebra , we obtain 52 order parameters as well as 12 renormalized band parameters as detailed in the appendix , which can be generally expressed as o_=^ _ [ eq : order ] where a summation over momentum and occupied states is implicit , and @xmath121 , with the pauli matrices , @xmath122 , acting on @xmath123 pseudo - spin ( valley ) space . note that , here we consider only the symmetry breaking orders in the particle - hole channel : superconductivity in the particle - particle channel due to weak onsite repulsion can only be achieved via beyond mf treatment , usually with exponentially small transition temperature@xcite . in this connection , we define the `` particle density '' at a given chemical potential @xmath124 as the density deviation from the total particle density where @xmath125 , namely , n=_k < k_^e()= -_k < k_^e_0()=0 , [ eq : density ] where @xmath126 ( @xmath127 ) is the eigenenergy of the non - interacting ( mf - decoupled interacting ) system ; @xmath128 is a cutoff momentum , which is chosen such that our results are insensitive to its magnitude . note that by any induced valley `` polarization '' , @xmath129 , we mean @xmath130 that is , the `` valley polarization '' due to external strain in the absence of interactions is subtracted in defining @xmath129 . versus @xmath7 phase diagram without strain . @xmath131 denotes the particle density at @xmath132 . solid and dashed lines mark phase boundaries of continuous and abrupt phase transitions , respectively . ] for a given @xmath6 and @xmath7 , the ground state can then be determined by minimizing the mf free energy with respect to various order parameters using the form detailed in the appendix . we numerically solve the resulting set of coupled equations self - consistently to obtain the zero - temperature @xmath6-@xmath7 phase diagrams both with and without the applied strain . note that , in order to make the underlying physics more transparent , we use the following model parameters in computations : @xmath133 , @xmath134 , @xmath135 , @xmath136 , and a cutoff energy , @xmath137 , which determines @xmath128 . however , we expect our results to be generic and relevant more generally , detailed effects of material - specific parameters notwithstanding . we begin by considering a system without strain , _ i.e. _ , @xmath138 . in the order parameter space we explored in this case , the leading orders ( aside from the metal phase ) are basically the ferromagnetic state ( fm ) , @xmath139 , and the @xmath18-orbital rich @xmath140 charge density wave ( @xmath18-cdw ) : = _ ( -1)^x+y[n_p_z,()+n_p_z , ( ) ] , where the length scale is in units of the lattice constant . the zero - temperature @xmath6-@xmath7 phase diagram is shown in fig . [ fig : pd1 ] . the @xmath18-cdw phase corresponds to a non - vanishing @xmath141 ( the leading piece ) , and the fm phase represents a non - vanishing value of @xmath139 ( the leading piece ) . finally , in the metal phase , all symmetries are preserved with vanishing values of all 52 order parameters . at @xmath142 , the `` @xmath18-cdw '' phase appears when the interaction strength @xmath6 is larger than the critical value @xmath143 . this phase involves a finite value of @xmath141 ; it is associated with broken translation symmetry , while the trs and both the mirror symmetries remain intact [ see schematic fig . [ fig : realspace](b ) ] , so that each dirac cone remains gapless . the existence of a critical value of @xmath6 can be understood by noticing that in the non - interacting limit , there are only four dirac points with zero density of states at the fermi level . the necessity of a threshold value of @xmath6 has also been predicted theoretically in 2d systems with linear energy dispersion , such as graphene and the surfaces of 3d weak tis@xcite . upon electron doping , the system assumes either the fm phase [ see fig . [ fig : realspace](a ) ] at relatively high particle densities , or the @xmath18-cdw phase [ see fig . [ fig : realspace](b ) ] at low particle densities . the fm phase is associated with broken trs as well as both the mirror symmetries simultaneously , resulting in gaps to open up at all the dirac points , although the corresponding ground state is still metallic with finite @xmath124 . note that the critical @xmath6 , above which the system enters into the gapless @xmath18-cdw phase , is less than the critical value for the fm phase , indicating a favorable free energy gain compared to the competing fm phase at low doping . -orbital rich , @xmath140 charge density wave state ( @xmath18-cdw ) . different colors label the two distinct atoms involved ; arrows depict the net out - of - the - plane spin polarization.,width=302 ] we further note several points in connection with the phase diagram of fig . [ fig : pd1 ] as follows : ( 1 ) the spin density wave ( sdw ) phase , as anticipated usually in dirac systems like graphene , does not occur here at @xmath142 for large @xmath6 . this could be attributed to the non - trivial orbital textures and strong spin - orbit coupling in our case ; ( 2 ) we find that the dynamically generated strain - like order , which could gap the spectrum , is relatively disfavored as the system tries to preserve the combined @xmath144 symmetry , although such a phase becomes favored in the presence of external strain ; ( 3 ) if we suppress inter - valley scattering , our preliminary results suggest that the cdw phase is also suppressed@xcite ; and finally , ( 4 ) as we pointed out already in sec . ii above , the density of states at an energy scale @xmath45 near a van hove singularity diverges like @xmath145 , indicating propensity for the occurrence of more symmetry breaking orders ( including those in the particle - particle channel ) . the present mean - field treatment would then be inappropriate , even in the weak - coupling limit . here we consider an applied strain with a moderate coupling strength , @xmath146 [ of order @xmath147 ) ] , which breaks the @xmath13-mirror symmetry . the resulting zero - temperature phase diagram is shown in fig . [ fig : pd2 ] . in contrast to the case without strain , a coexisting phase lying between the @xmath18-cdw and fm phases is now seen to emerge . suppression of the @xmath18-cdw phase with strain can be anticipated on physical grounds . the reason is that any @xmath140 order requires good nesting between the valleys , but the strain weakens the nesting tendency by opening gaps at dirac points along the @xmath13-mirror line . notably , in the strained system , spin polarization in the fm phase is mainly contributed by the @xmath19 orbital due to the broken mirror symmetry , which lifts the @xmath51 orbital to higher energies . versus @xmath7 phase diagram in the presence of a moderate strain . lines mark boundaries between various continuous phases . ] when the coupling strength of the applied strain @xmath89 becomes the largest energy scale in the system , all other degrees of freedom can be ignored at the @xmath104 valley as states around @xmath104 are all gapped out , leading to the phase diagram of fig . [ fig : pd3 ] . all @xmath140 orders are seen to disappear , and only the fm phase survives for @xmath148 at a given density @xmath7 . the presence of the concave boundary region separating the metallic and fm phases in fig . 8 can be understood from energetics : at any given density in this region , the chemical potential lies within the gap induced by the fm order at @xmath149 , yielding a gain in free energy and thus lowers @xmath150 . as the particle density increases , the chemical potential is eventually unable to locate within the gap and this argument fails . interestingly , at @xmath142 , although the ground state is insulating with a small gap due to trs breaking , the system does not exhibit quantum anomalous hall effect . this should be contrasted sharply with the case of a zeeman field applied perpendicular to the ( 001 ) surface of a tci in the non - interacting limit , where such a broken trs leads to non - vanishing chern number as shown in sec . iiib above . versus @xmath7 phase diagram in the case of large strain . solid line marks the boundary between the two continuous phases . ] some of the strains discussed in this study could be realized experimentally through chemical doping . since pbse and snse assume cubic and orthorhombic structures@xcite , respectively , strains or lattice distortions could likely be manipulated by adjusting chemical compositions . in fact , a ferroelectric - like lattice distortion has been observed in the tcis via stm / sts spectroscopy@xcite . mechanical tuning could provide another approach . for example , a piezoelectric layer could be grown on the surface of a tci to allow manipulation of strains via electric field as proposed in ref .. as to practical applications , we have shown clearly that the chern number of the surface states can be tuned through the interplay of applied strain and zeeman fields . this provides a potential new handle for controlling the topological conducting edge channels for low power - consuming , next generation electronic technologies . furthermore , our analysis shows that in the presence of electron - electron interactions , ferromagnetism could be turned on / off via strain fields , which , offering a novel pathway toward spintronics applications . in summary , we have systematically investigated the effects of strains , which break either one or both mirror symmetries without breaking time - reversal symmetry , on the ( 001 ) surface of the tcis . under mirror - symmetry breaking strains , the low energy dirac cones not only become massive , but also develop hedgehog - like spin textures . we show that the chern number of the surface states can be tuned via applied strain fields in the presence of a zeeman field . finally , we delineate the competing orders that can result from varying strengths of short - range ( repulsive ) electron - electron interactions at the mean - field level , and show that the resulting correlated phases are amenable to control through strain fields . the tunability of the interacting as well as the non - interacting electronic structures and topological states of the tcis via strain fields and their interplay with zeeman fields as revealed by our study suggests new pathways for developing spintronics and other applications platforms based on the tcis . we thank chen fang and hong yao for useful discussions . w.f.t . and c.y.h . acknowledge the support from most in taiwan under grant no.103 - 2112-m-110 - 008-my3 . the work at northeastern university was supported by the us department of energy ( doe ) , office of science , basic energy sciences grant number de - fg02 - 07er46352 , and benefited from northeastern university s advanced scientific computation center ( ascc ) and the nersc supercomputing center through doe grant number de - ac02 - 05ch11231 . h.l . acknowledges the singapore national research foundation for the support under nrf award no . nrf - nrff2013 - 03 . 99 m. z. hasan and c. l. kane , rev . phys . * 82 * , 3045 ( 2010 ) . x. l. qi and s. c. zhang , rev . * 83 * , 1057 ( 2011 ) . j. e. moore , nature ( london ) * 464 * , 194 ( 2010 ) . y. ando , j. phys . . jpn . * 82 * , 102001 ( 2013 ) . a. p. schnyder , s. ryu , a. furusaki , and a. w. w. ludwig , phys . rev . b * 78 * , 195125 ( 2008 ) . a. kitaev , aip conf . proc . * 1134 * , 22 ( 2009 ) . t. l. hughes , e. prodan , and b. a. bernevig , phys . rev . b * 83 * , 245132 ( 2010 ) . a. m. turner , y. zhang , r. s. k. mong , and a. vishwanath , phys . b * 85 * , 165120 ( 2012 ) . c. fang , m. j. gilbert , and b. a. bernevig , phys . rev . b * 86 * , 115112 ( 2012 ) . r. slager , a. mesaros , v. juricic , and j. zannen , nature physics * 9 * , 98 ( 2013 ) . k. shiozaki and m. sato , phys . b * 90 * , 165114 ( 2014 ) . l. fu , phys . * 106 * , 106802 ( 2011 ) . j. c. y. teo , l. fu , and c. l. kane , phys . rev . b * 78 * , 045426 ( 2008 ) . k . chiu , h. yao , and s. ryu , phys . b * 88 * , 075142 ( 2013 ) . t. h. hsieh , h. lin , j. liu , w. duan , a. bansil , and l. fu , nat . commun . * 3 * , 982 ( 2012 ) . xu , c. liu , n. alidoust , m. neupane , d. qian , i. belopolski , j. d. denlinger , y. j. wang , h. lin , l. a. wray , g. landolt , b. slomski , j. h. dil , a. marcinkova , e. morosan , q. gibson , r. sankar , f. c. chou , r. j. cava , a. bansil and m. z. hasan , nat . commun . * 3 * , 1192 ( 2012 ) . p. dziawa , b. j. kowalski , k. dybko , r. buczko , a. szczerbakow , m. szot , e. usakowska , t. balasubramanian , b. m. wojek , m. h. berntsen , o. tjernberg and t. story , nat . mater . * 11,1023 ( 2012 ) . y. tanaka , z. ren , t. sato , k. nakayama , s. souma , t. takahashi , k. segawa , and y. ando , nat . 8 * , 800 ( 2012 ) . a. gyenis , i. k. drozdov , s. nadj - perge , o. b. jeong , j. seo , i. pletikosic , t. valla , g. d. gu , and a. yazdani , phys . b * 88 * , 125414 ( 2013 ) . a. a. taskin , s. sasaki , k. segawa , and y. ando , phys . b * 89 * , 121302(r ) ( 2014 ) . y. okada , m. serbyn , h. lin , d. walkup , w. zhou , c. dhital , m. neupane , s. xu , y. j. wang , r. sankar , f. chou , a. bansil , m. z. hasan , s. d. wilson , l. fu and v. madhavan , science * 341 * , 1496 ( 2013 ) . i. zeljkovic , y. okada , m. serbyn , r. sankar , d. walkup , w. zhou , j. liu , g. chang , y. j. wang , m. z. hasan , f. chou , h. lin , a. bansil , l. fu , and v. madhavan , nature materials * 14 * , 318 ( 2015 ) . i. zeljkovic , y. okada , c .- y . huang , r. sankar , d. walkup , w. zhou , m. serbyn , f. chou , w .- f . tsai , h. lin , a. bansil , l. fu , m. z. hasan , and v. madhavan , nature physics * 10 * , 572 ( 2014 ) . m. iizumi , y. hamaguchi , k. komatsubara , and y. kato , j. phys . jpn . * 38 * , 443 ( 1975 ) . c. fang , m. j. gilbert and b. a. bernevig , phys . . lett . * 112 * , 046801 ( 2014 ) . m. serbyn and l. fu , phys . b * 90 * , 035402 ( 2014 ) . s .- xu , m. neupane1 , c. liu _ * 8 * , 616 ( 2012 ) . j. liu , t. h. hsieh , p. wei , w. duan , j. moodera and l. fu , nat . * 13,178 ( 2014 ) . a. h. castro neto , f. guinea , n. m. r. peres , k. s. novoselov , and a. k. geim . 81 * , 109 ( 2009 ) . t. o. wehling , e. sasioglu , c. friedrich , a. i. lichtenstein , m. i. katsnelson , and s. blugel , phys . lett . * 106 * , 236805 ( 2011 ) . importance of the coulomb interaction for relativistic quasiparticles may be assessed approximately by the dimensionless interaction strength , @xmath151 , where @xmath152 denotes the dielectric constant , and @xmath153 is the fermi velocity . for graphene , @xmath154 , while for snte @xmath155 is even higher due to a smaller value of @xmath153 . wang , w .- f . tsai , h. lin , s .- y . xu , m. neupane , m. z. hasan and a. bansil , phys . b * 87 * , 235317 ( 2013 ) . l. fu , c. l. kane , and e. j. mele , phys . * 98 * , 106803 ( 2007 ) . l. fu and c. l. kane , phys . b * 76 * , 045302 ( 2007 ) . g. yang , j. liu , l. fu , w. duan , and c. liu , phys . b * 89 * , 085312 ( 2014 ) . binghai yan , lukas mu chler , and claudia felser , phys . * 109 * , 116406 ( 2012 ) . qi and s .- c . zhang , physica e : low- dimensional systems and nanostructures * 44 * , 906 ( 2012 ) . c. fang and l. fu , phys . b * 91 * , 161105 ( 2015 ) . i. zeljkovic , d. walkup , b. assaf , k. l scipioni , r. sankar , f. chou , v. madhavan , arxiv:1501.01299 . b. andrei bernevig and taylor l. hughes , topological insulators and topological superconductors ( princeton university press , 2013 ) . w. kohn and j.m . luttinger , phys . . lett . * 15 * , 524 ( 1965 ) . s. raghu , x .- qi , c. honerkamp , and s .- c . zhang , phys . lett . * 100 * , 156401 ( 2008 ) . effects of long - range coulomb interaction should be examined further . a. j. strauss , phys . * 157 * , 608 ( 1967 ) . here we present further details of the mf theory used in sec . iv . in terms of the slowly varying field operators defined in eq . ( [ eq : ftc ] ) , the relevant number operators are given by _ p_z(r)&=&c_\|x , p_z^c_\|x , p_z+c_\|y , p_z^c_\|y , p_z + & + & ( -1)^x+y(c_\|x , p_z^c_\|y , p_z+c_\|y , p_z^c_\|x , p_z ) , + _ p_z(r)&=&c_\|x , p_z^c_\|x , p_z+c_\|y , p_z^c_\|y , p_z + & + & ( -1)^x+y(c_\|x , p_z^c_\|y , p_z+c_\|y , p_z^c_\|x , p_z ) , + _ p_x(r)&=&c_\|x , p_x^(r)c_\|x , p_x(r ) , + _ p_x(r)&=&c_\|x , p_x^(r)c_\|x , p_x(r ) , + _ p_y(r)&=&c_\|y , p_y^(r)c_\|y , p_y(r ) , + _ p_y(r)&=&c_\|y , p_y^(r)c_\|y , p_y(r ) , + _ p_z(r)&=&_p_z(r)+_p_z(r ) , + _ p_x(r)&=&_p_x(r)+_p_x(r ) , + _ p_y(r)&=&_p_y(r)+_p_y(r ) . we illustrate our mf treatment with the example of @xmath156 . for this purpose , we decouple bilinear fermion operators composed of @xmath157 or @xmath158 in the particle - hole channel into @xmath159 , where @xmath160 represents fluctuations away from the mean - field values , which are neglected . the condensation energy , @xmath161 , for @xmath156 is thus given by @xmath162 where the @xmath163-dependence is implicit . in fact , @xmath164 can be further rewritten as @xmath165 where @xmath166 was defined in eq . ( [ eq : order ] ) and @xmath167 is a unit area . notice that @xmath168 is the particle density defined in eq . ( [ eq : density ] ) . when the similar procedure is applied to other terms in eq . ( [ demo1 ] ) , it turns out that @xmath161 term becomes : @xmath169\notag\\ + s\frac{u}{64}&[(o_{001}-o_{031}+o_{301}-o_{331})^2\notag\\ & + ( o_{002}-o_{032}+o_{302}-o_{332})^2].\end{aligned}\ ] ] this allows us to obtain the mf decoupled @xmath156 as @xmath170\notag\\ -\frac{u}{32}&[(o_{001}-o_{031}+o_{301}-o_{331})\notag\\ & \times\hat{\psi}^{\dag}(\sigma_{001}-\sigma_{031}+\sigma_{301}-\sigma_{331})\hat{\psi}\notag\\ & + ( o_{002}-o_{032}+o_{302}-o_{332})\notag\\ & \times\hat{\psi}^{\dag}(\sigma_{002}-\sigma_{032}+\sigma_{302}-\sigma_{332})\hat{\psi}]\bigg\}\notag\\ -s\frac{u}{64}&[(o_{000}-o_{030}+o_{300}-o_{330})^2\notag\\ & -(o_{003}-o_{033}+o_{303}-o_{333})^2]\notag\\ + s\frac{u}{64}&[(o_{001}-o_{031}+o_{301}-o_{331})^2\notag\\ & + ( o_{002}-o_{032}+o_{302}-o_{332})^2].\end{aligned}\ ] ] by applying the same trick for the other interaction terms in eq . ( [ hint ] ) , we finally find 64 mf undetermined parameters ( summarized in table . [ t2 ] ) , and obtain the mf free energy per unit area at zero temperature as : f = d^2r^(h_(001)+h_int^mf)-n+e_con , where @xmath171 is the mf decoupled interaction hamiltonian for @xmath172 and @xmath173 is the mf condensation energy . the ground state can then be determined by solving the set of coupled mean - field equations resulting from minimizing @xmath174 with respect to all possible order parameters . m2.5cm\|m5.5 cm type & mf parameters + renormalized band parameters&@xmath175 , @xmath176,@xmath177 , , this is an order parameter for spontaneous mirror symmetry breaking . ] @xmath178,@xmath179 , , this is an order parameter for spontaneous mirror symmetry breaking . ] @xmath180,@xmath181,@xmath182,@xmath183 , , @xmath181 , @xmath182 and @xmath183 is nonzero , this would be an order parameter for spontaneous mirror symmetry breaking . ] @xmath129,@xmath184 , or @xmath184 is nonzero , this would be an order parameter for spontaneous @xmath49 symmetry breaking . ] @xmath168 + fm & @xmath185 , @xmath186 , @xmath187 , @xmath188 , @xmath139 , @xmath189 , @xmath190 , @xmath191 , @xmath192 , @xmath193 , @xmath194 , @xmath195 , @xmath196 , @xmath197 , @xmath198 , @xmath199 , @xmath200 , @xmath201 , @xmath202 , @xmath203 + cdw & @xmath204 , @xmath205 , @xmath206 , @xmath207 , @xmath208 , @xmath209 , @xmath210 , @xmath211 , @xmath212 , @xmath213 , @xmath214 , @xmath215 , @xmath216 , @xmath217 , @xmath218 , @xmath219 + sdw & @xmath220 , @xmath221 , @xmath222 , @xmath223 , @xmath224 , @xmath225 , @xmath226 , @xmath227 , @xmath228 , @xmath229 , @xmath230 , @xmath231 , @xmath232 , @xmath233 , @xmath234 , @xmath235
there is a very old question in the philosophy of space and time , concerning how and why bodies move in the particular way that they do in the absence of any external forces . the question originates with aristotle , and indeed , the puzzle is particularly acute when one thinks of it as the ancients might have . given some external influence on a body , it might seem clear why that body moves in one fashion rather than another : the external influence forces it to do so . but when there are no forces present , what does the work of picking one possible state of motion over any other ? consider planetary motion : there are no apparent forces acting on planets , and yet they proceed along fixed trajectories . why these orbits rather than others ? in aristotelian terms , what determines the `` natural motions '' of a body ? the modern answer to the question originates with galileo and descartes , but finds its canonical form in newton s first law of motion , which states that in the absence of external forces , a body will move in a straight line at constant velocity . this `` law of inertia , '' as newton called it , is preserved , _ mutatis mutandis _ , in general relativity , where inertial motion is governed by the _ geodesic principle_. the geodesic principle states that in the absence of external forces , the possible trajectories through four dimensional spacetime of a massive test point particle will be timelike geodesics i.e . , bodies will move along `` locally straightest '' lines , without acceleration . in standard presentations of general relativity , the geodesic principle is stated as a postulate ( cf . * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , much like newton s first law . however , shortly after einstein presented the theory , he and others began to suspect that one could equally well conceive of the geodesic principle as a theorem , at least in the presence of other standard assumptions of relativity theory @xcite . this shift from geodesic - principle - as - postulate to geodesic - principle - as - theorem has led to a widespread and deeply influential view that general relativity has a special explanatory virtue that distinguishes it from other theories of space and time : in the words of harvey brown , general relativity `` ... is the first in the long line of dynamical theories ... that _ explains _ inertial motion '' ( * ? ? ? the view holds that newtonian physics may answer the `` how '' part of aristotle s question , but there is a sense in which only general relativity answers the `` why '' part . although einstein s early attempts to prove the geodesic principle were not unambiguously successful , more recent efforts have shown that there is a precise sense in which the geodesic principle may be understood as a theorem of general relativity @xcite . however , it turns out that relativity is not unique in this regard . geometrized newtonian gravitation ( sometimes , newton - cartan theory ) is a reformulation of newtonian gravitation due to lie @xcite and kurt @xcite that shares many of the qualitative features of general relativity . in geometrized newtonian gravitation one represents space and time as a four dimensional spacetime manifold , the curvature of which depends dynamically on the distribution of matter on the manifold . gravitational influences , meanwhile , are not understood as forces , as in traditional formulations of newtonian gravitation ; rather , they are a manifestation of the curvature of spacetime . and in particular , inertial motion is governed by the geodesic principle : in the absence of external ( non - gravitational ) forces , bodies move along the geodesics of ( curved ) spacetime . recently , i have shown that the geodesic principle can be understood as a theorem of geometrized newtonian gravitation @xcite . mathematically , the newtonian theorem is nearly identical to the geroch - jang theorem . moreover , as i have argued elsewhere , when the background assumptions needed to prove these theorems are examined in the contexts of each theory , one can reasonably conclude that the geodesic principle has essentially the same status in both cases , though in neither theory is the situation as simple as one might have hoped @xcite . one consequence of this recent work is that einstein and others idea that the status of the geodesic principle in general relativity distinguishes the theory from other theories of space and time seems more difficult to hold on to . but it also raises a related issue . when one attends carefully to the details of these theorems , several complications arise concerning the strength and status of the assumptions necessary for proving them . given these complications , one might reasonably ask , do either of these theories _ explain _ inertial motion ? it is this second question that i will take up in the present paper . i will begin with a brief overview of geometrized newtonian gravitation , after which i will review the relevant theorems concerning the geodesic principle in that theory and general relativity . i will focus on the subtle ways in which the theorems differ , and on the complications that arise when one tries to interpret them . once this background material has been laid out , i will turn to the question at hand . the starting point for this discussion will be to observe that on one way of thinking about explanation in scientific theories , the answer to the question is `` no '' : neither of these theories explains inertial motion , at least if the assumptions going into the theorems have the character i describe . i want to resist this view , however , because i think it takes for granted that one can make clear distinctions between `` levels '' or `` tiers '' of fundamentality of the central principles of a theory . careful analysis of the geodesic principle theorems , meanwhile , suggests that there is another way of thinking about how the principles of a theory fit together . the alternative view i will develop i will call it the `` puzzle ball view '' or , perhaps more precisely , the `` puzzle ball conjecture''holds that the foundations of physical theories , or at least _ these _ physical theories , are best conceived as a network of mutually interdependent principles , rather than as a collection of independent and explanatorily fundamental `` axioms '' or `` postulates '' . on this view , one way to provide a satisfactory explanation of a central principle of a theory , such as the geodesic principle in general relativity or geometrized newtonian gravitation , would be to exhibit its dependence on the other central principles of the theory , i.e. , to show how the principle - to - be - explained is a consequence of the other central principles and basic assumptions of the theory . and this is precisely what the theorems i will describe do . and so , i will argue that there _ is _ a sense in which both theories explain inertial motion , though some care is required to say what is meant by `` explain '' in this context . i should be clear from the start : the language of explanation is a convenient one , but i am not ultimately interested in the semantics of the word `` explain '' . the goal is not to argue whether one thing or another is _ really _ an explanation . the dialectic , rather , is as follows . many people have suggested that general relativity provides an important kind of insight with regard to inertial motion , something to be valued and sought after in our physical theories . one might call this thing an `` explanation '' , or not . the point , though , is that when one looks in detail at just what one gets in relativity theory ( and in geometrized newtonian gravitation ) , it seems to work in a different way than one might have initially guessed it would . one response to this observation would be to say that we have not actually gotten what we were promised or , in the language above , that general relativity does _ not _ explain inertial motion . but another response is to try to better understand what we _ do _ get . my principal thesis is that if one takes this second path , an alternative picture emerges of how the foundations of theories work . and on this alternative picture , general relativity and geometrized newtonian gravitation both do provide an important and very useful kind of insight into inertial motion , and more , there are clear reasons why one should value and seek out this sort of insight . indeed , one might even think that what we ultimately get is what we should have wanted in the first place . i am inclined to use the word `` explanation '' for this sort of insight , but fully recognize that this usage may seem non - standard or incorrect to some readers . geometrized newtonian gravitation is best understood as a translation of newtonian gravitation into the language of general relativity , a way of making newtonian physics look as much like general relativity as possible , for the purposes of addressing comparative questions about the two theories . the result is a theory that is strikingly similar in many qualitative respects to general relativity , but which differs in certain crucial details . recall that in general relativity , a _ relativistic spacetime _ is an ordered pair @xmath0 , where @xmath1 is a smooth four dimensional manifold and @xmath2 is a smooth lorentzian metric on the manifold . in geometrized newtonian gravitation , meanwhile , one similarly starts with a smooth four dimensional manifold @xmath1 , but one endows this manifold with a different metric structure . specifically , one defines _ two _ ( degenerate ) metrics . temporal metric _ @xmath3 , has signature @xmath4 . it is used to assign temporal lengths to vectors on @xmath1 : the temporal length of a vector @xmath5 at a point @xmath6 is @xmath7 . vectors with non - zero temporal length are called _ timelike _ ; otherwise , they are called _ spacelike_. the second metric is a _ spatial metric _ @xmath8 , with signature @xmath9 . in general one requires that these two metrics satisfy an orthogonality condition , @xmath10 . it is important that the temporal metric is written with covariant indices and the spatial metric with contravariant indices : since both metrics have degenerate signatures , they are not invertible , and so in general one can not use either to raise or lower indices . in particular , this means that the spatial metric can not be used to assign spatial lengths to vectors directly . instead , one uses the following indirect method . given a spacelike vector @xmath5 , one can show that there always exists a ( non - unique ) covector @xmath11 such that @xmath12 . one then defines the spatial length of @xmath5 to be @xmath13 , which can be shown to be independent of the choice of @xmath11 . given a lorentzian metric @xmath2 on a manifold @xmath1 , there always exists a unique covariant derivative operator @xmath14 that is compatible with @xmath2 in the sense that @xmath15 . this does not hold for the degenerate newtonian metrics . instead , there are an uncountably infinite collection of derivative operators that satisfy the compatibility conditions @xmath16 and @xmath17 . this means that to identify a model of geometrized newtonian gravitation , one needs to specify a derivative operator in addition to the metric fields . thus , we define a _ classical spacetime _ as an ordered quadruple @xmath18 , where @xmath1 , @xmath3 , @xmath8 , and @xmath14 are as described , the metrics satisfy the orthogonality condition , and the metrics and derivative operator satisfy the compatibility conditions . a classical spacetime is the analog of a relativistic spacetime . note that the signature of @xmath3 guarantees that at any point @xmath6 , one can find a covector @xmath19 such that @xmath20 ; in cases where such a field can be defined globally , we call the associated spacetime _ temporally orientable_. in what follows , we will always restrict attention to temporally orientable spacetimes , and will replace @xmath3 with @xmath19 whenever we specify a classical spacetime . in both theories , timelike curves curves whose tangent vector field is always timelike represent the possible trajectories of point particles ( and idealized observers ) . and as in general relativity , matter fields in geometrized newtonian gravitation are represented by a smooth symmetric rank-2 field @xmath21 ( with contravariant indices ) . in general relativity , this field is called the _ energy - momentum tensor _ ; in geometrized newtonian gravitation , it is called the _ mass - momentum tensor_. the reason for the difference concerns the interpretations of the fields . in relativity theory , the four - momentum density of a matter field with energy - momentum tensor @xmath21 is only defined relative to some observer s state of motion : given an observer whose worldline has ( timelike ) tangent field @xmath5 , the four - momentum density @xmath22 as determined by the observer is given by @xmath23 . when @xmath22 is timelike or null , one can define the mass density @xmath24 of the field at a point , relative to the observer , as the length of @xmath22 . moreover , the four - momentum field can be further decomposed ( relative to @xmath5 ) as @xmath25 , where @xmath26 is the relative energy density as determined by the observer , and @xmath27 is the relative three - momentum density . thus , the field @xmath21 encodes the relative mass , relative energy , and relative momentum densities as determined by any observer . in geometrized newtonian gravitation , meanwhile , _ all _ observers make the same determination of the four - momentum density of a matter field at a point : for any observer , @xmath22 is given by @xmath28 . given a particular observer whose worldline has tangent field @xmath5 , though , one can decompose @xmath22 as @xmath29 , where @xmath30 is the ( observer - independent ) mass density associated with the matter field , and where @xmath31 is the relative three - momentum density of the matter field as determined by the observer . thus in geometrized newtonian gravitation , @xmath21 encodes the ( absolute ) mass density of a matter field , as well as its momentum relative to any observer . ) and relative energy density is @xmath32 , where @xmath5 is the tangent field to the observer s worldline . in geometrized newtonian gravitation , this distinction collapses : both ( absolute ) mass density and ( absolute ) energy density are given by @xmath33 . ] it is standard in both theories to limit attention to matter fields that satisfy several additional constraints . in particular , in both cases one assumes that matter fields satisfy the _ conservation condition _ , which states that their energy / mass - momentum fields are divergence free ( i.e. , @xmath34 ) . one also usually requires that such fields satisfy various _ energy conditions_. in geometrized newtonian gravitation , only one such condition is standard : it is the so - called _ mass condition_. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * mass condition * : a mass - momentum field satisfies the mass condition if , at any point , either @xmath35 or @xmath36 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ since @xmath37 is the mass density , this assumption states that whenever the mass - momentum tensor is non - vanishing , the associated matter field has positive mass . the situation is more complicated in general relativity , where there are several energy conditions that one may consider . i will mention a few because they are of particular interest for present purposes . one , called the _ weak energy condition _ , is ( at least _ prima facie _ ) quite similar to the mass condition it states that the energy density of a matter field as determined by any observer is always non - negative . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * weak energy condition * : an energy - momentum field satisfies the weak energy condition if , give any timelike vector @xmath5 at a point , @xmath38 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it is also common to consider stronger conditions . for instance , there are the _ dominant energy condition _ and the _ strengthened dominant energy condition _ : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * dominant energy condition * : an energy - momentum field satisfies the dominant energy condition if , given any timelike vector @xmath39 at a point , @xmath40 and @xmath41 is timelike or null . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * strengthened dominant energy condition * : an energy - momentum field satisfies the strengthened dominant energy condition if , give any timelike covector @xmath39 at any point in @xmath1 , @xmath42 and either @xmath35 or @xmath41 is timelike . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ if these two conditions obtain for some matter field , then not only do all observers take the field to have non - negative energy density , they also take its four - momentum to be causal or timelike ( respectively ) . in other words , these latter conditions capture the requirement that matter must propagate at or below the speed of light . the curvature of a classical spacetime is defined in the standard way : given a derivative operator @xmath14 , the _ riemann curvature tensor _ @xmath43 is the unique tensor field such that for any vector field @xmath5 , @xmath44}\xi^a$ ] . the _ ricci curvature tensor _ , meanwhile , is given by @xmath45 . in both contexts , one says that a spacetime is _ flat _ if @xmath46 ; in geometrized newtonian gravitation , one also says that a ( possibly curved ) spacetime is _ spatially flat _ if @xmath47 or , equivalently , @xmath48 . given these ingredients , one can state the sense in which in geometrized newtonian gravitation , the curvature of spacetime depends on the distribution of matter : namely , the central dynamical principle of the theory , the _ geometrized poisson equation _ , states that @xmath49 , where @xmath24 is the mass - density defined above . this expression explicitly relates the ricci curvature of spacetime to the distribution of matter . it is the newtonian analogue of einstein s equation , @xmath50 , where @xmath51 , or equivalently @xmath52 , where @xmath53 . there are a few points to emphasize here concerning the geometrized poisson equation . for one , if the geometrized poisson equation holds of a classical spacetime for some mass - momentum tensor @xmath21 , then the classical spacetime is spatially flat , since @xmath54 . this fact is a way of recovering a familiar feature of newtonian gravitation , namely that _ space _ is always flat , even though in the geometrized theory _ spacetime _ may be curved . second , in general relativity one can freely think of both the metric and the derivative operator as ( systemically related ) dynamical variables in the theory . in geometrized newtonian gravitation , this is not the case : instead , the metrical structure of a classical spacetime is fixed , and only the derivative operator ( or more specifically , the ricci curvature , which is defined in terms of the derivative operator ) is a dynamic variable . finally , there is a sense in which , given some matter distribution , the geometrized poisson equation `` fixes '' a derivative operator on a classical spacetime , but one has to be careful , as one can typically only recover a unique derivative operator satisfying the geometrized possion equation for a given matter distribution in the presence of additional boundary conditions or other assumptions . the geometrized poisson equation provides the sense in which in geometrized newtonian gravitation , spacetime is curved in the presence of matter ; the sense in which gravitational effects may be understood as a manifestation of this curvature is just the same as in general relativity . that is , a derivative operator allows one to define a class of geometrically privileged curves , the _ geodesics _ of the spacetime , which consist of all curves whose tangent fields @xmath5 satisfy @xmath55 everywhere . i have already said that the timelike curves of a spacetime represent the possible trajectories for massive particles ; the timelike geodesics , meanwhile , represent the possible _ unaccelerated _ trajectories of particles in both theories . the geodesic principle then connects these geometrically privileged curves with force - free motion . thus , in geometrized newtonian gravitation , as in general relativity , the distribution of matter throughout space and time affects the possible trajectories of massive point particles not by causing such particles to accelerate , but rather by dynamically determining a collection of unaccelerated curves . these features of geometrized newtonian gravitation provide the sense in which the theory is qualitatively similar to general relativity . but one might wonder what undergirds the implicit claim that geometrized newtonian gravitation is in some sense _ newtonian_. one sense in which the theory is newtonian is immediate : the degenerate metric structure of a classical spacetime captures the implicit geometry of space and time in ordinary newtonian gravitation , where one has a temporally ordered succession of flat three dimensional manifolds representing space at various times ( cf . but there is more to say . in standard formulations of newtonian gravitation , spacetime is flat . gravitation is a force mediated by a gravitational potential , which in turn is related to the distribution of matter by poisson s equation . in the present four dimensional geometrical language , this can be expressed as follows . we begin with a classical spacetime @xmath56 as before , but now we require that @xmath14 is flat , i.e. , @xmath46 . we again represent matter by its mass - momentum field @xmath21 , defined just as above , but we also define a scalar field @xmath57 , which is the gravitational potential . poisson s equation is written as @xmath58 where the index on @xmath59 is raised using @xmath8 , and where @xmath60 . and now the acceleration of a massive test point particle in the presence of a gravitational potential @xmath57 is given by @xmath61 , where @xmath5 is the tangent to the particle s trajectory . in other words , in standard newtonian gravitation matter accelerates in the presence of mass . it turns out that standard newtonian gravitation ( thus understood ) and geometrized newtonian gravitation are systematically related ( * ? ? ? 4.2 ) . specifically , given a classical spacetime @xmath56 with @xmath14 flat , a smooth mass density @xmath24 , and a smooth gravitational potential @xmath57 satisfying @xmath58 , there always exists a unique derivative operator @xmath62 such that @xmath63 is a classical spacetime , @xmath64 , and such that for any timelike vector field @xmath5 , @xmath61 if and only if @xmath65 . in other words , given a model of standard newtonian gravitation , there is always a model of geometrized newtonian gravitation with precisely the same mass density and allowed trajectories . additionally , the derivative operator @xmath62 will always satisfy two curvature conditions : @xmath66 and @xmath67 . this result is known as the trautman geometrization lemma ; it provides the sense in which one can always translate from standard newtonian gravitation into the geometrized theory . one can also prove a corresponding recovery lemma ( also due to trautman ) , allowing for translations back : namely , given a classical spacetime @xmath63 and smooth mass density @xmath24 satisfying @xmath64 , if @xmath66 and @xmath67 then there always exists a flat derivative operator @xmath14 and a gravitational potential @xmath57 such that @xmath56 is a classical spacetime , @xmath58 , and again for any timelike vector field @xmath5 , @xmath61 if and only if @xmath65 . note that this recovery result only holds in the presence of the two additional curvature conditions stated above ; moreover , in general the translation from geometrized newtonian gravitation to standard newtonian gravitation will _ not _ be unique . ( see figure [ translate ] . ) in general it is possible to translate between geometrized newtonian gravitation and standard newtonian gravitation , as depicted in this figure . on the left is a model of standard newtonian gravitation : one has a matter field represented by the world - tube of some body , such as the sun , and a curve orbiting this body , representing , say , a small planet . this curve corresponds to an allowed trajectory insofar as it is accelerating by the appropriate amount . on the right is the corresponding model of geometrized newtonian gravitation . one has precisely the same matter distribution , and the same allowed trajectory ( i.e. , the same orbit ) , but now we understand this trajectory to be allowed by the theory because it is a geodesic of a curved derivative operator , with curvature determined by the matter distribution . note that both theories have the same metrical structure , represented here by a succession of flat slices representing space at various times . ] with the background of the previous section in place , i can now state the precise sense in which the geodesic principle may be understood as a theorem in general relativity and geometrized newtonian gravitation . i will begin by stating both theorems , and then double back to the question of how one should interpret them . [ gj ] _ _ * @xcite__let @xmath68 be a relativistic spacetime , and suppose @xmath1 is oriented . let @xmath69 be a smooth imbedded curve . suppose that given any open subset @xmath70 of @xmath1 containing @xmath71 $ ] , there exists a smooth symmetric field @xmath21 with the following properties . 1 . [ sdec ] @xmath21 satisfies the _ strengthened dominant energy condition _ , i.e. , given any timelike covector @xmath39 at any point in @xmath1 , @xmath38 and either @xmath35 or @xmath41 is timelike ; 2 . [ cons]@xmath21 satisfies the _ conservation condition _ , i.e. , @xmath72 ; 3 . [ inside]@xmath73 ; and 4 . [ non - vanishing]there is at least one point in @xmath70 at which @xmath74 . then @xmath75 is a timelike curve that can be reparametrized as a geodesic . one can prove an almost identical theorem in geometrized newtonian gravitation . _ @xcite * _ [ w ] let @xmath18 be a classical spacetime , and suppose that @xmath1 is oriented . suppose also that @xmath76 . let @xmath69 be a smooth imbedded curve . suppose that given any open subset @xmath70 of @xmath1 containing @xmath71 $ ] , there exists a smooth symmetric field @xmath21 with the following properties . 1 . [ mass ] @xmath21 satisfies the mass condition , i.e. , whenever @xmath74 , @xmath36 ; 2 . [ cons2 ] @xmath21 satisfies the conservation condition , i.e. , @xmath72 ; 3 . [ inside2 ] @xmath73 ; and 4 . [ non - vanishing2 ] there is at least one point in @xmath70 at which @xmath74 . then @xmath75 is a timelike curve that can be reparametrized as a geodesic . as a first remark , it may not be obvious that either of these theorems should be understood to capture the geodesic principle at all , at least in a natural way . a principal difficulty in trying to derive the geodesic principle as a theorem concerns a kind of ontological mismatch between the geodesic principle and the rest of general relativity : namely , general relativity is a field theory , whereas the geodesic principle is a statement concerning point particles . one strategy for dealing with this problem is to try to model massive point particles as `` small '' bits of extended matter , and then try to show that under sufficiently general assumptions , the world tubes of such small bits of matter will contain timelike geodesics . but this turns out to be false in general geodesic motion only obtains in the idealized limit where the worldtube of a body collapses to a curve , in which case one can no longer represent matter as a smooth field on spacetime . the geroch - jang strategy , meanwhile , is different . instead of starting with some matter and asking what kind of trajectory it follows , one starts with a curve and asks under what circumstances that curve can be understood as a trajectory for arbitrarily small bits of extended matter . both theorems then state that the only curves along which arbitrarily small bits of matter can be constructed are timelike geodesics . importantly , one represents a `` small bit of matter '' by a smooth symmetric rank 2 tensor fields with support in some neighborhood of the curve . but a curve is not understood as a possible trajectory for a free massive test point particle if one can construct _ any _ smooth symmetric rank 2 tensor field in arbitrarily small neighborhoods of the curve rather , one limits attention to fields that satisfy additional constraints . the claim , then , is that these theorems capture the geodesic principle in both theories insofar as the additional constraints on the matter fields adequately capture what we intend by `` free massive test matter '' . this means that the interpretation of the theorems turns on the status of these conditions . and so , for a comparative study of the status of the geodesic principle in each theory , one wants to compare the status of each of these assumptions relative to their respective theories . two of the assumptions can be set aside immediately : in both theorems , assumptions ( 3 ) and ( 4 ) play the role of setting up the limiting process implicit in the theorems . assumption ( 3 ) limits attention to matter fields that vanish outside one s chosen neighborhood of the curve ( which captures the sense in which one is considering arbitrarily small bits of matter propagating along the curve ) , and assumption ( 4 ) indicates that the matter field must be non - vanishing somewhere along curve , ruling out the trivial case . these assumptions are identical in both cases , and neither is troublingly strong . there is also an obvious difference that can be safely ignored . in the newtonian theorem , we place an additional constraint on the curvature , namely @xmath76 . this is precisely the curvature condition needed to prove the trautman recovery theorem , allowing one to translate from a model of geometrized newtonian gravitation to a model of standard newtonian gravitation . for this reason , the curvature condition is naturally interpreted as a restriction to models of geometrized newtonian gravitation that are _ newtonian _ , in the sense that they admit translations back to models of standard newtonian gravitation . this presumably is the case of greatest interest , and so i am inclined to think of the assumption as benign . moreover , there is good reason to think that this assumption can be dropped , though to my knowledge , proving as much is still an open ( and perhaps interesting ) problem . the most striking difference between the two theorems concerns the respective assumptions ( 1 ) . in the newtonian theorem , this is the mass condition , i.e. , that whenever the mass - momentum field is non - vanishing , the mass density determined by any observer must be positive . this is the standard energy condition in geometrized newtonian gravitation , and more , it is natural in this context , as it captures the sense in which the bits of matter being represented are massive . in the geroch - jang theorem , meanwhile , one requires the strengthened dominant energy condition , which states that ( a ) all observers must assign non - negative ( mass-)energy density to the matter field ( the weak energy condition ) and ( b ) that if @xmath77 , then the four - momentum assigned to the matter field by any observer must by timelike . it seems natural to think that the _ weak _ energy condition , ( a ) , is playing the role played by the mass condition in the newtonian case : namely , it captures the sense in which the small bits of matter are massive , by requiring that they always have non - negative mass . but from this perspective , the second part of the condition , ( b ) , is a strong additional requirement . in newtonian gravitation , it would seem , one needs only to assume that mass is always positive to get timelike geodesic propagation , whereas in general relativity , one also needs to make an assumption about the timelike propagation of matter . however , the situation is not quite so simple as this . although the mass condition appears to be nothing more than an assumption about positive mass , it , too , contains an implicit assumption about timelike propagation . to see this , consider a different ( non - standard ) newtonian energy condition , which i will call the _ weakened mass condition_. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * weakened mass condition * : a mass - momentum field @xmath21 satisfies the weakened mass condition if at any point , @xmath78 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the weakened mass condition has a good claim on being the newtonian analogue of the weak energy condition and might similarly be understood as the claim that mass / energy density is always non - negative . but it is _ strictly weaker _ than the mass condition , since the weakened mass condition may be satisfied by mass - momentum fields that are spacelike , in the sense that @xmath77 but @xmath79 ( for example , consider @xmath80 , with @xmath81 a spacelike vector field ) . in other words , the mass condition amounts to the weakened mass condition plus the additional assumption that @xmath21 is timelike . we can make this explicit by defining an equivalent condition , the _ modified mass condition_. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * modified mass condition * : a mass - momentum field @xmath21 satisfies the modified mass condition if at any point , @xmath78 and either @xmath35 or @xmath82 is timelike . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the modified mass condition is equivalent to the mass condition , but would appear to be the natural translation of the strengthened dominant energy condition . on the other hand , one can also rewrite the weak energy condition as the _ strengthened weak energy condition_. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * strengthened weak energy condition * : an energy - momentum field satisfies the strengthened weak energy condition if , give any timelike vector @xmath5 at a point , either @xmath35 or @xmath83 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ stating that for all timelike @xmath5 , if @xmath77 , then @xmath84 . and _ this _ condition seems like the natural translation of the ( standard ) mass condition , but it is strictly weaker than the strengthened dominant energy condition ! this situation is summarized in table [ energyconditions ] . .[energyconditions ] this table summarizes the relationship between the various energy condition discussed in the text . single arrows represent `` apparently natural translations '' ; double arrows represent logical implications . [ cols="^,^,^",options="header " , ] there are thus two ways of thinking about the relationship between the energy conditions used in these theorems , depending on which `` natural translations '' one emphasizes . on one way of thinking , the mass condition is essentially the same as the strengthened weak energy condition . from this point of view , then , the strengthened dominant energy condition in the geroch - jang theorem is a strictly stronger assumption than the corresponding assumption in theorem [ w ] . more , one might be inclined to think that one gets something additional for free in the newtonian case , since the mass condition turns out to imply the ( apparently ) stronger modified mass condition , whereas the strengthened weak energy condition does not imply the strengthened dominant energy condition . meanwhile , on the other way of thinking about things , one argues that the strengthened dominant energy condition is essentially the same as the modified mass condition , which is fully equivalent to the mass condition . and so one concludes that the energy conditions required by the two theorems are essentially the same . there is also another possibility , which is to say that one can not perform simple translations between the energy conditions in these two theories at all . i am inclined to endorse this last option , though this raises new questions about how one should compare the theorems . there are a few things to say . first , irrespective of how one tries ( or does not try ) to translate these conditions , there are still two senses in which the strengthened dominant energy condition is stronger than the mass condition . one is that the timelike propagation clause of the strengthened dominant energy condition can be understood as the assumption that the instantaneous speed of matter , relative to any observer , must be strictly less than the speed of light . the corresponding clause of the ( modified ) mass condition , meanwhile , amounts to the assumption that matter can not propagate at _ infinite _ speed relative to any observer . and the assumption that a number must be less than a fixed finite value is stronger than the assumption that it must be finite , but not bounded . the second , more significant sense in which the strengthened dominant energy condition is stronger is that the _ only _ way in which matter in newtonian gravitation can be `` massive '' ( i.e. , have positive mass as determined by some observer ) is if it satisfies the mass condition . matter that satisfies the weakened mass condition but not the mass condition will necessarily have zero mass . and so one might argue that the mass condition is necessary to capture what is meant by `` massive '' in the context of newtonian gravitation . in general relatively , meanwhile , matter can be `` massive '' in two senses , _ without _ satisfying the strengthened dominant energy condition : it can be massive in the sense that it has positive ( mass-)energy density ( i.e. , it satisfies the weak energy condition ) , and it can be massive in the sense that some observers will assign it positive mass density ( i.e. , the relative four - momentum density as determined by some observers is timelike ) . this second sense trades on an important distinction between _ some _ observers assigning positive mass density and _ all _ observers assigning positive mass density . one might have thought that in order for a matter field to be massive , it would be sufficient if some observers , perhaps privileged say co - moving observers making determinations of `` rest mass density '' , when that makes sense and perhaps not , determine that the field has positive mass density . but the strengthened dominant energy condition requires considerably more than this . in geometrized newtonian gravitation , meanwhile , all of these distinction collapse . if anyone determines a matter field has positive mass , then everyone does . a final remark is that , understood within the context of the respective theories , the strengthened dominant energy condition is a more surprising assumption to have to make than the mass condition . one often thinks of relativity theory as _ forbidding _ superluminal propagation of matter , in the sense that somehow the geometric structure of the theory renders superluminal matter incoherent . but here , at least , it seems that we need to rule out superluminal propagation of matter as an additional assumption in order to derive the geodesic principle . this point can be made precise by asking whether one can drop or weaken the energy condition in the geroch - jang theorem and still derive the geodesic principle . and the answer is `` no '' . if one drops the energy condition altogether , it is possible to construct bits of matter that propagate along _ any _ timelike curve @xcite . and if one weakens the energy condition to the weak energy condition or the dominant energy condition , one can construct bits of matter that propagate along spacelike or null curves , respectively @xcite . to be sure , in the newtonian case the mass condition is similarly necessary ( the considerations offered in @xcite can be adapted to show that the weakened mass condition is not enough to get timelike geodesic motion in geometrized newtonian gravitation ) , but this does not seem as striking , since one does not expect newtonian gravitation to imply restrictions on the propagation of matter , even if it is standard to assume that matter can not propagate instantaneously in the theory . this leaves the conservation condition , assumption ( 2 ) in both theorems . the statement of the assumption is identical in both cases , namely that the tensor fields representing matter must divergence free . and in both theories , this assumption is a way of capturing that the bits of matter must be _ free _ in the sense of non - interacting . this interpretation is justified because in both theories there is a standard background assumption that at every point of spacetime , _ total _ energy / mass - momentum must be divergence free , and more , that a particular energy / mass - momentum field fails to be divergence free at a point just in case it is interacting with some other such field at that point . and so , to say that a particular field satisfies the conservation condition everywhere is to say that that field can not be exchanging energy / mass - momentum with any other fields . so far , it would seem that these assumptions have precisely the same status in both theorems . but this is too quick . although the assumptions are equally natural ways of capturing the desired sense of `` free '' in both theorems , they only have that interpretation in the presence of the background assumption regarding the local conservation of total energy / mass - momentum . and there is an argument to be made that this background assumption has a different status in general relativity than in geometrized newtonian gravitation . in general relativity , einstein s equation implies the conservation condition , at least for total source matter . this is because the equation can be written as @xmath85 , and it is a brute geometrical fact ( known as bianchi s identity ) that the right - hand side of this equation is always divergence free . thus , the left - hand side must also be divergence free . the geometrized poisson equation , however , does _ not _ imply the conservation condition . and so , if one has einstein s equation lurking in the background , one might be inclined to say that the background assumption that matter is conserved comes for free in general relativity , whereas it is an additional brute assumption in geometrized newtonian gravitation . there is an important caveat here the argument that einstein s equation implies the conservation condition only applies for source matter , whereas the geodesic principle is supposed to govern _ test _ matter , i.e. , matter that is not treated as a source in einstein s equation but nonetheless , one might think that the conservation condition has a special status even , to anticipate the discussion in the next section , a privileged explanatory status in general relativity because of its relation to einstein s equation . in the next section , i will turn to the question of whether either of these theorems should count as explanations of inertial motion . but before i do so , it will be helpful to sum up the discussion in the present section . i have now made precise the sense in which one can prove the geodesic principle as a theorem of both general relativity and geometrized newtonian gravitation . but , as i hope has become clear , interpreting and comparing these theorems is quite subtle . it is not quite right to say that the theorems have the same interpretation or significance : on the one hand , there is arguably a sense in which the conservation condition , necessary for both theorems , has a different and perhaps privileged status in general relativity ; and on the other hand , there are several senses in which the energy condition required for geroch - jang theorem is stronger than the condition required for the newtonian theorem , both in absolute terms and relative to the respective theories . despite these differences , however , there is at least one important sense in which the status of the geodesic principle is strikingly similar in both theories . in both cases , one _ can _ prove the geodesic principle as a theorem . but to do so , one needs to make strong assumptions about the nature of matter . the status of these assumption will play a central role in what follows . general relativity and geometrized newtonian gravitation , like any physical theory , involve a number of basic assumptions and central principles . for instance , general relativity begins with some background assumptions about matter and geometry : space and time are represented by a four dimensional , possibly curved lorentzian manifold ; matter is represented by its energy - momentum tensor , a smooth symmetric rank two field on spacetime . one then adds some additional assumptions , as principles indicating how to interpret and use the theory . one may stipulate that total energy - momentum at a point must satisfy the conservation condition . one assumes that matter fields satisfy various possible energy conditions , that idealized clocks measure proper time along their trajectories , and that free massive test point particles traverse timelike geodesics . we postulate a dynamical relationship between the geometrical structure of spacetime and the energy - momentum field . and so on . some of these assumptions involve stipulating kinematical structure ; others involve basic constraints and dynamical relationships ; others still tell us how to extract empirical content from the theory . all of them have some claim to centrality or fundamentality in the theory . but they are not necessarily independent . for instance , as i mention above , the conservation condition may be understood as a consequence of einstein s equation , at least for source matter . and so , at least in some contexts , one might want to think of the conservation condition as somehow subordinate to einstein s equation . one might even be inclined to say that it is einstein s equation that really deserves to be called the `` fundamental principle , '' while the conservation condition has some other , less fundamental status or , in other words , that einstein s equation _ explains _ why matter is locally conserved . we might even say that this is what it _ means _ to say that something like the conservations condition is explained by a theory : it can be derived from more fundamental principles in the theory . from this point of view , one might have thought that when einstein , eddington , and others have claimed that general relativity explains inertial motion , in the sense that one can prove the geodesic principle as a theorem , the claim would have been analogous to what i have just said about the conservation condition : namely , one can take some collection of other principles of the theory and use them to derive the geodesic principle . one might then think that the geodesic principle has the same subordinate status as the conservation condition . it may be central to the theory , but not truly fundamental . the fundamental principles are the ones that go into proving the geodesic principle . on this view , one thinks of the foundations of general relativity as a two - tiered system . on the top tier are the truly fundamental principles ; on the lower tier are the other central principles that can be derived from the top - tier principles . initially , perhaps , one thought that the geodesic principle and conservation condition were top - tier principles ; but the geroch - jang theorem and bianchi s identity show that they are really second - tier principles . thinking this way can lead to problems , however . the main moral of the last section was that although one can prove the geodesic principle as a theorem in both general relativity and geometrized newtonian gravitation , to do so one requires strong assumptions about the nature of matter . and so , if we want to move the geodesic principle to the lower tier , it would seem that we need to understand these assumptions as as top - tier principles . but this raises a question : why should we think of _ these _ principles as the truly fundamental ones ? or more specifically , why should we think of the conservation condition and the respective energy conditions as more fundamental than the geodesic principle itself ? if one were committed to the idea that the geodesic principle is a second - tier principle in one or both of these theories , perhaps one would be willing to include the assumptions needed to prove the geodesic principle among the truly fundamental principles of that theory . but it is hard to see how this is an appealing move on independent grounds . even if one were to argue that dynamical principles such as einstein s equation and the geometrized poisson equation are clearly more fundamental than the geodesic principle , it remains the case that the strong energy condition needed to prove the geroch - jang theorem is entirely independent of einstein s equation . ( and neither assumption follows from the geometrized poisson equation . ) more , there is a sense in which one can draw all of the inferential arrows in the opposite direction , at least in one important case . consider an energy / mass - momentum field of the form @xmath86 , for some smooth scalar field @xmath24 and smooth vector field @xmath5 . an energy / mass - momentum field of this form is the natural way of representing a matter field composed of mutually non - interacting massive point particles ( at least when @xmath24 is non - negative ) . and so , since the geodesic principle governs the behavior of free massive test point particles , we can use it to derive features of this matter field : specifically , the geodesic principle implies that the flow - lines of the field , which represent the trajectories of each speck of dust , must be timelike geodesics . these flow lines are just the integral curves of @xmath5 , and so it follows that @xmath5 must be timelike and geodesic ( i.e. , @xmath55 ) . but if @xmath5 is timelike , then @xmath21 satisfies the strengthened dominant energy condition ( or respectively , the mass condition in geometrized newtonian gravitation ) . and if it is geodesic , then @xmath21 is divergence free . thus the geodesic principle allows us to derive that matter fields consisting of non - interacting massive test point particles satisfy precisely the two conditions we need to assume in order to prove the geodesic principle . so perhaps we should not be so quick to declare the conservation condition and energy conditions top - tier in either theory . at very least , it is not perfectly clear that these assumptions are more fundamental than the geodesic principle . but thinking in this way might lead one to conclude that _ neither _ of the geodesic principle theorems has much explanatory significance , since ( the intuition might go ) explanations always proceed from more fundamental or basic facts to less fundamental facts . here , meanwhile , the arrows of fundamentality are muddled . and this would mean that not only is general relativity not special with regard to its explanation of inertial motion it does not explain inertial motion at all ! i do not find the argument i offer in the previous section compelling . it rests on a basic intuition : to explain something like the geodesic principle , one must begin with some truly fundamental principles and then provide an argument for why the principle - to - be - explained must follow from these more fundamental ones . this intuition takes for granted that we can make sense of a distinction between different tiers of fundamentality among the central principles of a theory like general relativity or geometrized newtonian gravitation . and i think that this is a mistake or at least , that there is a more compelling way of thinking about things . consider what the geodesic principle theorems _ do _ accomplish . in both of these cases , the theorems show how in the presence of other basic assumptions of the respective theories , the geodesic principle follows . or in other words , they show that given that one is committed to the rest of ( say ) general relativity , one must also be committed to the geodesic principle . one can not freely change the geodesic principle without also changing the rest of general relaitivty : one can not `` fiddle '' with the theory by ( merely ) replacing the geodesic principle with the assumption that free massive test point particles traverse some other class of curves uniformly accelerating curves , say , or spacelike curves . the geodesic principle is not modular , in the sense that one can not construct a collection of perfectly good theories that differ only with how they treat inertial motion . more , the theorems clarify precisely how it is that the geodesic principle `` fits in '' among the other central principles of general relativity . it seems to me that these reflections suggest a proposal . instead of thinking of the foundations of a physical theory as consisting of a collection of essentially independent postulates from which the rest of the theory is derived , one might instead think of the foundations of a theory as consisting of a network of mutually interdependent principles a collection of interlocking pieces , as in the spherical puzzle in figure [ puzzleball ] . the idea is that , as with the geodesic principle , one should generally expect that many of the central principles of a physical theory may be proved as theorems , given the rest of the theory . trying to make a distinction between the top - tier principles and the second - tier principles of a theory is not fruitful , then , since most , or even all , of the principles can be understood equally well as either postulate or theorem , and indeed , in different contexts it may well be desirable to think of them in different ways . importantly , theories are not modular , in the sense described above . we can not simply replace any given principle with some other one , at least not without changing the rest of the theory in possibly dramatic ways . and theorems like the geroch - jang theorem and its newtonian counterpart are of interest because they exhibit the details of these interdependencies . they show just how the pieces interlock . one way of thinking about the foundations of physical theories would have it that some of the central principles of a theory have a distinguished status as the `` truly fundamental '' principles . an alternative view , which i describe and advocate here , is that the foundations of a theory are better thought of as a network of mutually interdependent principles , interlocking like the pieces of a spherical puzzle . on this view , one would tend to expect that any of the central principles of a theory should be derivable from the rest of the theory with that principle removed , much like the overall shape of a puzzle ball constrains the shape of any individual piece . ] to be sure , nothing i have said thus far should count as an apodictic argument for the view i have described ( call it the `` puzzle ball view '' ) . nor will i give such an argument indeed , i am not sure what an argument for the claim that there are no truly fundamental principles of general relativity would look like . instead , i merely offer the view as an alternative way to conceptualize the kinds of interrelations between the principles of physical theories on display with theorems like the geroch - jang theorem . perhaps the proposal is best conceived as a conjecture , albeit one with some compelling early evidence , for the following reason : while the senses in which , for instance , the conservation condition and the geodesic principle follow from other standard assumptions of general relativity are now established , the senses in which other principles , such as einstein s equation or various energy conditions , are derivable from or constrained by the rest of the theory are less clear . and so we have the skeleton of a mathematical question : are _ all _ of the central principles of relativity theory and geometrized newtonian gravitation ( or other theories still ) indeed mutually interderivable in the way that i have suggested ? most or many of them ? or are the geodesic principle and conservation condition anomalies ? some important work has already been done on this topic : @xcite has shown a sense in which the geometrized poisson equation is the unique dynamical principle compatible with a collection of natural assumptions in newtonian gravitation ; similarly , @xcite and @xcite have shown that there is a sense in which the ( vacuum form of ) einstein s equation can be derived from ( in effect ) the geodesic principle , among other assumptions , and @xcite has argued that there is a sense in which the einstein tensor is the unique tensor that can appear on the left hand side of einstein s equation , even in the non - vacuum case . meanwhile , @xcite and @xcite have argued that even though the conservation condition in geometrized newtonian gravitation does not follow from the geometrized poisson equation , one can nonetheless derive it from other principles , at least if one considers lagrangian formulations of the theory . one might even understand newton s argument for universal gravitation as a kind of heuristic argument that the inverse square law of gravitation is the unique dynamical principle compatible with certain other central principles of standard newtonian gravitation ( including generalized empirical facts , such as the universality of elliptical orbits ) . results such as these provide tentative evidence for my basic hypothesis , that one should expect many or all of the central principles of these spacetime theories to be mutually interdependent . but i do not think these results are yet conclusive . specifically , what has not been done is to systematically study such results in order to try to characterize , in the way that has now been done for the geodesic principle theorems , ( 1 ) just what the assumptions going into these theorems are , ( 2 ) how natural the assumptions are in the contexts of the relevant theories , and ( 3 ) how these assumptions , in turn , depend on the other central principles of the theories ( if they do ) . and the attractiveness of the proposal presented here turns on the answers to these questions . it is only after a project of this form has been carried out that one can fully evaluate whether the central principles of of these theories are really as tightly intertwined as the puzzle ball view would have it . that said , if a careful study of this sort reveals that only some of the central principles of a theory are interconnected , it may still be fruitful to think about the foundations of theories in the way i propose here , since the discovery that some central principles of a theory ( say , energy conditions in general relativity ) are more peripheral than others need not imply that one can make sense of a unique or privileged collection of the most fundamental or basic principles . much will depend on just what the structure of the situation turns out to be . it is worth emphasizing that mapping out these kinds of relations between the central principles of a physical theory is of some independent interest , since understanding the extent to which the central principles of general relativity in particular are mutually interdependent could play an important role in the construction of future theories ( and in some ways , it already has ) . the reason has to do with this idea of `` fiddling '' with physical theories . there is a long - standing tradition of attempting to modify general relativity with small changes : for instance , in brans - dicke theory , one modifies einstein s equation to include an additional scalar field ; in teves gravitational theories , one additionally considers vector fields . in still other cases , one modifies general relativity by allowing derivative operators with torsion . in each of these examples ( and many others ) , one makes what appears to be a local change in the central principles of general relativity . but these small changes can have dramatic consequences : for instance , in einstein - cartan theory , a modification of general relativity with torsion , the conservation condition does not generally hold . one does not have a geodesic principle , at least in the ordinary sense , since in general the collection of self - parallel curves picked out by the derivative operator do not agree with the collection of extremal curves picked out by the metric , and free massive test point particles need not propagate along either class of curves . thus , apparently small tweaks can lead to a dramatically different theory , conceptually speaking . a clearer picture of just how the central principles of general relativity _ do _ fit together and constrain one another may provide important clarity into just what the ramifications of these `` small '' modifications to the theory are , and more , may help guide us in the search for alternative theories of gravitation , by indicating which principles are more or less tightly connected to which others . indeed , for this reason there is a sense in which the situation i described above , where some principles are very tightly interlocking and others turn out to be more loosely connected ( for instance , some principles play a role as assumptions in some theorems , but can not be proved in complete generality themselves ) is the most interesting , from the practical perspective of mapping out the space of possible future theories . in the next section , i will return to the question of explanation , now from the perspective of the present view . but before i do so , i want to clarify the puzzle ball view slightly , as the language i have used to describe it may call to mind two other well - known ideas . it seems to me that the view i have described is distinct from both . first , note that the present proposal involves a picture of theories on which one emphasizes the ways in which the principles cohere with one another . this way of thinking may be reminiscent of coherentism in epistemology , a variety of anti - foundationalism that holds that to justify a belief is to show how it coheres with one s other beliefs ( cf . * ? ? ? but there is at least one major difference . coherentism takes the coherence of one s beliefs to be a form of justification for those beliefs . nothing about the puzzle ball view should be taken to suggest that the justification for general relativity comes from the apparent fact that one can derive certain central principles from others rather , the justification for the theory is based on its empirical successes . or perhaps more precisely , our justification for general relativity is essentially independent of the relationship between the theory s central principles . to see the point most clearly , one might well expect both general relativity and geometrized newtonian gravitation to be coherent , in the sense of having mutually interdependent central principles . but this does not imply that they are equally well justified indeed , general relativity is better justified than geometrized newtonian gravitation even if the pieces of geometrized newtonian gravitation are more tightly interlocking . another view that the puzzle ball view may be reminiscent of is some variety of quinean holism ( cf . * ? ? ? quine famously used the `` web of belief '' metaphor when arguing for the interdependencies of our scientific beliefs , and against the analytic / synthetic distinction . one might worry that the puzzle ball picture above is just an alternative metaphor used to make a strikingly similar point indeed , the claim that we can not make a fruitful distinction between top - tier and second - tier principles sounds like an argument against an analytic / synthetic distinction , at least in the narrow domain of the foundations of certain physical theories . and perhaps it is right that i have recapitulated quine here , though if it is , i think the point deserves to be made again since it is relevant for the present discussion of the geodesic principle . still , while this chapter is not the occasion for detailed quine exegesis , i will point to two ways in what i have proposed is _ prima facie _ different from quine s holism , at least on the web - of - belief version . the first difference concerns just what the holism is supposed to be doing . quine uses the interdependencies between beliefs as an argument for a radical form of conventionalism : when faced with evidence that conflicts with our beliefs , we have considerable leeway in choosing which parts of the web of beliefs to revise . indeed , the web image is supposed to support a distinction between `` central '' or `` core '' beliefs and `` peripheral '' beliefs such that we can always accommodate challenges to our full collection of beliefs by modifying only the peripheral beliefs and leaving the core beliefs intact . but this is precisely the opposite of what i have argued here , at least with regard to the foundations of spacetime theories . instead , the idea is supposed to be that the foundations of physical theories are _ not _ modular , and that in general one has remarkably little latitude in how one revises a theory in light of new evidence . and this , i take it , is a desirable feature , since it provides a way out of the radical conventionalism i just described . since the various principles of a physical theory constrain one another , we have very few degrees of freedom for enacting minor changes in theories in light of new evidence . the second difference is related ( and relates , too , to coherentism as described above ) . quine s web of belief is supposed to be a ( descriptive ) metaphor for the sum total of one s beliefs . the view i have described here is much narrower in its scope . i do not claim that all of one s beliefs interlock in the way described ; nor do i claim that scientific knowledge as a whole can be characterized by a puzzle ball . the view does not even hold that particular scientific theories have this feature . the suggestion is that the _ central principles _ of some scientific theories are mutually interderivable , that the foundations of some physical theories should be though of in a certain way . i have been deliberately vague about just what is supposed to count as a central principle , in large part because i think that trying to list these principles in advance , even for well understood theories such as general relativity or geometrized newtonian gravitation , would be unproductive . in fact , one might expect that a full account of just what the central principles of a theory are may have to wait until one sees just what assumptions are necessary nodes when trying to map out the network of interconnected principles at the heart of a given physical theory . what i have done so far and what i think can be done at this stage is give examples of central principles of particular theories . and so one can say that among the central principles of general relativity , for instance , are things such as the conservation condition , einstein s equation , the geodesic principle , and various energy conditions . but the point is that a claim about a collection of principles of this specific character is quite different from a claim about human knowledge quite broadly . note that this last point means that there is still a robust sense in which one can think of some parts of a theory as having a special `` fundamental '' status , even on the puzzle ball view . specifically , one might take _ all _ of the central principles of a theory to be fundamental . this leaves quite a bit of a theory as non - fundamental for instance , particular predictions of a theory would not be among the central principles , and so these would not count as fundamental . if the puzzle ball view is to be viewed as anti - foundational , then , it is only with regard to determinations of relative fundamentality _ among _ the central principles of a theory . now that i have described the puzzle ball view in some detail , i can return to the question of principle interest in this paper : namely , is there a sense in which we should understand the geroch - jang theorem and its newtonian counterpart as explanations ? as a first remark , let me reiterate that if we are thinking in terms of the puzzle ball view , it does not make sense to think of theories in terms of `` top - tier '' principles and other , derived principles : in short , there is no way to make the distinction , at least among the central principles of the theory . none of the assumptions of a theory are distinguished as the truly basic or fundamental ones . and if if this is right , then the kind of explanation that we apparently can not get of the geodesic principle in general relativity and in geometrized newtonian gravitation is uninteresting . no , we can not derive the geodesic principle in either theory from more fundamental principles , but that is because it does not make sense of talk of unambiguously `` more fundamental '' principles in the first place . instead , what we can do is show how the geodesic principle in both of these theories fits into the rest of the puzzle ( as it were ) . this , too , may be understood as an answer to the question , `` why do bodies move in the particular way that they do in the absence of an external force ? '' these theorems reveal that in the absence of an external force , in the context of their respective theories , bodies _ must _ move along timelike geodesics . in other words , the other basic assumptions of the theory constrain the motion of ( small ) bodies . why timelike geodesic motion rather than any other ? because in general relativity , we understand matter to be conserved , and to be such that observers always attribute instantaneous subluminal velocities to it at any point . and it turns out that these assumptions , in the presence of the rest of the theory , imply that the only curves along which free massive test point particles can propagate are timelike geodesics . if we are committed to the rest of general relativity , then there is only one candidate principle for inertial motion . so do general relativity and/or geometrized newtonian gravitation explain inertial motion ? given the considerations just mentioned , i think the answer in both cases is `` yes '' , so long as one understands `` explain '' in the right way . at very least , these theorems provide deep insight and understanding into why bodies move in the particular way that they do in the absence of any external force which is precisely what we were after when we asked the question . moreover , the insight provided is that , in the context of the other central principles of the theories , the geodesic principle is _ necessary _ , the only principle governing inertial motion that is compatible with our other principles . it is a demonstration of precisely the ways in which the working parts of general relativity and geometrized newtonian gravitation constrain one another . that said , this kind of explanation differs in some important ways from other explanations that one may be accustomed to thinking about . in particular , if the puzzle ball view is correct , the kind of explanation i have just described need not be asymmetrical . that is , if general relativity might be said to explain inertial motion in the present sense by appealing to the fact that one can derive the geodesic principle from various other assumptions in the theory , one should _ not _ conclude that the geodesic principle can not play a role in other derivations that should also count as explanatory even derivations of the assumptions going into the geroch - jang theorem or its newtonian counterpart . indeed , one should expect that just as the geodesic principle is constrained by the other central assumptions of general relativity , so too are the conservation condition , the strengthened dominant energy condition , and even einstein s equation constrained . and by the same reasons i have offered above for the view that one might justly call the geroch - jang theorem an explanation of inertial motion , one might also say that explanations can be given for the conservation condition or einstein s equation , by showing how these principles of general relativity are derivable from the other central principles of the theory . in other words , general relativity explains inertial motion by appeal to einstein s equation , but it may equally well explains einstein s equation by appeal to the geodesic principle and other central assumptions of general relativity . this observation may give some readers pause . there is , by now , a long tradition of philosophers of science worrying about the so - called `` problem of explanatory asymmetry '' ( cf . * ) : intuitively , explanations appear to run in one direction and only one direction . the trajectory of a comet may explain why we see a bright light in the nighttime sky once every few hundred years , but a bright light in the nighttime sky can not explain the trajectory of a comet ; the height of a flagpole may explain the length of its shadow at sunset , but the length of the shadow does not explain the height of the flagpole . and so , many philosophers have argued , an account of explanation that allows symmetrical explanations situations where a explains b and b explains a is _ prima facie _ unacceptable . a few remarks are in order . first , @xcite has argued , i think correctly , that explanation should be understood as essentially pragmatic and in particular , that explanations should only be understood as responses to certain classes of question . to determine whether or not some particular response to a why question ( say ) should count as a satisfactory explanation depends on the context in which the question was asked and the particular demands of the questioner . while in some contexts we might want to say that a particular explanation runs in only one direction , there may well be other contexts in which the explanation would run in the other direction . if one is thinking in this way then the present example is simply a special case : if the question `` why do bodies move in the particular way that they do in the absence of an external force ? '' is understood as `` does general relativity require us to adopt the geodesic principle as the central principle governing inertial motion ? '' , then one is rightly satisfied by a response along the lines of the geroch - jang theorem , even if in other contexts i.e . , in response to other questions one might appeal to the geodesic principle to explain ( say ) einstein s equation . but there is also a more important point to make , here : i do not claim to be offering an `` account of explanation '' , or anything like it . i have not suggested that a necessary or even sufficient condition for being an explanation is to show how the thing to be explained `` fits in '' with the rest of a physical theory , in the sense that it is derivable from other central principles of a theory . the point , rather , is to try to spell out the sense in which a particular class of theorems that show how the central principles of a spacetime theory fit together might be understood as explanatory to say what , precisely , the theorems are doing , and why one might think of this as a kind of explanation , at least on the puzzle ball view . not all explanations work this way , nor do they need to in order for the story i have told here to be correct . and so , the fact that in _ some _ cases , we would want to say that if @xmath87 explains @xmath88 then @xmath88 can not explain @xmath87 in no way undermines the claim that the present explanations simply do not work that way . this point can be made most starkly by pointing to various other questions one can ask about inertial motion , even in general relativity , whose answers would be quite different from the geroch - jang theorem . consider , for example , a question concerning a particular instance of inertial motion . why , one might ask , does the perihelion of mercury s orbit precess ? one would answer this by appealing to some particular initial state of mercury and features of spherically symmetric solutions to einstein s equation to show that mercury s orbit is the only allowed trajectory for a body with certain properties in a solar system like ours . the geodesic principle may play a role in this argument , insofar as one might idealize mercury as a free massive test point particle , and einstein s equation may play a role , insofar as one would want to consider a spacetime that is a solution to the equation , but the argument would have nothing to do with the geroch - jang theorem . and moreover , one would expect this sort of explanation to be asymmetric : einstein s equation and the geodesic principle , along with some details concerning the state of the solar system and initial conditions for mercury , explain the precession of the perihelion of mercury ; the precession of the perihelion of mercury does not explain einstein s equation or the geodesic principle . but this is just the point . if the geroch - jang theorem and its newtonian counterpart should be countenanced as explanations , it is only because they are satisfactory answers to particular questions , and they are only explanatory in the context of _ those _ demands for explanation . a question concerning the orbit of mercury is quite different from a question concerning the nature of inertial motion generally . and these theorems answer only the most general version of the question : why _ this _ principle as opposed to any other ? this is no mean task , but it is a specific one , and it needs to be treated with care . versions of this work have been presented at cea - saclay ( paris ) , bergisch university wuppertal , university of pittsburgh , university of chicago , university of texas at austin , columbia university , brown university , new york university , university of western ontario ( twice ! ) , university of california berkeley , university of california irvine , yale university , and at the ` new directions ' conference in washington , dc . i am grateful to very helpful comments and discussion from all of these audiences , and particularly to ( in no special order ) john manchak , giovanni valente , craig callender , alexei grinbaum , harvey brown , david wallace , chris smeenk , wayne myrvold , erik curiel , ryan samaroo , john norton , john earman , howard stein , mike tamir , bryan roberts , shelly kagan , sahotra sarkar , josh dever , bob geroch , bill wimsatt , david albert , tim maudlin , sherri roush , josh schechter , and chris hill . special thanks are due to david malament , jeff barrett , kyle stanford , and pen maddy for many helpful discussions and comments on previous versions of this work . thank you , too , to dennis lehmkuhl for organizing the 2010 workshop on which this volume is based , and for editing the volume . damour , t. , 1989 . the problem of motion in newtonian and einsteinian gravity . in : hawking , s. w. , israel , w. ( eds . ) , three hundred years of gravitation . cambridge university press , new york , pp . 128198 . kvanvig , j. , 2007 . coherentist theories of epistemic justification . in : zalta , e. n. ( ed . ) , the stanford encyclopedia of philosophy . stanford university , stanford , ca , available at : http://plato.stanford.edu / entries / justep - coherence/. malament , d. , 2012 . a remark about the `` geodesic principle '' in general relativity . in : frappier , m. , brown , d. h. , disalle , r. ( eds . ) , analysis and interpretation in the exact sciences : essays in honour of william demopoulos . springer , new york , pp .
self - similar fractals are usually constructed as the compact invariant set of an iterated function system . analysis on these sets is still a rapidly growing field in mathematics . one approach is to study self - similar graphs as the discretization of fractals and then transfer the results on these graphs back to the fractals via rescaling . in this way one can construct brownian motion on fractals by starting with random walks on graphs , see for example barlow @xcite , lindstrm @xcite and the references therein . contracted images of a fractal as compact subsets of the fractal are called _ cells_. correspondingly , self - similar graphs have _ cell graphs _ as finite subgraphs which carry essential information of the whole graph . malozemov and teplyaev defined self - similarity of graphs axiomatically in @xcite . their definition was restricted to the case where the cells of the graph have exactly two boundary points . in @xcite one of the authors introduced self - similarity for an arbitrary number of boundary points . another axiomatic approach for self - similar graphs with more boundary points was chosen in @xcite . sabot constructed self - similar graphs using equivalence relations on word spaces , see @xcite . probably the most extensively studied self - similar graph is the sierpiski graph . in @xcite grabner and woess considered the green function of the simple random walk on the sierpiski graph which describes the returning of the random walk to a certain origin vertex . a combinatorial substitution based on path arguments was used to obtain a functional equation for this green function . iterating this equation they obtained the analytic continuation of the green function as a rapidly converging product of rational terms . the asymptotic behaviour and oscillation phenomena were computed by analysing the singularity of the green function in @xmath3 . these techniques had been introduced by flajolet and odlyzko in @xcite . _ symmetrically self - similar _ graphs were constructed in @xcite as class of self - similar graphs such that the combinatorial substitution mentioned above can be applied . sections [ sec : asymptotic ] to [ sec : transition ] of the present paper are a generalisation and a further development of the asymptotic analysis in @xcite . in section [ sec : ssg ] we recall from @xcite the axiomatic definition of self - similarity of graphs and a graph theoretic analogue to the banach fixed point theorem for self - similar graphs . heuristically , this theorem says that a self - similar graph has a `` centre '' , either as an origin vertex or as an origin cell . we will later focus on self - similar graphs with an origin vertex . our axiomatic approach to self - similarity of graphs is based on contracting a given infinite graph . on the other hand one can obtain self - similar graphs by a `` blow - up '' construction , starting with a finite cell graph . theorem [ thm : blowup ] yields this construction explicitly , the resulting graphs may have more irregular structures than discretized post critically finite self - similar ( pcfss ) sets , see @xcite for the notion of pcfss sets . it is proved that a symmetrically self - similar graph is bipartite if and only if its cell graphs are bipartite . section [ sec : green ] is devoted to the study of green functions . for the sierpiski graph , a functional equation @xmath4 is valid for the green function @xmath5 at the origin vertex , which was first observed by rammal and toulouse @xcite , see also @xcite . in @xcite this equation was generalised to the class of so called _ symmetrically _ self - similar graphs . one property of these graphs is that their cell graphs are all isomorphic . the functions @xmath6 and @xmath7 are rational functions associated with the simple random walk on the cell graph @xmath0 . the _ transition function _ @xmath6 is the generating function of the probabilities that the simple random walk , starting in the boundary of @xmath0 , hits another vertex in the boundary for the first time after exactly @xmath2 steps . the product @xmath8 converges on the fatou set @xmath9 of @xmath6 ( possibly with the exception of countably many points ) to the green function @xmath5 as a solution of the functional equation above . the point @xmath10 is the only attracting fixed point of @xmath6 and the fatou set @xmath9 contains @xmath11 . the simple random walk on a cell graph @xmath0 of a symmetrically self - similar graph @xmath12 is studied in section [ sec : geometric ] . the number @xmath13 of vertices in the boundary of a cell graph does not depend on the choice of the cell . the graph @xmath0 consists of @xmath14 amalgamated copies of the @xmath13-complete graph @xmath15 . this parameter @xmath14 corresponds to the usual _ mass scaling factor _ of self - similar sets . the number @xmath16 is called _ time scaling factor_. let this random walk start at a vertex @xmath1 in the boundary of @xmath17 . then @xmath18 is the average time until hitting another boundary vertex than @xmath1 . for symmetrically self - similar graphs with bounded geometry it is proved that @xmath19 , where @xmath20 is the expected number of returns to @xmath1 before hitting a new vertex in the boundary of @xmath0 and @xmath21 is the _ resistance scaling factor _ , see @xcite . for the proof we use the relation @xmath22 . in section [ sec : asymptotic ] we study the asymptotic behaviour of the @xmath2-step return probabilities . using the functional equation of section [ sec : green ] we derive a local singular expansion of @xmath5 in @xmath3 , @xmath23 for all @xmath24 such that @xmath25 and @xmath26 , where @xmath27 and @xmath28 . the function @xmath29 is @xmath30-periodic and holomorphic on some horizontal strip around the real axes . in @xcite this type of expansion was computed for the sierpiski graph with a weaker error term . using the method of singularity analysis we obtain the asymptotics of the @xmath2-step return probabilities @xmath31 to an origin vertex @xmath32 . for the case that @xmath12 is not bipartite we have @xmath33 where @xmath34 is a @xmath30-periodic , holomorphic function defined on some horizontal strip around the real axes . if @xmath12 is bipartite , then @xmath35 and @xmath36 satisfies the above asymptotic behaviour . this type of oscillation seems to be typical for random walks on self - similar graphs and brownian motion on fractals , see @xcite . it is proved in section [ sec : oscillation ] that the function @xmath34 is not constant if the julia set @xmath37 of @xmath6 is a cantor set . in section [ sec : transition ] it is shown that for any pair of vertices @xmath38 and @xmath39 , the transition probabilities @xmath40 satisfy the same asymptotic behaviour as @xmath31 . for this purpose we use a ratio limit theorem , see @xcite . section [ sec : line ] is devoted to the case where the cell graph is a finite line @xmath41 of length @xmath2 . the corresponding transition functions @xmath42 are conjugated to the chebychev polynomials @xmath43 via conjugacy map @xmath44 . their julia set is @xmath45 and the periodic function @xmath34 is constant . we study the probability for hitting a given vertex in the minimal number of steps for general , locally finite , reversible markov chains . this probability is maximal whenever the process can be projected on the simple random walk on @xmath41 . from this we deduce that the cell - graphs of symmetrically self - similar graphs are isomorphic to @xmath41 if and only if the transition function is conjugated to @xmath43 in the sense above . we are interested in characterisations of this type because we believe that these are the only symmetrically self - similar graphs with a constant periodic function @xmath34 . this would mean that in all other cases we may observe a non trivial oscillation of the asymptotic transition probabilities . in section [ sec : examples ] some examples are discussed . graphs @xmath46 with vertex set @xmath47 and edge set @xmath48 are always supposed to be undirected , without loops or multiple edges . vertices @xmath38 and @xmath39 are _ adjacent _ if @xmath49 is an edge in @xmath48 . the _ degree _ @xmath50 of @xmath38 is the number of vertices in @xmath47 being adjacent to @xmath38 . a graph has _ bounded geometry _ if the set of vertex degrees is bounded . the _ vertex boundary _ or _ boundary _ @xmath51 of a set @xmath17 of vertices in @xmath47 is the set of vertices in @xmath52 which are adjacent to some vertex in @xmath17 . a _ path of length @xmath53 _ from a vertex @xmath1 to a vertex @xmath54 is an @xmath55-tuple @xmath56 of vertices such that @xmath57 and @xmath58 are adjacent for @xmath59 . the path @xmath60 is _ closed _ if @xmath61 . the distance @xmath62 of @xmath1 and @xmath54 in @xmath12 is the length of the shortest path from @xmath1 to @xmath54 . a set @xmath17 of vertices is called _ connected _ if any pair of vertices in @xmath17 can be connected by a path in @xmath12 which does not leave @xmath17 . we write @xmath0 for the subgraph of @xmath12 spanned by @xmath63 . we briefly repeat the definition of self - similar graphs , see definitions 1 and 2 in @xcite or in @xcite : let @xmath64 be a set of vertices in @xmath47 and let @xmath65 be the set of connected components in @xmath66 . we define the _ reduced graph _ @xmath67 of @xmath12 be setting @xmath68 and connecting two vertices @xmath38 and @xmath39 in @xmath69 by an edge if and only if there exists a component @xmath17 in @xmath65 such that @xmath38 and @xmath39 are in the boundary of @xmath17 . [ def : selfsimilar ] a connected infinite graph @xmath12 is called _ self - similar _ with respect to @xmath70 and @xmath71 if 1 . no vertices in @xmath64 are adjacent in @xmath12 , 2 . the intersection of the boundaries of two different components in @xmath65 contains not more than one vertex and 3 . @xmath72 is an isomorphism between @xmath12 and @xmath67 . we will write @xmath73 instead of @xmath74 and @xmath75 instead of @xmath76 , where @xmath77 denotes the @xmath2-fold iterate of @xmath72 . components of @xmath78 are _ @xmath2-cells _ , 1-cells are also just called _ cells_. an _ origin cell _ is a cell @xmath17 such that @xmath79 . a fixed point of @xmath72 is called _ origin vertex_. see section [ sec : examples ] for examples . the following lemma is a graph theoretic analogue to the banach fixed point theorem , see ( * ? ? ? * theorem 1 ) or ( * ? ? ? * lemma 1 ) . [ lem : banach ] let @xmath12 be self - similar with respect to @xmath64 and @xmath72 . then @xmath12 is self - similar with respect to @xmath80 and @xmath81 , for any positive integer @xmath82 . either @xmath12 has exactly one origin cell and no origin vertex or exactly one origin vertex @xmath32 . if in this case @xmath12 is locally finite , then there is a positive integer @xmath2 such that the subgraphs @xmath83 of @xmath12 which are spanned by @xmath84 , for components @xmath85 in @xmath86 , are self - similar graphs with respect to @xmath87 the graphs @xmath83 have exactly one origin cell and @xmath32 is their origin vertex . let @xmath12 be a graph which is self - similar with respect to @xmath64 and @xmath88 . we call @xmath89 the _ @xmath2-cell graph _ @xmath90 , 1-cell graphs are also called _ cell graphs_. the graph @xmath12 is _ symmetric _ ( or _ doubly symmetric _ ) if it is locally finite and satisfies the following axioms : 1 . all cells are finite and for any pair of cells @xmath17 and @xmath91 in @xmath65 there exists an isomorphism @xmath92 of @xmath0 and @xmath93 such that @xmath94 . the automorphism group @xmath95 of @xmath0 acts doubly transitive on @xmath51 , which means that it acts transitively on the set @xmath96 where @xmath97 is defined as @xmath98 for any @xmath99 . if a self - similar graph @xmath12 satisfies ( s1 ) then let @xmath13 be the number of vertices in the boundary of a cell . axiom ( s1 ) implies that @xmath13 is independent of the choice of the cell @xmath17 . the next theorem describes a ` blow - up ' construction for self - similar graphs satisfying axiom ( s1 ) . [ thm : blowup ] let @xmath100 be a finite , connected graph and let @xmath101 be a subset of @xmath102 with the following properties : * the set @xmath101 has at least two elements , @xmath103 is connected and no pair of vertices in @xmath101 is adjacent in @xmath5 . * the graph @xmath5 consists of @xmath14 complete subgraphs @xmath104 which are isomorphic to the complete graph @xmath15 with @xmath13 vertices , such that @xmath105 and any two of these complete subgraphs have at most one vertex in common . then for each @xmath106 in @xmath107 there is a graph @xmath12 , a set @xmath108 and a map @xmath109 such that @xmath12 is self - similar with respect to @xmath64 and @xmath72 , and the following statements are true : the set @xmath103 is the unique origin cell of @xmath12 , @xmath5 is the corresponding cell graph and @xmath110 . for any cell @xmath17 there exists an isomorphism @xmath111 of @xmath0 and @xmath5 such that @xmath112 . a vertex @xmath32 is the origin vertex of @xmath12 if and only if it is a fixed point of @xmath72 in @xmath113 . in this case @xmath114 and @xmath115 is connected . we construct sequences of graphs @xmath116 , sets of vertices @xmath117 , where @xmath118 , and functions @xmath119 inductively such that they satisfy the following : * the function @xmath120 is an isomorphism between @xmath121 and the reduced graph @xmath122 . * the graph @xmath123 ( seen as union of graphs and not as set theoretic union ) is self - similar with respect to @xmath124 and @xmath125 , where @xmath72 is defined by @xmath126 , * and @xmath12 has the properties required in the theorem . first we set @xmath127 , @xmath128 , @xmath129 and @xmath130 . let @xmath131 be any bijective function . we suppose that there are graphs @xmath132 and sets of vertices @xmath133 , for @xmath134 , such that @xmath135 is a subgraph of @xmath132 and there is an isomorphism @xmath136 of @xmath135 and @xmath137 . let @xmath138 be a disjoint , isomorphic copy of @xmath139 and let @xmath140 be the corresponding isomorphism . in @xmath138 the image @xmath141 is now replaced by @xmath139 such that for any @xmath142 the vertex @xmath143 is identified with @xmath144 . in the resulting graph we replace every @xmath13-complete graph @xmath145 , which is not completely contained in @xmath139 , by an isomorphic copy of @xmath5 such that any different vertices in the copy , that correspond to vertices in @xmath101 , are amalgamated with different vertices in @xmath146 . in general this procedure of replacing the @xmath13-complete graphs is not unique . let @xmath147 be the resulting graph . we set @xmath148 . the function @xmath149 defined by @xmath150 is an isomorphism between @xmath139 and @xmath151 . the sequences @xmath152 and @xmath153 are increasing and @xmath123 is self - similar with respect to @xmath124 and @xmath125 defined by @xmath154 . this construction implies that ( a ) and ( b ) are satisfied . we have @xmath155 for the ` blow - up ' function @xmath72 and vertices @xmath38 and @xmath39 in @xmath156 and @xmath157 for its inverse contraction @xmath158 which is defined on @xmath64 , for details see ( * ? ? ? * theorem 1 ) . this implies that a fixed - point @xmath32 has must lie in @xmath159 and this intersection is a singleton . the set @xmath160 is connected in @xmath12 , thus @xmath115 is connected . by choosing @xmath161 every connected finite graph which is not complete can occur as cell graph of a self - similar graph satisfying ( s1 ) . any self - similar graph with a unique origin cell can be obtained as a blow - up in the sense of theorem [ thm : blowup ] , thus it characterises self - similar graphs which remain connected by removing an origin vertex , see lemma [ lem : banach ] . let @xmath12 be a symmetrically self - similar graph and let @xmath17 be any 1-cell of @xmath12 . then there exists exactly one 2-cell @xmath162 , such that @xmath163 . we denote the 1-cells being contained in @xmath162 by @xmath164 . the cell graph @xmath0 consists of @xmath14 amalgamated @xmath13-complete graphs @xmath165 , where @xmath166 and @xmath167 , see @xcite . for a vertex @xmath168 we call @xmath169 the _ branching number _ of @xmath1 . by axiom ( s2 ) , @xmath95 acts transitively on @xmath51 . thus all vertices @xmath1 of the boundary @xmath51 have the same branching number @xmath170 . furthermore , axiom ( s1 ) implies that all cell graphs are isomorphic and @xmath170 does also not depend on the choice of the cell @xmath17 . consequently , we define the branching number @xmath171 of a symmetrically self - similar graph as the branching number @xmath170 of any vertex @xmath1 in @xmath64 . this branching number characterises symmetrically self - similar graphs with bounded geometry : [ thm : bounded ] a symmetrically self - similar graph @xmath12 has bounded geometry if and only if @xmath172 . this theorem is an immediate consequence of theorem 3 in @xcite . the next theorem is a special case of corollary 1 in @xcite . [ thm : unbounded ] let @xmath12 be a symmetrically self - similar graph with unbounded geometry . then either there exists no origin vertex and @xmath12 is locally finite . or there is an origin vertex which is then the only vertex with infinite degree . in the following lemma we characterise those self - similar graphs satisfying ( s1 ) , which are bipartite . [ lem : bipartite ] let @xmath12 be a self - similar graph with respect to @xmath64 and @xmath72 satisfying axiom ( s1 ) . then @xmath12 is bipartite if and only if some cell graph is bipartite . furthermore , we have @xmath161 if @xmath12 is bipartite . if @xmath12 is bipartite then every cell graph is bipartite , since it is a subgraph of @xmath12 . now suppose that some cell graph is bipartite , whence all cell graphs are bipartite thanks to axiom ( s1 ) . a @xmath13-complete graph is bipartite if and only if @xmath173 . any cell graph @xmath0 consists of @xmath14 amalgamated copies of @xmath13-complete graphs , hence @xmath161 . every closed path in @xmath0 has even length . either all paths in @xmath0 connecting the two vertices in @xmath51 have even length or they have all odd length . suppose that @xmath12 is not bipartite . then there exists a closed path @xmath60 in @xmath12 with minimal odd length . this path can not be completely contained in one cell graph , hence it meets vertices in @xmath64 . let @xmath174 with @xmath175 and @xmath176 for all @xmath177 . let @xmath58 , @xmath178 be vertices in @xmath64 such that @xmath179 and @xmath180 for all @xmath181 . then @xmath182 , since @xmath60 is not contained in one cell graph . moreover , we have @xmath183 . otherwise @xmath184 would be a closed path in one cell graph and @xmath185 would be a closed path in @xmath12 with odd length strictly smaller than @xmath2 . let @xmath186 be all those indices , such that @xmath187 for @xmath188 . since the lengths of the paths @xmath189 for @xmath190 have same parity , @xmath191 is a closed path in @xmath12 of odd length strictly smaller than @xmath2 in contradiction to the minimality of @xmath60 . in this section we repeat and reformulate results and definitions from @xcite . let @xmath192 be the matrix of transition probabilities for the simple random walk on a symmetrically self - similar graph @xmath12 which is locally finite . the green function for vertices @xmath38 and @xmath39 in @xmath47 is defined as the generating function @xmath193 of the @xmath2-step transition probabilities @xmath40 from @xmath38 to @xmath39 for @xmath194 , where @xmath195 for any @xmath196 and @xmath28 . if the green function is considered as infinite dimensional matrix , we have @xmath197 let @xmath17 be a cell and let @xmath101 be a nonempty subset of @xmath51 . then @xmath198 denotes the transition matrix of the simple random walk on @xmath0 with absorbing boundary @xmath101 . this means that @xmath199 if @xmath200 is adjacent to a vertex @xmath39 in @xmath201 and @xmath202 otherwise . we define @xmath203 for all @xmath24 such that @xmath204 is invertible . the entries of the matrix @xmath205 are rational functions in @xmath206 which are holomorphic on @xmath207 for some @xmath208 . this follows from the fact that the simple random walk is absorbed by @xmath101 almost surely , since the graph @xmath0 is finite . let @xmath209 , then we define the rational functions @xmath6 and @xmath7 by @xmath210 the functions @xmath6 and @xmath7 are independent of the choice of @xmath1 because @xmath95 acts transitively on @xmath51 . furthermore , double transitivity implies @xmath211 for any vertex @xmath212 . now @xmath6 is the generating function of the probabilities that the simple random walk on @xmath0 starting in some vertex @xmath1 in @xmath51 hits a vertex in @xmath213 for the first time after exactly @xmath2 steps . whereas @xmath7 is the generating function of the probabilities that the random walk starting in @xmath1 returns to @xmath1 after @xmath2 steps without hitting a vertex in @xmath213 before . the start is counted as first visit , thus @xmath214 . we call @xmath6 the _ transition function _ and @xmath7 the _ return function _ of @xmath12 . we write @xmath215 instead of @xmath216 if @xmath12 is a self - similar graph with origin vertex @xmath32 . [ lem : funceq ] let @xmath12 be a symmetrically self - similar graph with origin vertex @xmath32 . then @xmath217 for all @xmath218 . this functional equation was shown for the sierpiski graph in @xcite and @xcite . it is a special case of lemma 6 in @xcite . the basic idea behind this identity is the substitution @xmath219 which corresponds to a combinatorial path substitution reflecting the self - similarity of the graph . for more details see section 4 in @xcite . the number of orbits of @xmath95 on the set @xmath220 is the number of ` types ' of transitions from one boundary vertex in @xmath51 to another . it is the number of variables that is needed for the functional equations of the green functions . similarly , this is the number of variables in the renormalization equations for constructing brownian motion on fractals ( see for example @xcite ) . axiom ( s2 ) ensures that only one variable is needed . our techniques would apply to more general self - similar graphs if the dynamics of multidimensional transition functions could be understood . let @xmath221 be the set of all poles of the function @xmath7 , and let @xmath222 be the set of all @xmath223 such that @xmath224 for some @xmath53 . the following lemma corresponds to the lemmata 9 , 11 , 12 and theorem 4 in @xcite . [ lem : dynamics ] let @xmath12 be a symmetrically self - similar graph with bounded geometry . the point @xmath3 is a repelling fixed point of the transition function @xmath6 . the point @xmath10 is a super - attracting fixed point of @xmath6 , and it is the only attracting fixed point . the fatou set @xmath9 of @xmath6 is the immediate basin of attraction of @xmath10 . the julia set @xmath37 of @xmath6 is a subset of @xmath45 . the set @xmath222 is a subset of @xmath45 . every green function @xmath225 is holomorphic in @xmath226 and can not be continued holomorphically to any point in @xmath37 . in @xcite this information was used for the description of the spectrum of the laplacian . here it is the basic tool for studying the transition probabilities . we recall that a point @xmath227 is called _ exceptional point _ of the map @xmath6 if the smallest completely invariant set containing @xmath227 is finite . let @xmath228 be an exceptional point of @xmath6 . then the smallest completely invariant set containing @xmath227 must also contain @xmath229 , since @xmath9 is the immediate basin of attraction of @xmath229 . since a rational function of degree greater or equal two has at most two exceptional points , we have @xmath230 and @xmath231 , see the proof of theorem 4.1.2 in the book of beardon @xcite . thus @xmath6 has at most the exceptional point @xmath10 . since @xmath232 the set of accumulation points of @xmath222 coincides with @xmath37 , see ( * ? ? ? * theorem 4.2.7 and 4.2.8 ) . hence @xmath233 and @xmath234 . we remark that either @xmath10 is exceptional and @xmath6 is conjugated to a polynomial or there are no exceptional points at all . actually , both cases can occur , as we will show by examples in sections [ sec : line ] and [ sec : examples ] . [ cor : continuation ] let @xmath12 be a symmetrically self - similar graph with bounded geometry and origin vertex @xmath32 . the green function @xmath5 has a unique holomorphic continuation to @xmath226 and @xmath235 for all @xmath236 . this convergence is uniform on any compact subset of @xmath226 . if @xmath12 is not bipartite then @xmath226 contains an interval @xmath237 $ ] for some @xmath238 . the first part is a consequence of lemma [ lem : dynamics ] . let @xmath12 be not bipartite . then , by lemma [ lem : bipartite ] , the cell graph @xmath0 is not bipartite . hence @xmath6 is neither even nor odd . therefore @xmath239 and @xmath226 contains @xmath237 $ ] for some @xmath208 . let @xmath12 be a symmetrically self - similar graph with respect to @xmath64 and @xmath72 . we call @xmath240 the _ time scaling factor _ of @xmath12 . then @xmath18 is the expected number of steps of the simple random walk on the cell graph @xmath0 for hitting @xmath213 , when starting in some @xmath209 . let @xmath162 be any 2-cell of @xmath12 . then the number of 1-cells in a 2-cell @xmath241 is called the _ mass scaling factor _ of @xmath12 . axiom ( s1 ) implies that @xmath14 is independent of the choice of the 2-cell @xmath162 . furthermore , we call @xmath242 the _ spectral dimension _ of @xmath12 . the spectral dimension was first introduced in physics literature using the density of states of the laplacian , see @xcite and the references therein . later on , the exponent @xmath243 was studied from several points of view , see @xcite . a measure @xmath244 on the set of vertices @xmath245 of a countable graph @xmath246 is _ invariant _ with respect to a transition matrix @xmath247 if @xmath248 , where @xmath244 is considered as row vector . written in coordinates this is @xmath249 for any @xmath250 . if @xmath246 is connected and finite , then the unique invariant probability measure of the simple random walk is given by @xmath251 [ lem : invmeasure ] let @xmath12 be a symmetrically self - similar graph , and let @xmath17 be any cell of @xmath12 . then @xmath252 defined by @xmath253 is the unique invariant probability measure of the simple random walk on @xmath0 . let @xmath38 be any vertex in @xmath201 . since the cell graph @xmath0 consists of @xmath14 amalgamated @xmath13-complete graphs , we have @xmath254 and @xmath255 . this implies the lemma . in the following we will make use of path arguments . for similar path decompositions see @xcite , @xcite , and especially @xcite . these constructions can be seen in the context of more general generating functions , see goulden and jackson @xcite . let @xmath17 be a cell of a symmetrically self - similar graph @xmath12 . for @xmath256 and @xmath257 let @xmath258 be the set of paths from @xmath1 to @xmath54 which do not hit any vertex in @xmath101 except for their start and their end vertex . the set of paths in @xmath258 with positive length is denoted by @xmath259 . the _ weight _ of a path @xmath260 in @xmath0 is defined by @xmath261 for @xmath262 . now @xmath263 is the probability , that the simple random walk on @xmath0 , starting in @xmath264 , follows the path @xmath60 in its first @xmath2 steps . for a set of paths @xmath265 we set @xmath266 for @xmath267 and @xmath268 we write @xmath269 for the set of all concatenations of paths in @xmath270 with paths in @xmath271 . then @xmath272 let @xmath1 again be a vertex in @xmath51 . we define @xmath273 and @xmath274 by @xmath275 then @xmath276 is the generating function of the probabilities that the simple random walk on @xmath0 , starting in @xmath1 , returns for the first time to @xmath1 after exactly @xmath2 steps , where @xmath277 . and @xmath278 is the generating function of the first return after @xmath2 steps without hitting any vertex in @xmath213 before . since @xmath95 acts transitively on @xmath51 , the functions @xmath276 and @xmath278 are independent of the choice of @xmath1 . [ lem : dfweight ] let @xmath17 be a cell of a symmetrically self - similar graph @xmath12 , and let @xmath1 and @xmath54 be two different vertices in @xmath51 . then @xmath279 and @xmath280 since @xmath281 is the set of all paths in @xmath0 from @xmath1 to @xmath1 which do not hit vertices in @xmath213 , we get @xmath282 every path in @xmath283 can be obtained by concatenating a finite number of paths in @xmath284 , therefore @xmath285 the definition of @xmath6 implies @xmath286 here we used the fact , that @xmath287 axiom ( s2 ) now yields the rest of the statement . [ lem : firstreturn ] let @xmath12 be a symmetrically self - similar graph , then @xmath288 let @xmath17 be any cell of @xmath12 and let @xmath1 be any vertex in @xmath51 . if we denote by @xmath289 those paths in @xmath290 , which hit a vertex in @xmath213 , then @xmath291 and therefore @xmath292 . for an integer @xmath2 in @xmath293 let @xmath294 be the set of all @xmath55-tuples @xmath295 in @xmath296 , such that @xmath297 for @xmath298 . if we write @xmath299 for the set @xmath300 then , by lemma [ lem : dfweight ] , @xmath301 now we can decompose @xmath289 in the following way : @xmath302 from this we obtain @xmath303 using lemma [ lem : dfweight ] and the fact , that @xmath304 . for the following statements we recall that @xmath172 if and only if the graph has bounded geometry , see theorem [ thm : bounded ] . let the simple random walk on @xmath0 start in a vertex @xmath1 in @xmath51 . then @xmath20 is the expected number of returns to @xmath1 ( the start is counted as the first visit ) before hitting another vertex in @xmath51 . resistance scaling factor _ is defined by the relation @xmath22 , see ( * ? ? ? * equation ( 6.29 ) ) . [ thm : dim ] let @xmath12 be a symmetrically self - similar graph . then @xmath305 . since @xmath306 and @xmath307 , we obtain @xmath308 using lemma [ lem : dfweight ] and [ lem : firstreturn ] . now @xmath309 is the expected number of steps of the first return of simple random walk on @xmath0 starting in vertex @xmath1 in @xmath51 . thus @xmath310 . a proof of this well known identity can be found in the book of brmaud , see ( * ? ? ? * theorem 3.2 ) . by lemma [ lem : invmeasure ] , we have @xmath311 completing the proof . let @xmath12 be a symmetrically self - similar graph . then @xmath312 . if @xmath12 is a symmetrically self - similar graph with bounded geometry then the last corollary implies @xmath313 . throughout the rest of this paper let @xmath12 be a symmetrically self - similar graph with bounded geometry and origin vertex @xmath32 . we recall that symmetrically self - similar graphs with an origin vertex have either bounded geometry or the origin vertex is the only vertex with infinite degree , see theorem [ thm : unbounded ] . since @xmath307 and @xmath314 there exists a holomorphic local inverse @xmath315 of @xmath6 in a neighbourhood of @xmath3 , which has a attracting fixed point at @xmath3 . for @xmath316 we write @xmath317 for the @xmath2-fold iterate of @xmath315 . let @xmath196 , @xmath318 and @xmath319 ; then we write @xmath320 for the set of all @xmath321 , such that @xmath322 and @xmath323 . in particular , we have @xmath324 . furthermore , we write @xmath325 for the horizontal strip @xmath326 [ lem : period ] let @xmath327 be a holomorphic function for some @xmath328 which satisfies the equation @xmath329 whenever @xmath206 in @xmath330 and @xmath331 . then there exists a @xmath30-periodic , holomorphic function @xmath29 defined on the strip @xmath332 and a @xmath333 , such that @xmath334 holds for @xmath335 for all @xmath27 . the function @xmath29 is constant if and only if @xmath145 is constant . this lemma was proved by one of the authors in ( * ? ? ? * lemma 5 ) . the proof of the lemma and the following theorem are based on ideas of de bruijn @xcite , odlyzko @xcite and grabner and woess @xcite . a detailed discussion of the methods can be found in the book ( * ? ? ? * section 16 ) . the point @xmath3 is a non - polar singularity of any green function of any recurrent random walk on an infinite , locally finite graph , see for example ( * ? ? ? * lemma 10 ) . lemma 3 in @xcite implies that the simple random walk on any symmetrically self - similar graph with bounded geometry is recurrent . [ thm : greenasymp ] let @xmath12 be a symmetrically self - similar graph with bounded geometry and origin vertex @xmath32 , then there exists a @xmath30-periodic , holomorphic function @xmath29 on the strip @xmath332 , such that the green function @xmath5 has the local singular expansion @xmath336 for @xmath335 and @xmath27 , where @xmath28 . we substitute @xmath337 into , where @xmath338 is holomorphic in @xmath339 and @xmath340 is some constant . this yields @xmath341 , where @xmath342 is given by @xmath343 and @xmath344 . furthermore we define @xmath345 since @xmath7 and @xmath342 are holomorphic in @xmath346 for some @xmath347 , the last product converges if @xmath348 . hence , we have to choose @xmath349 moreover , the convergence is uniform in @xmath346 , and therefore the function @xmath350 is holomorphic in @xmath346 . obviously , @xmath351 holds for all @xmath352 such that @xmath353 . let @xmath327 be defined by @xmath354 , then @xmath145 is holomorphic and satisfies @xmath355 whenever @xmath206 and @xmath356 are contained in @xmath330 . applying lemma [ lem : period ] to @xmath145 we get a @xmath30-periodic holomorphic function @xmath29 on the strip @xmath332 and a @xmath333 , such that @xmath334 holds for @xmath335 for all @xmath27 . since @xmath357 and @xmath350 is holomorphic in @xmath358 , we finally obtain @xmath359 for @xmath335 and @xmath27 . we will now use the method of ` singularity analysis ' due to flajolet and odlyzko @xcite , in order to obtain the asymptotic behaviour of @xmath31 . it should be mentioned here that a real tauberian theorem would not reveal the oscillating nature . [ thm : asymp ] let @xmath12 be a symmetrically self - similar graph with bounded geometry and origin vertex @xmath32 . if @xmath12 is not bipartite , then @xmath360 where @xmath361 is a @xmath30-periodic , holomorphic function given by its fourier series @xmath362 if @xmath12 is bipartite , then @xmath35 and @xmath363 since the @xmath30-periodic function @xmath29 of theorem [ thm : greenasymp ] is holomorphic in @xmath332 , we have @xmath364 for all @xmath365 and @xmath366 , where @xmath367 denotes the @xmath82-th fourier coefficient of @xmath29 . therefore it is possible to apply the method of ` singularity analysis ' . for the technical details of the transfer we refer to ( * section 5 ) and ( * ? ? ? * section 16 ) . if @xmath12 is not bipartite , then @xmath3 is the only singularity of @xmath5 at the boundary of the unit disk . the asymptotic of @xmath31 is now a consequence of the singularity analysis . if @xmath12 is bipartite , then @xmath5 is an even function . hence @xmath35 . furthermore , @xmath5 has the same singularity at @xmath368 as at @xmath3 . thus we have to add up the terms corresponding to the singularities @xmath3 and @xmath368 for the asymptotics of @xmath36 . we remark here that the function @xmath145 of the proof of theorem [ thm : greenasymp ] maps the interval @xmath369 to @xmath370 . by inspection of the proof of lemma [ lem : period ] in @xcite , we see that both @xmath29 and @xmath34 map @xmath370 to @xmath370 . hence the fourier coefficients of @xmath29 and @xmath34 satisfy @xmath371 for all @xmath365 . it is possible to compute the fourier coefficients of @xmath29 for a given graph @xmath12 numerically , see @xcite . let @xmath12 be a symmetrically self - similar graph with bounded geometry and origin vertex @xmath32 . we have shown that the asymptotics of @xmath31 carries an oscillating factor given by @xmath372 where @xmath34 is a @xmath30-periodic , holomorphic function given by the fourier series in theorem [ thm : asymp ] . in this section we will give a necessary condition for the case that the @xmath30-periodic function @xmath34 is constant . first of all we note that the following statements are equivalent : * the @xmath30-periodic function @xmath34 is constant . * the @xmath30-periodic function @xmath29 of theorem [ thm : greenasymp ] is constant . * the function @xmath145 given in the proof of theorem [ thm : greenasymp ] is constant . a closed interval in @xmath373 is denoted by @xmath374 $ ] , where @xmath375 . here @xmath376 and @xmath377 do not necessarily satisfy the inequality @xmath378 . for example , @xmath379 $ ] is the set @xmath380 . the same notation is used for open , respectively half open , intervals . [ thm : oscillation ] if @xmath34 is constant then the julia set @xmath37 of the transition function @xmath6 is a closed interval @xmath381 $ ] in the extended real line @xmath373 , where @xmath382 $ ] . the julia set @xmath383 is either a cantor set or a closed interval , see ( * ? ? ? * theorem 5.7.1 ) . we will use the notation of the proof of theorem [ thm : greenasymp ] . if @xmath34 is constant then the function @xmath145 is also constant . hence we obtain @xmath384 for @xmath385 , where @xmath17 is some constant and @xmath28 . since @xmath3 is not a polar singularity by lemma 10 in @xcite , we have @xmath386 . therefore @xmath5 has no analytic continuation to the interval @xmath387 and corollary [ cor : continuation ] implies that @xmath388 . furthermore , the set of accumulation points of @xmath222 coincide with @xmath37 . thus @xmath37 contains the interval @xmath387 and must therefore be itself a closed interval in @xmath373 . [ cor : oscillation ] if @xmath37 is a cantor set , then @xmath34 is non - constant . it is not clear , if the opposite direction of this corollary is also true . for rational functions with real julia set there is a general method to decide whether their julia set is an interval or a cantor set . therefore we refer to inninger s detailed discussion of rational functions in section 3 and section 4 of @xcite . however , for a given transition function @xmath6 the following observation may be useful as well : if we can find a real number @xmath389 which is smaller than some @xmath6-backwards iterate of @xmath3 together with an integer @xmath2 such that @xmath390 , then @xmath37 set is a cantor set . let @xmath12 be a symmetrically self - similar graph with bounded geometry and let @xmath391 be the matrix of transition probabilities for the simple random walk on @xmath12 . if @xmath12 is not bipartite , then @xmath391 is strongly aperiodic , that is , there exists a number @xmath392 such that @xmath393 for all @xmath394 . in general , @xmath395 is strongly aperiodic . since @xmath12 has bounded geometry and is connected , there exists a @xmath396 such that @xmath397 for all @xmath398 . thus @xmath395 is strongly aperiodic . if @xmath12 is not bipartite , then the cell graph @xmath0 of @xmath12 is not bipartite by lemma [ lem : bipartite ] . hence for any vertex @xmath168 there exists a closed path in @xmath0 from @xmath1 to @xmath1 of odd length . by adding a path of the form @xmath399 , where @xmath54 is adjacent to @xmath1 in @xmath0 , we can extend such a path for @xmath1 by an even number of edges . thus there is an odd number @xmath400 such that for any @xmath401 we can find a closed path from @xmath1 to @xmath1 of length @xmath400 . since any vertex @xmath398 is contained in at least one cell graph , there is a real number @xmath402 such that @xmath403 for all @xmath398 . as every @xmath404 can be written in the form @xmath405 with @xmath406 , strong aperiodicity follows . we remark here , that it is possible to choose @xmath407 , if @xmath408 , since the cell graph consists of @xmath14 copies of the @xmath13-complete graph . [ thm : offdiagonal ] let @xmath38 and @xmath39 be two vertices of a symmetrically self - similar graph @xmath12 with bounded geometry and origin vertex @xmath32 . if @xmath12 is not bipartite , then @xmath409 if @xmath12 is bipartite , let @xmath410 be an element of @xmath411 such that @xmath412 . then @xmath413 and @xmath414 if @xmath12 is not bipartite then @xmath415 by ( * ? ? ? * theorem 2 ) , since @xmath391 is irreducible , strongly aperiodic and recurrent . thus @xmath416 now theorem [ thm : asymp ] implies the statement . if @xmath12 is bipartite then we can partition @xmath47 into classes @xmath417 and @xmath418 , such that edges in @xmath48 only connect vertices in @xmath417 with vertices in @xmath418 . then @xmath395 restricted to @xmath419 for @xmath420 is irreducible , strongly aperiodic and recurrent . if @xmath421 , then @xmath38 and @xmath39 are in the same class . hence @xmath422 and @xmath423 if @xmath424 , then @xmath425 and @xmath426 note that the above sum is finite , since there are only finitely many vertices @xmath427 such that @xmath428 . furthermore , @xmath429 as @xmath34 is a @xmath30-periodic , holomorphic function , it is uniformly continuous on the real line . we have @xmath430 and therefore @xmath431 which completes the proof . jones computed estimates for the transition probabilities @xmath40 on the sierpiski graph which are uniform in space ( @xmath432 ) and time ( @xmath316 ) . his result yields the correct asymptotic type @xmath433 as @xmath434 , see @xcite . [ exmp : line ] let the cell graphs of a self - similar graph @xmath12 be isomorphic to the line @xmath41 of length @xmath435 ( figure [ fig : line]a ) . if the graph @xmath12 has an origin vertex @xmath32 , then it consists of an arbitrary number of one - sided infinite lines which are amalgamated in their initial vertex @xmath32 , see figure [ fig : line]b . if there is no origin vertex then @xmath12 is the two - sided infinite line . ( 290,80)(0,-35 ) [ fig : line ] ( 0,5)(0,0)(30,0)3 ( 60,0)(80,0 ) ( 110,0)(130,0 ) ( 130,0)(160,0 ) ( 130,0)(30,0)2 ( 0,0)(60,0 ) ( -3,-10)@xmath264 ( 27,-10)@xmath436 ( 57,-10)@xmath437 ( 127,-10)@xmath438 ( 157,-10)@xmath439 ( 80,10)@xmath41 ( 240,14 ) ( 100,0 ) ( 200,0 ) ( 93.97,34.2 ) ( 187.94,68.4 ) ( 76.6,64.28 ) ( 153.2,128.56 ) ( 93.97,-34.2 ) ( 187.94,-68.4 ) ( 0,0 ) ( -40,-10)@xmath32 ( 30,100)@xmath12 ( 187.94,68.4)(0,0)(153.2,128.56)(0,0)(200,0)(0,0)(187.94,-68.4 ) ( 187.94,68.4)(244.322,88.92 ) ( 153.2,128.56)(199.16,167.128 ) ( 187.94,-68.4)(244.322,-88.92 ) ( 0,0)(199.16,-167.128 ) ( 200,0)(260,0 ) ( 60,-35)_figure _ [ fig : line]a ( 238,-35)_figure _ [ fig : line]b the corresponding transition function @xmath42 is conjugated to the @xmath2-th chebychev polynomial with conjugacy map @xmath440 , @xmath441 see ( * ? ? ? * example 1 ) . as @xmath442 $ ] we get @xmath443 = ( \r \cup \{\infty\ } ) \setminus ( -1,1)$ ] . we have @xmath161 , @xmath444 , @xmath172 , @xmath445 and @xmath446 . furthermore , if there is an origin vertex , we have @xmath447 for @xmath448 . hence the function @xmath34 is constant . to prove that lines as cell graphs are characterised by its return and transition functions we state the following lemma . [ lem : lines ] let @xmath246 be a locally finite , connected graph , and let @xmath449 be the @xmath2-step transition probabilities of a reversible markov chain on @xmath246 which is of nearest neighbourhood type . then @xmath450 for different vertices @xmath38 and @xmath39 , which are at distance @xmath451 from each other . here equality holds , if and only if @xmath452 and @xmath453 for all @xmath454 , this is , the markov chain can be projected to the simple random walk on the line @xmath41 . we prove a more general statement which is more adequate for induction : for a vertex @xmath455 and a finite set of vertices @xmath456 of @xmath245 we write @xmath457 . let @xmath316 and let @xmath85 and @xmath101 be finite subsets of @xmath245 such that @xmath458 for all @xmath459 and @xmath460 . for @xmath461 we introduce the level set @xmath462 by @xmath463 these level sets are finite , @xmath464 and @xmath465 . furthermore , we define @xmath466 , and @xmath467 @xmath468 for a vertex @xmath38 in @xmath469 , where @xmath470 . we define @xmath471 for @xmath472 . then we claim that @xmath473 . equality holds , if and only @xmath474 for any vertex @xmath38 in @xmath469 , where @xmath475 . first of all we notice that the condition for equality indeed imply @xmath476 . we use induction over @xmath2 . for @xmath477 the inequality is immediate . now let @xmath478 and let the statement hold for all @xmath479 . factorizing with respect to the @xmath82-th step and using the reversibility yields @xmath480 for @xmath475 . now we use the inequality @xmath481 for @xmath482 , where @xmath475 . here equality holds if and only if @xmath474 . thus we obtain @xmath483 using cauchy - schwarz inequality for the second step . now reversibility together with the induction hypothesis for @xmath484 and @xmath485 implies that @xmath486 as we can factorize with respect to any @xmath475 we get the postulated conditions for equality . [ cor : lines ] let @xmath42 be the transition function and let @xmath487 be the return function of a cell graph which is isomorphic to the line @xmath41 of length @xmath435 . then there is no other cell graph of a self - similar graph with bounded geometry with transition function @xmath42 or return function @xmath487 . let the transition function of a cell graph be the function @xmath42 . then @xmath42 is either even or odd , whence the cell graph must be bipartite . by lemma [ lem : bipartite ] the self - similar graph itself must be bipartite and therefore @xmath161 . as we have bounded geometry , @xmath172 by theorem [ thm : bounded ] . the series expansion of @xmath42 in @xmath10 starts with @xmath488 . now lemma [ lem : lines ] implies that the cell graph @xmath0 must be the line @xmath41 . suppose @xmath487 is the return function of a cell graph @xmath0 which is not isomorphic to @xmath41 and let @xmath1 be a vertex in the boundary @xmath51 . since @xmath487 is even , we have @xmath161 and @xmath172 as before . there are vertices in @xmath17 which have edge degree strictly larger than 2 in the graph @xmath0 . let @xmath244 be the minimal distance of @xmath1 to a vertex @xmath489 with @xmath490 . then the probability of returning to @xmath1 in exactly @xmath491 steps is strictly smaller than the corresponding probability for the line @xmath41 . this means that the @xmath491-th coefficients of the series expansions of the return functions of @xmath0 and @xmath41 around @xmath10 are different . well known examples of symmetrically self - similar graphs are the viek graph and the sierpiski graph . both graphs do not belong to the following simple example of an infinite class of self - similar graphs . [ exmp : flake ] let the cell graph @xmath492 consist of @xmath2 copies of the @xmath2-complete graph which are amalgamated in one vertex @xmath38 , where @xmath435 . the boundary of the cell contains exactly one vertex from each of these copies , but not the vertex @xmath38 . starting with these cell graphs we can construct different symmetrically self - similar graphs in the sense of theorem [ thm : blowup ] . the four vertices of the cell graph @xmath493 in figure [ fig : flake4]a which constitute the boundary @xmath494 of the cell are drawn fat . figure [ fig : flake4]b shows the corresponding 4-cell graph @xmath495 . ( 345,235 ) [ fig : flake4 ] ( 0,85)(0,0 ) ( 40,40 ) ( 80,80 ) ( 80,0 ) ( 0,80 ) ( 5,45 ) ( 5,35 ) ( 75,35 ) ( 75,45 ) ( 35,5 ) ( 35,75 ) ( 45,5 ) ( 45,75 ) ( 0,0)(1,1)80 ( 0,80)(1,-1)80 ( 5,35)(75,45)(80,80)(45,75)(35,5)(0,0)(5,35)(35,5 ) ( 75,35)(5,45)(0,80)(35,75)(5,45 ) ( 35,75)(45,5)(80,0)(75,35)(45,5 ) ( 45,75)(75,45 ) ( 20,-20)_figure _ [ fig : flake4]a ( 120,0)(0,10 ) ( 90,0)_figure _ [ fig : flake4]b the transition function @xmath42 of @xmath492 is given by @xmath496 for @xmath435 . for the parameters we have @xmath497 , @xmath172 , @xmath498 and @xmath499 . we remark that for @xmath500 the cells are isomorphic to the line @xmath501 , @xmath502 , and @xmath503 $ ] , see example [ exmp : line ] . for @xmath504 we have @xmath505 here @xmath506 has to be read as @xmath507 if @xmath508 . as @xmath30 and @xmath506 belongs to @xmath509 , whereas @xmath229 and @xmath510 does not , the julia set is not connected in @xmath373 . hence @xmath509 is a cantor set for @xmath504 . in the case that there is an origin vertex we can conclude that the @xmath30-periodic function @xmath34 is not constant , using corollary [ cor : oscillation ] . many transition functions are of the form @xmath511 , where @xmath2 is the minimal number of steps from one point in the boundary of a cell @xmath17 to another ( in other words @xmath512 ) and @xmath391 is a polynomial with integer coefficients . as the following example shows , this is not true for all transition functions . [ exmp:34sierp ] the 3-dimensional , 4-scaled sierpiski graph consists of 20 amalgamated 4-complete graphs . the 1-cell and 2-cell can be seen in figure [ fig:34sierp ] . their boundary vertices are drawn fat . each 4-complete graph is represented by a solid tetrahedron . [ fig:34sierp ] _ figure _ [ fig:34sierp ] here we have @xmath513 and @xmath514 , @xmath515 , @xmath172 , @xmath516 and @xmath517 . as @xmath518 the julia set @xmath37 is a cantor set . if the constructed self - similar graph has an origin vertex then the @xmath30-periodic function @xmath34 is not constant . the authors want to thank peter grabner , klaus schmidt and wolfgang woess for fruitful discussions and financial support . especially , the results in section [ sec : geometric ] would not have been possible without the contributions of peter grabner and wolfgang woess .
the accretion processes are recognized to power many astrophysical sources , which are observable due to the release of energy when matter is spiralling down into the gravity well . the study of viscous accretion disks as underlying mechaisms for the extraction of potential gravitational energy began with the works of @xcite , @xcite . time - dependent problems for the accretion disks arise when describing a wide variety of outburst phenomena , observed in the accreting sources , i.e. binary systems with mass transfer and galactic nuclei , and systems with planet formation . it is usually possible to consider separately the vertical and radial structure of an accretion disk due to significantly different characteristic time - scales . in such a case the time - dependent radial structure of a disk can be described by a second - order partial differential equation , which is a consequence of the angular momentum and mass conservation . the energy conservation in a geometrically thin optically thick disk is provided by the local balance of the viscous heating and radiative cooling . the basic equation of the viscous evolution , eq . or below , describes the viscous angular momentum flux along the disk radius . this is a diffusion type equation with a variable ` diffusion ' coefficient . choice of an analytic approach to solve the equation depends on the nature of the viscosity involved . the preferential way of describing viscosity in astrophysical disks , although not exclusive , is to use @xcite , i.e. , the proportionality of the viscous stress tensor to the total pressure in the disk . one can generally assume that the kinematic viscosity @xmath4 is a product of two power functions : one function of radius and another of some local physical parameter , the latter usually represented by the surface density . the boundary conditions imposed on the accretion disk are important . while the outer boundary can be a freely expanding surface due to the spreading of the matter corresponding to the outward viscous angular momentum flux , in some situations certain conditions should be posed at some distance from the accreting object . disks in the binary systems are the main focus for such models . in a binary system , the tidal torques , acting inside the roche lobe of the accreting object , truncate the disk and provide a sink for the angular momentum . @xcite show that the effective tidal radius is close to the roche lobe : @xmath5 of the mean radius of the roche lobe , well coinciding with the largest non - intersecting periodic orbits in the restricted three - body problem @xcite . numerical models confirm that most of the tidal torque is applied in a narrow region at the edge of the disk , where perturbations become nonlinear and strong spiral shocks appear ( see * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) . in the present study , we leave aside the outer disk radius variations and other manifestations of the tidal interactions in the binary except for the disk truncation . we confine ourselves to searching for a solution to a problem of the viscous evolution of an accretion disk with the constant outer radius . we investigate a geometrically thin , axisymmetric keplerian finite accretion disk with a viscosity in the form @xmath6 , which is time - independent . we obtain green s functions ( [ eq : green_zerobi ] ) , ( [ eq : mdot_green_zerobi ] ) , ( [ eq : green_zeroai ] ) , and ( [ eq : mdot_green_zeroai ] ) , which are the kernels in the integrals ( [ eq : f_from_green ] ) , ( [ eq : mdot_from_green ] ) , and ( [ eq : sol_nonzeromdot_full ] ) to compute the viscous angular momentum flux of @xcite . ] @xmath7 and the accretion rate @xmath8 . the method of green s functions allow one to compute @xmath9 and @xmath10 for arbitrary initial surface density distributions . different boundary conditions are considered : no stress or no accretion at the inner boundary , which is located at the zero radial coorinate , zero or time - dependent mass inflow at the outer boundary . it is beleived that the mechanism of bursts in x - ray novae involves the thermal instability connected with opacity variations in the disks , as in dwarf novae . when a burst is ignited , and a sufficient portion of the disk is in the hot state , the viscous diffusion starts to govern the evolution , driving the characteristic rise and drop of the accretion rate onto the central object . it is a question , which probably allow no universal answer , why x - ray novae light curves have different profiles and pecularities . if the disk is not very large , as it happens in short - period x - ray transients , its whole body is likely to be engaged in the viscous evolution . it has been shown that an episodic mass input at the outer radius causes the fast - rise - exponential - decay ( fred ) profile due to the viscous evolution @xcite . using the green s function , one can explore the situation with an arbitrary initial distribution of the matter . actually , as we show in the present study after obtaining the green function , for realistic initial density distributions and for the case with the zero inner viscous stress , the evolution is generally a fred . it is interesting to compare the results of the model with what happens in the @xmath3-disks , the prevalent model for the disks turbulence in the stellar binaries . the second case , when the viscous stress is non - zero at the center , applies to the disks with the central source of the angular momentum , e.g. , rotating stars with magnetosphers . if the magnetosphere s radius is large enough , the matter has a super - keplerian angular momentum at its boundary . different regimes for such disks have been proposed : propellers and ` dead ' disks . using the green founction , the disks of constant mass that evolve to dead disks can be described . in binary systems the mass transfer rate is a key factor . at some stages it can be considered negligible comparing to the accretion rate in the disk . in other systems , mass transfer variations may cause the activity in the disk . the case of time - dependent mass transfer , which includes the constant transfer rate as a particular case , can be also studied analytically if @xmath11 . using the green s functions , it is possible to express @xmath9 and @xmath10 for any outer boundary conditions ( for any @xmath12 ) as well as for any initial conditions . in section 2 , we present the viscous evolution equation and review its known solutions for accretion disks . in section [ sec : ev_fin_disk ] , the green s functions are found for different inner and outer boundary conditions . an analytic way to calculate the disk evolution with the variable mass transfer at the outer radius is presented ; the features of a radiating dead disk are considered . in section [ s.decay ] , a stage of decaying accretion is dealt with and a comparison to the @xmath3-viscosity models is made . application to the fred light curves of short - period x - ray transients is presented in section [ s.xnovae ] . the last sections are dedicated to the discussion and summary . the equation of the viscous evolution in the accretion disk is the diffusion - type equation ( e.g. , * ? ? ? * ) : @xmath13\ , , \label{eq : time - dep - acc1}\ ] ] where , among the standard designations of the time , radius , angular speed , and the surface density we have the component of the viscous stress tensor @xmath14 , integrated over the full thickness of the disk @xmath15 where @xmath16 is the half - thickness of the disk at radius @xmath17 , and the stress tensor is related to the differential rotation of gas masses by the kinematic viscosity @xmath4 as follows : @xmath18 for the keplerian disks @xmath19 , and we rewrite @xmath20 with the formula for the surface density @xmath21 assuming that @xmath4 is independent on @xmath22 , we have @xmath23 it is convenient to introduce a new independent variable the specific angular momentum . then , for the keplerian disk , equation ( [ eq : time - dep - acc1 ] ) can be rewritten as : @xmath24 the acknowledged approach is to solve the disk evolution equation for the variable proportional to the full viscous torque acting between the adjacent rings in the disk , @xmath25 . this way , the boundary conditions are easy to express . from the above expressions , the following relation between @xmath26 and the accretion rate can be obtained : @xmath27 where the radial velocity @xmath28 is negative . one chooses a procedure to solve ( [ eq : basic_tda_kepler ] ) provided the form of @xmath29 is known . this form is determined by the physical conditions in the disk . for example , the vertical structure can be solved for the @xmath3-disk , and the required relation between @xmath4 and @xmath30 can be found . let us give a perspective of some studies ( mainly analytic ) of the viscously evolving accretion disks . they differ in respect of the viscosity prescription and boundary conditions . the analytic solutions are the benchmarks for more sophisticated and involved numerical models . for the kinematic viscosity in the form @xmath6 , the equation of the viscous evolution becomes a linear partial differential equation . in @xcite @xcite obtained particular solutions to the equation proposed by his teacher von weizscker ( @xcite ) and described the principles to determine the general solution for the infinite and finite problems . for the accretion disks that can extend _ infinitely _ , @xcite ( hereafter lp74 ) find , for the two types of the inner boundary conditions , the green s functions that provide means to describe the whole development of the disk . the inner radius of the disk is zero in their exact solution . the long - term self - similar evolution of the accretion rate goes as a power - law @xmath31 with a parameter @xmath32 ( see also appendix a ) . @xcite considers the properties of an infinite accretion disk with a central source of the angular momentum using the green - function technique . a related problem is solved by @xcite who finds the green s function for an infinite disk with the boundary conditions posed at the finite inner radius . @xcite consider a _ finite _ disk with the viscosity constant over the radius and in time and derive the exponential decay of the accretion rate in the disk . @xcite study the mass flow rate through a disk resulting from a varying mass - supply rate , derive the green s function for the accretion - rate and its fourier transform . by means of numerical simulations , they investigate a _ finite _ disk , residing between non - zero @xmath33 and @xmath34 , with the non - zero accretion rate at the outer radius . @xcite find general green s function for an explicit dependence of the inner boundary @xmath33 on time and a non - zero mass across @xmath33 . in the present work , we study analytically a finite accretion disk with the zero @xmath33 and find an analytic solution describing its whole evolution , from the rise to decay . for the case of the zero stress tensor in the disk center and the mass deposition at the outer disk edge , the green s function of the problem is found in @xcite . similarity solutions of the first kind of the non - linear differential equation @xcite have been found at the `` decaying - accretion stage , '' when the total angular momentum of an _ infinite _ disk is constant @xcite . accretion rate is found to decrease as @xmath36 for the thomson opacity , and as @xmath37 for the kramers opacity ( see also * ? ? ? * ; * ? ? ? @xcite obtain self - similar solutions that describe separately three stages of the disk evolution . the first two are the self - similar solutions of the second kind : the initial movement of the inner edge of a ring of matter toward the center and the rise of the accretion rate in the center ( @xmath38 for the kramers opacity and @xmath39 for the thomson opacity ) . the third stage is the decay of the accretion and corresponds to the solution of @xcite . the first two stages of @xcite may be applicable to the _ finite _ disks when the conditions at the distant outer boundary do not affect the behaviour of the disk near the center . @xcite derive an analytic self - similar solution for a disk subject to the gravitational instability , which transfers mass and angular momentum in the disk , for which @xmath40 , and a numerical solution for the whole evolution . @xcite found a self - similar solution for @xmath41 , if the viscosity is due to the action of convectively driven turbulent viscous stresses ( @xmath42 ) . accretion disks with a central source of angular momentum and no mass transfer through the inner boundary were considered by @xcite who derived a self - similar solution ( see also @xcite ) . @xcite in the extensive study of the circumbinary disks around supermassive black holes obtained self - similar solutions for the disks with a possible mass transfer across the orbit of the secondary black hole . for a _ finite _ disk , the radial and temporal solution is obtained by @xcite yielding @xmath43 for the kramers opacity and @xmath44 for the thomson opacity . in the context of the self - similarity solutions , we would like to note that for an advection - dominated infinitely expanding accretion flow , @xcite obtains by similarity methods a time - dependent solution with the conserved total angular momentum . if the kinematic viscosity @xmath4 is not a function of @xmath30 and is a function of the radius alone , we arrive at the linear differential equation . we write the kinematic viscosity as @xmath45 then eq . ( [ eq : basic_tda_kepler ] ) for the dependent variable @xmath7 is as follows @xmath46 where @xmath47 we rewrite in the form similar to that in lp74 : @xmath48 where the constants are defined as follows : @xmath49 there is a degeneracy of the index @xmath50 at @xmath51 . but it is only apparent as @xmath50 enters equation ( [ eq : diff_lyn - pringle ] ) in the denominators . value @xmath51 corresponds to the change of the class of functions , which satisfy the differential equation , and is left out of the following consideration . in the case of @xmath51 the viscous time is constant over the disk radii @xmath52 . we limit ourselves to the finite positive values of @xmath50 and the keplerian disks . in the astrophysical hot accretion disks , the turbulent viscosity is at work . according to the prandtl concept @xcite , the kinematic turbulent viscosity is an averaged product of the two random values : @xmath53 , where @xmath54 and @xmath55 are the path and speed of the turbulent motion . for the component of the viscous stress tensor we write @xmath56 . following the prandtl consideration , , which leads to . @xcite proposed the @xmath3-disks : @xmath57 . from the hydrostatic balance one derives the sound speed @xmath58 and obtains for a keplerian disk @xmath59 . if the relative half - thickness of the @xmath3-disk is invariable then @xmath60 and @xmath61 . this value of @xmath62 takes place for the ` @xmath63-viscosity disks ' with @xmath64 , suggested by @xcite to act in the protoplanetary and galactic disks . the standard model of the @xmath3-disk implies that @xmath65 varies with @xmath17 and @xmath30 . let us find the green s function of the linear equation ( [ eq : diff_lyn - pringle ] ) with the specific boundary conditions at the boundaries . a green s function or a diffusion kernel is a reaction of the system to the delta impulse , in other words , the green s function is the solution to the differential equation with the initial condition @xmath66 , where @xmath67 is the dirac delta function . we search a solution by separating the variables , , where @xmath68 is a constant and @xmath69 is the outer disk radius . substituting this into ( [ eq : diff_lyn - pringle ] ) , we get @xmath70 where @xmath71 . this is a lommel s transformation of the bessel equation , which can be derived by introducing a new independent variable @xmath72 . a particular solution is represented as @xmath73\,\ ] ] @xcite , where @xmath74 and @xmath75 are the bessel functions of a non - integer order . if @xmath50 is an integer number , the bessel functions of the second kind should be used instead of @xmath75 . for @xmath51 , the solution is no more a bessel function ( see appendix i of * ? ? ? the value @xmath51 marks a situation when the characteristic viscosity time @xmath76 is constant over the radius . we do not consider this case in the present study . a general solution for an infinite problem is a superposition of particular solutions and is expressed by lp74 as an integral over all positive values of @xmath77 : @xmath78 for a problem inside a finite interval , a general solution is a superposition of particular solutions and is expressed as an infinite sum @xcite : @xmath79\ , , \label{eq : gen_sol_finite}\ ] ] where constants @xmath80 , @xmath81 . and @xmath82 are to be defined from two boundary conditions and one initial condition . the dimensionless coordinate @xmath83 lies in the interval @xmath84 $ ] , while the physical coordinate @xmath85 ( specific angular momentum ) lies inside @xmath86 $ ] . the outer boundary condition , which sets the accretion rate at the outer radius of the disk , can be written as follows @xmath87 in the simplest case , the homogeneous outer - boundary condition corresponds to the zero mass - inflow rate at the outer boundary . as this happens , the viscous angular momentum flux @xmath7 takes some non - zero value , which implies withdrawal of the angular momentum at the outer boundary . this corresponds to the situation of the disk in the binary system , when the angular momentum is removed from the outer edge of the disk by the tidal action of the secondary . at , one trivial inner boundary condition approximates the structure for the so - called ` standard ' disk , for which the absence of the viscous stresses at the disk inner edge is assumed . another simple case , @xmath88 at @xmath89 , describes the situation of no accretion through the inner disk edge . this case is applicable for the stars with strong magnetosphere , when the alfvn radius is greater than the corotation radius , impeding the accretion of the matter . we consider these two cases separately . first , we consider the case of the zero viscous stress at the inner boundary . approximately , this corresponds to the solution @xmath90 at @xmath89 and only the coefficients at @xmath74-terms in ( [ eq : gen_sol_finite ] ) remain non - zero . thus , sum ( [ eq : gen_sol_finite ] ) at @xmath91 becomes @xmath92 we assume the zero accretion rate at @xmath34 . such boundary condition approximates the situation of the low mass - transfer from the secondary in the binary system , for example , when the accretion rate in the disk is much higher than the mass transfer rate during an outburst . dropping the time part of the solution , we arrive at condition ( [ eq : zero_dotm_out ] ) required for every term in the sum ( [ eq : gen_sol_finite_zerobi_t0 ] ) , which can be rewritten as follows : @xmath93 thus , the general solution for the case of zero viscous stress tensor at the center and zero accretion rate at the outer boundary is the sum ( [ eq : gen_sol_finite ] ) , where @xmath94 and @xmath80 are the positive roots of the transcendental equation ( [ eq : out_boundary ] ) . to define @xmath81 from the initial condition , the finite hankel transforms are used . the series @xmath95 with condition ( [ eq : out_boundary ] ) are known as the dini s series ( see * ? ? ? function @xmath96 can be expressed as the dini s expansion if the function satisfies the dirichlet conditions in the interval and the coefficients are defined as @xmath97 @xcite with the finite hankel transform @xmath98 where @xmath99 . we look for a solution for the @xmath100-function as the initial condition : @xmath101 . using the properties of the dirac function , we obtain @xmath102 thus , for the specific boundary conditions , we derive the green s function , which is the solution to ( [ eq : diff_lyn - pringle ] ) with @xmath100-function as the initial distribution @xmath103 where @xmath80 are the positive roots of eq . ( [ eq : out_boundary ] ) and @xmath104 . the function is plotted in fig . [ fig.green_finite ] for consecutive times . the curve for time @xmath105 corresponds to the maximum accretion rate in the center . it is plotted at @xmath106 given by below . , @xmath107 , @xmath108 , @xmath109 , @xmath110 . initial spike was at position @xmath111 . constants @xmath112 , @xmath61 . [ fig.green_finite ] , scaledwidth=45.0% ] given the dimensional initial distribution @xmath113 in the finite accretion disk , the distribution of @xmath7 at any @xmath114 is @xmath115 the accretion rate can be found from @xmath116 where the green s function for the accretion rate @xmath117 functions @xmath118 and @xmath119 are found by @xcite for the case @xmath120 along with an analytical form obtained from the asymptotic forms for small and large @xmath121 . one can express the initial @xmath7 , proportional to the disk mass , from the initial distribution of the surface density , using ( [ eq : stress_kepl_av ] ) and ( [ eq : consts_lb ] ) : @xmath122 where @xmath123 and @xmath124 . let the initial mass be located at some radius @xmath125 where the keplerian specific angular momentum is @xmath126 , and the mass of the infinitely thin ring is @xmath127 , @xmath69 is the maximum possible specific angular momentum of the matter . the surface density of the @xmath100-ring is @xmath128 we take into account that @xmath129 and @xmath130 . applying ( [ eq : f_sigma_l ] ) , we obtain @xmath131 and the accretion rate evolution is as follows @xmath132 where we designate @xmath133 the time @xmath134 of the maximum and its accretion rate in the disk center @xmath135 are of order of those for the infinite disks . using the solution of lp74 , we find the time of the accretion rate peak for the infinite disk @xmath136 and value @xmath137 . the numerical factor in is presented in table [ tab : zeroes ] for various , physically motivated values of @xmath50 . in fig . [ fig.cmax1 ] the ratios of maximum times and accretion rates for the finite and infinite disk are shown for several cases of @xmath50 . the ratios approach @xmath138 for @xmath139 the disk ` does not feel ' the boundary at the time when the accretion rate peaks . in the disk , whose inner accretion rate evolution is shown in fig . [ fig.mdotin ] . top - hats are located at @xmath140 ( solid line ) and @xmath141 of @xmath34 ( dashed line ) and the wide distribution , corresponding to @xmath142 , goes in @xmath143 ( dotted line ) . the disk mass is the same in all cases and equals 1 . the outer specific angular momentum @xmath144 . the top - hat distributions are constructed using @xmath145 . parameter @xmath61 . [ fig.g_init ] , scaledwidth=45.0% ] if the initial surface density is distributed as a power - law with radius @xmath146 , we use the following expression in eq . ( [ eq : f_sigma_l ] ) : @xmath147 assuming the mass of the disk is enclosed between the radii @xmath148 and @xmath149 . in fig . [ fig.mdotin ] the variation of the accretion rate at the disk center is presented for three different initial distributions of @xmath7 ( see fig . [ fig.g_init ] ) : two narrow top - hats ( @xmath150 ) with different @xmath148 and @xmath149 and the one corresponding to @xmath151 . all three initial distributions are constructed for the same initial mass of the disk . the evolution of a thin ring located far from the outer disk edge has a characteristic power - law interval , when the disk behaves as if it was infinite . the interval lasts about @xmath152 $ ] . if @xmath151 , the mass of the disk at @xmath91 is enclosed mainly in the outer parts , because it grows as the cube of the radius . as expected , the accretion rate s evolution after the peak depends weakly on the peculiarities of the initial distribution of mass in the disk ( compare the dotted and solid line in fig . [ fig.mdotin ] ) , as long as its bulk was similarly positioned . the viscous heating in the disk ( its half ) is found from @xmath7 : @xmath153 which can be equated to @xmath154 assuming the local energy balance . in the center of the disk , the quasi - stationary distribution develops soon and the accretion rate is constant over radii : @xmath155 . the zone of constant @xmath8 expands over time . the viscous heating with the quasi - stationary distribution of @xmath7 diverges at the zero coordinate . in the real astrophysical disks , the inner radius of the disk is a finite value @xmath156 . note that integrating over the disk surface from @xmath157 to @xmath158 also yields the incorrect result for the total power released in the disk . ( as a remark , the integration of gives the correct result for the case of the zero accretion rate at the center , see [ s.deaddisk ] ) . in fact , setting @xmath159 at the finite radius @xmath157 leads to a time - independent correction of @xmath7 near @xmath157 . this correction is found for the infinite disks by @xcite and represents the classical factor @xmath160 . naturally , the correction holds for a finite disk . it follows that the bolometric luminosity can be calculated via the classical formula @xmath161 at times @xmath162 with the first term in the correction of order of @xmath163 for earlier times . for the disks in the stellar binary systems , usually , ratio @xmath164 , and we are safe to calculate @xmath165 by the classical formula with @xmath166 found from . 0.5 cm condition @xmath88 at necessitates that all @xmath167 in , which becomes for @xmath91 : @xmath168 where @xmath80 are the positive roots . in the previous section , the zero root contributed nothing to ( [ eq : gen_sol_finite_zerobi_t0 ] ) . here the zero root provides a term like @xmath169 , which must contribute something , because the bessel function @xmath75 behaves as @xmath170 in the vicinity of zero . the method to find @xmath82 for @xmath171 is analogous to that for @xmath81 , since the dini s series are convergent for the order of the bessel function @xmath172 @xcite @xcite . ] . we consider only non - integer @xmath173 and obtain @xmath174 it is known about the dini s series of the negative order that an additional first term is to be added to the sum , if the equation defining roots @xmath80 has particular coefficients=0 , the the additional term appears if @xcite . ] . just this case occurs for the properly rewritten outer boundary condition ( [ eq : zero_dotm_out ] ) : @xmath175 the additional term in the dini s series is expressed as @xmath176 where @xmath22 is a free variable @xcite . thus , adopting @xmath177 , from we obtain the green s function for the equation of the non - stationary accretion with the zero inner accretion rate and the finite outer radius : @xmath178 where @xmath80 are the positive roots of eq . ( [ eq : out_boundary_neg ] ) and @xmath179 . formulas ( [ eq : f_from_green ] ) and ( [ eq : mdot_from_green ] ) can be applied without changes . this green function is illustrated in fig . [ fig.green_finiteb1 ] . at the late stages , the accretion disks forget all information about the initial distribution of the viscous stresses . for the disks without accretion on to the center , the uniform distribution of @xmath7 develops , with its magnitude being proportional to the mass stored in the disk . we take the derivative of ( [ eq : green_zeroai ] ) and obtain the green function for the accretion rate : @xmath180 it approaches zero when @xmath181 for any time @xmath121 . this corresponds to our boundary condition at the center : @xmath182 . this disk , without sink of mass and additional mass supply , emits radiation even when it reaches the ` end ' of its evolution , as the viscous heating does not stop when the flow of matter ceases . if fig . [ fig.lumin_zeroai ] its bolometric luminosity is plotted for the two initial positions of the ring ( the details are given in the next subsection ) . the stable configuration of the confined ` dead ' disk ( we adopt the name from the work by @xcite ) has the following viscous angular momentum flux at each @xmath17 : @xmath183 where is used and @xmath184 as before . the total luminosity of the dead disk is a constant value @xmath185 . it can be obtained by the integration of over the disk surface between @xmath157 and @xmath158 . unlikely to the case of the zero torque at @xmath186 , the late - time value of @xmath7 does not depend on the location at which condition is set and solely depends on the evolution at the large radii . this was shown for the infinite disks by @xcite . value of @xmath157 affects the limunosity only by a change in the size of the emitting surface . there is still a time - dependent correction obtained by @xcite having an effect at early times @xmath187 . the dead - disk state has to end sometime , as the disk can not receive infinitely the angular momentum from the central object . if the central star has a magnetosphere , its radius @xmath188 can be estimated from the equality of the viscous and magnetic torque at the magnetosphere boundary ( lp74 ) : @xmath189 where @xmath190 is the magnetic dipole of the star , @xmath191 is a dimensionless factor of order of unity @xcite . within some factor , this condition is equivalent to the equality of the gas and magnetic pressure at the inner disk boundary @xcite . the accretion is inhibited if the inner disk radius , @xmath188 , is greater than the corotation radius , because the drag exerted by the magnetic field is super - keplerian : @xmath192 where @xmath193 , @xmath194 and @xmath195 are the mass , angular speed and period of the revolution of the central star . in the last expression the mass is normalized by @xmath196 m@xmath197 . thus we obtain for a final steady configuration @xmath198 the ratio can be rather close to unity . the normalizations are @xmath199 g for the disk mass , @xmath200 [ g@xmath201 for the magnetic dipole , 100 days for the viscous time , r@xmath197 for the outer disk radius . situation of @xmath202 is studied by @xcite , who argue that no considerable expulsion of matter from the disk is expected in such regime , as contrasted to the propeller scenario . using the magnetosphere radius as the inner disk radius , we find the steady disk luminosity : @xmath203 for @xmath204 from the both sides , for typical @xmath205 ( see table [ tab : zeroes ] ) . the upper limit on the luminosity of the disk with conserved mass is estimated when the inner radius equals the corotation radius : @xmath206 the regime can be sustained for sufficiently strong magnetic dipole and fast rotation of the neutron star : @xmath207 . the disk starts from the outer radius ( the solid line ) or from @xmath208 ( the dashed line ) , corresponding to the green s functions in fig . [ fig.green_finiteb1 ] . [ fig.lumin_zeroai ] , scaledwidth=43.0% ] the flux of the angular momentum @xmath7 at small radii and the inner radius of the disk always manage to adjust themselves to the outer variations that proceed on longer viscous times . the steady value @xmath209 could be low due to a degradation of @xmath210 in the outer disk with low luminosity . then the normalized viscous time in the formulae above may be greater accordingly to @xmath211 . luminosity during the spreading of the ring ( fig . [ fig.green_finiteb1 ] ) is calculated by integrating expression over both surfaces of the disk confined between the outer and inner radius ; it is plotted in fig . [ fig.lumin_zeroai ] . the inner radius of the disk is varying according to eq . and depends on the star s magnetic dipole . its typical value @xmath212r@xmath197 . the torque transfers the angular momentum of the central star to the disk and , via the disk , to the orbital motion of the binary . the corotation radius increases gradually and , eventually , accretion on the center begins @xcite . the characteristic brake time is the angular momentum of the star @xmath213 divided by the torque @xmath214 : @xmath215 where @xmath216 g @xmath217 and a typical value of @xmath50 is substituted . if the mass of the disk grows due to the matter income , the braking time of the star shortens , while the magnetosphere radius decreases at the same time . we return now to the standard accretion disks with zero stress tensor at the center . for large @xmath121 , the first term of the sum dominates , and a single exponential dependence on time prevails : @xmath222 this corresponds to the well - known exponential dependence on time of the accretion rate in the disk with the time - independent viscosity @xmath4 . the characteristic time of the exponential decay is @xmath223 where we use @xmath224 . in table [ tab : zeroes ] we give locations @xmath225 of the first zeroes of equation ( [ eq : out_boundary ] ) for several values of @xmath50 . the table also lists the numerical factors in the expressions for the rise and decay time , and . at late times , the profile of @xmath226 is self - similar , because @xmath227 . the distribution @xmath228 is also self - similar . actually , this profile develops already at the very beginning of the exponential decay . this is illustrated in fig . [ fig.mdisk_mdot ] by the fact that the ratio of the inner accretion rate to the disk mass , which is the integral of the surface density , approaches a constant value that depends only on the form of the radial distribution . in table [ tab : zeroes ] , the values of the parameter @xmath229 can be found . parameter @xmath230 describes the self - similar profile of @xmath231 and was calculated by @xcite for the non - stationary finite @xmath3-disks . for the stationary disks with non - zero accretion rate , @xmath232 . what does the presence of this profile mean for the released energy in the disk ? using the energy balance equation for the optically thick disks , we obtain that at the outer radius @xmath233 , i.e. the outer ring luminosity is by factor @xmath234 less for the viscously evolving disk ( @xmath235 ) than in the case of the stationary disk . the factor @xmath236 for @xmath237 . . [ fig.mdisk_mdot ] , scaledwidth=45.0% ] let us now compare the finite - disk solution for the time - independent @xmath4 and that obtained for the turbulent viscosity in the general form @xmath238 . for the latter form of viscosity , eq . acquires the following view : @xmath239 embracing a dimension constant @xmath240 and dimensionless parameters @xmath241 and @xmath242 : @xmath243 for a finite disk with such type of viscosity the inner accretion rate decays as a power law @xcite : @xmath244 where @xmath245 is the accretion rate at the time @xmath91 , which can be attributed to any moment of the decay stage . . describes the viscous evolution of the @xmath3-disks . the characteristic time of the solution @xmath246 , where @xmath247 is a numerical constant that can be calculated for specific @xmath248 and @xmath62 . using the expression for @xmath249 and relation @xmath250 , we obtain @xmath251 for the @xmath3-disk in the kramers regime of the opacity , parameters @xmath252 , @xmath253 , and @xmath254 @xcite should be used in eqs . . in this case , the dimensionless factor in is close to unity . value of @xmath255 depends on the choice of the zero time . ccccccl 0 & @xmath256 & 1.0585 & @xmath257 & @xmath258 & 1.267 & constant @xmath4 + @xmath259 & @xmath260 & 1.2430 & @xmath261 & @xmath262 & 1.363 & @xmath263disks with constant @xmath65 ; @xmath264disks + @xmath265 & @xmath266 & 1.2927 & @xmath267 & @xmath268 & 1.392 & @xmath263disks , thomson opacity ( @xmath269 ) + @xmath270 & @xmath271 & 1.3793 & @xmath272 & @xmath273 & 1.444 & @xmath263disks , kramers opacity ( @xmath274 ) + 1 & @xmath259 & 1.5708 & @xmath275 & @xmath276 & 1.571 & @xmath277 + 2 & @xmath278 & & & & & @xmath279 does not depend on @xmath17 + + @xmath271 & @xmath280 & 3.4045 & @xmath281 & & 0 & dead @xmath263disks , thomson opacity + @xmath265 & @xmath266 & 3.3425 & @xmath282 & & 0 & dead @xmath263disks , kramers opacity in the shakura sunyaev disks , the viscous stress is proportional to the total pressure in the disk , and this proportionality is parametrized by the @xmath3-parameter @xcite : @xmath283 then the kinematic viscosity can be expressed using ( [ eq : stress_kepl_av ] ) as @xmath284 or @xmath285 to approximate the viscous evolution of the @xmath3-disk by the disk with steady viscosity , one has to estimate the appropriate parameter @xmath62 for eq . and . this can be effectively done using the relation @xmath286 . the solution for the stationary pressure - dominated stationary regions @xcite gives the relative half - thickness @xmath287 , consequently , @xmath288 or @xmath289 in , for the free - free opacity regime . table [ tab : zeroes ] lists the parameters approximating in a similar way the standard @xmath3-disk with the thomson opacity and the ` dead ' disks models @xcite . ( @xmath290 , the solid line ) and @xmath291 ( kramers opacity , @xmath252 , @xmath253 , the dashed line ) . disk parameters @xmath34 and @xmath292 are the same in the both models . [ fig.teordep ] , scaledwidth=45.0% ] in fig . [ fig.teordep ] the two laws are plotted : exponential @xmath293 for @xmath235 and power - law for @xmath294 . viscosity at the outer radius and the outer radius itself are the same . if applied to the real bursts during the time spans of order of @xmath295 , the models can not be easily discriminated ( this was also pointed at by * ? ? ? this fact provides a possibility to estimate @xmath3 using the @xmath1-folding time of the light curve . in the next section , we show how it could be done using the observations of the disks in x - ray novae , whose light curves in many cases show exponential decays . in the x - ray novae , binary systems with a compact star and , typically , a less - than - solar - mass optical star , transient outbursts are thought to originate due to a dwarf nova - type instability ( e.g. , * ? ? ? the temperature rises in a portion of the ` cold ' neutral disk , which becomes ionized , and the viscosity @xmath4 rises . moreover , the turbulent parameter @xmath3 changes from @xmath296 to higher @xmath297 . the heating front moves outward from some ignition radius with the speed @xmath298 . it stops at a radius , where the surface density of the disk is too low to sustain the stable ( on the thermal time - scale ) hot state @xcite . in the dwarf novae , the hot state accretion is quenched swiftly ( comparing to the viscous time ) by the matter returning to the cold state . it was suggested that in the x - ray novae the illumination from the center has a stabilizing effect on the outer disk , keeping the temperatures above the hydrogen ionization temperature @xcite . @xcite propose that in order for the disk to undergo the viscous evolution an intensive irradiation must be included . a canonical light curve of an x - ray nova is described as a fred @xcite . theoretically , if the disk is truncated from outside , a steeply decaying light curve is produced @xcite . the main property is that the disk does not expand infinitely , which is a consequence of its being in a binary system , and its radius does not change on a viscous time scale . scenario by @xcite produces exponential decays due to the steadiness of the parameter @xmath4 in time . more realistic temporal dependence of @xmath4 leads to a steep power - law decay @xcite , which is close to an exponential one during few viscous times ( see the previous section ) . the possibility to model the x - ray novae bursts , which have the fred profile , by the viscously evolving disk with the steady viscosity and after a descrete mass - transfer event , was shown by @xcite ( see also the modeling in @xcite ) . if a burst is due to an instability affecting a slab of matter extended along the radii , the green s function obtained in the present work should be used . effectively , after a heating front has passed , the dwarf - nova type instability leads to some distribution of the hot - ionized matter with high viscosity . in the most trivial case , when the bulk mass is concentrated near @xmath158 , the evolution of the disk approaches the one found by @xcite . let us compare the typical time of the heating front propagation @xmath299 and the time @xmath134 it takes the inner accretion rate to peak , which is of order of @xmath300 ( see and fig . [ fig.cmax1 ] ) . relating @xmath3 and @xmath4 by , we obtain : @xmath301 where the gas temperature is normalized by @xmath302 k , central star mass by a solar mass , and radius by @xmath303 cm , @xmath304 is the molecular weight of the hot matter . radius @xmath125 is the characteristic radius where the bulk mass resides at the beginning of an outburst . this ratio indicates that we are safe to model an x - ray nova outburst at time @xmath305 using the green s function obtained . if one assumes the commonly suggested surface density distribution in the quiescent state in the x - ray novae , @xmath306 @xcite , then , as it is shown in fig . [ fig.mdotin ] , after the peak of the burst , the only important parameters are the disk mass and the outer radius . for such density distribution , the bulk mass of the disk resides close to the outer radius . as we show in sect . [ s.decay ] , already at the maximum of an outburst the disk acquires the self - similar radial distribution , which is not dependent on the initial distribution of matter . the necessary condition for this is that the bulk mass of the viscously evolving disk is located near the outer radius . this is consistent with the absence of a power - law interval before the exponential decline on the the light curves of the x - ray novae that show the fred behaviour . times the square relative half - thickness of the disk at @xmath158 for the six bursts of x - ray novae . lower panel : corresponding estimates of the viscosity parameter @xmath4 at @xmath307r@xmath197 . [ fig.alpha_p ] , scaledwidth=45.0% ] eqs . and yield the ratio @xmath308 taking into account that @xmath134 is very close to @xmath309 for @xmath310 . the value of @xmath311 is depicted in fig . [ fig.cmax1 ] , where @xmath134 is the peak time for the finite disk . for @xmath235 ( see table [ tab : zeroes ] ) we obtain @xmath312 . this provides a rough test whether the particular light curve is produced by the viscously evolving disk made of hot ionized matter and with approximately constant outer radius . the viscous - disk relation @xmath313 for the hot - disks with the kramers opacity in the outer parts leads to a conclusion that to obtain a secondary maximum on the x - ray light curve of an x - ray nova , the disk has to provide an extra mass input at the radius that is determined by the time of the input . the secondary maximuma are usually observed around @xmath314 after the peak . if the extra mass input happens at the time of the burst maximum then it should happen at least at a radius about @xmath315 ( from @xmath316 and @xmath317 ) , without mass input from the intermediate radii . another possibility is that the secondary maximum is triggered later after the peak and then it can happen at a radius @xmath318 ( as in the model of @xcite ) . if an x - ray nova has the viscously evolving accretion disk with constant @xmath158 during the interval of the ` exponential decay ' ( usually , first @xmath319 days after the peak ) , we can infer the value of the turbulent parameter @xmath3 , using the closeness of the model with steady visosity and the @xmath3-disk . as the irradiation , which provides the steadiness of the hot zone outer radius , does not change the vertical structure of the disk @xcite , we can adopt @xmath235 applicable for the hot ionized standard @xmath3-disks . combining eqs . and , we get @xmath320 which relates the @xmath3-parameter to the observed @xmath321 , the time of the bolometric luminosity or the inner accretion rate . the last formula resembles quite a few others found in the literature . similarly , the relation of @xmath3 to the half - thickness of the disk can not be eluded . however , an advatage is that contains the e - folding time which can be inferred quite accurately from observations . we have sampled six fred light curves in different x - ray novae ( fig . [ fig.srcs_lc ] ) and fitted them with the steady-@xmath4 viscously - evolving disk model . for particular values of @xmath322 , the model curves agree well with the peak - normalized count - rates before the second maximum , which is observed in most of the cases . for a0620 - 00 , we use the value of @xmath321 , estimated by @xcite from the @xmath323 kev light curve during the decline . variation of the initial inner radius of the hot zone leads to slight changes of the model curve for the particular initial distribution of @xmath228 , as can be seen in the panel for . in fig . [ fig.alpha_p ] we plot the estimates for @xmath3-parameter times the normalized square of the relative half - thikness at the outer radius of the disk . theses estimates are obtained using with the values @xmath322 indicated in fig . [ fig.srcs_lc ] . in the lower panel , the viscosity @xmath4 is plotted , found accordingly to . as @xmath4 is not constant over radius , we choose to present its value at the same radius r@xmath197 for all cases . the considered x - ray novae have known orbital periods . the masses of the primary and secondary components , mass ratios are from @xcite . for black hole candidate xtej1753.5 - 0127 with @xmath324 hr @xcite we set rather arbitrarily @xmath325 m@xmath197 and @xmath326 m@xmath197 . the roche lobes @xmath327 are calculated using the formula by @xcite , and @xmath328 . the vertical barrs in fig . [ fig.alpha_p ] reflect the errors related to the uncertainties of the binary parameters and do not take into account uncertainty of @xmath322 . @xcite give the half - thickness of the outer parts for the stationary @xmath3-disks when the hydrogen is completely ionized : @xmath329 with the parameters normalizations equal to @xmath330gs@xmath331 and @xmath332 cm . it can be rewritten for the factor in fig . [ fig.alpha_p ] as follows @xmath333 where @xmath3 is normalized to 0.1 and @xmath334 is the conventional rate of accretion producing the critical luminosity : @xmath335 ergs@xmath331 . for the black hole masses in the range @xmath336 m@xmath197 , accretion rates in @xmath337 , @xmath3 in @xmath338 , and @xmath158 in @xmath339r@xmath197 , the above factor changes between extremes @xmath340 and @xmath341 . for 4u1543 - 47 with parameters @xmath342 @xcite , @xmath343 m@xmath197 @xcite , @xmath344 r@xmath197 @xmath345 and @xmath346 one obtains the factor @xmath347 . it follows from fig . [ fig.alpha_p ] that @xmath3 can be less than 1 in this case , but this is still not a self - consistent value . apparently , the disk in this longer - period x - ray nova was not entierly involved in the viscous evolution that produced the fred light curve in 2002 . levels of @xmath3 of the other five bursts ( fig . [ fig.alpha_p ] , the upper panel ) agree with the estimates for the fully ionized , time - dependent accretion disks obtained by various approaches in x - ray and dwarf novae @xcite , in the decretion disks of be stars @xcite . the values of @xmath322 , indicated in fig . [ fig.srcs_lc ] , are to be considered with caution . note that the observed light curves are limited to an energy band . spectral modeling might be required in many cases to obtain accurate @xmath322 . if an x - ray nova during the burst is in the _ high / soft _ state , when the disk emission dominates the spectrum , the spectral modeling can provide @xmath348 in a straightforward way . we expect @xmath322 not to alter by a factor larger than 2 for the x - ray flux dominated by the disk ( preliminary resluts for 2 - 10 kev ) . on the other hand , some x - ray novae bursts proceed entirely in the _ low / hard state _ , such as the burst of xtej1753.5.-0127 @xcite . still , the form of the corresponding x - ray light curves is informatve . even if the optically thick disk does not reach the innermost orbits , it is very probable that the radiated energy is proportional to the accretion rate , because the released potential gravitational energy is the ultimate source to power the source . second , the innermost disk peculiarities do not influence the behavior of @xmath349 as the latter is defined by the process of the viscous matter redistribution in the outer disk . equation is written without the tidal - stress term , following the proposition that it should be negligible everywhere in the disk , except in a thin ring near the tidal truncation radius @xcite . according to @xcite , effects near the tidal - torque radius can be approximated with an effective boundary condition @xmath182 . numerical models confirm that most of the tidal torque is applied in a narrow region at the edge of the disk , where perturbations become non - linear and strong spiral shocks appear ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) . as @xcite notes , in the case of dwarf novae , during outbursts , the observed values of the outer radius appear to be approximately consistent with the theoretical predictions @xcite argue that the action of the tidal torques are important also well inside the tidal radius . they notice at the same time that observations do not allow discriminating between the rapidly growing and smoother tidal stresses , as the model light curves are not strongly affected . the tidal radius for all mass ratios can be approximated as @xmath350 of the average roche lobe radius of a given component in circular binaries @xcite . the disk can extend beyond this radius due to the high inclination of the binary orbit @xcite . effects of the high eccentricity of the binary orbit on the disk truncation radius have been studied by various methods ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . the gravitational influence of the secondary on the disk in a binary system leads to a number of interactions , non - resonant ( or tidal ) and resonant , any of those transferring angular momentum from the disk to the binary @xcite . the tidal distortions generally lead to the truncation of the disk , whereas the resonances of different strength impact or do not the disk structure , depending on their growth rate and other conditions @xcite . an example of a resonance action is the debated explanation of the superhamps in su uma stars ( see , e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in our mathematical set - up , we have presumed a fixed outer radius of the disk without going into details of its actual value . thus the model is appplicable to the coplanar disks in low - eccentricity binaries . the outer radius of the disk can also change due to intrinsic processes . reproducing the whole evolutionary cycles of the dwarf novae and x - ray novae , @xcite , @xcite should have taken into account the variations of the outer disk radius , but these changes are significant during the time spans much greater than the viscous times . under some astrophysical conditions , viscously evolving disks formed in binary systems are effectively truncated from outside . in the present work , we use the method proposed by @xcite to find the green s functions for the linear viscous evolution equation for the disk with the finite outer radius . green s functions are obtained for the viscous angular momentum flux @xmath7 and the accretion rate @xmath8 . two inner boundary conditions at the zero coordinate are considered . they correspond to the accreting disks and to the disks with no mass sink at the center . the green s functions allow one to compute @xmath9 and @xmath351 for arbitrary initial surface density profiles that develop into self - similar distributions on sufficiently long time scale . the analytic formula to calculate the disk evolution with the variable outer mass inflow is derived . a solution , which can be found with the use of the green s function , has the properties defined by the main equation of the viscous evolution and thus provides the basic description of the transient phenomena in the viscous disk . the green s functions are found in the form of quickly converging series and can be easily reproduced with standard computer methods . the c - code written for the case of the accreting disc with the use of the gnu scientific library can be downloadedgalja / lindisk.tgz ] ] . we present the relations between the rising time , @xmath1-folding time , and disk viscosity . for six bursts in the x - ray novae , which are of fred type , we show that the model describes well the peak - normalized light curves before the second maximum . this favors the mainly viscous nature of their evolution during this period and enables us to obtain an estimate of viscosity @xmath4 , which depends on the outer disk radius . the models with time - independent viscosity are shown to approximate well the evolution of @xmath3-disks during time intervals comparable to the viscous time . consequently , estimates of the @xmath3 parameter may be derived for the disks in the binaries with known period and masses . these estimates rely on the relative half - thickness of the disk at the outer radius . the estimates of @xmath3 for the five short - period x - ray novae outbursts , groj0422 + 32 in 1992 , a0620 - 00 in 1975 , gs1124 - 68 in 1991 , gs2000 + 25 in 1998 , and xte j1753.5 - 0127 in 2005 , are in line with the values estimated so far for the hot viscous disks @xcite . another green s fucntion is found for the disk that has zero accretion rate at the inner radius and acquires angular momentum from the central star . in the steady state , the mass of the disk can not be very high and the disk has low luminosity . it radiates most of the rotational power transferred from the central star . this disk works as a gear transmitting the angular momentum of the central star to the orbital motion . the ` dead ' stage is ended by an abrupt fall or trickle of the matter on to the star after the centrifugal barrier has moved close enough or beyond the magnetosphere s radius . the author is grateful to n. i. shakura , k. stempak , j. l. varona , m. perez and the anonymous referee for useful comments . the work is supported by the russian science foundation ( grant 14 - 12 - 00146 ) . green s function for the infinite disk , obtained by lp74 , can be written in the dimensionless form with our designations ( c.f . formula ) as follows : @xmath352 where @xmath353 is the modified bessel function of the first kind , @xmath354 , @xmath355 is some characteristic value of the specific angular momentum , which we use instead of @xmath69 . to find the physical distribution of @xmath226 in the course of the evolution of the initial narrow ring of matter , initially located at the coordinate @xmath356 , one takes the integral of @xmath357 ( [ eq : fring ] ) with the kernel @xmath358 on the interval @xmath359 $ ] : @xmath360 where @xmath127 is the initial mass of the ring . the accretion rate at the center can be found from @xmath361 ( see ( [ eq : mdot_from_green ] ) ) . one finds @xmath362 or @xmath363 where the typical times @xmath364 and @xmath365 are introduced , @xmath366 we find that the accretion rate peaks at the time @xmath367 with the value @xmath368 for infinite disks , see @xcite and @xcite , who have obtained the green s function solution to the viscous evolution of a disk with mass sources / sinks , which are distributed in a fashion along the disk radius . we have the following problem with the inhomogeneous boundary and initial conditions : @xmath369 let us substitute function @xmath231 by a function @xmath370 using the relation @xmath371 this substitution is used , for example , to solve the problem of finding the temperature distribution in a cylinder , on a surface of which a thermal flux is defined @xcite . for the new @xmath370 the problem is the following : @xmath372 where @xmath373 the corresponding sturm liouville problem with a free variable @xmath374 is investigated in sect . [ sec : ev_fin_disk ] : it consists of eq . ( [ eq : sturm - liuv ] ) @xmath375 where @xmath376 or @xmath377 plays a role of an eigenvalue , and the boundary condition @xmath378 at @xmath379 . changing to the free variable @xmath380 , we can write for a non - integer @xmath50 @xmath381\ , , \ ] ] where @xmath382 and @xmath383 depend on the inner - boundary condition . according to the steklov theorem , the solution to ( [ eq : problim_inh2 ] ) can be obtained in the form : @xmath384 where coefficients @xmath385 are to be found from the inhomogeneous first - order differential equation over @xmath121 . let us substitute ( [ eq : steklov ] ) into ( [ eq : sturm - liuv1 ] ) and ( [ eq : problim_inh2 ] ) , multiply by @xmath386 and integrate over @xmath83 from 0 to @xmath138 , using the orthogonality property of the eigenfunctions . we arrive at @xmath387 with designations @xmath388 where @xmath389 is the norm of eigenfunction @xmath390 . solution to is as follows @xmath391 let us write down the solution to ( [ eq : problim_inh2 ] ) for the case @xmath392 , that is , @xmath393 : @xmath394 the last expression is equivalent to . in sect . [ sec : ev_fin_disk ] , we find the coefficients @xmath395 and @xmath82 for the dirac delta function as the initial condition and express @xmath226 using the green s function . in the analogy , the solution to the problem with @xmath396 has the view @xmath397 where @xmath220 is given by ( [ eq : green_zerobi ] ) or ( [ eq : green_zeroai ] ) . substituting ( [ eq : sol_nonzeromdot ] ) into ( [ eq : substit ] ) and using functions @xmath398 and @xmath399 from ( [ eq : problim_inh2 ] ) , we finally arrive at ( [ eq : sol_nonzeromdot_full ] ) . the coordinate integral in its second term can be taken analytically ; it involves the lommel functions and can be tabulated beforehand for the particular values of @xmath50 and @xmath80 for the sake of the computational efficiency . the second term of for @xmath94 can be expressed as : @xmath400 \mathrm{d } \tau~,\ ] ] where @xmath401 , a. t. 2007 , in astronomical society of the pacific conference series , vol . 367 , massive stars in interactive binaries , ed . n. st .- louis & a. f. j. moffat , 485 , astronomical society of the pacific conference series
the abstract state machine ( asm ) thesis , first proposed in @xcite and elaborated in @xcite , asserts that every algorithm is equivalent , on its natural level of abstraction , to an abstract state machine . beginning in @xcite and continuing in @xcite , @xcite , @xcite , and @xcite , the thesis has been proved for various classes of algorithms . in each case , the class of algorithms under consideration was defined by postulates describing , in very general terms , the nature of the algorithms . in each case the main theorem was that all algorithms of this class are equivalent , in a strong sense , to appropriate asms . the work in @xcite directly leading to the present paper is briefly overviewed in the following section . the present paper continues this tradition , but with an important difference . previously , the standard syntax of asms , as presented in @xcite , was adequate , with only very minor modifications . our present work , however , requires a significant extension of that syntax . the extension allows an asm program to refer to the order in which the values of external functions are received from the environment , and it allows the program to declare a step complete even if not all external function values have been determined . the main purpose of this paper is to extend the analysis of interactive , small - step algorithms , begun in @xcite , by removing the restriction to `` ordinary '' algorithms . here we provide postulates and definitions describing a general notion of interactive , small - step algorithm . in a companion paper @xcite we also extend the syntax and semantics of abstract state machines ( asms ) so that non - ordinary algorithms become expressible . there we provide syntax and semantics for asms incorporating interaction that need not be ordinary , verification that asms satisfy the postulates , and proof that every algorithm satisfying the postulates is equivalent , in a strong sense , to an asm . the algorithms considered in this paper proceed in discrete steps and do only a bounded amount of work in each step ( `` small - step '' ) but can interact with their environments during a step , by issuing queries and receiving replies . such algorithms were analyzed in @xcite , subject to two additional restrictions , which we expressed by the word `` ordinary . '' first , they never complete a step until they have received replies to all the queries issued during that step . second , the action of the algorithm the queries issued and the next state produced at any step is completely determined by the algorithm , the current state , and the function that maps the current step s queries to the environment s answers . in other words , this answer function is the only information from the environment that the algorithm uses . in particular , the order in which the environment s answers are received has no bearing on the computation . in the present paper , we lift these restrictions . we allow an algorithm to complete a step even while some of its queries remain unanswered . we also allow the algorithm s actions to depend on the order in which answers were received from the environment . see the discussion in section [ post ] and particularly remark [ time - stamp ] for a discussion of why the order of arrival of answers is of special importance and why it can not be subsumed in answer functions . it was shown in @xcite that ordinary algorithms are equivalent to asms of the traditional sort , essentially as described in @xcite . in order to similarly characterize the more general algorithms of the present paper , we shall , in @xcite , extend the syntax and semantics of asms . in particular , we provide a way to refer to the timing of the evaluations of external functions and a way to terminate a step while some queries remain unanswered . see subsection [ hist ] for a discussion of the intuitive picture of interactive algorithms that leads to these particular extensions and indicates why they suffice . at the referees request , we comment briefly on the impact to be expected from this work . numerous issues are involved here , which would be better addressed in a broad discussion of the asm thesis rather than in this introduction , but it seems appropriate to say a few words about future prospects . probably the most immediate impact that we expect from the asm approach is on the foundations of software engineering . clarification of commonly used notions like state , level of abstraction , interaction is long overdue . closer to theoretical computer science , the asm computation model is the most general computation model that we know . it supports for example computing with structures rather than strings ; see @xcite for particular examples leading to interesting theoretical results and questions . finally it may be disappointing that , unlike the church - turing thesis , the asm thesis has not been used for negative results ; will it be ? the attention of theorists is moving gradually toward the analysis of the overall behavior of algorithms and not only their input - output behavior . the previous work most closely related to the present paper is the series of papers @xcite , in which the asm thesis is proved for other classes of algorithms small - step algorithms in @xcite , parallel algorithms in @xcite ( both without external intra - step interaction , but with communication between subprocesses in the parallel case ) , and ordinary interactive small - step algorithms in @xcite . also , andreas glausch and wolfgang reisig @xcite adapted the postulates of @xcite to describe a restricted but important class of `` small - step '' distributed algorithms . in this section , we briefly overview the work in @xcite directly leading to the present paper , and we also briefly discuss other work that is related to our goals and analysis , though it applies rather different methods . many algorithms are naturally understood on a high level of abstraction ; for example states could be relational structures ( databases ) . the step to a simulation operating over string representations of the abstract entities may lower the abstraction level of the algorithm ; see section [ sec : other ] for discussion . such a step may be less trivial than meets the eye : for instance it is not known whether there is a polynomial - time algorithm that , given two adjacency matrices , determines whether they represent the same graph . the abstract state machine ( asm ) model of @xcite intended to capture arbitrary algorithms at their natural levels of abstraction . the asm thesis of @xcite asserts that every algorithm is equivalent , on its natural level of abstraction , to an abstract state machine . subsequent experimentation provided confirmation of the thesis @xcite . paper @xcite was the first of a series of papers offering _ speculative justification _ for the thesis , for particular classes of algorithms . they all follow the same general pattern ; 1 . describe axiomatically a class @xmath0 of algorithms . 2 . define behavioral equivalence of @xmath0 algorithms . 3 . define a class @xmath1 of abstract state machines . prove the following characterization theorem for @xmath0 : @xmath2 , and every @xmath3 is behaviorally equivalent to some @xmath4 . the characterization provides a theoretical programming language for @xmath0 and opens the way for more practical languages for @xmath0 . the justification of the asm thesis thus obtained is speculative in two ways : * the claim that @xmath0 captures the intuitive class of intended algorithms is open to criticism . * definition of behavioral equivalence is open to criticism . but the characterization of @xmath0 by @xmath1 is precise , and in this sense the procedure _ proves _ the asm thesis for the class of algorithms @xmath0 modulo the chosen behavioral equivalence . so @xmath0 should be as broad as possible within the intended class of algorithms , and behavioral equivalence should be as fine as possible , even if coarser notions of equivalence might suffice for many purposes . the finer the behavioral equivalence , the stronger the characterization theorem . in this subsection we briefly overview the realization of this program for isolated small - step algorithms in @xcite and ordinary interactive small - step algorithms in @xcite . the algorithms of @xcite are executed by a single sequential agent and are isolated in the following sense : there is no interaction with the environment during the execution of a step . the environment can intervene in between algorithm s steps . but we concentrate on step - for - step simulation , and so inter - step interaction with the environment can be ignored . this class of algorithms is axiomatized by three simple postulates . the * sequential time postulate * says that an algorithm defines a deterministic transition system , a ( not necessarily finite - state ) automaton . more explicitly , the algorithm determines * a nonempty collection of states , * a nonempty subcollection of initial states , and * a state - transition function . the algorithm is presumed to be deterministic . nondeterministic choices involve interaction with the environment ; see @xcite for discussion . the term state is used in a comprehensive way . for example , in case of a turing machine , a state would include not only the control state but also the head position and the tape contents . the postulate does not mention a notion of final state , see @xcite for discussion . sequential time already suffices to define _ behavioral equivalence _ of algorithms : two algorithms are behaviorally equivalent if they have the same * states , * initial states , and * state - transition functions . this is as fine as possible , though it may be too fine for many purposes . but , as already noted above , the finer the equivalence , the stronger the characterization theorem . the * abstract state postulate * says that * all states are first - order structures of a fixed vocabulary , * the transition function does not change the base set of a state , and * isomorphism of structures preserves everything , which here means states , initial states and the transition function . it reflects the vast experience of mathematics and mathematical logic according to which every static mathematical situation can be adequately represented as a first - order structure . the idea behind the second requirement is that , even when the base set seems to increase with the creation of new objects , those objects can be regarded as having been already present in a `` reserve '' part of the state . what looks like creation is then regarded as taking an element from out of the reserve and into the active part of the state . ( the nondeterministic choice of the element is made by the environment . ) see @xcite and the next section for discussion . the idea behind the third requirement is that all relevant state information is reflected in the vocabulary : if your algorithm can distinguish red integers from green integers , then it is not just about integers . the * bounded exploration postulate * expresses the idea that a sequential algorithm ( in the traditional meaning of the term ) `` computes in steps of bounded complexity '' @xcite . more explicitly , it asserts that the values of a finite set @xmath5 of terms ( also called expressions ) , that depends only on the algorithm and not on the input or state , determine the state change ( more exactly the set of location updates ) for every step ; see @xcite or the next section for precise definitions of locations and updates . thus , there is a uniform upper bound on the total size of locations that an algorithm needs to explore at any state . bounded exploration is the reason why all algorithms considered , here and in the rest of this paper , are called ` small - step ' . the characterization theorem of @xcite establishes the asm thesis for the class @xmath0 of algorithms defined by sequential time , abstract state , and bounded exploration and the class @xmath1 of machines defined by the basic asm language of update rules , parallel rules and conditional rules @xcite . while the intent in @xcite is to capture algorithms executing steps in isolation from the environment , a degree of _ intra - step interaction _ is accommodated in the asm literature since @xcite : ( i ) using the import command to create new elements , and ( ii ) marking certain functions as _ external _ and allowing the environment to provide the values of external functions . one pretends that the interaction is inter - step . this requires the environment to anticipate some actions of the algorithm . also , in @xcite , nesting of external functions was prohibited ; the first study of asms with nested external functions was @xcite . the notion of _ run _ determines whether _ inter - step _ interaction is allowed , see @xcite for precise definitions and discussion . in @xcite it is argued at length why the inter - step form of interaction can not suffice for all modeling needs . as a small example take the computation of @xmath6 where @xmath7 is an external call ( a query ) with argument 7 , whose result @xmath8 is used as the argument for a new query @xmath9 . an attempt to model this as inter - step interaction would force splitting the computation into substeps . but at some level of abstraction we may want to evaluate @xmath6 within a single step . limiting interaction to the inter - step mode would necessarily lower the abstraction level . thus @xcite sets modeling _ intra - step _ interaction as its goal . different forms of interaction , such as message - passing , database queries , remote procedure calls , inputs , outputs , signals , all reduce to a single universal form : a single - reply zero - or - more - arguments not - necessarily - blocking _ query_. all arguments and the reply ( if any ) should be elements of the state if they are to make sense to the algorithm . for a formal definition of queries see @xcite ; a reminder is given in the next section . for a detailed discussion and arguments for the universality of the query - reply approach see @xcite . articles @xcite limit themselves to interactive algorithms which are _ ordinary _ in the sense that they obey the following two restrictions : 1 . the actions of the algorithm depend only on the state and the replies to queries , and not on other aspects , such as relative timing of replies , and 2 . the algorithm can not complete its step unless it has received replies to all queries issued . the first restriction means that an algorithm can be seen as operating on pairs of form @xmath10 where @xmath11 is a state and @xmath12 an _ answer function _ over @xmath11 : a partial function mapping queries over @xmath11 to their replies . the second restriction means that all queries issued are _ blocking _ ; the algorithm can not complete its step without a reply . ( some uses of non - blocking , asynchronous queries can still be modeled , by assuming that some forms of queries always obtain a default answer . but this is an assumption on environment behavior . ) the present paper lifts both restrictions , and thus extends the theory to _ general _ interactive algorithms . several of the postulates of @xcite cover the same ground as the postulates of @xcite , but of course taking answer functions into account . the most important new postulate is the * interaction postulate * , saying that the algorithm , for each state @xmath11 , determines a _ causality relation _ @xmath13 between finite answer functions and queries . the intuition behind @xmath14 is this : if , over state @xmath11 , the environment behaves according to answer function @xmath12 then the algorithm issues @xmath15 . the causality relation is an abstract representation of potential interaction of the algorithm with the environment . but not all answer functions can actually arise in such interaction . context _ for @xmath16 is a minimal ( with respect to graph inclusion ) answer function @xmath12 with the following closure property : if @xmath17 for some subfunction @xmath18 , then @xmath19 . intuitively , this means that @xmath12 gives answers for exactly those queries that the algorithm would ask , given @xmath12 . this incorporates both the idea of blocking all queries that are issued must be answered and the more basic idea that an algorithm can not see replies to queries that it did nt issue . the * updates postulate * asserts that , for every context @xmath12 for @xmath16 , the algorithm either _ fails _ or produces the next state . this refines the transition relation of sequential time of @xcite . the possibility of explicit failure is new here ; the algorithm may obtain replies that are absurd or inconsistent from its point of view , and it can fail in such a case . the next state , if there is one , is defined by an update set , which can also contain trivial updates : `` updating '' a location to the old value . trivial updates do not contribute to the next state , but in composition with other algorithms can contribute to a clash , see @xcite and also the next section for discussion . the inductive character of the context definition is unwound and analyzed in detail in @xcite . the answer functions which can occur as stages in the inductive construction of contexts are called _ well - founded_. this captures the intuition of answer functions which can actually arise as records of interaction of an algorithm and its environment . two causality relations ( over the same state ) are _ equivalent _ if they make the same answer functions well - founded . equivalent causality relations have the same contexts but the converse is not in general true : intermediate intra - step behavior matters . two ordinary interactive algorithms are _ behaviorally equivalent _ if * they have the same states and initial states , * at all states they have equivalent causality relations , and * at all state - context pairs , either they both fail or they produce the same update sets as required by updates . the * bounded work postulate * of @xcite extends bounded exploration of @xcite to queries . as a consequence every well - founded answer function is finite . furthermore , there is a uniform bound on the size of well - founded answer functions . queries can be expressed by external functions @xcite , but there is a complication . in some cases different textual occurrences of the same external function symbol applied to the same arguments can be reasonably assumed to define the same query , and thus to represent the same reply . in some cases , for instance with object - creation operators of oop languages , different occurrences of the same external function symbol with the same arguments are always assumed to represent different queries which always obtain different fresh replies . there are also many intermediate possibilities . most of @xcite is devoted to a discussion of and a general solution to this problem using the device of _ query - templates _ , which we reuse here . in @xcite the syntax and semantics of ordinary interactive asms is defined . the syntax is essentially the same as the asm language of @xcite with the addition of external function symbols , a subset of the asm language of @xcite . what is new is that the restriction of @xcite , forbidding nesting of external function symbols , can now be safely lifted : the query semantics and the analysis of causality provided by @xcite did not exist at the time @xcite was written . thus the semantics of @xcite was constrained to inter - step interaction only . the characterization theorem for the class @xmath0 of algorithms defined in @xcite and the class @xmath1 of machines defined in @xcite is proved in @xcite . efforts to characterize algorithms mathematically began with the work of alonzo church @xcite and alan turing @xcite in the 1930 s . turing gave a careful analysis and a gradual simplification of what can happen during an algorithmic computation performed by a person . the result was the turing machine model of computation and a mathematical characterization , now universally accepted , of the functions ( from finite strings to finite strings ) that are computable in principle , disregarding limitations of time and memory . the simplifications , however , typically lower the level of abstraction in a drastic way ; just compare any algorithm in @xcite or @xcite with a turing machine implementation of that algorithm . the latter involves details the data representation on the tape , indications how the read / write head should find the right place for the next `` real '' work , etc . that are not essential to the algorithm ( but essential for the turing machine ) . thus , although turing s work successfully characterized the ( in principle ) computable functions , it characterized the algorithms that compute them only to the extent that one ignores levels of abstraction . ( `` in principle '' means that there are no bounds on resources , so `` computable '' does not mean that we are able carry out the computation in the real world . ) the asm thesis , in contrast , asserts that algorithms can be described at their natural levels of abstraction by asms . the textbook algorithms can be written as asms without introducing irrelevant details of implementation . a detailed implementation , such as a turing machine or an assembly language program , can also be written as an asm . there are some other differences between the turing machine model and the asm model related to lowering of the abstraction level in the turing machine model . in the turing model , inputs and outputs are strings . asms deal naturally with computations involving abstract entities . examples include the gauss elimination algorithm over arbitrary fields and the ruler - and - compass constructions of geometry . particularly important examples include computations with finite structures , e.g. relational databases . databases can be implemented as strings . to do that , arbitrary choices may be necessary ( to order the underlying sets ) . thus string encoding exposes the implementation while database queries are supposed to be implementation independent . furthermore , even when an algorithm works with strings , its parallel or interactive character can be lost in a turing machine simulation . slightly before turing s paper @xcite , church @xcite proposed another mathematical characterization of computability , in terms of definability in lambda - calculus . church s proposal turned out to be equivalent to turing s ; that is , the lambda - definable functions are just those that are computable by turing machines . unlike turing s proposal , however , church s was not accompanied by an attempt to analyze the general notion of algorithm ; it therefore lies considerably farther from our work . andrei n. kolmogorov @xcite has given another mathematical description of computation , presumably motivated by the physics of computation rather than by an analysis of the actions of a human computer . for a detailed presentation of kolmogorov s approach , see @xcite . also see @xcite and the references there for information about research on pointer machines . like turing s model , these computation models also lower the abstraction level of algorithms . yiannis moschovakis @xcite proposed that the informal notion of algorithm be identified with the formal notion of _ recursor_. a recursor is a monotone operator over partial functions whose least fixed point includes ( as one component ) the function that the algorithm computes . the approach does not seem to scale to algorithms interacting with an unknown environment . see ( * ? ? ? * section 4.3 ) for a critique of moschovakis s computation model . an approach to interactive computing was pioneered by peter wegner and developed in particular in @xcite . the approach is based on special interactive variants of turing machines called _ persistent turing machines _ , in short ptms . interactive asms can step - for - step simulate ptms . goldin and wegner assert that `` any sequential interactive computation can be performed by a persistent turing machine '' @xcite . but this is not so if one intends to preserve the abstraction level of the given interactive algorithm . in particular , ptms can not step - for step simulate interactive asms @xcite . the topic of the present paper is intra - step interaction between a single sequential - time algorithm and its environment ( that is , the rest of the world ) . we are not aware of any previous literature on intra - step interaction . on the other hand , the communication mechanism that emerges from our study is not limited to intra - step interaction . our investigation did not happen in a vacuum . it is hard for us to indicate particular sources that influenced our thinking , but surely consciously or unconsciously our thinking was influenced by the theory of distributed algorithms , by programming languages attempting to provide means for communication between computing agents , and by computing technology . a few references that seem relevant to us are @xcite . this section is devoted to the description of interactive , small - step algorithms by means of suitable definitions and postulates . some parts of this material are essentially the same as in @xcite , which dealt with small - step algorithms that interact with the environment only between steps ; some other parts are as in @xcite , which dealt with small - step algorithms that interact with the environment within steps but only in the `` ordinary '' manner described in the introduction . we shall present these parts again here , without repeating the explanations and motivations from @xcite and @xcite , but see section [ sec : overview ] . for the genuinely new material , dealing with the non - ordinary aspects of our algorithms interaction with the environment , we shall present not only the definitions and postulates but also the reasons and intuitions that lie behind them . in the following presentation of the postulates and their intuitive motivation , we use the word `` algorithm '' ( or `` interactive , small - step algorithm '' ) in an informal sense . our discussion amounts to an analysis of this intuitive notion , and our postulates are a distillation of the results of the analysis . afterward , in definition [ alg - def ] , we give the phrase `` interactive small - step algorithm '' a formal meaning , based on the postulates . thus , definition [ alg - def ] implicitly asserts a thesis , namely that the postulates and the formal notion defined from them adequately capture the intuitive notion that we analyzed . in @xcite , we shall show that algorithms in our formal sense are behaviorally equivalent to asms . in the light of this result , the thesis implicit in definition [ alg - def ] is really the asm thesis for interactive small - step algorithms . throughout the definitions and postulates that follow , we consider a fixed algorithm @xmath20 . we may occasionally refer to it explicitly , for example to say that something depends only on @xmath20 , but usually we leave it implicit . our first postulate is a cosmetic improvement of the state postulate of @xcite , and most of it is assembled from parts of postulates in @xcite . we refer the reader to @xcite for a careful discussion of the first three parts of the postulate and to @xcite for the last part . states postulate : the algorithm determines a finite vocabulary @xmath21 , a nonempty set s of _ states _ which are all @xmath21-structures , a nonempty subset @xmath22 of _ initial states _ , a finite set @xmath23 of _ labels _ ( to be used in forming queries ) . as in the cited earlier papers , we use the following conventions concerning vocabularies and structures . [ vocab ] a vocabulary @xmath21 consists of function symbols with specified arities . some of the symbols in @xmath21 may be marked as _ static _ , and some may be marked as _ relational_. symbols not marked as static are called _ dynamic_. among the symbols in @xmath21 are the logic names : nullary symbols @xmath24 , @xmath25 , and @xmath26 ; unary @xmath27 ; binary equality ; and the usual propositional connectives . all of these are static and all but @xmath26 are relational . an @xmath21-structure consists of a nonempty base set and interpretations of all the function symbols as functions on that base set . in any @xmath21-structure , the interpretations of @xmath24 , @xmath25 , and @xmath26 are distinct . in any @xmath21-structure , the interpretations of relational symbols are functions whose values lie in @xmath28 . in any @xmath21-structure @xmath11 , the interpretation of @xmath27 maps @xmath29 and @xmath30 to @xmath29 and everything else to @xmath30 . in any @xmath21-structure @xmath11 , the interpretation of equality maps pairs of equal elements to @xmath29 and all other pairs to @xmath30 . in any @xmath21-structure @xmath11 , the propositional connectives are interpreted in the usual way when their arguments are in @xmath28 , and they take the value @xmath30 whenever any argument is not in @xmath28 . we may use the same notation @xmath11 for a structure and its base set . we may omit subscripts @xmath11 , for example from @xmath24 and @xmath25 when there is no danger of confusion . a _ potential query _ in state @xmath11 is a finite tuple of elements of @xmath31 . a _ potential reply _ in @xmath11 is an element of @xmath11 . here @xmath31 means the disjoint union of @xmath11 and @xmath23 . so if they are not disjoint , then they are to be replaced by disjoint isomorphic copies . we shall usually not mention these isomorphisms ; that is , we write as though @xmath11 and @xmath23 were disjoint . interaction of an algorithm with its environment has been analyzed at length in @xcite . the analysis established that all forms of interaction reduce to queries and replies . the algorithms of @xcite are insensitive to the relative timing of replies within a step , and the environment interaction during a step could be expressed there as an answer function mapping queries to their replies . the following simple example shows that the expressive power of answer functions is not sufficient in general . [ ex : broker ] for a simple example , consider a broker who has a block of shares to sell . he asks two clients whether they want to buy the shares . both of them reply that they want to buy the whole block . the broker will sell the shares to the client whose message reaches him first . [ ans - fn ] an _ answer function _ for a state @xmath11 is a partial map from potential queries to potential replies . a _ history _ for @xmath11 is a pair @xmath32 consisting of an answer function @xmath33 together with a linear pre - order @xmath34 of its domain . by the _ domain _ of a history @xmath35 , we mean the domain @xmath36 of its answer function component , which is also the field of its pre - order component . recall that a _ pre - order _ of a set @xmath37 is a reflexive , transitive , binary relation on @xmath37 , and that it is said to be _ linear _ if , for all @xmath38 , @xmath39 or @xmath40 . the equivalence relation defined by a pre - order is given by @xmath41 the equivalence classes are partially ordered by @xmath42\leq[y]\iff x\leq y,\ ] ] and this partial order is linear if and only if the pre - order was . the _ length _ of a linear pre - order is defined to be the order type of the induced linear ordering of equivalence classes . ( we shall use this notion of length only in the case where the number of equivalence classes is finite , in which case this number serves as the length . ) we also write @xmath43 to mean @xmath39 and @xmath44 . ( because a pre - order need not be antisymmetric , @xmath43 is in general a stronger statement than the conjunction of @xmath39 and @xmath45 . ) when , as in the definition above , a pre - order is written as @xmath34 , we write the corresponding equivalence relation and strict order as @xmath46 and @xmath47 . the same applies to other subscripts and superscripts . we use histories to express the information received by the algorithm from its environment during a step . the notion of answer function comes from @xcite , where these functions , telling which queries have received which replies , represented the whole influence of the environment on the algorithm s work . when algorithms are not required to be ordinary , then additional information from the environment , namely the relative order in which replies were received , becomes relevant to the computation . this information is represented by the pre - order part of a history . if @xmath48 and @xmath49 , this means that the answer @xmath50 to @xmath51 was received strictly before the answer @xmath52 to @xmath15 . if @xmath53 , this means that the two answers were received simultaneously the rest of this subsection is devoted to explaining in more detail the intuition behind this formalization of the intuitive notion of the history of an algorithm s interaction with its environment during a step . we do not , however , repeat here sections 2 and 4 of @xcite . the first of these two sections explains in detail our reasons for using queries and replies as our model of the interaction between an algorithm and its environment . the second explains the reasons for our specific definitions of ( potential ) queries and replies . here , we presuppose these explanations and adopt the query - reply paradigm of @xcite ; see also section [ sec : overview ] . our task now is to explain what is added to the picture from @xcite when we remove the restriction to ordinary algorithms . much of the information provided by an environment can and should be viewed as being part of its replies to queries . this includes not only the information explicitly requested by the query but also such varied information as `` how hard did the user ( who is part of the environment ) bang on the keyboard when typing this input '' or `` at what time was this input provided '' if such information is relevant to the algorithm s execution . thus , we can view such information as being included in the answer function @xmath54 , without referring to the second component of a history , the pre - order . the purpose of the pre - order , on the other hand , is to represent the order in which the algorithm becomes aware of the environment s answers . even if the environment provides a time - stamp as part of each reply , this order can not be read off from the replies . the algorithm may become aware of the replies in an order different from that indicated by the time stamps . in the broker example [ ex : broker ] , the broker will sell the shares to the client whose message reaches him first , even if the message from the other client is sent earlier . [ time - stamp ] could the environment provide , as part of its replies , time stamps that tell when the algorithm becomes aware of the replies ? if so , could we then proceed as in @xcite , where algorithms get information from the environment only via the answer function ? in other words , could we dispense with the pre - order component in histories ? the idea of a time stamp telling not when the reply was sent but when it is received is not as absurd as it might appear . the factors affecting the time of receipt , such as the speed of connections over which the reply travels and the speed at which the algorithm is executed , are not specified by the algorithm and the state , so they are part of the environment , albeit a different part from that ordinarily involved in answering queries . so the complete environment could conceivably assemble replies that include the time that the algorithm sees them . such an environment would be quite unusual , and of course it would have to be subject to the constraint that the time stamps accurately reflect at least the order in which replies are received . but , if such an environment can be imagined , could we use it , at least for theoretical purposes , to avoid the need for the pre - order compnent in histories ? if so , then the present paper could stay closer to the work previously done in @xcite , and we would have been quite happy to avoid the extra work . the pre - orders are , however , needed , even for theoretical purposes . consider the broker example above . part of the broker s algorithm tells him what to do if both clients reply . to model that in the style of @xcite , assuming that replies include accurate information about their arrival times , the algorithm would have to wait for both replies , read both time stamps , and act accordingly . but in fact , the algorithm does not wait for both replies . if it receives one reply , it acts on the basis of that reply without waiting for the other . in effect , this algorithm and others like it decide the relative order of receipt of two replies as soon as one of the replies is received . if one is received and the other is nt , then the received one is earlier . this common sense approach is unavailable if the timing comparisons have to be based on reading time stamps embedded in the replies . thus , even if we constrained our environments to always provide time stamps that accurately indicate the time a reply is seen by the algorithm , we would still need machinery like that in the present paper to decide the relative order of two replies even if only one has been seen . even though the pre - order @xmath34 is about the replies , the formal definition says that it pre - orders the domain of the answer function @xmath54 , i.e. , the set of queries . the reason is a technical one : different queries may receive the same reply , and in that case the single reply could occur at several positions in the ordering . each query , on the other hand , is issued just once because , in accordance with the conventions of @xcite , if what appears to be the same query is issued repeatedly , then we regard the repetitions as distinct queries ( since they can receive different replies ) . thus , the order in which replies are received can be conveniently represented by pre - ordering the associated queries . it may be more intuitive to think of pre - ordering the set of pairs ( query , reply ) , i.e. , the graph of the answer function . this view of the situation would make no essential difference , since these pairs are in canonical bijection with the queries . we emphasize that the timing we are concerned with here is logical time , not physical time . that is , it is measured by the progress of the computation , not by an external clock . if external , physical , clock time is relevant , as in real - time algorithms , it would have to be provided separately by the environment , for it is not part of the program or of the state . the relevant values of the physical time could be regarded as the replies to repeated queries asking `` what time is it ? '' in particular , we regard a query as being issued by the algorithm as soon as the information causing that query ( in the sense of the interaction postulate below ) is available . this viewpoint would be incorrect in terms of physical time , for the algorithm may be slow to issue a query even after it has the prerequisite information . but we can regard a query as being logically available as soon as information causing it is present . this is why we include , in histories , only the relative ordering of replies . the ordering of queries relative to replies or relative to each other is then determined . the logical time of a query is the same as the logical time of the last of the replies causing that query . our use of pre - orders rather than partial orders , i.e. , our allowing two distinct replies to be received simultaneously , also reflects our concern with logical rather than physical time . one can argue that no two events are exactly simultaneous in the physical sense ( though they could be so nearly simultaneous that one should treat them as simultaneous ) , but logical simultaneity is certainly possible . it means merely that whenever the computation had access to one of the replies it also had access to the other . the linearity of the pre - ordering , i.e. , the requirement that every two replies be received either in one order or in the other or simultaneously , formalizes the following important part of our view of sequential - time ( as opposed to distributed ) algorithms . ( it is not a postulate of our formal development but an informal principle that underlies some of our definitions and postulates . ) one executor principle a small - step algorithm is executed by a single , sequential entity . even if there are ( boundedly many ) subprocesses running in parallel , they all operate under the control of a single , master executor . the aspect of this principle that is relevant to the linearity of histories can be summarized in the following informal principle , in which the first part emphasizes the role of the master executor in the interaction with the environment , while the second part is a consequence of the sequentiality of the executor . holistic principle the environment can send replies only to ( the executor of ) the algorithm , not directly to parts of it ; similarly it receives queries only from the algorithm , not from parts of it . any two replies are received by the algorithm either simultaneously or one after the other . the pre - orders in our histories are intended to represent the order in which replies are received by the master executor . if there are subprocesses , then the master may pass the replies ( or information derived from them ) to these subprocesses at different times , but that timing is part of the algorithm s internal operation ( possibly also influenced by the replies to other queries like `` what time is it ? '' ) , not part of the environment interaction that histories are intended to model . the linearity of our pre - orders reflects the fact that they represent the ordering as seen by a single , sequential entity ; this entity sees things in a sequence . in more detail , we picture the execution of one step of an algorithm as follows . first , the algorithm ( or , strictly speaking , its executor ) computes as much as it can with only the information provided by the current state . this part of the computation , the first phase , will in general include issuing some queries . then the algorithm pauses until it is `` awakened '' by the environment , which has replied to some ( not necessarily all ) of the queries from phase 1 . the algorithm proceeds , in phase 2 , to compute as much as it can using the state , the new information from the environment , and the information recorded during phase 1 . then it again pauses until the environment has provided some more replies ( possibly to queries from phase 2 and possibly to previously unanswered queries from phase 1 ) and awakens the algorithm . then phase 3 begins , and this pattern continues until the algorithm determines , in some phase , that it is ready to complete the current step , either by executing certain updates ( computed during the various phases ) of its state or by failing ( in which case the whole computation fails and there is no next state ) . the logical ordering of replies refers to the phases at which replies were received . that is , if @xmath55 , this means that the reply @xmath56 to @xmath57 was received at an earlier phase than the reply @xmath58 to @xmath59 . similarly , @xmath60 means that these two replies were received at the same phase . we have assumed , in the preceding description of phases , that the algorithm is awakened to begin a new phase only when some new reply has been provided . one may worry that this is not general enough , that the environment can awaken the algorithm even when no new replies are available . the rest of this remark is devoted to discussing this issue and justifying our assumption ; it can safely be skipped by readers not worried about such subtleties . in such a case , it is natural to assume that the algorithm , having determined that it has no new information with which to advance the computation , simply resumes its pause until awakened again . in order for the algorithm to be a small - step algorithm , such fruitless awakenings must happen only a bounded number of times per step , with the bound depending only on the algorithm . the reason is that , even if all the algorithm does when awakened is to observe that no new replies have arrived , this observing is work , and a small - step algorithm can do only a bounded amount of it per step . it seems possible , however , for an algorithm to admit a few fruitless awakenings per step , and the results of the computation could even depend on these awakenings . consider , for example , an algorithm that works as follows . it begins by issuing a query @xmath15 and pausing . when awakened , it outputs 1 and halts if there is a reply to @xmath15 ; otherwise , it pauses again . when awakened a second time , it outputs 2 if there is now a reply to @xmath15 , and , whether or not there is a reply , it halts . notice that , if @xmath15 receives the reply @xmath61 , then whether this is seen at the first or the second awakening does nt affect the history , which consists of the function @xmath62 and the unique pre - order on its domain @xmath63 . so the history fails to capture all the information from the environment that is relevant to the computation . there are two ways to correct this discrepancy . one , which we shall adopt , is to regard the fruitless awakening as amounting to a reply to an implicit query , of the form `` i m willing to respond to awakening . '' ( for several fruitless awakenings , there would have to be several such queries , distinguished perhaps by numerical labels . ) now the two scenarios considered above , where @xmath15 has received a reply at the first awakening or only at the second , are distinguished in the histories , because the reply to the new , implicit query will be simultaneous with the reply to @xmath15 in the one scenario and will strictly precede the reply to @xmath15 in the other scenario . an alternative approach would be to avoid introducing such implicit queries but instead to replace the pre - order constituents of histories by slightly more complicated objects , `` pre - orders with holes . '' the idea is that the scenario where an answer to @xmath15 is available only at the second awakening would be represented by the answer function @xmath62 as above , but the pre - order would be replaced with something that says `` first there s a hole and then @xmath15 . '' formalizing this is actually quite easy , as long as the pre - order is linear and the set @xmath37 to be pre - ordered is finite ( as it will be in the situations of interest to us ) . a linear pre - order of @xmath37 with holes is just a function @xmath51 from @xmath37 into the natural numbers . for any @xmath64 , the elements of @xmath65 constitute the @xmath66 equivalence class , which may be empty in case of a hole . ordinary pre - orders correspond to the special case where the image @xmath67 is an initial segment of n. although the approach using pre - orders with holes corresponds more directly to intuition , we prefer the first approach , with implicit queries , for two reasons . first , it allows us to use the standard terminology of pre - orders rather than introducing something new . second , its chief disadvantage , the need for implicit queries , is mitigated by the fact that we need other sorts of implicit queries for other purposes . for example , in ( * ? ? ? * section 2 ) , implicit queries modeled the algorithm s receptiveness to incoming ( unsolicited ) messages and to multiple answers to the same query . notice that , in all cases , our implicit queries represent the algorithm s willingness to pay attention to something provided by the environment . we next define initial segments of histories , which will serve to model the interaction with the environment part way through a step . let @xmath68 be a pre - order of a set @xmath37 . an _ initial segment _ of @xmath37 with respect to @xmath68 is a subset @xmath69 of @xmath37 such that whenever @xmath39 and @xmath70 then @xmath71 . an _ initial segment _ of @xmath68 is the restriction of @xmath68 to an initial segment of @xmath37 with respect to @xmath68 . an _ initial segment _ of a history @xmath72 is a history @xmath73 , where @xmath69 is an initial segment of @xmath36 with respect to @xmath34 . ( we use the standard notation @xmath74 for the restriction of a function or a relation to a set . ) we write @xmath75 to mean that the history @xmath76 is an initial segment of the history @xmath35 . notice that any initial segment with respect to a pre - order @xmath68 is closed under the associated equivalence @xmath77 . notice also that if @xmath78 then @xmath79 is an initial segment of @xmath34 . we also point out for future reference that , if two histories @xmath80 and @xmath81 are initial segments of the same @xmath35 , then one of @xmath80 and @xmath81 is an initial segment of the other . intuitively , if @xmath72 is the history of an algorithm s interaction with the environment up to a certain phase , then a proper initial segment of this history describes the interaction up to some earlier phase . if @xmath68 pre - orders the set @xmath37 and if @xmath82 , then we define two associated initial segments as follows . @xmath83 the following postulate is almost the same as the postulate of the same name in @xcite . the only difference is that we use histories instead of answer functions . this reflects the fact that the decision to issue a query can be influenced by the timing of the replies to previous queries . interaction postulate for each state @xmath11 , the algorithm determines a binary relation @xmath13 , called the _ causality relation _ , between finite histories and potential queries . the intended meaning of @xmath84 is that , if the algorithm s current state is @xmath11 and the history of its interaction so far ( as seen by the algorithm during the current step ) is @xmath35 , then it will issue the query @xmath15 unless it has already done so . when we say that the history so far is @xmath35 , we mean not only that the environment has given the replies indicated in @xmath33 in the order given by @xmath34 , but also that no other queries have been answered . thus , although @xmath35 explicitly contains only positive information about the replies received so far , it also implicitly contains the negative information that there have been no other replies . of course , if additional replies are received later , so that the new history has @xmath35 as a proper initial segment , then @xmath15 is still among the issued queries , because it was issued at the earlier time when the history was only @xmath35 . this observation is formalized as follows . for any state @xmath11 and history @xmath35 , we define sets of queries @xmath85 thus , @xmath86 is the set of queries that have been issued by the algorithm , in state @xmath11 , by the time the history is @xmath35 , and @xmath87 is the subset of those that have , as yet , no replies . [ descr - causes ] we have described the causality relation in terms of detailed causes , histories @xmath35 that contain the algorithm s whole interaction with environment up to the time the caused query is issued . a text describing the algorithm , either informally or in some programming language , would be more likely to describe causes in terms of partial information about the history . for example , it might say `` if the query @xmath51 has received the reply @xmath8 but @xmath88 has no reply yet , then issue @xmath15 . '' this description would correspond to a large ( possibly infinite ) set of instances @xmath89 of the causality relation , namely one instance for every history @xmath35 that fits the description , i.e. , such that @xmath90 and @xmath91 . more generally , whenever we are given a description of the conditions under which various queries are to be issued , we can similarly convert it into a causality relation ; each of the conditions would be replaced by the set of finite histories that satisfy the condition . the reverse transformation , from detailed causes representing the whole history to finite descriptions , is not possible in general . the difficulty is that the finite descriptions might have to specify all of the negative information implicit in a history , and this might be an infinite amount of information . in most situations , however , the set of queries that might be issued is finite and known . then any finite history @xmath35 can be converted into a finite description , simply by adding , to the positive information explicit in @xmath35 , the negative information that there have been no replies to any other queries among the finitely many that could be issued . for semantical purposes , as in the present section , it seems natural to think of causes as being the detailed histories . from the behavior of an algorithm , one can easily determine whether @xmath84 ( at least when @xmath35 is attainable as in definition [ att - def ] below and we ignore causes of @xmath15 that have proper initial segments already causing @xmath15 ) ; just provide the algorithm with replies according to @xmath35 and see whether it produces @xmath15 at that moment ( and not earlier ) . it is not easy to determine , on the basis of the behavior , whether @xmath15 is issued whenever a description , like `` @xmath51 has reply @xmath8 and @xmath88 has no reply '' is satisfied , as this involves the behavior in a large , possibly infinite number of situations . for syntactic purposes , on the other hand , descriptions seem more natural than detailed histories . and indeed , the syntax of our asms ( in @xcite ) will not involve histories directly but will involve descriptions like `` @xmath51 has reply @xmath8 and @xmath88 has no reply '' in the guards of conditional rules . notice that there is no guarantee that @xmath92 , although this will be the case for attainable histories ( defined below ) . in general , a history as defined above may contain answers for queries that it and its initial segments do nt cause . it may also contain answers for queries that would be issued but only at a later phase than where the answers appear . in this sense , histories need not be possible records of actual interactions , between the algorithm and its environment , under the causality relation that is given by the algorithm . the following definition describes the histories that are consistent with the given causality relation . informally , these are the histories where every query in the domain has a legitimate reason , under the causality relation , for being there . a history @xmath35 is _ coherent _ , with respect to a state @xmath11 or its associated causality relation @xmath13 , if @xmath93 , and the linear order of the @xmath46-equivalence classes induced by @xmath34 is a well - order . the first requirement in this definition , which can be equivalently rewritten as @xmath94 says that , if a query @xmath15 receives an answer at some phase , then it must have been issued on the basis of what happened in strictly earlier phases . in particular , the queries answered in the first phase , i.e. , those in the first @xmath46-class , must be caused by the empty history . the second requirement is needed only because we allow infinite histories ; finite linear orders are automatically well - orders . the purpose of the second requirement is to support the usual understanding of `` cause '' by prohibiting an infinite regress of causes . it will follow from the bounded work postulate below in conjunction with the interaction postulate above , that we can confine attention to histories whose domains are finite . then clause 2 in the definition of coherence would be automatically satisfied . nevertheless , we allow a coherent history @xmath35 to have infinite domain and even to have infinitely many equivalence classes , with respect to @xmath34 , provided their ordering induced by @xmath34 is a well - ordering . there are two reasons for this generality . first , it will allow us to state the bounded work postulate in a relatively weak form and then to deduce the stronger fact that the histories we really need to consider ( the attainable ones ) are finite . second , it avoids formally combining issues that are conceptually separate . finiteness is , of course , an essential aspect of computation , especially of small - step computation , and the bounded work postulate will formalize this aspect . but the notions of coherent and attainable histories are conceptually independent of finiteness , and so we do not impose finiteness in their definitions . [ cpt1 ] as mentioned above , the notion of coherence is intended to capture the idea of a history that makes sense , given the algorithm s causality relation . here we implicitly use the fact that we are describing an entire algorithm , not a component that works with other components to constitute an algorithm . if we were dealing with a component @xmath95 , working in the presence of other components , then it would be possible for queries to be answered without having been asked by @xmath95 , simply because another component could have asked the query . see @xcite for a study of components . it follows immediately from the definitions that any initial segment of a coherent history is coherent . a history @xmath35 for a state @xmath11 is _ complete _ if @xmath96 . thus , completeness means that all queries that have been issued have also been answered . it can be regarded as a sort of converse to coherence , since the latter means that all queries that have been answered have also been issued earlier . if a complete history has arisen in the course of a computation , then there will be no further interaction with the environment during this step . no further interaction can originate with the environment , because no queries remain to be answered . no further interaction can originate with the algorithm , since @xmath35 and its initial segments do nt cause any further queries . so the algorithm must either proceed to the next step ( by updating its state ) or fail . at the end of a step the algorithm will , unless it fails , perform some set of updates . the next postulate will express this property of algorithms formally . first , we adopt the formalization of the notion of update that was used in earlier work , starting with @xcite . a _ location _ in a state @xmath11 is a pair @xmath97 where @xmath98 is a dynamic function symbol from @xmath21 and @xmath99 is a tuple of elements of @xmath11 , of the right length to serve as an argument for the function @xmath100 interpreting the symbol @xmath98 in the state @xmath11 . the _ value _ of this location in @xmath11 is @xmath101 . an _ update _ for @xmath11 is a pair @xmath102 consisting of a location @xmath103 and an element @xmath104 of @xmath11 . an update @xmath102 is _ trivial _ ( in @xmath11 ) if @xmath104 is the value of @xmath103 in @xmath11 . we often omit parentheses and brackets , writing locations as @xmath105 instead of @xmath106 and writing updates as @xmath107 or @xmath108 instead of @xmath109 or @xmath110 . the intended meaning of an update @xmath107 is that the interpretation of @xmath98 is to be changed ( if necessary , i.e. , if the update is not trivial ) so that its value at @xmath99 is @xmath104 . because our algorithms are not required to be ordinary , they may finish a step without reaching a complete history . the following postulate says that they must finish the step when they reach a complete history , if they have not already done so earlier . step postulate part a the algorithm determines , for each state @xmath11 , a set @xmath111 of _ final histories_. every complete , coherent history has an initial segment ( possibly the whole history ) in @xmath111 . intuitively , a history is final for @xmath11 if , whenever it arises in the course of a computation in @xmath11 , the algorithm completes its step , either by failing or by executing its updates and proceeding to the next step . since incoherent histories can not arise under the algorithm s causality relation , it will not matter which such histories are final . we could have postulated that only coherent histories can be final , but there is no need to do so . it does no harm to allow incoherent final histories , though they are irrelevant to the algorithm s computation . by allowing them , we may slightly simplify the description , as algorithms subject to our postulates , of programs written in some programming language ; if the language allows one to say `` finish the step '' then this can be directly translated into final histories , without having to check for coherence . similar comments apply to the remaining parts of the step postulate below , where we require update sets or failure for coherent histories but allow them also for incoherent ones . [ att - def ] a history for a state @xmath11 is _ attainable _ ( in @xmath11 ) if it is coherent and no proper initial segment of it is final . since initial segments of coherent histories are coherent , it follows that initial segments of attainable histories are attainable . the attainable histories are those that can occur under the given causality relation and the given choice of final histories . that is , not only are the queries answered in an order consistent with @xmath13 ( coherence ) , but the history does not continue beyond where @xmath111 says it should stop . our earlier statements to the effect that incoherent histories do nt matter can be strengthened to say that unattainable histories do nt matter . thus , for example , although we allow one final history to be a proper initial segment of another ( even when both are coherent ) , the longer of the two will be unattainable , so its being final is irrelevant . step postulate part b for each state @xmath11 , the algorithm determines that certain histories _ succeed _ and others _ fail_. every final , attainable history either succeeds or fails but not both . we write @xmath112 for the set of successful final histories and @xmath113 for the set of failing final histories . non - final or unattainable histories may succeed or fail or both , but this will be irrelevant to the behavior of the algorithm . if we restrict attention to attainable histories , then @xmath112 and @xmath113 partition @xmath111 . the intended meaning of `` succeed '' and `` fail '' is that a successful final history is one in which the algorithm finishes its step and performs a set of updates of its state , while a failing final history is one in which the algorithm can not continue the step ends , but there is no next state , not even a repetition of the current state . such a situation can arise if the algorithm computes inconsistent updates . it can also arise if the environment gives inappropriate answers to some queries . for more about failures , see the discussion following the update postulate in ( * ? ? ? * section 5 ) . there is , however , one difference between our present situation and that in @xcite . there , the algorithm depended upon getting answers to all its queries , so it is reasonable to expect that an inappropriate answer always results in failure . in our present situation , however , the algorithm may be willing to finish its step without getting the answer to a certain query . in this case , an inappropriate answer to that query need not force failure . when a history in @xmath114 arises during the computation , the algorithm should perform its updates and proceed to the next step . we formalize this in the third and last part of the step postulate , keeping in mind that only attainable histories can arise . the notation @xmath115 for the update set is taken from @xcite . the superscript @xmath116 refers to the fact that trivial updates can be included in the update set . although they do not affect the next state , they can affect whether clashes occur when our algorithm is run in parallel with another . step postulate part c for each attainable history @xmath117 for a state @xmath11 , the algorithm determines an _ update set _ @xmath118 , whose elements are updates for @xmath11 . it also produces a _ next state _ @xmath119 , which has the same base set as @xmath11 , has @xmath120 if @xmath121 , and otherwise interprets function symbols as in @xmath11 . the step postulate is the analog , for our present situation , of the update postulate of @xcite . there are two differences between these postulates , reflecting the two requirements of ordinariness in @xcite that have been lifted in the present work . first , since we now consider algorithms that can finish a step without waiting for all queries to be answered , we include in the notion of algorithm the information about when to finish a step . this is the notion of final history introduced in part a of the step postulate . if an algorithm never finishes a step while queries are still pending , then its final histories could be defined as being the complete histories ( or the complete , coherent histories ) , and there would be no need to specify f separately in the algorithm . the complete , coherent histories correspond to the contexts , as defined in @xcite . second , because we allow an algorithm to take into account the order in which it receives replies to its queries , decisions about failures and updates depend not only on the answer function @xmath33 but on the whole history @xmath35 , including the ordering information contained in @xmath34 . the intuition and motivation behind the other aspects of the step postulate are discussed in detail in @xcite and @xcite . in notations like @xmath111 , @xmath114 , @xmath122 , @xmath118 , and @xmath119 , we may omit @xmath11 if only one @xmath11 is under discussion . we may also add the algorithm @xmath20 as a superscript if several algorithms are under discussion . notice that the step postulate requires @xmath118 and @xmath119 to be defined when @xmath35 is an attainable , final , successful history , but it allows them to be defined also for other histories . any values they may have for other histories , however , will not affect the algorithm s computation . notice also that the next state is completely determined by the current state and the update set . so when describing an algorithm , we need not explicitly describe @xmath123 if we have described @xmath115 . if @xmath118 clashes , i.e. , if it contains two distinct updates of the same location , then the description of @xmath119 is contradictory , so the next state can not exist . thus , if such a @xmath35 is attainable and final , it must be failing . that is , clashes imply failure . [ ex;broker : re ] we revisit the broker example [ ex : broker ] in terms of our definitions . part of the broker s algorithm is easy to describe : the empty history causes two queries , @xmath124 and @xmath125 . query @xmath126 expresses the take - it - or - leave - it offer to client @xmath127 to buy a block of @xmath8 shares of stock @xmath128 at price @xmath51 . a history that contains a positive reply from just one client is final and successful and produces updates recording a sale to that client . note that , if both clients reply positively but at different times , then the history as an initial segment containing only the reply from the client who answered first , so the stock is sold to this client . it is also easy to handle histories in which both clients answer negatively ; such a history is successful and final but results in no updates ( or perhaps in an update recording that these clients had the chance to buy the stock and turned it down ) . note that , although this situation represents a failure of the broker to accomplish his aim , it should not be considered a failing history , as it does not mean that the broker goes out of business . what if the broker receives positive replies from both clients at the same time ? it is tempting and perhaps realistic to assume that this situation can not occur , but that assumption amounts to a constraint on the environment , and we do not consider in this paper such a restriction of the theory to some `` admissible '' histories . furthermore , when the broker really can not receive two replies simultaneously , this will be because of some aspect of his receiving mechanism , and it seems that such aspects can be modeled using one of the following ideas . the broker could systematically prefer one of the clients . then a history with simultaneous positive replies from both clients would be final and successful and would produce updates recording a sale to the preferred client . alternatively , the broker could break ties non - deterministically . that is , a history with simultaneous positive replies from both clients would cause a new query to the environment , asking for a non - deterministic choice of a client . any reply naming a client would result in a successful final history with a sale to that client . to finish the specification of the algorithm , we must say what is to be done if netiher client replies , or if one replies negatively and the other does nt reply . one approach is to declare such histories not to be final . the broker simply waits , perhaps forever . a more realistic approach is that the broker s offers to the clients include a time limit . an easy way to model this is to have the empty history cause , in addition to the offers to the clients , a third query @xmath129 , to be thought of as a time - out query , i.e. , a request to be informed by the environment , when the time limit has been reached . then any history with a reply to this query but no positive reply from a client will be final and successful and result in no update ( or an update recording that there was no sale and why ) . the algorithm can still hang , if the environment fails to respond to the time - out query , but that seems a reasonable result for a broker who imposes a time - limit but ca nt find out when it has been reached . although the primary subject of this paper is intra - step interaction between an algorithm and its environment , it is not our intention to prohibit inter - step interaction of the sort described in @xcite and mentioned in [ sec : overview : seqth ] . inter - step interaction is , however , quite easy to describe ; the environment is permitted to make arbitrary changes to the state s dynamic functions between the steps of the algorithm . thus , a _ run _ of an algorithm is a ( finite or infinite ) sequence of states , starting with an initial state , in which each state but the first is obtained from its predecessor either by an intervention of the environment or by a step of the algorithm . in the intervention case , the successor state has the same base set and static functions as the predecessor , but the dynamic functions can be arbitrarily altered . in the case of an algorithm step , there is a successful final history @xmath35 , describing the environment s replies to the algorithm s queries during the step , and if the state before the step was @xmath11 then the state after the step is @xmath119 . since what happens in the intervention steps is quite arbitrary , our concern in this paper is to analyze what happens in the algorithmic steps . the notion of run , therefore , occurs only in peripheral remarks like this one , not in the main development . a referee asked why both intra - step and inter - step interaction are needed . in fact , one could eliminate inter - step interaction , but it would require considerable cooperation from the environment because of the following complications . note that there is no a priori bound on the number of locations that the environment could update in an inter - step action . so that action can not be simulated by a small - step algorithm in one step . this problem can be circumvented by having the algorithm ask about these environmental updates only when they are needed for the computation . this means that the environment must `` remember '' these updates until they are needed ; furthermore , it must keep track of whether these updates are overwritten by the algorithm s own updates . as in previous work , starting with @xcite , we require that the information relevant to the algorithm s computation that is given by the state must be explicitly given by the structure of the state , not implicitly given by the particular nature of its elements . formally , this means that the computation must be invariant under isomorphisms . any isomorphism @xmath130 between states can be extended in an obvious , canonical way to act on queries , answer functions , histories , locations , updates , etc . we use the same symbol @xmath127 for all these extensions . isomorphism postulate suppose @xmath11 is a state and @xmath130 is an isomorphism of @xmath21-structures . then : @xmath131 is a state , initial if @xmath11 is . @xmath127 preserves causality , that is , if @xmath84 then @xmath132 . @xmath127 preserves finality , success , and failure , that is , @xmath133 and @xmath134 . @xmath127 preserves updates , that is , @xmath135 for all histories @xmath35 for @xmath11 . in the last part of this postulate , and throughout this paper , we adopt the convention that an equation between possibly undefined expressions is to be understood as implying that if either side is defined then so is the other . we have required that isomorphisms preserve even the irrelevant parts of the algorithm , like update sets for unattainable or non - final histories . this requirement could be dropped , thereby weakening our postulate , without any damage to our results . nevertheless , it seems a natural requirement in a general description of algorithms . the intuition behind it is that , even if an algorithm includes irrelevant information , the states should still be abstract ; not even the irrelevant information should be able to `` see '' the particular nature of the elements of the state . for example , if @xmath136 is an isomorphism , and @xmath35 is an unattainable ( hence irrelevant ) history in @xmath111 , then our postulate requires @xmath137 ( which is equally unattainable and thus irrelevant in @xmath131 ) to be in @xmath138 . to allow the contrary would be to allow the algorithm to refer to the difference between @xmath11 and @xmath131 , a difference that , because of the isomorphism , involves only the particular elements constituting these states , not their abstract structure the final postulate formalizes the requirement that a small - step algorithm can do only a bounded amount of work in any one step . the bound depends only on the algorithm . work includes assembling queries ( from elements of the state and labels ) and issuing them , computing what queries to issue , deciding whether the current history suffices to finish the step , deciding whether the computation has failed , and computing updates . our formalization of this closely follows the corresponding postulate in ( * ? ? ? * section 5 ) ; the use of a set of terms to represent the bounded part of the state that the algorithm looks at goes back to @xcite . bounded work postulate there is a bound , depending only on the algorithm , for the lengths of the tuples in @xmath86 , for all states @xmath11 and final , attainable histories @xmath35 . there is a bound , depending only on the algorithm , for the cardinality @xmath139 , for all states @xmath11 and final , attainable histories @xmath35 . there is a finite set @xmath5 of @xmath21-terms ( possibly involving variables ) , depending only on the algorithm , with the following property . suppose @xmath11 and @xmath140 are two states and @xmath35 is a history for both of them . suppose further that each term in @xmath5 has the same value in @xmath11 as in @xmath140 when the variables are given the same values in @xmath141 . then : if @xmath84 then @xmath142 ( so in particular @xmath15 is a query for @xmath140 ) . if @xmath35 is in @xmath112 or @xmath113 , then it is also in @xmath143 or @xmath144 , respectively . @xmath145 . this completes our list of postulates , so we are ready to define the class of algorithms to be treated in this paper and @xcite . [ alg - def ] an _ interactive , small - step algorithm _ is any entity satisfying the states , interaction , step , isomorphism , and bounded work postulates . since these are the only algorithms under consideration in most of this paper , we often omit `` interactive , small - step '' ; on the other hand , when we want to emphasize the difference between these algorithms and the `` ordinary algorithms '' treated in @xcite , we may refer to the present class of algorithms as `` general , interactive , small - step algorithms '' . as we explained near the beginning of this section , definition [ alg - def ] provides a formal meaning for `` ( interactive , small - step ) algorithm '' , whereas the preceding discussion , leading up to the postulates , was based on an informal notion of ( interactive , small - step ) algorithm . the definition expresses our belief that the formal notion accurately represents the informal one , i.e. , that our postulates adequately describe the intuitive notion . the remainder of this section is devoted to some terminology and results connected with the bounded work postulate . a set @xmath5 with the property required in the third part of the bounded work postulate is called a _ bounded exploration witness _ for the algorithm . two pairs @xmath146 and @xmath147 , consisting of states @xmath11 and @xmath140 and a single @xmath35 that is a history for both , are said to _ agree _ on @xmath5 if , as in the postulate , each term in @xmath5 has the same value in @xmath11 as in @xmath140 when the variables are given the same values in @xmath141 . the first two parts of the bounded work postulate assert bounds for final , attainable histories . they imply the corresponding bounds for all attainable histories , thanks to the following lemma . let @xmath11 be a state . any coherent history for @xmath11 is an initial segment of a complete , coherent history for @xmath11 . any attainable history for @xmath11 is an initial segment of a final , attainable history for @xmath11 . since @xmath11 is fixed throughout the proof , we omit it from the notation . to prove the first assertion , let @xmath35 be a coherent history . according to the definition , its order - type ( meaning , strictly speaking , the order - type of the ordering induced by @xmath34 on the equivalence classes under @xmath46 ) is some ordinal number @xmath12 . we shall inductively define a sequence of coherent histories @xmath148 , starting with @xmath149 . here @xmath64 will range over either the set n of all natural numbers or the set of natural numbers up to some finite @xmath150 to be determined during the construction . after @xmath148 is defined , if it is complete , then stop the construction , i.e. , set @xmath151 . if @xmath148 is incomplete , this means that the set @xmath152 is nonempty . extend the answer function @xmath153 by adjoining @xmath154 to its domain and assigning it arbitrary values ( in @xmath11 ) there . call the resulting answer function @xmath155 , and make it into a history @xmath156 by pre - ordering its domain as follows . on @xmath157 , @xmath158 agrees with @xmath159 . all elements of @xmath160 are @xmath158 all elements of @xmath154 . elements of @xmath154 are @xmath158 only each other . in other words , we extend the ordering of @xmath157 by adding @xmath154 at the end , as a single equivalence class . it is straightforward to check that each @xmath148 ( if defined , i.e. , if the sequence hasnt ended before @xmath64 ) is a coherent history ; the order - type of its equivalence classes is @xmath161 . if @xmath148 is defined for only finitely many @xmath64 , then this is because the last @xmath148 was complete , and so we have the required result . it remains to consider the case where @xmath148 is defined for all natural numbers @xmath64 . in this case , let @xmath162 be the union of all the @xmath148 s . ( more formally , if we regard functions and orderings as sets of ordered pairs , then @xmath163 is the union of the @xmath153 s , and the pre - order @xmath164 is the union of the @xmath159 s . ) then @xmath162 is also coherent . indeed , if @xmath165 then there is some @xmath64 such that @xmath166 . as @xmath148 is coherent , @xmath167 . but , since each @xmath148 is an initial segment of the next , and therefore of @xmath162 , we have @xmath168 and so @xmath169 . furthermore , the order - type of @xmath162 is @xmath170 ( where @xmath171 is the order - type of the natural numbers ) , so it is well - ordered . to finish the proof of the first part of the lemma , we need only check that @xmath162 is complete . suppose , toward a contradiction , that @xmath172 . so there is an initial segment @xmath76 of @xmath162 such that @xmath173 . by the interaction postulate , @xmath76 is finite , so it is an initial segment of @xmath148 for some @xmath64 . then @xmath174 that contradicts the assumption that @xmath172 , so we have shown that @xmath162 is complete . thus , @xmath175 is an initial segment of the complete , coherent history @xmath162 , and the first assertion of the lemma is proved . to prove the second assertion , let @xmath35 be an attainable history . in particular , it is coherent , so , by the first assertion , it is an initial segment of a complete , coherent history @xmath162 . by part a of the step postulate , @xmath162 has an initial segment @xmath76 that is a final history . if @xmath162 has several initial segments that are final histories , then let @xmath76 be the shortest of them ; thus no proper initial segment of @xmath76 is final . since both @xmath35 and @xmath76 are initial segments of @xmath162 , one of them is an initial segment of the other . since @xmath35 is attainable and @xmath76 is final , @xmath76 can not be a proper initial segment of @xmath35 . therefore , @xmath35 is an initial segment of @xmath76 . furthermore , @xmath76 is coherent , because it is an initial segment of the coherent history @xmath162 . it follows , since no proper initial segment of @xmath76 is final , that @xmath76 is attainable , as desired . the following corollary extends the first assertion in the bounded work postulate to histories that need not be final . [ bd - query ] there is a bound , depending only on the algorithm , for the lengths of the tuples in @xmath86 for all states @xmath11 and all attainable histories @xmath35 . the bound on lengths of queries issued by final , attainable histories , given by the bounded work postulate , applies to all attainable histories , because these are , by the lemma , initial segments of final ones . the next corollary similarly extends the second assertion of the bounded work postulate , and adds some related information . [ bd - dom ] there is a bound , depending only on the algorithm , for @xmath139 , for all states @xmath11 and attainable histories @xmath35 . the same number also bounds @xmath176 for all attainable @xmath35 . by the lemma , any attainable history @xmath35 is an initial segment of a final , attainable history @xmath162 . then @xmath177 . the bound provided by the bounded work postulate for @xmath178 thus applies to @xmath35 as well . this proves the first assertion of the corollary , and the second follows because @xmath35 , being attainable , is coherent , which implies @xmath92 . one of our principal aims in this paper and its sequel @xcite is to show that every algorithm , in the sense defined above , is behaviorally equivalent , in a strong sense , to an asm . of course , this goal presupposes a precise definition of the notion of behavioral equivalence , and we devote the present section to presenting and justifying that definition . as in earlier work on the asm thesis , beginning in @xcite and continuing in @xcite , the definition of equivalence is intended to express the idea that two algorithms behave the same way in all possible situations . we must , of course , make precise what is meant by `` behave the same way '' and by `` possible situations . '' much of what needs to be said here was already said in ( * ? ? ? * section 6 ) in the more restricted context of ordinary interaction , and we shall refer to some of that discussion in motivating our definition of behavioral equivalence . part of the definition of equivalence of algorithms is straightforward . as in previous work , we require equivalent algorithms to have the same states , the same initial states , the same vocabulary , and the same labels . the requirement that they agree as to states and initial states is clearly necessary for any comparison at all between their behaviors , specifically the aspect of behavior given by the progression of states in a run . the requirement that they agree as to vocabulary is actually a consequence of agreement as to states ( and the requirement , in the definition of algorithm , that there be at least one state ) , because any structure determines its vocabulary . the requirement that they agree as to labels ensures that the algorithms have , in any state , the same potential queries ; this is needed for any comparison between their behaviors , since issuing queries is observable behavior . the requirements just discussed say that equivalent algorithms agree as to all the items introduced in the states postulate . it is tempting to go through the remaining postulates , look for statements of the form `` the algorithm determines '' such - and - such , and require equivalent algorithms to have the same such - and - such . unfortunately , this approach , which in @xcite would produce the correct notion of equivalence , is too restrictive when applied to interactive algorithms in @xcite and the present paper . the difficulties were already pointed out in ( * ? ? ? * section 6 ) in connection with the causality relation . the examples given there exhibit the following two sorts of problems . first , a causality relation could have instances @xmath84 whose @xmath35 could never actually occur in the execution of the algorithm , for example because @xmath179 contains queries that the algorithm would never issue . second , as in ( * ? ? ? * example 6.4 ) , the @xmath35 in an instance @xmath84 of causality could contain redundant elements , such as a query - reply pair that would have to be present in order for another query in @xmath179 to be issued . algorithms whose causality relations differ only in such irrelevant ways should count as equivalent . that is , we should care only about what queries the algorithm issues in response to histories that can actually occur when this algorithm runs . similar comments apply to the remaining ingredients of an algorithm , in which histories are used . the notions of final , successful , and failing histories and the update sets should not be required to agree completely when two algorithms are equivalent ; it suffices that they agree on those histories that can actually occur . [ cpt2 ] we emphasize again that we are dealing here with a situation where only the algorithm and its environment are involved . when several algorithms interact ( and we do not choose to consider each as a part of the environment for the others ) , the situation would be more complex . consider , for example , two algorithms that function as components in a larger computation . the first of these components might issue a query whose reply is used by the second . in that case , instances @xmath84 of the second component s causality relation can be relevant even if that component would never issue the queries in @xmath179 . some aspects of this situation will arise in @xcite when we discuss the `` do in parallel '' construct of asms ; a thorough discussion will be provided in @xcite . to formalize the preceding discussion , we must still say something about the notion of a history that can actually appear . we have already introduced a precise version of this notion , namely the notion of an attainable history . this notion , however , depends on the causality relation and the notion of finality given with the algorithm . when we define equivalence of two algorithms , which one s attainability should we use ? there are two natural choices : require the algorithms to agree ( as to queries issued , finality , success , failure , and updates ) on those histories that are attainable for both algorithms , or require agreement on all histories that are attainable under at least one of the algorithms . ( there are also some less natural choices , for example to require agreement on the histories attainable under the first of the two algorithms ; this threatens to make equivalence unsymmetric . ) fortunately , these options lead , as we shall prove below , to the same notion of equivalence of algorithms . having advertised our notion of equivalence as a strong one , we take the apparently stronger of the natural definitions as the official one and then prove its equivalence with the apparently weaker one . [ equiv - def ] two algorithms are _ behaviorally equivalent _ if they have the same states ( therefore the same vocabulary ) , the same initial states , and the same labels , in each state , they have the same attainable histories , for each state @xmath11 and each attainable history @xmath35 , they have the same set @xmath86 , for each state , the two algorithms have the same attainable histories in @xmath112 and @xmath113 ( and therefore also in @xmath111 ) . for each state and each attainable , final , successful history , they have the same update sets . the following lemma shows that the notion of equivalence is unchanged if we delete the second item in the definition and weaken the subsequent ones to apply only to histories that are attainable for both algorithms . [ equiv - la ] suppose two algorithms have the following properties . they have the same states ( therefore the same vocabulary ) , the same initial states , and the same labels , for each state @xmath11 and each history @xmath35 that is attainable for both algorithms , they have the same set @xmath86 , for each state , any history that is attainable in both algorithms and is in @xmath112 or @xmath113 for one of the algorithms is also in @xmath112 or @xmath113 , respectively , for the other . for each state and each history that is attainable for both and final and successful ( for one and therefore both ) , they have the same update sets . then these two algorithms are equivalent . comparing the hypotheses of the lemma with the definition of equivalence , we see that it suffices to prove that , under the hypotheses of the lemma , any history that is attainable in a state @xmath11 for either algorithm must also be attainable in @xmath11 for the other . indeed , this would directly establish the second clause of the definition , and it would make the subsequent clauses in the definition equivalent to the corresponding hypotheses in the lemma . let @xmath180 and @xmath181 be two algorithms satisfying the hypotheses of the lemma . fix a state @xmath11 for the rest of the proof ; we shall suppress explicit mention of @xmath11 in our notations . to prove that @xmath180 and @xmath181 have the same attainable histories , we first recall that any attainable history ( for either algorithm ) has finite domain by corollary [ bd - dom ] and therefore has finite length . we can therefore proceed by induction on the length of histories . since the empty history vacuously satisfies the definitions of `` coherent '' and `` attainable '' , we need only verify the induction step . consider , therefore , a nonempty history @xmath35 that is attainable for @xmath180 . let the last equivalence class in @xmath179 be @xmath182 and let @xmath76 be the initial segment of @xmath35 obtained by deleting @xmath182 from the domain . then @xmath76 is attainable for @xmath180 ( because @xmath35 is ) and therefore also for @xmath181 ( by induction hypothesis ) . furthermore , the linear order of the equivalence classes of @xmath35 , being finite , is certainly a well - ordering . so to complete the proof that @xmath35 is attainable for @xmath181 , it suffices to verify that it satisfies , with respect to @xmath181 , the first clause in the definition of coherence ( that every @xmath183 is caused by an initial segment that ends before @xmath15 ) and that no proper initial segment of it is final . for the first of these goals , observe that the induction hypothesis gives what we need if @xmath184 , so we need only deal with the case that @xmath185 . in this case , we know that @xmath186 because @xmath35 is coherent for @xmath180 . but since @xmath76 is attainable for both algorithms , the second hypothesis of the lemma applies , and we infer that @xmath187 , as required . for the second goal , we already know that @xmath76 is attainable for @xmath181 and so none of its proper initial segments can be final . the only proper initial segment of @xmath35 not covered by this observation is @xmath76 , so it remains only to show that @xmath76 is not final for @xmath181 . since @xmath76 is not final for @xmath180 ( because @xmath35 is attainable for @xmath180 ) and since @xmath76 is attainable for both algorithms , the third hypothesis of the lemma immediately gives the required conclusion . a referee asked why equivalence of algorithms is not defined in a more usual way , by defining a notion of _ run _ and then defining two algorithms being equivalent if they have the same runs . the alternative definition of equivalence is possible but it is not simpler . a run would not be merely a sequence of states as in @xcite ; it would have to include the queries and replies as well . note also that our definition of equivalence requires that the two algorithms have the same behavior at every state @xmath11 whether @xmath11 is reachable or not . it is thus independent of the set of initial states . to match this , the run - based definition of equivalence should allow runs to start at arbitrary states . we have formally defined the class of interactive small - step algorithms ( definition [ alg - def ] ) by means of postulates that generalize the postulates of @xcite and @xcite , and we have explained why we believe that this definition matches the intuitive notion of interactive small - step algorithm . we have also defined a strong notion of behavioral equivalence for such algorithms . this work provides the essential prerequisites for an analysis of these algorithms in terms of abstract state machines and for the proof of thhe asm thesis in thee small - step interactive case . that analysis and proof will be carried out in @xcite , beginning with the definition of the appropriate class of asms . nick benton , luca cardelli and cdric fournet , `` modern concurrency abstractions for c # , '' in _ proceedings of the 16th european conference on object - oriented programming ( ecoop 2002 ) . june 10 - 14 , 2002 university of mlaga , spain _ , b. magnusson , ed . , springer - verlag , lecture notes in computer science 2374 . andreas blass , yuri gurevich , dean rosenzweig , and benjamin rossman , `` interactive small - step algorithms ii : abstract state machines and the characterization theorem '' ( tentative title ) , in preparation . yuri gurevich , `` evolving algebras : an introductory tutorial '' , bull . eatcs 43 ( feb . 1991 ) , 264284 . reprinted with slight revisions in _ current trends in theoretical computer science : essays and tutorials _ , eds . g. rozenberg and a. salomaa world scientific , 1993 , 223394 . alan m. turing , `` on computable numbers , with an application to the entscheidungsproblem '' , _ proceedings of the london mathematical society _ , series 2 , vol . 42 ( 19361937 ) , 230265 ; correction , _ ibidem _ , vol . 43 , 544546 . reprinted in @xcite and available online at @xmath188
one of the key technologies for quantum information and computation is purification / distillation of quantum states @xcite . particular pure states , such as entangled states , often play significant roles there , but it is not easy to find such `` clean '' states in nature . it is therefore required to prepare them out of mixed states ; otherwise , we can not carry out any interesting ideas of quantum information and computation . a new purification mechanism has recently been proposed @xcite . it is shown that repeated measurements on a system , say a , result in a purification of another system , say b , in interaction with a @xcite . that is , the state of b is driven to a pure state irrespectively of its ( generally mixed ) initial state , if certain conditions are satisfied . remarkably , if appropriate adjustment of the relevant parameters is possible , the maximal yield , which is prescribed by the initial mixed state of b and its target pure state , can be attained , while keeping the maximal fidelity , by a finite number of measurements on a ( an optimal purification " ) . this constitutes a remarkable contrast to the standard purification protocol @xcite , in which it is generally difficult to realize both a non - vanishing yield and the maximal fidelity at the same time . since an entangled state is one of the pure states of two quantum systems , say a and b , one can think of a possibility of extracting the entangled state between a and b by repeatedly performing measurements on x which interacts with both a and b. this possibility has already been pointed out @xcite and explored to show that one of the bell states can be extracted when this mechanism is applied to a three - qubit system , where qubits a and b always interact with the other qubit x on which one and the same measurement is repeatedly performed . notice that in this case the two systems a and b are not spatially separated , because they are supposed to locally interact with x. on the other hand , it is often required , e.g. , in the ideas of quantum teleportation and communication @xcite , to establish an entanglement between two quantum systems that are located at or at least can be sent , without loosing the entanglement , to distant places . in this respect , it would be worth remembering that interesting ideas of generating an entanglement between two cavities @xcite and of transferring an entanglement between two modes in a cavity to that between other two modes in different cavities @xcite have been proposed . in the former a two - level atom is sent to interact successively with the two cavities resulting in the generation of an entanglement between the two , and in the latter the entanglement is shown to be transferred by a two - level atom which passes through the two cavities and interacts with the relevant cavity modes . in this paper , this kind of successive interaction with two quantum systems is incorporated within the framework of the new purification mechanism @xcite to show that an entanglement can be established between the states of the two systems spatially separated ( or that can be separated ) . notice that the entangled state is distilled from an arbitrary initial state that is in general mixed , while in the generation of entanglement in ref . @xcite the initial state should be prepared in an appropriate pure state and in the transfer of entanglement in ref . @xcite the state is assumed to be initially entangled . after a brief review of the new purification mechanism in sec . [ sec : rev ] , a scheme of successive interaction is introduced in a three - qubit system , a+b+x , in which system x is assumed to interact first with system a and then with b , in sec . [ sec : qqx ] . system x is prepared in an initial pure state and is measured after it has interacted with a and b. then only those events in which system x is found in the initial state are kept . this process will be repeated many times . it is shown that an optimal entanglement purification is actually realizable for a particular choice of interaction and by properly adjusting interaction times and strengths between a and x and b and x. in sec . [ sec : ccx ] , another example of an entanglement purification is examined in a physical system where a two - level atom x is injected successively to the two cavities a and b `` back and forth , '' interacts with their cavity modes under the rotating - wave approximation and the state of x is repeatedly measured in a prescribed way . it is explicitly shown that , under certain conditions , a particular entangled state between the lowest two modes of each cavity is extracted , irrespectively of the initial cavity states . finally we summarize the results obtained and give future perspectives in sec . [ sec : sum ] . let the total system consist of two parts , system a and system b , and the dynamics be described by the total hamiltonian @xmath0 where @xmath1 stands for the interaction between the two ( sub)systems . we initially prepare the system in a product state @xmath2 at @xmath3 . notice that system b can be in an _ arbitrary mixed _ state @xmath4 . we perform measurements on a at regular intervals @xmath5 to confirm that it is still in the state @xmath6 @xcite , while the total system a+b during the time @xmath5 evolves unitarily in terms of the total hamiltonian @xmath7 . since the measurement is performed only on system a , the action of such a ( projective , for simplicity ) measurement can be conveniently described by the following projection operator @xmath8 thus the state of system a is set back to @xmath6 every after @xmath5 , while that of b just evolves dynamically on the basis of the total hamiltonian @xmath7 . we repeat the same measurement , represented by ( [ eq : calo ] ) , @xmath9 times and collect only those events in which system a has been found in the state @xmath6 consecutively @xmath9 times ; other events are discarded . the state of system b is then described by the density matrix @xmath10 where @xmath11 is an operator acting on b and @xmath12\nonumber\\ & = { \mathop{\mathrm{tr}}\nolimits}_\text{b}\bigl[\bigl(v_\phi(\tau)\bigr)^n\rho_\text{b}(0 ) \bigl(v_\phi^\dagger(\tau)\bigr)^n\bigr ] \label{eq : ptaun}\end{aligned}\ ] ] is the probability for these events to occur ( yield ) . this normalization factor appearing in ( [ eq : rhobn ] ) reflects the fact that only the right outcomes are collected in this process . in order to examine the asymptotic state of system b for large @xmath9 , consider , assuming its existence , the spectral decomposition of the operator @xmath13 , which is not hermitian , @xmath14 . we therefore need to set up both the right- and left - eigenvalue problems @xmath15 the eigenvalue @xmath16 is complex valued in general , but its absolute value is bounded @xcite @xmath17 this reflects the unitarity of the time evolution operator @xmath18 . these eigenvectors are assumed to form a complete orthonormal set in the following sense @xmath19 ( we normalize @xmath20 as @xmath21 , while the norm of @xmath22 has been fixed by the above relations and is not necessarily unity . ) the operator @xmath13 itself is now expanded in terms of these eigenvectors @xmath23 it is now easy to see that the @xmath9th power of this operator is expressed as @xmath24 and therefore it is dominated by a single term for large @xmath9 @xmath25 provided the largest ( in magnitude ) eigenvalue @xmath26 is _ discrete , nondegenerate and unique_. if these conditions are satisfied , the density operator of system b is driven to a pure state @xmath27 with the probability @xmath28 the pure state @xmath29 , which is nothing but the right - eigenvector of the operator @xmath13 belonging to the largest ( in magnitude ) eigenvalue @xmath26 , is thus distilled in system b. this is the purification scheme proposed in @xcite . a few comments are in order . first , the final pure state @xmath29 toward which system b is to be driven is dependent on the choice of the state @xmath6 on which system a is projected every after measurement , on the measurement interval @xmath5 and the hamiltonian @xmath7 , but does not depend on the initial state of system b at all . in this sense , the purification is accomplished irrespectively of the initial ( mixed ) state @xmath4 . second , as is clear in the above exposition , what is crucial in this purification scheme is the repetition of one and the same measurement ( more appropriately , spectral decomposition ) and the measurement interval @xmath5 need not be very small @xcite . it instead remains an adjustable parameter . third , if we can make other eigenvalues than @xmath26 much smaller in magnitude @xmath30 by adjusting parameters , we will need fewer steps ( i.e. , smaller @xmath9 ) to purify system b. it is now evident that the purification can be made _ optimal _ , if the conditions ( [ eq : quick ] ) and @xmath31 are satisfied . this condition ( [ eq : noloss ] ) assures that we can repeat as many measurements as we wish without running the risk of losing the yield @xmath32 in order to make the fidelity to the target state @xmath29 , @xmath33 , \label{eq : f}\ ] ] higher . actually , the yield @xmath32 decays like @xmath34 and the condition ( [ eq : noloss ] ) can bring us with the non - vanishing yield @xmath35 even in the @xmath36 limit . therefore the condition ( [ eq : noloss ] ) makes the two ( sometimes not compatible ) demands , i.e. , _ higher fidelity and non - vanishing yield _ , achievable , with fewer steps when the condition ( [ eq : quick ] ) is met . in this sense , the purification is considered to be optimal . it would be desirable if an optimal purification can be realized by an appropriate choice of the state @xmath6 and/or tuning of the measurement interval @xmath5 and parameters in a given hamiltonian . a few simple systems have already been examined @xcite to show how such optimal purifications are made possible . since an entangled state of system a+b is one of the pure states , there is a possibility that we apply the purification mechanism described above to distill the initial ( generally mixed ) state of a+b to a desired entangled state . this possibility has already been pointed out and it has been explicitly demonstrated that one of the bell states @xmath37 of the two qubit systems a+b can actually be extracted if we repeatedly measure one and the same state of another qubit system x , the interactions of which are symmetrical with respect to a and b , resulting with the maximal yield @xcite . one of the limitations of this model is that the entanglement can be established only when the two systems a and b locally interact with the same system x at the same time and therefore it does not seem to allow to establish an entanglement between two systems that are spatially separated . a new scheme is certainly needed . when the two systems a and b are spatially separated , their local interactions with the other system x can not take place simultaneously . this means that the interactions are considered to become effective one by one , i.e. , system x first interacts with , say system a and then with b @xcite . this kind of process can be conveniently described by a time - dependent total hamiltonian . in this section , a total system composed of three qubits ( or three spin @xmath38 systems ) , a , b , and x , is considered . the two qubits a and b interact with the other qubit x successively , the state of x , initially prepared in a particular state , say in up state , is measured after the interactions with a and b , and only those events in which the state of x is found in up state will be retained and other events are just discarded . the process will be repeated many times and we are interested in the resulting state of a+b . assume that the three - qubit system a+b+x is described by a time - dependent total hamiltonian @xmath39 . qubit x , which is initially prepared in up state @xmath40 , is first brought to interaction with qubit a for time interval @xmath41 . the hamiltonian in this period is @xmath42 next , after a free time evolution under the free hamiltonian @xmath43 for time duration @xmath44 , qubit x interacts with another qubit b , which has no direct interaction with a , for @xmath45 . the hamiltonian for this period reads @xmath46 after another free time evolution for @xmath47 , the state of x is measured and only those cases in which qubit x is found in its initial up state @xmath40 shall be retained . the whole process , i.e. , @xmath48 shall be repeated many times and we are interested in the final state of a+b . see fig . 1 . , is brought to interaction with qubits a and b successively and then its state is measured . if it is found in @xmath40 , the whole process is repeated again ; other events are discarded.,scaledwidth=48.0% ] in order to describe the above process explicitly and to show the possibility of entanglement distillation by this process , we consider the following free and interaction hamiltonians @xmath49 where @xmath50 s are pauli matrices acting on the hilbert space of system a and so on , and @xmath51 and @xmath52 are real ( assumed , for definiteness , to be positive ) coupling constants . the free hamiltonians for a , b , and x are assumed to be the same for simplicity and are characterized solely by the common energy gap @xmath53 . it is clear that the relevant evolution operator for the whole process ( [ eq : wp ] ) is given by @xmath54 it is an elementary task to evaluate this operator , since each factor on the right - hand side is easily evaluated in terms of the eigenstates of the hamiltonian in the exponent . indeed , we have @xmath55 , \label{eq : rf}\end{gathered}\ ] ] and similarly @xmath56\\ { } + \langle{\downarrow}_\text{x}\| \{-i\sin\varphi_\text{b}\sin2\theta_\text{b } \|{\uparrow}_\text{b}\rangle\langle{\downarrow}_\text{b}\|-i\sin ( g_\text{b}t_\text{b } ) \|{\downarrow}_\text{b}\rangle\langle{\uparrow}_\text{b}\|\ } \bigr ] , \label{eq : lf}\end{gathered}\ ] ] where the angles @xmath57 and @xmath58 are defined as [ eq : angles ] @xmath59 let us introduce a parity operator @xmath60 whose eigenvalues @xmath61 and @xmath62 single out two subspaces of the product hilbert space @xmath63 invariant under the action of the operator @xmath64 . the two states @xmath65 and @xmath66 generate the even parity subspace and the following @xmath67 matrix @xmath68 with its elements @xmath69\\ \mathcal{m}_{22 } = { } & \cos ( g_\text{a}t_\text{a})\cos ( g_\text{b}t_\text{b } ) \label{eq : m22}\end{aligned}\ ] ] allows to completely characterize the action of @xmath64 in this subspace as follows @xmath70 = e^{-i\omega(t_\text{a}+\tau_\text{a}+t_\text{b}+\tau_\text{b})}{\mathcal{m } } \left[\begin{matrix } \|{\uparrow}_\text{a}{\uparrow}_\text{b}\rangle\\ \noalign{\smallskip } \|{\downarrow}_\text{a}{\downarrow}_\text{b}\rangle \end{matrix}\right ] . \label{eq : p+}\ ] ] we proceed in the same way for the odd parity subspace spanned by the states @xmath71 and @xmath72 . to this end we define the @xmath67 matrix @xmath73 with its elements @xmath74 so that the action of @xmath64 is represented as @xmath75 = e^{-i\omega(t_\text{a}+t_\text{b}+2\tau_\text{b})}\mathcal{n } \left[\begin{matrix } \|{\uparrow}_\text{a}{\downarrow}_\text{b}\rangle\\ \noalign{\smallskip } \|{\downarrow}_\text{a}{\uparrow}_\text{b}\rangle \end{matrix}\right ] . \label{eq : p-}\ ] ] in order to show explicitly that the process ( [ eq : wp ] ) with the particular choice of interaction ( [ eq : hab ] ) admits an entanglement distillation for qubit - system a+b , it turns out to be enough to consider a much simpler case . let us treat systems a and b symmetrically , except for the ordering of their interactions with system x. we choose the same parameters for a and b , i.e. , @xmath76 , @xmath77 and @xmath78 ( @xmath79 and @xmath80 ) . for the parity - odd states , the matrix @xmath73 now is simplified to be @xmath81 it is easy to find the condition under which an entangled state of the form @xmath82 is a real parameter , is an eigenstate of this matrix @xmath73 ( and therefore , of the operator @xmath64 ) . a straightforward calculation shows that if the parameters @xmath83 , @xmath84 and @xmath5 are so chosen that the following relation @xmath85 is satisfied , the state @xmath86 with @xmath87 is indeed an eigenstate of @xmath64 belonging to the eigenvalue @xmath88 . [ there is another possibility of optimal distillation of the above entangled state ( [ eq : psi ] ) , but with a different @xmath89 , i.e. , @xmath90 . this case is realized under the condition ( [ eq : opt1 ] ) with the replacement @xmath91 ; the corresponding eigenvalue is also given by the shifted one , i.e. , @xmath92 . ] notice that we are not allowed to set @xmath93 because it would result in a degenerate ( in magnitude ) eigenvalue . observe that we have essentially two conditions ( [ eq : opt1 ] ) , while we have three independent combination of parameters @xmath94 , @xmath95 and @xmath96 . we , therefore , have a possibility of an optimal distillation of the entangled state @xmath86 , if the magnitudes of the other eigenvalues of @xmath64 are made smaller than unity . the remaining eigenvalue of @xmath73 under the conditions in ( [ eq : opt1 ] ) reads @xmath97\\ = e^{-i\omega(t+\tau ) } [ -\cos^2(gt)+\sin^2(gt ) ] , \label{eq : ev'}\end{gathered}\ ] ] and its absolute value can not be made unity when @xmath98 . on the other hand , matrix @xmath68 is expressed as @xmath99 and the absolute values of the eigenvalues of this matrix can not reach unity if @xmath100 and @xmath98 . this means that , under these conditions on @xmath94 , @xmath95 and @xmath96 satisfying the relation ( [ eq : opt1 ] ) , an optimal purification ( i.e. , distillation ) of the entangled state @xmath101 is possible . it is in fact easily shown that the left - eigenstate of @xmath64 belonging to the eigenvalue @xmath88 is expressed as @xmath102 and therefore the yield @xmath32 approaches asymptotically , as @xmath9 becomes large , a finite value @xmath103\\ { } \times\rho_\text{ab}(0 ) \bigl[\|{\uparrow}_\text{a}{\downarrow}_\text{b}\rangle+e^{i\chi}\|{\downarrow}_\text{a}{\uparrow}_\text{b}\rangle\bigr ] . \label{eq : yield}\end{gathered}\ ] ] this is nothing but the probability of finding the target entangled state @xmath104 in the initial state @xmath105 . in the previous section , the possibility of realizing an entanglement distillation is demonstrated for the three - qubit system a+b+x . the particular form of interaction ( [ eq : hab ] ) is shown to be suitable for this purpose following the procedure ( [ eq : wp ] ) . in this section , another application of the purification mechanism @xcite is explored in a system composed of a two - level system ( e.g. , an atom ) interacting with two single - mode cavities . the two cavities may be located at spatially distant places ( or may be near and separated later ) and we aim at extracting an entanglement between the two - cavity states by repeatedly bringing the two - level atom into interaction with them and then selecting a particular state of the atom by measurements . the ideas of generating @xcite and of transferring @xcite entanglement in two - cavity system have already been proposed and studied in the context of the cavity quantum electrodynamics ( cqed ) @xcite , spectacularly developed over the last two decades both in the microwave @xcite and optical @xcite domains . entanglement is generated from a properly prepared pure state in the former @xcite and an initially prepared entanglement in one cavity is transformed into another entanglement between the two cavities in the latter @xcite , via successive interactions with a two - level atom . the atom plays the role of a `` mediator '' or `` transformer '' of entanglement . a similar , but more complicated role is sought for the atom in the present scheme , because its interactions with the cavities and the measurements of its state are expected to enable us to extract an entangled state , that is to produce an entanglement distillation , _ irrespectively of the initial states of the two cavities_. for simplicity , suppose that the two cavities , a and b , are identical and their interaction with a two - level atom x is well described by the jaynes cummings hamiltonian @xcite . let @xmath106 and @xmath107 indicate the annihilation operators of the modes of the two cavities a and b , respectively . the free and interaction hamiltonians are @xmath108 a state where the atom x is in up(down ) state and the modes @xmath106 and @xmath107 are in the @xmath109th and @xmath110th levels , respectively , is denoted as @xmath111 ( @xmath112 ) . since our purpose is to extract a pure state not in product form but entangled , it turns out that simple processes like ( [ eq : wp ] ) for the three - qubit system would not work . indeed , because of the choice of the jaynes cummings ( rotating - wave ) interactions ( [ eq : hab ] ) , the number operator @xmath113 commutes with any of @xmath43 , @xmath114 , and @xmath115 , and therefore any state of the two - cavity system of the form @xmath116 , which is a product state , is easily seen to be an eigenstate of the time evolution operator constructed analogously to ( [ eq : v ] ) . ( if the down state of x is measured , product states @xmath117 are found to be eigenstates of the relevant time evolution operator . ) thus a process different from ( [ eq : wp ] ) would be necessary for our purpose . the above consideration would suggest that , with interaction given by ( [ eq : hab ] ) , it would be better to select after measurement a state of x different from the initial state . however , at the same time , we need a procedure that can be repeated many times within the present framework of the purification mechanism . thus we choose a procedure that can be described schematically as @xmath118\\ & \quad\to\text{free evolution for $ \tau_\text{a}$}\nonumber\\ & \quad\to\text{interaction between x and b for $ t_\text{b}$}\nonumber\\ & \quad\to\text{free evolution for $ \tau_\text{b}$}\nonumber\\ & \quad\to\text{projection to\ } \|{\uparrow}\rangle\nonumber\\ & \quad\to\text{free evolution for $ \tau_\text{b}$}\nonumber\\ & \quad\to\text{interaction between x and b for $ t_\text{b}$}\nonumber\\ & \quad\to\text{free evolution for $ \tau_\text{a}$}\nonumber\\ & \quad\to\text{interaction between x and a for $ t_\text{a}$}\nonumber\\ & \quad\to\text{projection to\ } \|{\downarrow}\rangle . \label{eq : wp2}\end{aligned}\ ] ] see also fig . [ fig:2 ] . , is brought to interaction with cavity modes @xmath106 and @xmath107 in the two cavities a and b successively , and its state is measured after the interactions . atoms that are found in state @xmath119 will be sent back to the cavities in the reversed order . the state of atom x is again measured , and if it is found in @xmath120 , the whole process is repeated ; other events are discarded.,scaledwidth=48.0% ] this is clearly a generalization of the purification process , `` projection '' @xmath121 `` time evolution '' @xmath121 `` projection . '' indeed , in the above scheme , time evolution " is not meant in the usual sense , that is described by a total hamiltonian . it is instead interrupted by another projection . however , the condition under which the purification mechanism does work is essentially the same as in the ordinary cases and all what we have to do here is to investigate the relevant evolution operator corresponding to the above process ( [ eq : wp2 ] ) . it would be important to notice that the above choice of the initial and projected states for system x is not arbitrary . in fact if it were prepared in the up state , the procedure analogous to that described before does not work . the vacuum state of the two cavities , which is a product state , would indeed turn out to be an eigenstate . the relevant evolution for the above process ( [ eq : wp2 ] ) is represented by products of the time - evolution and projection operators and each of them is easily evaluated . the only non - trivial operators are @xmath122 and @xmath123\\ { } -i\sin\varphi_\text{b}^{(m+1)}\|{\downarrow},m+1\rangle\langle{\uparrow},m\| + e^{i\omega t_\text{b}}\cos\varphi_\text{b}^{(m ) } \|{\downarrow},m\rangle\langle{\downarrow},m\| \bigr ) , \label{eq : utb}\end{gathered}\ ] ] where angles @xmath124 and @xmath125 are defined as @xmath126 it is an elementary task to evaluate in this case the relevant evolution operator @xmath127 , analogously to , @xmath128 and its explicit expression reads @xmath129 , \label{eq : vcexp}\end{aligned}\ ] ] where @xmath130 . it is manifest from ( [ eq : vcexp ] ) that in the product hilbert space of the two cavities there are sectors of @xmath127 within which the action of @xmath127 is closed . these invariant sectors are characterized by the number @xmath131 . we have @xmath132 states @xmath133 for the ( @xmath131)th sector . notice that the singlet state ( vacuum state ) @xmath134 ( @xmath135 ) belongs to zero - eigenvalue of @xmath127 and we need not consider it , for @xmath136 . this is closely related to the choice of the initial state of x and is the reason why we must prepare x in the down state @xmath120 . let us turn our attention first to the doublet subspace ( @xmath137 ) . the action of @xmath127 on this subspace is easily read from ( [ eq : vcexp ] ) as @xmath138 observe that the determinant of this matrix always vanishes , which means that one of the eigenvalues is zero and the other is given by the trace of the matrix @xmath139 . therefore we have a possibility of obtaining the largest ( in magnitude ) eigenvalue by adjusting the parameter @xmath140 so that @xmath141 ( the possibility @xmath142 would result , not in an entanglement distillation , but in a product - state purification . ) in such a case , the above eigenvalue equation is simplified to @xmath143\\ & { } \times\left ( \begin{matrix } \cos\varphi_\text{b}^{(1)}&\pm\sin\varphi_\text{b}^{(1)}\\ \noalign{\smallskip } \mp\sin\varphi_\text{b}^{(1)}&\cos\varphi_\text{b}^{(1 ) } \end{matrix } \right ) \left ( \begin{matrix } \|1,0\rangle\\ \noalign{\medskip } \|0,1\rangle \end{matrix } \right ) , \label{eq : dblt2}\end{aligned}\ ] ] from which it is clear that the entangled state @xmath144 can be extracted with the maximal probability by this setup . notice that it still remains the freedom to adjust the value of @xmath145 . the remaining task is to check whether there are other eigenstates of @xmath127 belonging to eigenvalues with unit magnitude , under the condition ( [ eq : ccond ] ) . consider the invariant sector characterized by @xmath146 that is composed of @xmath147 states @xmath148 . the action of @xmath127 on this sector is represented by the following matrix [ see ( [ eq : vcexp ] ) ] @xmath149 where matrix elements @xmath150 can be read from ( [ eq : vcexp ] ) as @xmath151 it is important to notice that the condition ( [ eq : ccond ] ) implies that the element @xmath152 vanishes , irrespectively of @xmath153 and thus the sector further splits into two subsectors @xmath154 and @xmath155 . furthermore , it is easily seen that the entangled state in the latter subspace of the form @xmath156 has the eigenvalue @xmath157 , while , as shown in the appendix , no eigenstate in the former subspace @xmath154 belongs to a unit ( in magnitude ) eigenvalue ( if @xmath158 is smaller than 9 ) . we have seen that there are , for any @xmath158-sector , many entangled states @xmath156 ( @xmath159 ) ( that increase , in general , in higher @xmath158-sectors ) extracted with the optimal probabilities by the process ( [ eq : wp2 ] ) . repeated interactions of the two - level atom x in the cavities a and b and the prescribed measurements ( projections ) certainly bring us with a statistical ( classical ) mixture of these entangled states . the situation would not be considered completely satisfactory , since we would not be able to distill a single entangled state by the process ( [ eq : wp2 ] ) . there is , however , a way out of this difficulty . we may prepare such an initial ( mixed ) state of a+b that contains only those sectors with relatively small @xmath158s . such a preparation of the initial state would effectively eliminate the possibility of obtaining other states than @xmath160 after performing the process ( [ eq : wp2 ] ) . for example , we may consider the following preparation procedure , which is nothing but a purification process applied to cavity b. we send a two - level atom prepared in down state @xmath120 to cavity b. after its interaction , which is again assumed to be of the form ( [ eq : hab ] ) , with b , the atom is measured and only those events in which it is found in the state @xmath120 are retained . this process is to be repeated many times and the resulting state of a+b , which will be used as the initial state for the following entanglement distillation process ( [ eq : wp2 ] ) , would be dominated by the state @xmath161 , since the vacuum state of system b is the unique eigenstate of the relevant evolution operator @xmath162 belonging to eigenvalue unity if no fine tunings are made on the parameters . after having prepared the state @xmath105 , we repeat the process ( [ eq : wp2 ] ) under the condition ( [ eq : ccond ] ) . we would finally end up with the single entangled state @xmath163 , because our initial state @xmath105 satisfies @xmath164 for @xmath165 , where @xmath166 is the left - eigenstate corresponding to @xmath160 . concluding this section we wish to give typical values of relevant parameters under the aspect of the possibility of implementing our proposal in laboratory . we concentrate on the estimation of the total duration @xmath167 of the experiment . to this end , we choose to be in the context of microwave cqed where both the geometrical arrangement of the experimental set up and the intensity of the atom - field coupling regime , seem more favorable to our proposal . let us first note that the typical atom - field coupling constant @xmath83 ( @xmath51 or @xmath52 ) can be chosen in such a way that @xmath168@xmath169 @xcite . moreover the lifetime of a rydberg atom is @xmath170 @xcite . as for the quality factor @xmath171 of the cavities currently used in laboratory , we quote typical values of the order of @xmath172@xmath173 @xcite corresponding to a cavity damping time @xmath174@xmath175 . considering that in our case @xmath176 , the total duration @xmath167 of the experiment may be estimated and turns out to be compatible with the entanglement distillation proposed in this section . in this paper , the idea of extracting entangled states among systems located at spatially separated places , irrespectively of their initial states , has been proposed and applied to simple systems to show the potentiality of a novel measurement - based purification scheme @xcite . the establishment of entanglement distillation relies on the successive interactions between the systems under consideration and the so - called mediator " quantum system . in the first example , it is demonstrated that the entanglement between the two qubit states is possible via their interactions with another qubit , which plays the role of the entanglement mediator , with the maximal yield ( optimal distillation ) . in the case of distillation of cavity - mode entanglement , however , a modification of the original simple scheme , that is , `` interaction @xmath121 measurement @xmath121 interaction @xmath177 , '' is required and the modified procedure ( [ eq : wp2 ] ) turns out to result in the entangled state @xmath163 , after an appropriate preparation of the initial state . we stress that there would also be a possibility of obtaining an entangled state not only in the lowest sector @xmath178 but also in the higher @xmath158 sector , provided an appropriate initial state be prepared . it would be worth stressing that in spite of such a modification required in the second example , the underlying notion is still the same : the action of a measurement ( represented by a projection operator , for simplicity ) causes an essential and critical dynamical change , not only in the system measured , but also in the others interacting with the former . since the notion is so general , one can devise various applications of this scheme in many different situations . the examples explored in this paper are just two of them and further applications will be reported elsewhere . finally we add some comments on the practical setup of our proposal . in both schemes reported in the previous sections , the entanglement mediator is an atom appropriately prepared before entering in interaction with the two subsystems to be entangled and subjected to a conditional measurement of its internal state at the end of the two successive coupling processes . we wish to stress that to realize in practice the required `` many crosses '' scheme one may synchronize the injection of the @xmath179th atom of the sequence into the process with the successful outcome of the internal state measurement of the @xmath180th one . this kind of experimental setup might be preferable to the conceptually simpler one based on the idea of using always the same atom reversing its direction of motion at exit to re - inject it into the process . thus the representations in figs . 1 and 2 , as far as the aspect under scrutiny is concerned , have been reported only for the sake of simplicity . the authors ( h.n . , m.u . , and k.y . ) acknowledge useful and helpful discussions with prof . is grateful for the warm hospitality at universit di palermo . this work is partly supported by a grant for the 21st century coe program ( physics of self - organization systems ) at waseda university and a grant - in - aid for priority areas research ( b ) ( no . 13135221 ) , both from the ministry of education , culture , sports , science and technology , japan , by a grant - in - aid for scientific research ( c ) ( no . 14540280 ) from the japan society for the promotion of science , by a waseda university grant for special research projects ( no . 2002a-567 ) and by the bilateral italian - japanese project 15c1 on `` quantum information and computation '' of the italian ministry for foreign affairs . is investigated with particular attention on its eigenstates belonging to the eigenvalues with unit magnitudes . since this matrix is real and symmetric , its eigenvalues are all real , and the eigenvalues of relevance in the framework of our procedure here are @xmath184 . for the first possibility @xmath61 , let us consider the determinant @xmath185 ( @xmath186 ) defined by @xmath187 it is easy to see that the particular form of @xmath185 and the definitions of @xmath188 and @xmath189 in ( [ a : cjdj ] ) lead to a recursion relation @xmath190 for @xmath191 . this is further reduced to @xmath192 if we set @xmath193 , then @xmath194 is found to be positive semi - definite and to satisfy @xmath195 this relation is easily solved to yield the explicit form of @xmath196 with since each term in ( [ a : pk ] ) has the same sign to each other , vanishing of @xmath198 , which is nothing but the condition for the matrix ( [ a : matrix ] ) to possess eigenstates that belong to eigenvalue unity , is equivalent to that of each term . this means that there are @xmath158 conditions for three parameters @xmath199 , @xmath200 and @xmath158 and it seems impossible to have a vanishing @xmath198 in general , unless most of the conditions are simultaneously satisfied . if we choose a particular value for @xmath199 , say @xmath201 as in ( [ eq : ccond ] ) , however , all @xmath198 with @xmath202 vanish because each term in ( [ a : pk ] ) contains either @xmath203 or @xmath204 , each of which vanishes . on the other hand , the matrix ( [ a : matrix ] ) is shown to have no eigenstate belonging to the eigenvalue @xmath62 . consider the following determinant @xmath205 where @xmath188 and @xmath189 are again given in ( [ a : cjdj ] ) . since @xmath206 and therefore @xmath207 and @xmath208 , let us assume that all @xmath209 s are positive definite for @xmath210 up to @xmath211 . then it follows that @xmath212\\ > { } & c_ij_{i-1}-d_i^2j_{i-2}\nonumber\displaybreak[0]\\ = { } & \sin^2\varphi_\text{a}^{(i)}\cos^2\varphi_\text{b}^{(k - i+1)}\nonumber\\ & { } \times \bigl ( j_{i-1}-\cos^2\varphi_\text{a}^{(i-1)}\sin^2\varphi_\text{b}^{(k - i+1)}j_{i-2 } \bigr)\nonumber\\ & { } + \cos^2\varphi_\text{a}^{(i)}\sin^2\varphi_\text{b}^{(k - i)}j_{i-1}. \label{a : irr}\end{aligned}\ ] ] this relation recursively yields inequalities @xmath213 the last factor on the right most hand is shown to be positive definite @xmath214 which means that the quantity on the left - hand side in ( [ a : irr1 ] ) is positive definite . we conclude that @xmath215 is positive definite and therefore @xmath216 does not vanish , which completes the proof . 99 c. h. bennett , g. brassard , s. popescu , b. schumacher , j. a. smolin , and w. k. wootters , phys . lett . * 76 * , 722 ( 1996 ) ; * 78 * , 2031(e ) ( 1997 ) ; c. h. bennett , d. p. divincenzo , j. a. smolin , and w. k. wootters , phys . rev . a * 54 * , 3824 ( 1996 ) ; j. i. cirac , a. k. ekert , and c. macchiavello , phys . * 82 * , 4344 ( 1999 ) . for reviews , see _ the physics of quantum information _ , edited by d. bouwmeester , a. ekert , and a. zeilinger ( springer - verlag , heidelberg , 2000 ) ; a. galindo and m. a. martn - delgado , rev . * 74 * , 347 ( 2002 ) , and references therein . h. nakazato , t. takazawa , and k. yuasa , phys . lett.90 , 060401 ( 2003 ) ; k. yuasa , h. nakazato , and t. takazawa , j. phys . jpn . * 72 * suppl . c , 34 ( 2003 ) . in this respect , the present scheme seems to be closely related to the so - called quantum zeno effect , though the mechanism is quite different , as has been pointed out in ref . @xcite . as for the quantum zeno effect , see , for example , b. misra and e. c. g. sudarshan , j. math . phys . * 18 * , 756 ( 1977 ) ; k. kraus , found . phys . * 11 * , 547 ( 1981 ) ; h. nakazato , m. namiki , and s. pascazio , int . j. mod b * 10 * , 247 ( 1996 ) ; d. home and m. a. b. whitaker , ann . * 258 * , 237 ( 1997 ) ; p. facchi and s. pascazio , in _ progress in optics _ , edited by e. wolf ( elsevier , amsterdam , 2001 ) , vol . 42 , p. 147 . k. yuasa , h. nakazato , and m. unoki , j. mod . opt . * 51 * , 1005 ( 2004 ) ; h. nakazato , m. unoki , and k. yuasa , phys . a * 70 * , 012303 ( 2004 ) . wu , d. lidar , and s. schneider , e - print quant - ph/0402209 . for reviews , see c. h. bennett and d. p. divincenzo , nature ( london ) * 404 * , 247 ( 2000 ) ; m. a. nielsen and i. l. chuang , _ quantum computation and quantum information _ ( cambridge university press , cambridge , 2000 ) , and references therein . p. meystre , in _ progress in optics _ , edited by e. wolf ( elsevier science , new york , 1992 ) , vol . 30 , p. 261 ; j. a. bergou and m. hillery , phys . a * 55 * , 4585 ( 1997 ) ; a. messina , eur . j. d * 18 * , 379 ( 2002 ) ; d. e. browne and m. b. plenio , phys . a * 67 * , 012325 ( 2003 ) . a. napoli , a. messina , and g. compagno , fortschr . * 51 * , 81 ( 2003 ) . h. mabuchi and a. doherty , science * 298 * , 1372 ( 2002 ) . e. hagley , x. matre , g. nogues , c. wunderlich , m. brune , j. m. raimond , and s. haroche , phys . * 79 * , 1 ( 1997 ) ; m. weidinger , b. t. h. varcoe , r. heerlein , and h. walther , _ ibid . _ * 82 * , 3795 ( 1999 ) ; a. rauschenbeutel , g. nogues , s. osnaghi , p. bertet , m. brune , j. m. raimond , and s. haroche , _ ibid . _ * 83 * , 5166 ( 1999 ) ; j. m. raimond , m. brune , and s. haroche , rev . phys . * 73 * , 565 ( 2001 ) . r. j. thompson , g. rempe , and h. j. kimble , phys . lett . * 68 * , 1132 ( 1992 ) ; c. j. hood , t. w. lynn , a. c. doherty , a. s. parkins , and h. j. kimble , science * 287 * , 1447 ( 2000 ) . e. t. jaynes and f. w. cummings , proc . ieee * 51 * , 89 ( 1963 ) ; see also m. o. scully and h. walther , phys . rev . a * 39 * , 5229 ( 1989 ) . h. walther , in _ quantum communication , computing , and measurement 3 _ , edited by p. tombesi and o. hirota ( kluwer academic / plenum publishers , new york , 2001 ) , p. 463 . s. haroche , in _ fundamental systems in quantum optics _ , edited by j. dalibard , j .- raimond , and j. zinn - justin ( north - holland , amsterdam , 1992 ) .
the formation mechanism of the thick disk is one of the most noticeable grey points of our current understanding of the galactic formation and evolution process . this uncertainty is particularly unfortunate because , being the thick disk a ubiquitous feature among spiral galaxies @xcite , its formation must be a common stage in the early evolution of late - type galaxies . during the nearly three decades since its discovery @xcite , many models have been proposed . the merging scenario , in which the early merging of a small satellite galaxy heats a primordial galactic disk producing an old , thick , and kinematically hot disk population @xcite , has been the most favored model in the last decade , following the evidence that the thin and thick disks are chemically distinct @xcite . nevertheless , even this scenario is not free of problems ( see , for example , * ? ? ? * ) , and alternative models have recently been drawn to attention ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? it is clear that the general properties of the galactic thick disk , such as its mean metallicity or mean kinematics , are not enough to distinguish between the models proposed for its formation . moreover , the merging scenario has many free parameters , such as the mass of the merging satellite and the inclination of its orbit with respect to the galactic disk , and the observations must constrain them if the quality of the model is to be finally assessed . in the last few years , theoretical simulations have started to cast predictions of the stellar distribution of stars in the multi - dimensional spatial - kinematical - chemical space ( e.g. , * ? ? ? * ; * ? ? ? for example , @xcite have shown that , within the merging scenario , the time of impact , and the mass and orbit of the infalling satellite can be deduced from the distribution of the expected kinematical clumps in the @xmath7@xmath8 plane , while @xcite , @xcite , and @xcite find that the @xmath4/@xmath5 ratio and its variation with galactocentric distance are excellent indicators of the inclination angle of the merging event . @xcite and @xcite have also shown that the observed decrease of rotation velocity with distance from the plane points to a low - latitude merging . at the same time , the observations are gathering information about the spatial variations of the chemical composition and velocity distribution ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . detailed results of this kind are strongly needed because , through comparison with the expectations from the different models , they can be key to solve the puzzle of the galactic thick disk formation . we are undertaking an extensive survey aimed to reveal the kinematical and chemical vertical structure of the galactic thick disk @xcite . preliminary kinematical results were presented by @xcite and @xcite , while the sample was also used to investigate the galactic dark disk @xcite , and lithium - rich stars in the galactic thick disk @xcite . in this paper , we focus on the kinematical results , studying the trend of kinematical quantities as a function of distance from the galactic plane . in later papers of this series , the collected spectra will be used to measure the metallicity and chemical abundances of the sample , to study the variation of the thick disk chemistry with galactic height . our investigation is based on the sample of @xmath61200 red giants defined by @xcite , vertically distributed with respect to the galactic plane in a cone of fifteen degrees radius centered on the south galactic pole . the sample covers a large range of galactic heights , from the plane to @xmath95 kpc , and it is volume - complete up to @xmath10=3 kpc . all the objects have 2mass photometry @xcite , and absolute proper motions from the spm3 catalog @xcite . the sample was defined by the color cut 0.7@xmath111.1 , to isolate intermediate - metallicity stars . main sequence ( ms ) dwarfs were excluded both by a sloped cut at fainter magnitudes , parallel to the ms , which excludes all but the nearest ( d@xmath1263 pc ) dwarfs , and by conservative kinematical criteria imposing a stellar velocity lower than the local escape velocity ( see * ? ? ? * for more details ) . l l c l c run & date & n. nights & instrument & n. stars + coralie1 & 2005 , september & 4 & coralie@euler & 39 + feros1 & 2005 , september & 6 & feros@2.2 & 183 + dupont1 & 2005 , october & 6 & echelle@dupont & 168 + coralie2 & 2005 , october & 3 & coralie@euler & 25 + feros2 & 2006 , august & 6 & feros@2.2 & 161 + dupont2 & 2006 , september & 6 & echelle@dupont & 119 + mike1 & 2006 , aug.nov . & 5x0.5 & mike@magellan & 77 + mike2 & 2007 , october & 2 & mike@magellan & 52 + we collected high - resolution echelle spectra for two - thirds of the girard et al.s sample ( 824 stars ) . all the stars in the brighter half of the sample were spectroscopically observed , while the completeness decreases with magnitude for fainter objects . we found a high contamination ( @xmath635% ) by nearby dwarf in the faintest end of the distribution , hence stars with @xmath13 were given lower priority after the first observations . the distribution of proper motions and colors of the observed sub - sample is shown in figure [ f_sample ] . the comparison with the complete sample reveals that no selection effect was introduced : the slight overabundance of red stars in the observed sample is due to the higher completeness at brighter magnitudes , where stars are on average redder . as discussed later , we will restrict our investigation to stars with galactic height @xmath141.3 kpc , to avoid a strong thin disk contamination . this reduced the sample under study to 514 stars . we visually inspected all the spectra , identifying 46 probable dwarf stars . as expected , they show on average low radial velocities but high u and v components , and only one is found at @xmath159 . we also found 22 stars with [ fe / h]@xmath161.5 , as deduced from comparisons with synthetic spectra . they were considered probable halo contaminants and were excluded from further analysis . we note that the metal - poor tail of the thick disk extends to much lower metallicities @xcite , but below this limit thick disk stars are outnumbered by halo members by a factor of nearly ten @xcite . finally , we also excluded from analysis 34 stars that , in the galactic cylindrical coordinate system , had velocity components ( defined in section [ s_disp ] ) outside the range @xmath17u@xmath18300 km s@xmath3 , @xmath19v@xmath12300 km s@xmath3 , and @xmath17w@xmath18150 km s@xmath3 . the cut in @xmath20 was aimed to remove the residual halo contamination , while the other components were restricted to exclude stars with wrong distance or bad proper motion . the cut in @xmath8 was offset toward negative values to take into account the vertical shear ( see section [ s_resmean ] ) . our final sample thus comprised 412 stars . the interstellar reddening e(@xmath21 ) was derived for each target from the @xcite maps , and the @xmath22 magnitude and ( @xmath23 ) color were de - reddened by means of the transformations of @xcite . the distance to each star was then estimated through a color - absolute magnitude relation calibrated on 47tucanae , a disk globular cluster @xcite whose stellar population is very similar to the galactic thick disk both in age and metallicity @xcite . the fit of the cluster red giant branch yields the relation ( l. miller 2008 , priv . comm . ) : @xmath24 which were translated into absolute magnitude and de - reddened color assuming @xmath25=13.50@xmath260.08 and e(@xmath21)=0.024@xmath260.004 for the cluster distance modulus and reddening , respectively @xcite . inspecting the 2mass photometric data of 47tuc used to derive equation ( [ eq_dist ] ) , we found that the uncertainty on @xmath27 is of the order of @xmath60.2 magnitudes . this is only a marginal contribution to the final error in distance , because the main source of uncertainty is the relatively wide range of age and metallicity covered by thick disk stars . indeed , the 2mass photometric errors have only negligible impact , because they do not exceed 0.03 magnitudes for our fainter targets ( @xmath2810.6 ) . we estimated the effect of the age and metallicity distribution on the derived absolute magnitudes by means of yale - yonsei isochrones @xcite . we assumed a scatter of 0.3 dex in metallicity , that should include the bulk of thick disk stars @xcite when excluding the scarcely - populated low- and high- metallicity tails @xcite , and a scatter of 2 gyr in age @xcite . they were considered uncorrelated , because the age - metallicity relation for the thick disk is very weak @xcite . we finally estimated the error on distance to be @xmath620% , quadratically summing all the relevant contributions . thin disk stars do not follow the age and metallicity distribution assumed to derive equation ( [ eq_dist ] ) , and their distances should be systematically biased . indeed , younger , more metal - rich red giants are intrinsically fainter than our estimate . comparing the absolute magnitudes calculated from equation ( [ eq_dist ] ) with yale - yonsei isochrones between 2 and 8 gyr , and metallicity following the age - metallicity relation of @xcite and @xcite , we found that the distance of thin disk stars would be overestimated by 1020% . this systematic error is small , but it has a relevant consequence on the contamination of the sample , which is artificially stretched to larger heights from the plane . our sample is contaminated by the thin disk , because the targets were selected through a color cut that efficiently eliminates low - metallicity stars , but which excludes only a small fraction of metal - rich objects , as it can be deduced from figure 1 of @xcite . in figure [ f_contam ] we plot an estimate of the fraction of thin disk stars in the sample , assuming 300 and 900 pc as thin and thick disk scale height , respectively , and a thick disk local normalization of 12% @xcite . we also calculated the contamination in the case of a distance overestimate of 15% for thin disk objects . the curves can be considered an upper limit , because the color cut , unaccounted for in the calculation , could have removed part of the contamination . in the same figure we indicate the results of a rough empirical estimate , obtained fitting a double gaussian to the distribution of the vertical velocity component ( @xmath20 ) in five bins of 50 stars . the dispersions of the two gaussian components were kept fixed , and the only free parameter of the fit was the number of thin disk stars in the bin . the error bars show the results when varying @xmath5 in the range 14 - 16 and 35 - 40 km s@xmath3 for the thin and the thick disk , respectively . this test was performed only for @xmath291 - 2 kpc , where both populations contribute by more than 20% of the objects , i.e. 10 stars in each bin . the empirical errorbars agree well with the curve obtained when the distance bias is considered , although the observed contamination fall - off with @xmath1 seems steeper than the modeled one . in conclusion , our sample is contaminated by thin disk stars , whose distances are overestimated . this affects even the estimate of their kinematics , which is also biased . therefore , we will not analyze the results for the thin disk , and we will limit our study to @xmath141.3 kpc , to avoid the strong contamination ( @xmath3050% ) of the nearest subsample . @xcite estimated that the halo contamination in the sample should be about 8% , hence we would expect 41 halo members among the 514 targets with @xmath141.3 kpc . the cut in @xmath20 removed 10 probable halo contaminants and , according to the statistics of @xcite , 18 of the excluded low - metallicity targets are expected to be halo stars . the cuts in @xmath7 and @xmath8 could also have removed some halo objects , but the effect of this cut is harder to quantify , because an unknown fraction of the outliers are probably objects with problematic proper motion or wrong distance . the residual halo contamination in our sample of 412 targets should therefore be lower than @xmath63% ( 13 stars ) , and it can be safely neglected . the spectra were collected during six observing runs between 2005 and 2007 , at la silla and las campanas observatories . the details of the observations and data reduction were presented in @xcite , and we will outline here only the most relevant points . four different telescopes were used , because the stars span a wide range in magnitude ( @xmath8=516 ) , but the instruments had similar characteristics . the basic information of the runs is given in table [ t_obs ] . the spectra covered the whole visual range from the atmospheric cutoff on the blue side to @xmath69000 , except for coralie data which only reached 6800 on the red edge . we did not reduce the spectra collected with the mike blue arm because of too low stellar flux , and mike data were thus limited to 4850 on the blue end . the spectral resolution varied between 32000 and 50000 , depending on the spectrograph . in particular , the 0@xmath317 slit was always used for mike runs ( r=32000 ) , while at dupont both the 0@xmath3175 and 1@xmath32 slits were used ( r=40000 and 30000 , respectively ) , depending on weather conditions . during each run we secured the spectra of up to seventeen bright stars with accurate parameters from the literature ( radial and rotational velocities , temperature , gravity , metallicity ) . exposure times were chosen according to weather conditions , to reach s / n=70100 for all the targets . comparison lamp arcs were acquired only during daytime calibrations for the fiber spectrographs coralie and feros . the second fiber of these two instruments was allocated to the sky background , because the spectra were usually collected next to full moon . lamp arcs were collected at intervals of about 2 hours during dupont and mike runs , and each spectrum was calibrated with the average of the two lamps temporally enclosing it , linearly weighted by the temporal distance between science and calibration frames . spectra were reduced with standard iraf[multiblock footnote omitted ] tasks , and we kept the procedures as much as possible identical for all the data . we a posteriori verified that the reduced feros and coralie spectra were of the same quality as the products of their dedicated pipelines . we analyzed the dark frames collected for all the runs , but we always found a negligible instrumental dark current and no clear 2d pattern , hence no dark correction was applied . the wavelength calibration lamp spectra were extracted in the same position on the ccd as science targets , to avoid systematics introduced by the curvature of the orders . the spectrum of a bright , well - exposed star was used to trace the orders on the ccd in each observing night , allowing only for rigid shifts among the frames . then , the spectra were extracted with an optimum extraction algorithm @xcite , normalized , and finally merged . the radial velocity ( rv ) for all the targets was measured with a cross - correlation ( cc ) technique @xcite as implemented in the iraf _ fxcor _ task . the procedure was described in detail in @xcite , and we give only a brief summary here . the spectrum of each star was cross - correlated with three template stars observed in the same run , namely hd180540 , hd223559 , and hd213893 ( this last replaced by hd224834 for coralie data ) , encompassing the temperature range of the targets . their rv was fixed by a cc with a synthetic spectrum generated with the spectrum code @xcite , because we found poor agreement between the available literature sources . the analysis of the seventeen standard stars acquired during observations , and of the solar spectra collected each night , revealed that the rv zero - point thus defined was biased by 0.3 - 0.7 km s@xmath3 , depending on the instrument , and this offset was removed . the three measurements were averaged , although they never differed by more than 0.2 km s@xmath3 . the cc was restricted to the range 4600 - 6800 ( 5000 - 6800 for mike data ) , to avoid the poor - quality blue section and the telluric bands at longer wavelengths . all rvs were corrected to heliocentric velocities , then the position of the telluric band at 6875 was used to correct the rvs for instrumental effects , mainly caused by a displacement on ccd between the lamp and science spectra , and an off - center position of the star inside the slit ( see , for example , the analysis of * ? ? ? corrections of up to 2 km s@xmath3 were applied , but with little scatter ( @xmath60.5 km s@xmath3 ) within each observing night . the final rv errors were estimated as the quadratic sum of the five relevant uncertainties introduced in the procedure : the cc and wavelength calibration error , the uncertainty on the zero - point and its offset , and the error on the estimate of the correction for instrumental effects . the resulting errors were typically in the range 0.4 - 0.7 km s@xmath3 . the final rvs of all the 824 stars will be made available at the cds website . l c c reference & n. stars & @xmath33 + & & km s@xmath3 + @xcite & 145 & 0.8@xmath262.5 + @xcite & 9 & @xmath341.5 + @xcite & 9 & 0.4@xmath261.8 + @xcite & 8 & @xmath352.1 + @xcite & 6 & @xmath363.8 + @xcite & 6 & @xmath370.9 + @xcite & 6 & 1.8@xmath264.2 + we found 211 previous rv measurements for the stars in our sample . our results agree excellently with literature sources : the mean difference ( in the sense ours@xmath38literature ) is 0.4@xmath262.7 km s@xmath3 , where the uncertainty is the rms of the differences . the comparison with the works that share with us more than five stars in common is given in table [ t_lit ] . the mean rv difference is always of the order of 1 km s@xmath3 , except for @xcite , whose rvs are higher than ours by a non - negligible amount ( @xmath65 km s@xmath3 ) . however , the number of stars in common is too small to conclude that this offset is significant . the proper motion , radial velocity , and distance of each target were combined to derive its ( @xmath39 ) velocity components in the galactic cylindrical reference frame , where @xmath7 is positive toward the galactic center , @xmath8 is directed in the sense of galactic rotation , and @xmath20 points toward the north galactic pole . the error on these velocities was derived propagating the uncertainty on the observed quantities . we assumed a proper motion error of 3 mas yr@xmath3 for all the stars , as this value is more realistic than the uncertainties quoted in the spm3 catalog ( t. girard , 2009 , priv . see also @xcite ) . the velocities were corrected subtracting the solar peculiar motion ( @xmath40)=(10.0 , 5.1 , 7.2 ) km s@xmath3 @xcite . @xcite recently proposed the revised values ( @xmath40)=(11.0 , 12.2 , 7.3 ) km s@xmath3 , but we preferred to adopt the older ones , for sake of continuity with previous works . in any case , the definition of the solar motion does not affect the velocity dispersions nor the off - diagonal terms of the dispersion matrix ( equation [ e_cross ] ) , while the effects on the mean values are discussed in section [ s_resmean ] . the sample was then divided into several bins , in which we calculated the mean velocities , the dispersions ( @xmath41 ) , and the non - diagonal terms of the dispersion matrix @xmath42 where the sum is extended to all the stars in the bin , and @xmath43 . the results are given in table [ t_res ] . the bins were defined by the requirement that their centers were equally spaced by 0.1 kpc from @xmath1=1.5 to 4.5 kpc , to uniformly sample the variation with @xmath1 of the kinematical quantities , while the width was imposed by the fixed number of stars per bin ( see below ) . this implied a partial overlap of the bins , increasing with distance from the plane due to the decreasing density of observed stars . however , @xcite have shown that the binning definition does not alter the results , and in fact our results are very similar to that work , despite the very different binning schemes . the bins with @xmath442.4 kpc comprised 50 stars each , while at lower heights , where the number of observed stars is larger , the bin size was increased to 80 targets for 2.1@xmath452.4 kpc , and 100 targets for @xmath462.1 kpc . we thus ensured a good statistic of thick disk stars in bins contaminated by the thin disk . in each bin , the mean value and dispersion of each velocity component were measured by means of the analysis of the corresponding probability plot @xcite . this powerful tool is highly insensitive to outliers , and it can be employed even in case of poorly populated bins . the data are first ordered with increasing value of the velocity . each point @xmath47 is then assigned a value @xmath48 , equal to the deviation from the mean expected for the @xmath49 point of a normal distribution , in units of the standard deviation . when the underlying distribution is gaussian , the data follow a linear relation in the @xmath48-velocity plane , whose intercept and slope are given by the mean value and the dispersion , respectively . the probability plots were fitted with a straight line , weighting each data point by its error , two examples are shown in figure [ f_probplot ] . the intrinsic velocity dispersions were then obtained quadratically subtracting the mean velocity error in the bin . the wings of the distribution showing any deviation from linearity were excluded from the fit , suspected to be affected by objects with wrong distance or problematic proper motion , or by residual halo members . the analysis of artificial data samples , generated through monte - carlo simulations and analyzed as the real data , indicated that the velocity dispersion is underestimated by 1 - 2 km s@xmath3 when the thin disk contamination approaches 10% . the contamination was therefore neglected for @xmath442.5 kpc , where it is expected to be lower than 5% ( i.e. two stars per bin , see figure [ f_contam ] ) . following @xcite , in the presence of a non - negligible thin disk contamination ( @xmath462.5 kpc ) we derived the mean velocity and the dispersion of thick disk stars fitting only the wings of the probability plot , and only the negative one for @xmath8 . the cut excluded the ranges @xmath17u@xmath1850 km s@xmath3 , @xmath17w@xmath1830 km s@xmath3 , and @xmath5060 km s@xmath3 , approximatively 1.5 times the expected thin disk dispersion , thus removing about 90% of the contaminants . the two wings of the probability plot should have the same slope but a different intercept , and the mean velocity was obtained averaging the two intercepts obtained from the fit . although the mean values thus derived agree well with the trends observed in the contamination - free bins at higher @xmath1 , we regard them as less reliable . the formal errors of the least - square procedure used to obtain the intercept and the slope in the probability plots mainly indicate the deviations from gaussianity of the observed distribution . hence , they are not a good estimate of the real uncertainties . the errors were therefore derived by means of monte - carlo simulations . for each bin , we generated one thousand artificial samples of 50 stars , changing the proper motion , distance , and rv of each star assuming a gaussian random noise with dispersion equal to the estimated errors ( 3 mas yr@xmath3 , 20% , and 0.4 - 0.7 km s@xmath3 , respectively ) . the measurements were repeated in each artificial bin as done for the real data , and the dispersion of these artificial measurements was assumed as an estimate of the observational errors . the final errors are given in table [ t_res ] . the non - diagonal terms of the dispersion matrix , @xmath51 , @xmath52 , and @xmath53 , were calculated by means of equation [ e_cross ] . the errors were estimated from monte - carlo simulations , as done for the velocity dispersions . for each bin , one thousand artificial data sets of 50 values were simulated , drawn from a distribution characterized by the same dispersion matrix as the real data . in each simulation , a gaussian random noise with a dispersion equal to the observational errors was added , and the non - diagonal terms were estimated by means of equation [ e_cross ] . the errors were defined by the scatter of these simulated measurements . with the same simulations we also evaluated the systematic error introduced by the observational uncertainties ( e.g. * ? ? ? * ) , comparing the results when the random noise was considered or not . we found that the expected systematic thus introduced is negligible , being about one order of magnitude lower than the uncertainties on the measured values . therefore , we did not correct the observed non - diagonal terms for observational errors . c r r r r r r r r r @xmath1 & @xmath54 & @xmath55 & @xmath56 & @xmath4 & @xmath57 & @xmath5 & @xmath52 & @xmath51 & @xmath53 + kpc & km s@xmath3 & km s@xmath3 & km s@xmath3 & km s@xmath3 & km s@xmath3 & km s@xmath3 & km@xmath58 s@xmath59 & km@xmath58 s@xmath59 & km@xmath58 s@xmath59 + 1.5 & @xmath3831@xmath264 & @xmath3846@xmath264 & 17@xmath261 & 81@xmath265 & 57@xmath265 & 38@xmath261 & 570@xmath26310 & 400@xmath26470 & 10@xmath26210 + 1.6 & @xmath3820@xmath264 & @xmath3855@xmath264 & 5@xmath261 & 77@xmath265 & 59@xmath265 & 37@xmath261 & 840@xmath26290 & @xmath38180@xmath26460 & @xmath3820@xmath26220 + 1.7 & @xmath3828@xmath264 & @xmath3861@xmath265 & 3@xmath261 & 79@xmath265 & 64@xmath265 & 38@xmath261 & 930@xmath26310 & @xmath38130@xmath26510 & @xmath3840@xmath26250 + 1.8 & @xmath3824@xmath264 & @xmath3869@xmath265 & 8@xmath261 & 79@xmath265 & 61@xmath266 & 40@xmath261 & 810@xmath26330 & @xmath38760@xmath26500 & @xmath3880@xmath26240 + 1.9 & @xmath3816@xmath264 & @xmath3877@xmath265 & 3@xmath261 & 78@xmath265 & 60@xmath265 & 40@xmath261 & 850@xmath26330 & @xmath381060@xmath26500 & @xmath38120@xmath26240 + 2.0 & @xmath3810@xmath265 & @xmath3885@xmath264 & 4@xmath261 & 83@xmath266 & 55@xmath266 & 39@xmath261 & 940@xmath26340 & 30@xmath26460 & @xmath3890@xmath26210 + 2.1 & @xmath3822@xmath265 & @xmath3878@xmath265 & 3@xmath261 & 80@xmath266 & 58@xmath266 & 38@xmath261 & 1060@xmath26320 & 470@xmath26470 & @xmath38130@xmath26220 + 2.2 & @xmath3816@xmath265 & @xmath3876@xmath267 & @xmath3813@xmath261 & 77@xmath266 & 63@xmath267 & 42@xmath261 & 690@xmath26330 & 0@xmath26550 & 230@xmath26260 + 2.3 & @xmath381@xmath266 & @xmath3881@xmath265 & @xmath389@xmath261 & 81@xmath267 & 58@xmath266 & 40@xmath261 & 750@xmath26340 & @xmath38500@xmath26470 & 10@xmath26230 + 2.4 & @xmath3812@xmath265 & @xmath3885@xmath266 & @xmath3812@xmath261 & 80@xmath266 & 59@xmath266 & 40@xmath261 & 410@xmath26330 & @xmath38710@xmath26480 & 180@xmath26240 + 2.5 & @xmath3818@xmath266 & @xmath3885@xmath266 & @xmath382@xmath261 & 78@xmath267 & 63@xmath267 & 42@xmath261 & 630@xmath26330 & @xmath381870@xmath26600 & @xmath3840@xmath26270 + 2.6 & @xmath3825@xmath266 & @xmath3890@xmath266 & 1@xmath261 & 71@xmath267 & 66@xmath267 & 42@xmath261 & 530@xmath26310 & @xmath382390@xmath26660 & @xmath38460@xmath26290 + 2.7 & @xmath3823@xmath266 & @xmath3898@xmath266 & 7@xmath261 & 72@xmath267 & 62@xmath267 & 39@xmath261 & 870@xmath26290 & @xmath381150@xmath26520 & @xmath381360@xmath26300 + 2.8 & @xmath3829@xmath266 & @xmath3895@xmath267 & @xmath382@xmath261 & 76@xmath267 & 62@xmath267 & 41@xmath261 & 880@xmath26320 & @xmath381740@xmath26580 & @xmath381310@xmath26300 + 2.9 & @xmath3817@xmath266 & @xmath38115@xmath267 & 5@xmath261 & 83@xmath267 & 68@xmath267 & 40@xmath261 & 840@xmath26340 & @xmath38820@xmath26590 & @xmath381230@xmath26300 + 3.0 & @xmath387@xmath266 & @xmath38126@xmath267 & 8@xmath261 & 90@xmath267 & 67@xmath268 & 42@xmath261 & 1440@xmath26410 & @xmath3850@xmath26610 & @xmath38460@xmath26280 + 3.1 & @xmath386@xmath267 & @xmath38129@xmath268 & 1@xmath261 & 101@xmath268 & 67@xmath268 & 43@xmath261 & 1610@xmath26450 & @xmath381150@xmath26690 & @xmath38650@xmath26290 + 3.2 & 2@xmath266 & @xmath38131@xmath268 & @xmath381@xmath261 & 99@xmath268 & 63@xmath268 & 42@xmath261 & 1470@xmath26440 & @xmath381820@xmath26650 & @xmath38440@xmath26280 + 3.3 & 8@xmath267 & @xmath38140@xmath268 & 0@xmath261 & 101@xmath268 & 66@xmath268 & 44@xmath262 & 2180@xmath26490 & @xmath382610@xmath26740 & @xmath38310@xmath26300 + 3.4 & 12@xmath267 & @xmath38140@xmath268 & 1@xmath261 & 98@xmath268 & 63@xmath269 & 43@xmath262 & 2100@xmath26480 & @xmath381880@xmath26660 & 260@xmath26280 + 3.5 & 18@xmath267 & @xmath38135@xmath269 & 5@xmath261 & 95@xmath269 & 63@xmath269 & 44@xmath262 & 1950@xmath26460 & @xmath383390@xmath26740 & 260@xmath26280 + 3.6 & 29@xmath267 & @xmath38137@xmath268 & 9@xmath261 & 101@xmath269 & 64@xmath269 & 44@xmath262 & 2290@xmath26510 & @xmath382170@xmath26700 & 360@xmath26290 + 3.7 & 7@xmath267 & @xmath38132@xmath269 & 0@xmath262 & 91@xmath269 & 61@xmath2610 & 44@xmath262 & 2180@xmath26460 & @xmath382150@xmath26610 & 610@xmath26280 + 3.8 & 15@xmath267 & @xmath38134@xmath2610 & 1@xmath262 & 92@xmath269 & 68@xmath2610 & 43@xmath262 & 1990@xmath26450 & @xmath382120@xmath26700 & @xmath3830@xmath26300 + 3.9 & 5@xmath268 & @xmath38139@xmath269 & @xmath386@xmath262 & 94@xmath2610 & 66@xmath2610 & 43@xmath262 & 1540@xmath26430 & @xmath38520@xmath26620 & 210@xmath26280 + 4.0 & 7@xmath268 & @xmath38143@xmath2610 & @xmath381@xmath262 & 93@xmath269 & 66@xmath2611 & 41@xmath262 & 1910@xmath26420 & @xmath38100@xmath26610 & 10@xmath26280 + 4.1 & 17@xmath268 & @xmath38145@xmath2610 & 1@xmath262 & 92@xmath2610 & 68@xmath2610 & 44@xmath262 & 1570@xmath26440 & 720@xmath26640 & @xmath3830@xmath26300 + 4.2 & 22@xmath268 & @xmath38142@xmath2611 & 2@xmath262 & 95@xmath2610 & 72@xmath2611 & 45@xmath262 & 1480@xmath26460 & 880@xmath26710 & 10@xmath26320 + 4.3 & 12@xmath269 & @xmath38148@xmath2610 & @xmath382@xmath262 & 94@xmath2610 & 69@xmath2611 & 48@xmath262 & 2320@xmath26510 & 1170@xmath26680 & 180@xmath26330 + 4.4 & 17@xmath269 & @xmath38152@xmath2611 & 0@xmath262 & 96@xmath2610 & 72@xmath2611 & 46@xmath262 & 2330@xmath26500 & 1370@xmath26710 & @xmath3850@xmath26330 + 4.5 & 19@xmath2610 & @xmath38158@xmath2611 & @xmath382@xmath262 & 93@xmath2611 & 76@xmath2611 & 46@xmath262 & 2180@xmath26480 & 3960@xmath26930 & 240@xmath26360 + the trend of the mean velocities with distance from the plane is shown in figure [ f_resmean ] . @xmath56 is consistent with no significant departure from zero in the whole range . some deviations are observed for @xmath602.5 kpc , but in this range the results are affected by large errors and , as already commented , not very trustworthy . this result agrees with previous studies , that found no net vertical motion up to various kpc from the galactic plane @xcite . on the contrary , @xmath61 has a more complex behavior : between @xmath1=1.5 and 3 kpc we detect a non - null mean velocity directed toward the galactic anticenter , with an average value @xmath54=@xmath623 km s@xmath3 . beyond 3 kpc from the plane @xmath54 abruptly increases and changes sign , and the net velocity between 3.5 and 4.5 kpc is 15@xmath262 km s@xmath3 toward the galactic center . another possible interpretation of the results is that @xmath54 linearly increases with @xmath1 , and a s - shaped feature is superimposed to this trend between 2.5 and 3.5 kpc , as observed for @xmath4 and other kinematical quantities ( see section [ s_comove ] ) . after the exclusion of this range , the fit returns a vertical increase of 15@xmath63 km s@xmath3 kpc@xmath3 . adopting the new values of @xcite for the solar peculiar motion causes not a relevant change on the results , as @xmath54 would be higher by about 1 km s@xmath3 only . previous studies have shown evidence that stars more distant than 1 kpc from the plane show a non - null net radial motion of the order of @xmath69 km s@xmath3 toward the galactic anticenter @xcite . our results between @xmath1=1.5 and 3 kpc agree with their conclusion , although our mean value is larger by about a factor of two . the sign flip observed at larger @xmath1 , on the contrary , has never been reported in the literature . nevertheless , @xcite analyze halo stars , @xcite do not reach these heights above the plane , and @xcite have too few stars in this range . very interestingly , @xcite detect a small positive mean radial component out to @xmath64 kpc , and a negative value in their more distant bins , although the authors conclude that these results are consistent , within errors , with a null net motion . @xcite proposed that an inward motion of the local standard of rest ( lsr ) of the order of 10 km s@xmath3 is responsible for the observed non - null mean value of @xmath54 , as suggested by the fact that this is observed among both disk and halo stars more distant than 1 kpc from the sun . this hypothesis can not account for any vertical trend of @xmath54 other than a constant non - zero value at any @xmath1 . thus , while it is not contradicted by our results , at least another effect must be invoked . for example , an outward motion of @xmath625 km s@xmath3 of the stars between 3.5 and 4.5 kpc , coupled with the mentioned lsr motion , could explain our results . this clumpy kinematical distribution would not be surprising , because perturbations produced by the bar and the spiral arms are expected to form groups of stars with velocity offset as large as 50 km s@xmath3 @xcite . @xcite studied the proper motions of our same stellar sample , and their results for @xmath61 are overplotted to ours in the upper panel of figure [ f_resmean ] , after correcting for the solar peculiar motion and changing the sign of @xmath7 to account for the different direction of the reference axis . the agreement with our results is poor : while their nearly flat profile roughly coincides with our mean value ( @xmath387.7 km s@xmath3 ) , they did not detect any change of sign , nor a steep positive gradient . the different approach to the same data must have caused the different results . for example the features observed by us could have been masked in @xcite by their smoothed , statistical distance determination , or by dwarf stars and halo contaminants , removed in our work . in the middle panel of figure [ f_resmean ] we show the measured trend of @xmath65 , compared to other results from the literature . the mean rotational velocity of thick disk stars decreases with distance from the plane , due to the well - known vertical shear first detected by @xcite . the fit of the data , after the exclusion of the less reliable bins at @xmath602.5 kpc , yields @xmath66 ( km s@xmath3 ) . as can be seen in figure [ f_resmean ] , this solution excellently matches the results of @xcite , and even the data points of @xcite , at @xmath67 kpc , are well aligned with the derived linear relation . both these investigations measure a vertical shear of @xmath3830 km s@xmath3 kpc@xmath3 , as also recently found by @xcite , and @xcite . had we fitted all the data points down to @xmath1=1.5 kpc , we would have found a steeper slope ( @xmath68 km s@xmath3 kpc@xmath3 ) , at the upper limit of the range spanned by literature values , which vary from @xmath694 @xcite to @xmath701 km s@xmath3 @xcite . the revised values for the solar peculiar motion proposed by @xcite cause an upward revision of the results by 7.1 km s@xmath3 . @xcite and @xcite measured a shallower slope ( @xmath712.1 and @xmath622 km s@xmath3 kpc@xmath3 , respectively ) between 0.7 and 2.8 kpc . while our results are compatible with theirs in our nearest bins , we find a lower mean rotational velocity beyond @xmath1=2 kpc . nevertheless , the different vertical rotational gradient found by these studies is not necessarily a disagreement , because they sample a different @xmath1-range , and the underlying shear is not required to be strictly linear . in fact , @xcite have proposed the non - linear relation @xmath72 km s@xmath3 from the analysis of sdss data . the combined data points of @xcite , @xcite , @xcite , and of the present work , closely follow this equation ( see figure [ f_vall ] ) , and a fit in the form @xmath73 km s@xmath3 returns a very similar solution , with @xmath74 km s@xmath3 , @xmath75 km s@xmath3 kpc@xmath3 , and @xmath761.23 . very noticeably , the results of four surveys finding a different linear relation are all well reproduced by a single non - linear curve proposed by an independent work . the underlying vertical trend of the thick disk rotational velocity is therefore most probably close to but not exactly linear , and better represented by a low - order power law . the vertical profile of the velocity dispersions is shown in figure [ f_disp ] , compared to other works from the literature . we excluded from comparison the results of @xcite , because they do not distinguish between different disk components , thus finding steeper gradients as a result of the mix of thin and thick disk stars . caution must also be taken when comparing the results obtained in different ranges of @xmath1 , because the underlying trend of the dispersions is not necessarily linear , and the gradient can assume different values . for example , the models of @xcite indicate that the vertical profile should be progressively flatter at increasing @xmath1 . the linear fit of the trends with @xmath1 yields : @xmath77 @xmath78 @xmath79 where @xmath1 is in kpc . the quoted errors were obtained restricting the fit to a subset of eight non - overlapping bins , to avoid the underestimate induced by the correlation between the data points . the results are identical to those of @xcite , despite the different binning scheme , except for @xmath57 for which we derive a slightly smaller gradient , but the difference is not significant ( 0.4 km s@xmath3 kpc@xmath3 ) . the gradients are small , and the three dispersions increase by only @xmath67% between 2.5 and 3.5 kpc . this explains why the change of the thick disk kinematics with distance from the plane has not been clearly detected for about two decades after its discovery . the results for @xmath5 are the most precise and reliable , because @xmath690% of the information on @xmath20 comes from rvs , whose errors are an order of magnitude smaller than those in proper motions . @xmath80 shows a small but clear and constant increase , with small scatter around the derived linear relation . the vertical gradient is smaller than the one found by @xcite by about a factor of two , but the results are consistent at the 1@xmath48 level . @xcite propose an even steeper profile ( @xmath610 km s@xmath3 kpc@xmath3 ) in the range @xmath1=02 kpc , but the same authors suspect that this could be due to contamination by halo stars increasing with galactic height . the dispersion of the rotational velocity component , @xmath57 , also shows a clear increase with @xmath1 , although the data points are affected by larger errors and are more scattered . the vertical gradient is smaller than the results of both @xcite and @xcite , by a factor of three and two , respectively , and the difference is at the 2@xmath48 level in both cases . the vertical profile of @xmath4 shows a very peculiar behavior , with large deviations from linearity between 2.5 and 3.5 kpc . this feature will be discussed in section [ s_comove ] . outside this range , the data points closely follow a linear relation , and the solution given in equation [ e_su ] was obtained excluding this interval from the fit . the derived vertical gradient of @xmath4 is very similar to that found by @xcite , but their solution is offset toward higher values by about 10 km s@xmath3 , while the gradient measured by @xcite is 4@xmath48 times higher than ours . in conclusion , we confirm that the velocity dispersions of the galactic thick disk increase with distance from the plane , as suggested by previous investigations @xcite , but we derive vertical gradients that are in general smaller than other studies . the differences with @xcite can be at least in part due to the aforementioned expected decrease of the gradient with distance from the plane , because the results for @xmath5 are consistent , and the solutions proposed for the other two components overlap in the @xmath0-range in common ( 1.52.2 kpc ) . on the contrary , @xcite studied our same sample , and the different data analysis must be the cause of the disagreement . since they measured both higher dispersions and steeper vertical gradients than us , their results could have been affected by the dwarf stars and halo contaminators removed by us , and/or the thin disk contamination , that they did not take into account . the three dispersions increase with distance from the plane proportionally , and the anisotropy is approximatively constant at all @xmath1 . in figure [ f_ratio ] the vertical profile of the ratios @xmath4/@xmath5 and @xmath4/@xmath57 is shown , where the errors come from propagation of the uncertainties on the dispersions . the mean value of @xmath4/@xmath5 is 2.08@xmath260.06 , where the error was estimated from the statistical error on the mean for eight uncorrelated bins , as before . the linear fit of the data points returns a negligible gradient ( 0.09@xmath260.07 kpc@xmath3 ) , and the data are consistent with a flat profile . literature values for @xmath4/@xmath5 span a wide range from 1.1 @xcite to 1.9 ( e.g. , * ? ? ? * ) , and our results are at the upper end of this distribution . as shown in the upper panel of figure [ f_ratio ] , the results of @xcite agree with our measurements in the z - range in common , but they deviate from our measurements if extrapolated to higher distance from the plane . the results for @xmath4/@xmath57 also show no significant gradient ( 0.06@xmath260.05 kpc@xmath3 ) and a mean value of 1.36@xmath260.05 . this is lower than the value predicted by the epicyclic approximation ( 1.49 , cfr . equation 3 - 76 of * ? ? ? * ) , indicating that the population under analysis can not be assumed in nearly circular orbits . our results are well within the range spanned by literature values , which vary from @xmath811 ( e.g. , * ? ? ? * ; * ? ? ? * ) , to @xmath811.6 @xcite . as in the case of @xmath4/@xmath5 , the results of @xcite overlap with our data points in the range in common , but their extrapolation to higher @xmath1 do not . the analytical expressions given by @xcite for @xmath4 and @xmath57 , on the contrary , return a value of their ratio much lower than our data points at any galactic height . this is probably due to their much higher @xmath57 , as shown in the middle panel of figure [ f_disp ] . the non - diagonal term of the dispersion matrix @xmath82 gives a measurement of the correlation between the @xmath47-th and @xmath83-th velocity components . in fact , if two principal axis of the dispersion ellipsoid are aligned with the @xmath47-th and @xmath83-th axis , we have @xmath82=0 , and the two velocity components are uncorrelated . the orientation of the dispersion ellipsoid of an old , dynamically relaxed population is related to the shape of the galactic potential @xcite , and can indicate the presence of non - axisymmetric structures in the disk @xcite . moreover , @xcite and @xcite have shown that the vertical tilt of the ellipsoid allows to estimate the flattening of the dark halo . however , calculating the expectation of a specific halo model through integration of orbits is beyond the scope of the paper , and here we will only present the observational results , for use of future works . our results for the three cross - terms are shown in figure [ f_cross ] . the profile of @xmath51 closely follows a decreasing linear relation up to @xmath1=4 kpc , but abruptly jumps to positive values at larger heights . this is probably due to the sign flip of @xmath54 analyzed in section [ s_comove ] . in any case , the results indicate that the @xmath7 and @xmath8 velocity components are correlated , and the velocity ellipsoid is titled in the radial - longitudinal plane . the rotation angle , i.e. the vertex deviation @xmath84 , was calculated in each bin by means of the relation : @xmath85 @xcite . the results are shown in the upper panel of figure [ f_angles ] . we measure a non - negligible vertex deviation , increasing from nearly zero at @xmath1=1.5 kpc to @xmath86 at 3.5 kpc . the linear fit in the range 1.54 kpc yields the relation @xmath87 . our results agree with @xcite , who measured @xmath88 at @xmath0=1.1 kpc . previous investigations evidenced that the vertex deviation decreases from @xmath89 for young populations to nearly zero for old , metal - poor disk stars @xcite . our results are not at variance with this conclusion if the vertex deviation increases with @xmath1 , as suggested by our observations , because previous studies were limited to small galactic heights . @xmath53 shows no significant deviation from zero in the whole range of @xmath1 , except between 2.5 and 3.5 kpc , as discussed in section [ s_comove ] . this indicates that the @xmath8 and @xmath20 velocity components are not correlated . on the contrary , @xmath52 is significantly different from zero , and steadily increases with @xmath1 . proposing a reliable linear expression for @xmath90 is not straightforward , because it shows irregular fluctuations . as expected by @xcite and @xcite , @xmath52 is always bound between zero and @xmath91 , the value assumed when the velocity ellipsoid is aligned with the spherical coordinate system , because we find that @xmath92 km@xmath93 s@xmath59 at any galactic height . we also find that the expression proposed by @xcite , obtained under the assumption that the ellipsoid points toward the galactic center and its axis ratio is constant in the spherical coordinate system , is a relatively good approximation of the measured value up to about 3 kpc from the plane , but the predicted slope is too shallow and the agreement with observations degrades with @xmath1 ( see figure [ f_cross ] ) . the tilt angle in the @xmath7-@xmath20 plane can be calculated , analogously to the vertex deviation , from the equation : @xmath94 and the results are shown in figure [ f_angles ] . we measure a slight increase of @xmath95 with @xmath1 , and the linear fit of the data , after the exclusion of two deviating points , yields the relation @xmath96 $ ] . the results agree with previous works at @xmath1=11.5 kpc @xcite , but not with the measurements of @xcite at @xmath971 kpc , whose sample is most probably dominated by thin disk stars . the dashed curve in figure [ f_angles ] indicates the value of @xmath95 when the dispersion ellipsoid points toward the galactic center , assuming r@xmath98=8 kpc . the tilt angle is constantly lower than this in the range 1.54.5 kpc , and the ellipsoid thus is directed toward a point located behind the galactic center , at a galactocentric distance increasing with @xmath1 from r@xmath99=2.3 to 9 kpc . orbit integration studies indeed predict this result @xcite , but the inclination is noticeably higher than the expectations , because theoretical calculations return r@xmath99=510 kpc at @xmath1=1.1 kpc . this could indicate that the galactic potential used in these studies needs to be refined to match the observations . on the contrary , the increase of @xmath100 modeled by @xcite is too steep , and their expectation @xmath95(3.5 kpc)=26@xmath101 is at variance with our results . @xcite showed that the vertical trend of the tilt angle is an excellent observational signature of the underlying gravity law , and it can be used as a test for the modified newtonian dynamics theory ( mond , * ? ? ? * ) , because its expectation diverges from that of the newtonian dynamics with distance from the plane . unfortunately , the calculations of @xcite are limited to @xmath972 kpc , and the range of overlap with our data is very narrow . the value of @xmath95 expected by @xcite at @xmath1=2 kpc is 12@xmath101 for newtonian dynamics and 10@xmath101 for mond . we find @xmath95(2 kpc)=@xmath102 and the mean of the five measurements in the range 1.8 - 2.2 kpc is @xmath103 , in excellent agreement with mond expectations . the linear fit presented in section [ s_resangles ] also implies @xmath95(2 kpc)=@xmath104 . however , we derive a vertical gradient ( @xmath105 kpc@xmath3 ) much shallower than the predictions of @xcite for both mond ( 5@xmath101 kpc@xmath3 ) and newtonian dynamics ( 6@xmath101 kpc@xmath3 ) . in conclusion , our observations agree better with the models that @xcite derived for mond than with those from newtonian dynamics , but the extension of @xcite calculations to higher @xmath1 , and to specific initial conditions for the thick disk , are required to perform a reliable test of the gravitational law . the vertical profile of @xmath106 shows a puzzling behavior between 2.5 and 3.5 kpc from the galactic plane ( top panel , figure [ f_disp ] ) , where the dispersion deviates from the linear trend . a similar feature is clearly observable in the trend of other quantities related to the radial velocity component @xmath7 , like @xmath54 and @xmath52 ( as already noted by * ? ? ? * ) , but deviations from the linear trend in the range @xmath1=2.53 kpc could be present even in other profiles , such as @xmath65 , @xmath107 , @xmath80 , and @xmath108 . the origin of this behavior is unclear . it is possible that a group of comoving stars , forming a sub - structure in the galactic thick disk , is affecting the measured kinematics between 2.5 and 3.5 kpc from the plane . in particular , the similarity of the profile of two totally unrelated quantity such as @xmath108 and @xmath106 is instructive , because it excludes the possibility that this behavior is only due to some bad measurements . in any case , the existence of a kinematical substructure among our stars can not be claimed on the basis of these results only . our targets are distributed in a very narrow range of @xmath109 , and the data do not provide any direct information about the radial behavior of the kinematical quantities . nevertheless , some indirect indication can be derived . manipulating the radial jeans equation in cylindrical coordinates in steady state , with the radial component of the force expressed as f@xmath110 , we obtain : @xmath111 where we also assumed that @xmath112 , @xmath113/@xmath114=0 , and that the density decays exponentially with both @xmath109 and @xmath0 , with scale length @xmath115 and scale height @xmath116 . inserting the observed quantities in the right hand side of equation [ e_jeans ] we can thus estimate , as an exercise , the radial behavior of @xmath4 . the large uncertainties involved prevent a precise measurement , but the comparison with the expectations provides a consistency check for the kinematical data here presented , because an unphysical result could indicate a problem with them . the results are shown in figure [ f_jeans ] , where we assumed @xmath115=3.6 kpc and @xmath116=0.9 kpc @xcite , r@xmath98=8 kpc , and @xmath117=220 km s@xmath3 . the data points are scattered around the mean value @xmath118 km s@xmath3 kpc@xmath3 , following the behavior of @xmath106 discussed in section [ s_comove ] , with no clear vertical trend . in the nearest bins the results are partially consistent , but more negative , than what found by @xcite and ( * ? ? ? * @xmath119 km s@xmath3 kpc@xmath3 ) , although neese & yoss refer to the thin disk . the observations show @xcite that @xmath5 should exponentially decay in the radial direction , with the same scale length of the mass density . this is often considered valid even for @xmath4 , under the assumption of a constant anisotropy , and @xcite found that this should be the best approximation at the solar position . the results also roughly agree with the expectation of this model , as shown in figure [ f_jeans ] . in conclusion , the presented kinematics implies a radial behavior of @xmath4 consistent with both previous observations and with the expectations of theoretical predictions . the presence of kinematical gradients in the galactic thick disk is a powerful diagnostic to discriminate between the various models of its formation @xcite , but unfortunately the model expectations concerning the vertical gradient have not been investigated so far . it would be interesting to compare , in the near future , the predictions of the various models with the observed trend with @xmath0 of the velocity dispersions . on the contrary , the radial gradient of the dispersions has been modeled in the context of thick disk formation through disk heating by the merging of minor satellites @xcite . the observations of @xcite agree with the results of these simulations , provided that a low inclination orbit of the merging satellite is assumed . a similar conclusion is drawn comparing our measurements of @xmath4/@xmath5 with the theoretical expectations of the merging scenario . in fact , the simulations showed that this ratio is strongly linked to the inclination angle of the orbit @xcite , and our result ( @xmath4/@xmath5=2.08 ) favors a small inclination angle , @xmath120030@xmath101 ( see figure 15 of * ? ? ? * ) . a low inclination orbit of the infalling satellite is also required to reproduce the rather large vertical gradient of the rotational velocity , found in this work as in previous investigations @xcite . on the contrary , the radial migration model predicts a much shallower gradient ( @xmath3817 km s@xmath3 kpc@xmath3 , * ? ? ? * ) , incompatible with the observations . in conclusion , the models of thick disk formation have so far provided only fragmentary predictions about its kinematical properties , and our observations can not still be used to fully discriminate between them . however , we find that our results are consistent with the scenario where the thick disk formed through dynamical heating of a pre - existing galactic disk , induced by the merging of a minor satellite . moreover , all the kinematical evidence shows that , if this is the correct model , a low - latitude ( @xmath121 ) merging event is strongly preferred . we have analyzed a sample of @xmath6400 thick disk stars , measuring the variation of their kinematical properties as a function of distance from the galactic plane , from @xmath1=1.5 to 4.5 kpc . our results can be summarized as follows : * while the mean vertical velocity component @xmath56 shows no significant deviation from zero in the whole range , between 1.5 and 3 kpc , we find a net radial motion of about 20 km s@xmath3 directed toward the galactic anticenter . other authors have recently found evidence for a similar behavior , proposing an inward motion of the lsr , although our mean velocity is larger than their proposed value by a factor of two . however , we find that @xmath54 changes sign for @xmath1223 kpc , and that a radial motion of the lsr , although not excluded , can not alone explain this behavior . * the mean rotational velocity of the thick disk decreases with distance from the galactic plane , as found by many previous investigations . the linear fit of our data returns a gradient of @xmath2 km s@xmath3 kpc@xmath3 , our data points and the results of three other previous works are globally better represented by a power - law with index 1.25 , very similar to what has recently been proposed by @xcite . * all the velocity dispersions steadily increase with distance from the galactic plane , closely following a linear relation . the gradients we found are , however , smaller than those proposed by previous works . * while the velocity dispersions increase with @xmath1 , the ratios @xmath123 and @xmath124 show no significant vertical trend . the observations thus indicate a substantial constancy with @xmath1 of the anisotropy . * we find a non - negligible vertex deviation , increasing with @xmath1 from values close to zero to @xmath86 at @xmath1=3.5 kpc . this is consistent with previous investigations , which found a very small vertex deviation of old stellar population close to the galactic plane . * the tilt angle steadily increases with distance from the galactic plane . as expected , the orientation of the velocity ellipsoid in the @xmath7-@xmath20 plane results , at any @xmath1 , intermediate between alignment with the cylindrical and spherical coordinate systems . according to calculations by @xcite , the tilt angle at @xmath1=2 kpc coincides with the expectation of mond , although the extension of their models to higher galactic heights is required to perform a conclusive test of the underlying gravitational law . * the vertical trend of many kinematical quantities show deviations from linearity between 2.5 and 3.5 kpc . the origin of these features is unknown , but it could indicate the presence of a sub - structure at this galactic height , such as a comoving group of stars . * the results are fully consistent with the model of thick disk formation through dynamical heating of a pre - existing galactic disk . if this is the correct scenario , a low inclination angle of the merging event is strongly preferred . however , not all the models proposed so far could be tested by our observations , and more simulations are required to obtain a detailed comparison able to discriminate between them . c.m.b . and r.a.m . acknowledge support from the chilean centro de astrofsica fondap no . 15010003 , and the chilean centro de excelencia en astrofsica y tecnologas afines ( cata ) basal pfb/06 . c.m.b . also thanks f. mauro for useful discussions . all authors acknowledge partial support from the yale university / universidad de chile collaboration . the spm3 catalog was funded in part by grants from the us national science foundation , yale university and the universidad nacional de san juan , argentina . we warmly thank w. f. van altena , v. i. korchagin , t. m. girard , and d. i. casetti - dinescu for their help and suggestions . allende prieto , c. , beers , t. c. , wilhelm , r. , et al . 2006 , apj , 636 , 804 amendt , p. , & cuddeford , p. 1991 , apj , 368 , 79 assmann , p. , fellhauer , m. , kroupa , p. , brns , r. c. , & smith , r. 2011 , mnras , 415 , 1280 beers , t. c. , & sommer - larsen , j. 1995 , apjs , 96 , 175 bekki , k. , & tsujimoto , t. 2011 , apj , 738 , 4 bensby , t. , feltzing , s. , & lundstrm , i. 2003 , a&a , 410 , 527 bensby , t. , zenn , a. r. , oey , m. s. , & feltzing , s. , 2007 , apj , 663 , l13 bienaym , o. 1999 , a&a , 341 , 86 bienaym , o. 2009 , a&a , 500 , 781 bienaym , o. , famaey , b. , wu , x. , zhao , h. s. , & aubert , d. 2009 , a&a , 500 , 801 binney , j. , & merrifield , m. 1998 , galactic astronomy ( princeton , nj : princeton university press ) , 1998 binney , j. , & spergel , d. 1983 , the nearby stars and the stellar luminosity , ed . philip , & u. schenectady ( n.y . : davis press ) , iau col . , 76 , 259 bochanski , j. j. , munn , j. a. , hawley , s. l. , et al . 2007 , aj , 134 , 2418 bond , n. a. , ivezi , z. , sesar , b. , et al . 2010 , apj , 716 , 1 bournaud , f. , elmegreen , b. g. , & martig , m. 2009 , apj , 707 , l1 brown , w. r. , beers , t. c. , wilhelm , r. , et al . 2008 , aj , 135 , 564 cardelli , j. a. , geoffrey , c. c. , & mathis , j. s. 1989 , apj , 345 , 245 carney , b. w. , latham , d. w. , & laird , j. b. 1989 , aj , 97 , 423 carollo , d. , beers , t. c. , chiba , m. , et al . 2010 , apj , 712 , 692 carraro , g. , van altena , w. f. , moni bidin , c. , et al . 2005 , aas , 37 , 1378 casetti - dinescu , d. i. , girard , t. m. , korchagin , v. i. , & van altena , w. f. 2011 , apj , 728 , 7 chiba , m. , & beers , t. c. 2000 , aj , 119 , 2843 cuddeford , p. , & amendt , p. 1992 , mnras , 256 , 166 dalcanton , j. j. , & bernstein , r. a. 2002 , aj , 124 , 1328 dehnen , w. , & binney , j. 1998 , mnras , 298 , 387 duflot , m. , figon , p. , & meyssonnier , n. 1995 , a&as , 114 , 269 evans , l. t. 1970 , mnras , 149 , 311 feltzing , s. , bensby , t. , & lundstrm , i. 2003 , a&a , 397 , l1 fuchs , b. , dettbarn , c. , rix , h .- w . , et al . 2009 , aj , 137 , 4149 fuhrmann , k. 1998 , a&a , 338 , 161 flynn , c. , & freeman , k. c. 1993 , a&as , 97 , 835 gilmore , g. , & reid , n. 1983 , mnras , 202 , 1025 girard , t. m. , dinescu , d. i. , van altena , w. f. , et al . 2004 , aj , 127 , 3060 girard , t. m. , korchagin , v. i. , casetti - dinescu , d. i. , et al . 2006 aj , 132 , 1768 gmez , f. a. , minchev , i. , villalobos , a. , oshea , b. w. , & williams , m. e. k. 2011 ( arxiv:1105.4231 ) gratton , r. g. , bragaglia , a. , carretta , e. , et al . 2003 , a&a , 408 , 529 gratton , r. g. , carretta , e. , matteucci , f. , & sneden , c. 2001 , a&a , 358 , 671 gray , r. o. , & corbally , c. j. 1994 , aj , 107 , 742 hamaker , h. c. 1978 , appl . stat . , 27 , 7 haywood , m. 2001 , mnras , 325 , 1365 hori , l. , & liu , t. 1963 , pasj , 15 , 100 horne , k. 1986 , pasp , 98 , 609 ivezi , z. , sesar , b. , juri , m. , et al . 2008 , apj , 684 , 287 jones , d. h. p. 1972 , apj , 178 , 467 juri , m. , ivezi , z. , brooks , a. , et al . 2008 , apj , 673 , 864 kent , s. , & de zeew , t. 1991 , aj , 102 , 1994 kobayashi , c. , & nakasato , n. 2011 , apj , 729 , 16 kuijken , k. , & gilmore , g. 1989 , mnras , 239 , 571 kuijken , k. , & tremaine , s. 1991 , in dynamics of disc galaxies , varberg castle , sweden , 71 loebman , s. r. , roskar , r. , debattista , v. p. , et al . 2011 , apj , 737 , 8 lutz , t. e. , & hanson , r. b. 1992 , asp conf . series , 25 , 257 lynden - bell , d. 1962 , mnras , 124 , 95 majewski , s. r. 1992 , apjs , 78 , 87 majewski , s. r. 1993 , ara&a , 31 , 575 majewski , s. r. , kunkel , w. e. , law , d. r. , et al . 2004 , aj , 128 , 245 milgrom , m. 1983 , apj , 270 , 371 monaco , l. , villanova , s. , moni bidin , c. , et al . 2011 , a&a , 529 , a90 moni bidin , c. , moehler , s. , piotto , g. , et al . 2006 , a&a , 451 , 499 moni bidin , c. , girard , t. m. , carraro , g. , et al . 2009 , rmxac , 35 , 109 moni bidin , c. 2009 , phd thesis , univ . de chile , 1 moni bidin , c. , carraro , g. , mndez , r. a. , & van altena , w. f. 2010 , apj , 724 , l122 neese , c. l. , & yoss , k. m. 1988 , aj , 95 , 463 ollongren , a. 1962 , baas , 16 , 561 purcell , c. w. , bullock , j. s. , & kaplinghat , m. 2009 , apj , 703 , 2275 quillen , a. c. , doughery , j. , bagley , m. b. , minchev , i. , & comparetta , j. 2010 , mnras , submitted ( arxiv:1010.5745 ) quinn , p. j. , hernquist , l. , & fullagar , d. p. 1993 , apj , 403 , 74 reddy , b. e. , lambert , d. l. , & allende prieto , c. 2006 , mnras , 367 , 1329 rocha - pinto , h. j. , maciel , w. j. , scalo , j. & flynn , c. 2000 , a&a , 358 , 850 roskar , r. , debattista , v. p. , stinson , g. s. , et al . 2008 , apj , 675 , l65 rybka , s. p. , & yatsenko , a. i. 2010 , kpcb , 25 , 309 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , apj , 500 , 525 schnrich , r. , & binney , j. 2009 , mnras , 396 , 203 schnrich , r. , binney , j. , & dehnen , w. 2010 , mnras , 403 , 1829 seth , a. c. , dalcanton , j. j. , & de jong , r. s. 2005 , aj , 130 , 1574 shapiro , k. , gerssen , j. , & van der marel , r. 2003 , apj , 126 , 2707 siebert , a. , bienaym , o. , binney , j. , et al . 2008 , mnras , 391 , 793 soubiran , c. , bienaym , o. , & siebert , a. 2003 , a&a , 398 , 141 spagna , a. , lattanzi , m. g. , re fiorentin , p. , & smart , r. l. 2010 , a&a , 510 , 4 skrutskie , m. f. , cutri , r. m. , stiening , r. , et al . 2006 , aj , 131 , 1163 smith , m. c. , evans , n. w. , belokurov , v. et al . 2009 , mnras , 399 , 1223 tonry , j. & davis , m. 1979 , aj , 84 , 1511 vallenari , a. , pasetto , s. , bertelli , g. , et al . 2006 , a&a , 451 , 125 van der kruit , p. c. , & searle , l. 1981 , a&a , 95 , 105 van der kruit , p. c. , & searle , l. 1982 , a&a , 110 , 61 veltz , l. , bienaym , o. , freeman , k. c. , et al . 2008 , a&a , 2008 , 480 , 753 villalobos , a. , & helmi , a. 2008 , mnras , 391 , 1806 villalobos , a. , & helmi , a. 2009 , mnras , 399 , 166 villalobos , a. , kazantzidis , s. , & helmi , a. 2010 , apj , 718 , 314 wyse , r. f. g. & gilmore , g. 2005 , astro - ph/0510025 yi , s. k. , kim , y .- c . , & demarque , p. 2003 , apjs , 144 , 259 yoss , k. m. , neese , c. l. , & hartkopf , w. i. 1987 , aj , 94 , 1600 yoss , k. m. , & griffin , r. f. 1997 , japa , 18 , 161 walker , i. r. , mihos , j. c. , & hernquist , l. 1996 , apj , 460 , 121 zinn , r. 1985 , apj , 293 , 424 zwitter , t. , siebert , a. , munari , u. , et al . 2008 , aj , 136 , 421
transformations between different solid states of a material are essential for many technological and scientific applications , and understanding their kinetics is therefore not only interesting because of the scientific variety and beauty , but also important for tailoring of new materials with specific properties . often , these processes are accompanied by elastic deformations , which can be e.g. due to density differences or structural changes , which provoke stresses at interfaces between adjacent phases . the influence on the thermodynamics of the transitions between different phases has been thoroughly discussed in the literature @xcite , whereas the kinetics of these processes are still far less understood . facing this lack of understanding , we would like to mention that very fruitful concepts have been developed to describe the dynamics of solidification or melting processes as moving boundary problems . it means that the evolution of the different phases is described on the very detailed level of the motion of the interfaces between them , for which suitable equations of motion have to be posed . a particular application is the famous example of dendritic growth , where heat or component diffusion is the rate limiting process . at the interface between the solid and melt phase the temperature or concentration is close to its phase equilibrium value . the concentration jump or the release of latent heat due to partitioning leads to a jump in the corresponding fluxes at the interfaces , and the magnitude of this jump is directly related to the local interface velocity . altogether , this provides a closed description for the motion of each interface point , and many efforts have been done towards analytical and numerical solutions of this problem . it turns out that the questions concerning existence , shape and growth velocity of steady state patterns in solidification crucially depend on selection mechanisms . in this spirit the influence of many different physical effects on solidification or melting processes has been studied throughout the years @xcite . for example , the effect of isotropic surface tension was proven not to serve as a selection mechanism for a solution , whereas anisotropic surface tension leads to a unique solution . also , the appearance of triple junctions can provide a selection mechanism , and can therefore lead to a solution of the free boundary problem @xcite . of course , the situation considering solid - solid transformation kinetics differs significantly from the solidification or melting problem , where elastic effects often play only a minor role . still , we follow here the same successful concepts to formulate a free boundary description for the transformation kinetics in solids , now including the important effect of elasticity . in real systems many other effects apart from elastic effects come into play like inhomogeneous compositions , the mullins - sekerka instability , crystal anisotropies , polycrystalline structures , etc . and also quite successful attempts have been made to model this complex behavior ( see for example in @xcite and references therein ) . here , in contrast , we focus on gaining a fundamental understanding of the elastic influence on pattern formation processes , and we restrict our investigations therefore to simple and clean " model systems . correspondingly , we consider only a single diffusion field , namely the thermal diffusion in pure materials . the formulation of such a model system requires to write down explicit expressions for the boundary conditions and suitable equations of motion at the propagating interfaces , which therefore have to be tracked during the entire dynamical process . finally , it provides a detailed description of the microstructure evolution on the level of individual grains in a polycrystalline structure . as for the dendritic growth , we focus here on the regime of diffusion limited growth . then , propagation of a front generates or consumes latent heat , which has to diffuse through the system , and we assume that this is the rate - limiting process . this is in contrast to isothermal solid - solid transformations that were discussed in @xcite , which are driven by local interface kinetics , therefore allowing fast front propagation also on the scale of the sound speed . many important properties result from the influence of elastic effects due to structural transformations and density differences . apart from sharp interface methods , that are based on green s function techniques , we also use phase field descriptions , which are nowadays widely used . more importantly , however , is that we supplement these numerical investigations by analytical calculations which allow to understand these processes and the relevant parameters on a deeper basis . specifically , we consider the growth of an equilibrium solid phase @xmath0 from another metastable solid phase @xmath1 , as depicted in fig . [ fig : geometry ] . at a temperature @xmath2 , both stress - free phases are in bulk equilibrium with each other , whereas below this value the new @xmath0 phase is thermodynamically favorable . far ahead of the propagating front , the system is therefore exposed to a lower temperature @xmath3 , and we define a dimensionless temperature @xmath4 , where @xmath5 denotes the latent heat of the transition and @xmath6 the heat capacity . within the solid phases the evolution of the temperature field obeys a diffusion equation , @xmath7 where @xmath8 is the thermal diffusivity , which we assume to be equal in both phases , as well as the heat capacity ( symmetrical model ) . the motion of the interfaces follows from energy conservation principles , which account here for the net heat flux and the generation or absorption of latent heat due to the local motion of the interface with normal velocity @xmath9 , @xmath10}-\nabla w_{int}^{[\alpha ] } ) , \label{eq : heat - conservation - equation}\end{aligned}\ ] ] where @xmath11 is the interface normal and the superscripts denote the phases . elastic effects enter into the problem via a modification of the local phase equilibrium condition at the moving interfaces . in contrast to previous investigations @xcite , we now assume the phases to be coherently connected , i.e. we neglect the appearance of any slips , detachments or defect formation at the interfaces , which is reasonable for small misfits . it is important to mention here , that with the assumption of coherency we formulate a completely different model being capable to describe qualitatively different types of elastic effects . for example the well - known effect of elastic hysteresis @xcite , which was proven to disappear without coherency @xcite , is now included in the model . in a lagrangian description of linear elasticity , the displacement field is denoted by @xmath12 , and the coherency condition reads , @xmath13}=\mathbf{u}^{[\beta]}$ ] at the interface . the strain field is then given by the symmetrical spatial derivative , @xmath14 . this implies that tangential and shear strains @xmath15 are continuous at the interface with @xmath16 being the two tangential directions . force balance demands the continuity of the normal and shear stresses , @xmath17 , where @xmath18 is the normal direction . here , the stresses are defined as the derivative of the free energy with respect to the strains , @xmath19 . , of a thermodynamically favored phase @xmath0 into a metastable phase @xmath1 , which is initially at temperature @xmath20 . the motion of the interface is locally described by the interface - normal velocity @xmath21 , which is given by the local heat conservation eq . ( [ eq : heat - conservation - equation ] ) . the dashed lines at the two side borders indicate that we will discuss two different types of outer boundary conditions : free growth , where the strains decay for @xmath22 , and channel growth , with fixed displacements and thermal insulation at the walls . in the growth direction @xmath23 the system is assumed to be infinite in both cases . [ fig : geometry],scaledwidth=55.0% ] the interface temperature for a planar , stress free interface is given by @xmath24 , with @xmath25 . capillary corrections induce a curvature dependent term , and for simplicity we discuss only the case of isotropic surface tension @xmath26 . finally , elastic deformations lead to an additional change of the interface temperature . altogether , this leads to the local equilibrium boundary condition @xcite @xmath27 with @xmath28 being the capillary length and @xmath29 the local curvature of the interface , counted positive for a convex phase @xmath0 ( see fig . [ fig : geometry ] ) ; notice that due to the coherency constraint @xcite , the elastic shift of the equilibrium temperature is proportional to the difference of a new potential @xmath30 } - \tilde{f}_{el}^{[\beta]}$ ] , which is related to the elastic free energy @xmath31 via the legendre transformation @xmath32 . here , we use the sum convention for repeated indices for the sake of brevity . from now on , we restrict our considerations to linear isotropic elasticity . taking the relaxed state of the @xmath1 phase as reference , the elastic contribution to the free energy there reads @xmath33 } & = \frac{e}{2(1+\nu)}\left(\frac{\nu}{1 - 2\nu}\epsilon_{ii}^{2}+\epsilon_{ik}^{2}\right),\label{eq : free - energy - reference}\end{aligned}\ ] ] where @xmath34 and @xmath35 are young s elastic modulus and poisson ratio , respectively . in contrast , the elastic contribution to the free energy density of the new phase @xmath0 reads @xmath36 } & = \frac{e}{2(1+\nu)}\left(\frac{\nu}{1 - 2\nu}\left(\epsilon_{ii}-\epsilon_{ii}^{0}\right)^{2}+\left(\epsilon_{ik}-\epsilon_{ik}^{0}\right)^{2}\right),\label{eq : free - energy - new}\end{aligned}\ ] ] which has an other state of zero elastic energy due to the lattice strain @xmath37 assigned to the phase transformation ( see for example @xcite ) . in the following we will call @xmath37 the eigenstrain of the transformation . particular cases of eigenstrains and their interpretation will be discussed in the next section . the mechanical equilibrium and coherency conditions provide expressions for the discontinuous jumps of the strains at the interface @xmath38}-\epsilon_{nn}^{[\alpha]}= & \epsilon_{nn}^{0}+\frac{\nu}{1-\nu}(\epsilon_{ss}^{0}+\epsilon_{\tau\tau}^{0}),\label{eq : strain - jump - n}\\ \epsilon_{n\tau}^{[\beta]}-\epsilon_{n\tau}^{[\alpha]}= & \epsilon_{n\tau}^{0},\quad\epsilon_{ns}^{[\beta]}-\epsilon_{ns}^{[\alpha]}=\epsilon_{ns}^{0},\label{eq : strain - jump - tau / s}\end{aligned}\ ] ] which are only related to the imposed eigenstrain . defining an eigenstress tensor via hooke s law @xmath39 we obtain after a few straightforward algebraic manipulations for the elastic contribution to the local equilibrium condition eq . ( [ eq : phase - equilibrium - condition ] ) @xmath40 } - \frac{e}{2(1-\nu^{2})}\left((\epsilon_{\tau\tau}^{0})^{2}+(\epsilon_{ss}^{0})^{2}+2\nu\epsilon_{\tau\tau}^{0}\epsilon_{ss}^{0}+2(1-\nu)(\epsilon_{\tau s}^{0})^2\right ) . \label{eq : phase - equilibrium - elast - cont}\end{aligned}\ ] ] apparently , this expression depends on the strain state of the @xmath1 phase at the interface . as already implicitly assumed above , the elastic degrees of freedom relax fast on diffusive timescales , and therefore the application of static elasticity is legitimate . therefore , newton s second law becomes @xmath41 the above set of equations now has to be supplemented by boundary conditions at the external system boundaries , and we will discuss different scenarios below . they complete the self - sonsistent description of the moving boundary problem , which require the simultaneous solution of the elastic and diffusion equations , coupled via the boundary conditions at the propagating interfaces , leading to a complicated nonlinear and nonlocal problem . we note that we will later also discuss more complicated situations with more than one growing phase . then , also similar boundary conditions and equations have to be set up for the interfaces between them . typically , then also triple junctions will appear , where the contact angles are given by young s law . the simplest case of a characteristic lattice strain is to assume the bond length of the new phase @xmath0 to be uniformly longer or shorter in all directions in comparison to the reference phase , i.e. @xmath42 , which implies a density change . notice , that this kind of transformations , to which we will refer to as _ dilatational eigenstrain _ , only changes the volume of an elementary cell , and not its relative dimensions or its scaled shape . the opposite case are transformations , where the volume of the unit cell is conserved and only the shape is changed . these _ shear transformations _ are characterized by a traceless eigenstrain tensor . a particular transition involving shear strain can occur e.g. in hexagonal crystals , where the transition leads to a lowering of the symmetry from @xmath43 to @xmath44 . this is the case for example in hexagonal - orthorombic transitions in ferroelastics ( see @xcite and references therein ) . let the principal axis @xmath45 be oriented in @xmath46 direction . for simplicity we neglect all other possible strains with higher ( axial ) symmetry . by proper choice of the crystal orientation around the main axis in the initial phase , we obtain the new phase in three possible states in the @xmath47 plane due to the original hexagonal symmetry ( see fig . [ fig : hexagonaltoothorombic ] ) . visualisation of the hexagonal to orthorhombic transformation . from the symmetry its obvious , that the orientation angle @xmath48 of the new orthorhombic phase can have only the three different values @xmath49.,scaledwidth=60.0% ] then the nonvanishing components of the eigenstrain tensor @xmath50 are @xmath51 and @xmath52 , where the orientation angle @xmath48 is either @xmath53 or @xmath54 . we also consider a more generic mixed mode case , where the self - strain of the transformation is a linear combination of the above two different transformations . we will discuss the following two mixing cases @xmath55 and refer to them as positive or negative mixing . it is important to mention that switching from positive to negative mixing corresponds to a rotation of the eigenstrain tensor by an angle @xmath56 . since our model is isotropic in all respects except the choice of the eigenstrain tensor , the only predefined direction is given by the growth velocity therefore , we can interpret the negative mixing case to represent physically the same system as the positive mixing case but with a growth direction perpendicular to it . as a first application , we discuss the growth of the stable @xmath0 phase from the metastable @xmath1 phase in an infinitely large system , which is exposed to the temperature @xmath3 far away from the interface . for simplicity , we discuss an effectively two - dimensional infinite system , assuming translational invariance in @xmath46 direction . in particular , we assume a plane strain situation from point of view of elasticity . also , we choose the elastic boundary conditions such that all stresses decay far away from the interface . seeking for steady state solutions in free space , we use green s function methods for solving both the elastic and the diffusional problem , to finally obtain self - consistently the shape and the velocity of the growing front . while the derivation of the integral representation of the thermal field is well - known , see e.g. @xcite , we present the calculation of @xmath57 in more detail @xcite . to this end , we express the elastic problem using the eigenstrain - independent stress tensor @xmath58 ; for the @xmath1 phase it is the same as the nomimal stress , @xmath59}=\sigma_{ik}^{[\alpha]}$ ] , whereas in the @xmath0 phase it is @xmath60}=\sigma_{ik}^{[\beta]}+\sigma_{ik}^{0}$ ] . in terms of the new stress tensor @xmath58 , the mechanical equilibrium equation @xmath61 introduces a force density which is localized on the interface and vanishes in the bulk , @xmath62}-\tilde{\sigma}_{in}^{[\alpha]}=\sigma_{in}^{0}. \label{eq : artificial - interface - forces}\end{aligned}\ ] ] this means that the problem of two coherently connected solids with different reference state is equivalent to a monolithic material with a distribution of point forces along the interface . we can therefore use an integral representation of the displacement field , @xmath63 where the integration is performed along the interface , and the free - space green s tensor @xmath64 of the elastic problem is given by @xcite @xmath65 this function is the displacement response at @xmath66 to a point force located in the origin . as a result , the mismatch provokes long - ranged elastic deformations , and therefore induces a nonlocal character to the problem . the above expressions can then be used to calculate the elastic influence on the interface temperature @xmath67 . far behind the tip the interface profile becomes parabolic and is described by the asymptotic ivantsov parabola @xcite . in this region , the elastic strains have decayed , and only the constant local contribution to @xmath68 ( second term in eq . [ eq : phase - equilibrium - elast - cont ] ) remains . consequently , the effective driving force @xmath69 consists not only of the thermal undercooling @xmath70 , which is controlled by the applied temperature @xmath3 , but also a contribution @xmath71 from the elasticity , @xmath72}{2(1-\nu^{2})l^2}. \label{eq : definition - deltael}\end{aligned}\ ] ] we can consider @xmath71 as a parameter quantifying the strength of the elastic effects . the elastic effects therefore induce a shift of the equilibrium temperature , and the effective undercooling @xmath69 is lower than the actually applied thermal driving force , @xmath73 which reflects the elastic hysteresis @xcite . by elimination of the thermal field @xcite we can derive the steady state equation for the shape of the solid - solid interface in a closed dimensionless representation , which reads in the co - moving frame of reference @xcite , @xmath74k_{0}(p\mu(x , x ' ) ) . \label{mainpartclosedequation}\end{aligned}\ ] ] here , @xmath75 denotes the peclet number ( @xmath76 is the steady state velocity ) , and it is related to the driving force by the implicit relation @xmath77 . furthermore , @xmath78 is the modified bessel function of third kind in zeroth order , and @xmath79^{1/2}$ ] . in the above expression we rescaled all lengths by the radius of curvature @xmath80 of the asymptotic ivantsov parabola . consequently , the central equation ( [ mainpartclosedequation ] ) , together with the expression for the elastic contribution ( [ eq : phase - equilibrium - elast - cont ] ) , where the strains are expressed via the integral representation ( [ formal_solution_displacements ] ) results in a complicated integro - differential equation for the interface shape . it differs from conventional dendritic growth by the appearance of the elastic term . to better understand its influence , we briefly review the solvability conditions for classical dendritic growth @xcite , for situations with pure dilatational and shear strain @xcite , and cases with a triple junction @xcite : in classical dendritic growth , any solution for isotropic surface tension with nonzero solvability parameter @xmath81 must exhibit a finite negative tip kink angle @xmath82 . for this reason , dendritic growth is not possible for isotropic surface energy , since no smooth dendrite profiles exist . a dilatational eigenstrain simply shifts the equilibrium interface temperature , and therefore suffers from the same missing selection as the dendrite problem without surface tension anisotropy . the presence of a shear eigenstrain , which will also be discussed in more detail below , leads to the propagation of a bicrystal . in this case , the elastic effects operate as new selection mechanism . independent of this elastically induced selection mechanisms , the presence of a triple junction alone also works as a selection mechanism . we solve the eigenvalue problem eq . ( [ mainpartclosedequation ] ) numerically for the steady state shape and the stability parameter @xmath83 in the spirit of ref . @xcite , and the results are presented in section [ results ] . the growth of the @xmath0 phase in a channel of finite width @xmath84 differs significantly from the free growth scenario discussed above . the specific behavior depends of course on the boundary conditions , and we assume fixed grip conditions on the long sides of the channel , i.e. the displacement is constant . phases growing with constant velocity @xmath76 into the metastable @xmath1 phase in a finite channel of width @xmath84 . the shapes in @xmath85 and @xmath86 correspond to the dilatational case ( @xmath87 ) , with @xmath88 and an undercooling , which is chosen such that according to eq . ( [ eq : fraction - of - beta - phase ] ) it would correspond to an asymptotic phase fraction @xmath89 . while @xmath85 shows the shape of the symmetrical finger , @xmath86 shows the shape of the sightly faster parity broken finger ( see fig . [ fig : diagonaleigenstrain]a ) . in panel @xmath90 we have plotted the shape in the case of negative mixing eigenstrain , @xmath91 , which in that case corresponds to growth in the privileged direction . the parameters are @xmath92 and @xmath93 ( for the resulting velocity see fig . [ fig : diagonaleigenstrain]b ) . , scaledwidth=65.0% ] again , we discuss an effectively two - dimensional system by the assumption of translational invariance in @xmath46 direction ( see fig . [ dilatationalsetup ] for the setup ) . furthermore , we assume that the @xmath1 phase far in front of the growing tip is fully relaxed , which is in agreement with the boundary conditions @xmath94 and @xmath95 along the boundaries of the whole , infinitely long strip . from a thermal point of view , we assume thermal insulation along these boundaries , so no temperature flux through the surfaces , @xmath96 on the outer boundaries . far ahead of the tip , the temperature is given by the dimensionless undercooling @xmath70 . due to the phase transition , the release of the latent heat leads to a heating of the system , therefore the temperature far behind the tip @xmath97 is higher than in front . under the given circumstances it is quite obvious that in this tail region a homogeneous state is reached . in particular , since the interfaces do not move there , the temperature becomes indeed constant . for simplicity , we discuss here only the specific case of a purely diagonal eigenstrain , i.e. @xmath98 for @xmath99 , to demonstrate the analytical procedure . in contrast , in the case of an eigenstrain tensor with nonzero off - diagonal elements , bicrystal patterns emerge , and we will briefly discuss this situation below ; it is quite obvious that the analysis can be performed in an analogous way . due to the confinement of the system , phase coexistence is possible in a finite range of undercoolings , and we can calculate the asymptotic volume fraction @xmath100 of the new phase ( see fig . [ dilatationalsetup ] ) . notice , that in contrast to the isothermal situation discussed in @xcite , the temperature in the tail region is not a control parameter but has to be found self - consistently , due to the thermally insulating boundaries . in terms of the asymptotic phase fraction @xmath100 , the width of the @xmath0 phase in the tail region is written as @xmath101 . mechanical compatibility requires @xmath102}+{\lambda}\epsilon_{xx}^{[\beta ] } & = 0,\label{eq : mixedmode - coherency - cond_yy}\end{aligned}\ ] ] since the average lateral strain must vanish due to the boundary conditions for the displacement . together with the mechanical equilibrium condition of force balance , we can then calculate the elastic fields in the tail region , and consequently the elastic energy . then , we calculate the energy excess @xmath103 , due to the appearance of a finite fraction of the new phase @xmath100 . it is defined as the difference between the energy for @xmath104 without any phase @xmath0 and the energy with an arbitrary but finite fraction , @xmath105}+(1-{\lambda})f_{el}^{[\alpha]}\right)-2\gamma+w\int\limits _ { 0}^{\lambda}\left(\frac{l(t_{eq}-t(\lambda'))}{t_{eq}}\right)d\lambda'.\end{aligned}\ ] ] initially , there is only the elastically relaxed but metastable phase @xmath1 at the constant temperature @xmath70 . as soon as we have a finite phase fraction @xmath100 , the energy increases by elastic and capillary contributions . finally , the integral appears due to thermal insulation , since an increase of the amount of the phase @xmath0 by phase transformation processes generates latent heat , and therefore causes an increase of the temperature . using the heat conservation condition we find the relation between this temperature and the phase fraction to be @xmath106 therefore , we obtain for the energy excess as a function of the asymptotic phase fraction @xmath100 @xmath107 where the parameter @xmath71 , as already introduced above in eq . ( [ eq : definition - deltael ] ) , defines the strength of elastic effects , and @xmath108 is a parameter depending on the type of eigenstrain . the maximum of this energy excess , @xmath109 , provides the equilibrium fraction , which expresses the thermodynamic balance between elastic deformation and free energy release due to the phase transition , @xmath110 if the fraction @xmath111 is in the range @xmath112 , we obtain phase coexistence in the asymptotic regime far behind the tip . we would like to point out that one can derive this equation also by using the phase equilibrium condition eq . ( [ eq : phase - equilibrium - condition ] ) . then growth of the @xmath0 phase demands that the energy excess ( [ abc123 ] ) for the equilibrium fraction ( [ eq : fraction - of - beta - phase ] ) has to be positive , @xmath113 this is equivalent to the condition @xmath114 , where @xmath115 is given by @xmath116 we note that for the shear transformations discussed above , the most favorable process is the growth of a bicrystal , where two twinned phases @xmath0 and @xmath117 with @xmath118 grow together , because this is energetically most favorable @xcite . for the bicrystal configuration , the analysis can be performed in a similar way , and then also the @xmath119 grain boundary energy appears in the energy balance analogous to eq . ( [ qwert ] ) . considering the channel geometry , we use a phase field formulation to solve the full free boundary problem numerically . this method is very versatile and allows to study even more complicated scenarios , e.g. the full dynamical evolution and not only the steady state regime , three - dimensionsal situations with full anisotropy of elasticity and surface tension or simulations beyond the symmetrical model , where the two phases have different diffusion constants . nevertheless , for brevity of presentation , we confine our investigations to the simple , isotropic , two - dimensional and thermally symmetric case that we also used for the green s function approach in the free space geometry and the above analytical calculations . the phase field method introduces an additional numerical lengthscale @xmath120 , the width of the transition region between the different solid phases . this means that an order parameter , which discriminates between the different phases , has a profile @xmath121/2 $ ] in normal direction through an interface . we operate here in the sharp interface limit , which means that this lengthscale is significantly smaller than all other physical lengthscales present in the problem . then the description recovers the desired sharp interface equations that were proposed above . hence , we introduce the phase field @xmath82 having a value @xmath122 in the metastable initial phase , and a value @xmath123 in the new phase . we start from a free energy functional @xmath124 & = \int\limits _ { v}\left(f_{s}+f_{dw}+f\right)dv,\label{eq : phasefieldmodel - free - energy - functional}\end{aligned}\ ] ] where @xmath125 is the gradient energy density and @xmath126 is the double well potential , guaranteeing that the free energy functional has two local minima at @xmath123 and @xmath122 corresponding to the two distinct phases of the system . the form of the double well potential and the gradient energy density are chosen such that the phase field parameter @xmath120 defines the interface width and the parameter @xmath26 corresponds to the interface energy of the sharp interface description @xcite . the free energy density @xmath127 of our phase field description is chosen to be an interpolation between the two distinct free energies of the model , @xmath128}_{el}h(\phi)+f^{[\beta]}_{el}(1-h(\phi ) ) + l\frac{t - t_{eq}}{t_{eq } } ( 1-h(\phi)),\label{eq : phasefieldmodel - free - energy - interpolation}\end{aligned}\ ] ] where we choose the interpolation function @xmath129 to be @xmath130 . it is the simplest polynom satisfying the necessary interpolation conditions @xmath131 and @xmath132 , and having a vanishing slope at @xmath123 and @xmath122 , in order not to shift the bulk states . the evolution equation of the phase field is given by the variational expression @xmath133 for large values of the mobility @xmath134 we recover the desired case of diffusion limited growth . the elastodynamic equations read @xmath135 which recover static elasticity for slowly propagating interfaces in comparison to the sound speed @xmath136 , with @xmath137 being the mass density . for the temperature field we have the usual thermal diffusion equation with the motion of the phase field or interface being a source of latent heat , @xmath138 where the prime denotes the derivative with respect to @xmath82 . the phase field model presented here is very similar to the model in @xcite . however , there only the very special case of melting or crystallization of a confined sphere with elastic effects was discussed . for the simulations we typically use parameters @xmath139 and @xmath140 . apart from the theoretical investigations above , we also performed numerical simulations for both the free growth in an infinite system and in a channel or narrow strip . as already mentioned before , we observe the growth of a single new phase if the eigenstrain tensor is purely diagonal , and twin structures for shear transformations with @xmath118 . also , we consider mixed mode situations which are between these two extreme cases . we first consider the situation of free growth . here , the simplest case is that of a dilatational eigenstrain where all diagonal elements have the same value , i.e. @xmath141 ; this case is obtained from eq . ( [ mixedmodeeigenstraindefinition ] ) by setting @xmath142 . as already mentioned before , selection of a steady solution in the absence of elastic effects is not possible for isotropic surface tension . it is quite remarkable that also the inclusion of elastic effects due to the dilatational eigenstrain of the transition does not change this situation ; the reason is that the elastic effects can be mapped to a change of the undercooling only , and therefore we recover again the missing selection of conventional free dendritic growth @xcite . it turns out that also for the volume - preserving shear strain transitions with @xmath143 , where all other components @xmath144 vanish , i.e. @xmath145 and @xmath53 , steady state solution do not exist either . ( see eq . ( [ eq : stabilityparameterdefinition ] ) ) as function of the mixing parameter @xmath146 for the privileged growth direction with an eigenstrain tensor @xmath147 when @xmath148 . on the right panel we show the stability parameter @xmath149 for two different values of mixing @xmath146.,title="fig:",scaledwidth=45.0% ] ( see eq . ( [ eq : stabilityparameterdefinition ] ) ) as function of the mixing parameter @xmath146 for the privileged growth direction with an eigenstrain tensor @xmath147 when @xmath148 . on the right panel we show the stability parameter @xmath149 for two different values of mixing @xmath146.,title="fig:",scaledwidth=45.0% ] under these circumstances it is very suprising , that for values of the mixing parameter in between zero and one @xmath150 , @xmath53 we do find steady state solutions for free growth and negative mixing . this is shown in fig . [ mixedmoderesultsinglecrystal2]a , where the selected `` stability parameter '' @xmath151 , which is a dimensionless measure for the growth velocity , is plotted as function of the mixing parameter @xmath146 . in general , the eigenvalue @xmath152 depends on @xmath153 and @xmath154 , but in the regime of small peclet numbers the elastic driving force and @xmath154 appear only in the combination @xmath155 . [ mixedmoderesultsinglecrystal2]b shows the dependence of the selected stability parameter on @xmath155 for two different values of @xmath146 . for the other type of mixing @xmath156 corresponding to the unprivileged direction of growth , we do not find steady state solutions in free space . we note that from the selected stability parameter @xmath157 one can extract the steady state growth velocity @xmath76 via the relation @xmath158 , where we choose a dilatational eigenstrain @xmath159 . for small driving forces we obtain velocities and shapes corresponding to a symmetric solution ( see fig . [ dilatationalsetup]@xmath160 ) . for higher driving forces a first order transition towards asymmetrical solutions takes place ( see parity broken finger morphology in fig . [ dilatationalsetup]@xmath161 ) . in the right panel the comparison of growth velocities between phase field and the free space calculations is shown , for the case of negative mixing corresponding to the privileged growth direction @xmath162 . the agreement is reasonably good , especially for driving forces @xmath100 slightly above the critical phase fraction @xmath163 of zero growth velocity.,title="fig:",scaledwidth=45.0% ] , where we choose a dilatational eigenstrain @xmath159 . for small driving forces we obtain velocities and shapes corresponding to a symmetric solution ( see fig . [ dilatationalsetup]@xmath160 ) . for higher driving forces a first order transition towards asymmetrical solutions takes place ( see parity broken finger morphology in fig . [ dilatationalsetup]@xmath161 ) . in the right panel the comparison of growth velocities between phase field and the free space calculations is shown , for the case of negative mixing corresponding to the privileged growth direction @xmath162 . the agreement is reasonably good , especially for driving forces @xmath100 slightly above the critical phase fraction @xmath163 of zero growth velocity.,title="fig:",scaledwidth=45.0% ] in contrast , for the growth in a channel , steady state solutions do also exist for the case of a dilatational eigenstrain , @xmath87 ( see fig . [ fig : diagonaleigenstrain]a ) . as expressed through eqs . ( [ eq : free - energy - gain ] ) and ( [ eq : delta - equilibrium ] ) , the phase @xmath0 can not grow for phase fractions @xmath100 below the threshold @xmath163 . beyond this value the growth of a symmetrical finger ( dendrite ) is the most favorable solution . we note that in situations where free growth solutions do not exist , the still existing solutions in the channel must have a scale that is determined by the size @xmath84 of the channel . for higher driving forces an asymmetric finger grows with higher velocities , and is therefore more likely to be observed ; typical finger shapes are shown in fig . [ dilatationalsetup ] . for the situation of @xmath164 , where steady state solutions exist also in the free space , we performed a quantitative comparison of the results of the two methods in fig . [ fig : diagonaleigenstrain]b . for this comparison we directly relate the fraction @xmath165 to the driving force @xmath69 for free growth via eq . ( [ eq : fraction - of - beta - phase ] ) and eq . ( [ eq : shifted - undercooling ] ) . then , in the limit of small peclet numbers , the latter is related to @xmath69 according to @xmath166 . finally , via the calculated function @xmath149 and eq . ( [ eq : stabilityparameterdefinition ] ) we obtain the growth velocity as a function of @xmath100 , which is then compared to the phase field results . the results for the growth velocity agree well in a range of intermediate fractions of the new phase @xmath0 . in particular , we confirmed that the steady state solutions which were obtained by the green s function method are indeed stable , as the dynamical phase field method tracks only these stable branches . for low phase fractions , growth is not possible below the threshold @xmath163 , which tends to zero for infinitely wide channels . for very high driving forces , the growing finger is influenced by the geometrical confinement of the channel width and also affected by kinetic corrections to the sharp interface equations of motion , which therefore lead to a deviation from the free growth result which was obtained by the green s function method . it is interesting to compare the situation to a case without elasticity , where the problem reduces to classical growth in a channel @xcite . it is known that there even in wide channels symmetric solutions do not exist below a fraction @xmath167 , whereas we found that the eigenstrain leads to the existence of solutions at already significantly lower driving forces @xmath100 , see fig . [ fig : diagonaleigenstrain]a . shows the single tip solution and @xmath86 the twinned finger morphology , which corresponds to the most favorable solution branch for higher driving forces . the positive mixing chosen for @xmath85 and @xmath86 leads to growth along the privileged direction and the parameters are @xmath168 . the panel @xmath90 shows the shape of the slow growth direction for @xmath169 . [ bicrystalshapes],scaledwidth=65.0% ] in contrast to the cases studied above , the presence of off - diagonal eigenstrain entries lead to the preferential growth of bicrystal patterns . this behavior can easily be understood intuitively , since the eigenstrain in the @xmath0 phase favors the presence of off - diagonal strain . then the boundary conditions of fixed displacements require that the surrounding @xmath1 phase is sheared in the opposite direction . this implies a finite elastic energy which is stored far behind the tip . if , on the other hand , a second finger @xmath117 of the same size with an off - diagonal eigenstrain with opposite sign appears , the elastic energy contribution that is associated with the off - diagonal elements of the strain tensor vanishes . consequently , the total elastic energy in the tail region is reduced , and the growth of the bicrystal therefore thermodynamically favored . an additional energy contribution emerges from this bicrystal configuration , since the boundary between @xmath0 and @xmath117 generates a grain boundary energy @xmath170 . we consider in particular the case of vanishingly small twin boundary energy , @xmath171 . this affects also the geometrical situation at the tip , which is now a triple junction , and the contact angle @xmath82 becomes zero , so the overall finger has a smooth contour . typical shapes of such bicrystal pattern are shown in fig . [ bicrystalshapes ] , which are here related to the growth in a channel , but they look similar for free growth . we note that we simulated only half of the channel using the symmetry of the growing structures @xcite . : comparison of growth velocities obtained from the boundary integral technique for the free space geometry @xcite with the velocities from phase field simulations for the channel geometry ( @xmath172 ) . [ fig : bicrystalpureshearcomp],scaledwidth=60.0% ] considering the free space geometry , the first important observation is that in contrast to the single crystal scenario , steady state solution do exist also for pure shear , @xmath145 , which is already discussed in @xcite . new is here that we found the same morphology also in the channel geometry , showing surprisingly good agreement in the comparison of growth velocities , done in the same way as was done for the single crystal described above ( see fig . [ fig : bicrystalpureshearcomp ] ) . again , we would like to point out that the stability of the bicrystal solution obtained from the green s function technique is confirmed by the phase field simulations . we also did investigations of the bicrystal free growth considering mixed eigenstrains , and the results are summarized in fig . [ bicrystalsigmaofeta ] and [ bicrystalpositiveandnegativedilatinshearsigmadeltael ] . in the first graph , the stability parameter @xmath157 , which is again a measure for the dimensionless velocity , is shown as function of the admixture parameter @xmath146 for fixed ratio @xmath155 . as before , the elastic driving force and the @xmath154 appear only the combination @xmath155 for small peclet numbers . as function of the admixture parameter @xmath146 , for the two different types of mixing and @xmath173 . the solid line corresponds to the privileged direction of growth , i.e. positive mixing , while the dotted line shows the dependence for the unprivileged negative mixing configuration.,scaledwidth=60.0% ] as function of the control parameter @xmath155 , for positive mixing , whereas the in the right panel @xmath149 is plotted for negative mixing . additionally , both graphs contain the @xmath149 dependence for the pure shear configuration @xmath145 , as a reference to compare both types of mixings , which was taken from @xcite . , title="fig:",scaledwidth=45.0% ] as function of the control parameter @xmath155 , for positive mixing , whereas the in the right panel @xmath149 is plotted for negative mixing . additionally , both graphs contain the @xmath149 dependence for the pure shear configuration @xmath145 , as a reference to compare both types of mixings , which was taken from @xcite . , title="fig:",scaledwidth=45.0% ] as already mentioned above it is quite interesting that the appearance of an additional dilatational strain leads to an effective anisotropy of the system : whereas in the case of a positive admixture @xmath174 the velocity grows with increasing @xmath146 , the opposite happens for negative admixture @xmath175 . this suggests that in free growth the bicrystal structure will eventually grow in the direction with the higher propagation velocity . for the fast branch , we did not find solutions beyond the point @xmath176 . in fig . [ fig : bicrystallmixedmodecompchannelandfreegrowth ] we show the comparison of the two geometries for the bicrystal case . again , the phase field results confirm the stability of the solution branch for low driving forces . in this region , we obtain a convincing quantitative agreement between the growth velocities obtained for the free growth scenario using the green s function method , and the channel geometry with the phase field method . in the left panel the favored direction is shown ( positive admixture ) for @xmath177 . in this case we again observe dendritic solutions for small driving forces @xmath100 ( see also fig . [ bicrystalshapes]a ) . for the chosen parameters in the channel geometry , all solutions on the upper branch are dendritic . on the lower branch , however , the steady state shapes are double - fingers , which means here the curvature directly at the point of symmetry has changed its sign and we obtain two tips ( see fig . [ bicrystalshapes]b ) . the velocity of these branched solutions is here always smaller than for the single - tip structures . for the other direction , i.e. negative admixture , and a mixing parameter @xmath91 , we show the comparison of growth velocities in the right panel of fig . [ fig : bicrystallmixedmodecompchannelandfreegrowth ] . here , we obtain a convincing agreement between the different methods and associated geometries at least for small driving forces , indicating dynamical stability also for this type of solution . obviously , the growth velocities are here substantially lower than for the previous case , and for fully free growth an initial bicrystal seed would therefore preferentially grow in the fast " direction . for bicrystal growth with mixed eigenstrain : comparison of free growth results obtained by green s function methods to phase field simulations for the growth in a channel . left panel : positive admixture with @xmath178 . right panel : negative admixture , @xmath164.,title="fig:",scaledwidth=45.0% ] for bicrystal growth with mixed eigenstrain : comparison of free growth results obtained by green s function methods to phase field simulations for the growth in a channel . left panel : positive admixture with @xmath178 . right panel : negative admixture , @xmath164.,title="fig:",scaledwidth=45.0% ] in summary , we investigated the diffusion - limited kinetics of solid - solid transformations in the presence of lattice strain . we studied in particular the free growth with green s function methods and the growth in a channel by phase field techniques , exhibiting a comparable behavior . it turns out that the elastic effects have a strong influence on the selection of the arising patterns and the corresponding steady state velocities , which differ crucially from conventional dendritic growth . in particular , we find that the selected stability parameter and the corresponding growth velocity can be much higher compared to the classical case of dendritic growth , where the solutions are selected by the tiny effect of the anisotropy of surface tension . different structural transitions , which are described by dilatational and shear eigenstrains , lead to a rich behavior , and we find very different structures already for the simple cases that were discussed here . in comparison to the green s function method , which is computationally very efficient , the phase field technique has the advantage of higher flexibility . in particular , it can easily be extended to three - dimensional cases , it can be used also for non - steady state situations or anisotropic or non - symmetrical situations with different material parameters in the phases .
the end products of stellar evolution for 98% of all stars will be white dwarfs . this final state results because most stars are not massive enough to ignite c and o , so that nuclear reactions cease with the formation of a degenerate core of he or a combination of he , c , and o. with the cessation of nuclear reactions , white dwarfs emit light only via their cooling and become dimmer as time passes . these remnants have very thin surface layers of h and/or he , the bulk of the envelope having been blown away during relatively quiescent stages of mass loss in the progenitor star . for the oldest stellar populations , @xmath215% of the entire present day mass of the system is tied up in white dwarfs , and therefore their distribution and detailed properties represent a unique link to explore the nature of the now - evolved population ii progenitors . such studies are enhanced in star clusters where all of the remnants are co - spatial , have the same total age , and evolved from stars with the same metallicity , yet over a range in initial mass . milky way globular star clusters have been dated to a formation time roughly 12 gyr ago ( e.g. , krauss & chaboyer 2003 ) , and therefore represent the first structures to form in the galaxy . these systems are one of the most important tools that we have to probe the formation and evolution processes of galaxies @xcite . given their ages , the present day main - sequence turnoff mass in globulars is about 0.8 @xmath0 , and therefore the bulk of their population has evolved through post main - sequence evolution , the majority having formed faint white dwarf stars . through almost 400 orbits of hubble space telescope imaging observations this decade , our team has uncovered the complete white dwarf cooling sequences of the two nearest globular clusters , m4 and ngc 6397 @xcite , and will target the third nearest cluster , 47 tuc , in cycle 17 ( go-11677 ) . a shallower study of the massive globular omega cen by @xcite and @xcite has successfully uncovered over 6500 remnants in this single cluster , roughly a quarter of the total known white dwarf population in the entire milky way from all other studies combined ( e.g. , eisenstein et al . 2006 , and subsequent sloan digital sky survey work ) . the color - magnitude diagrams ( cmds ) of these clusters , extending to impressively faint magnitudes , are presented in figure [ fig : globswds ] ( see also bedin et al . 2009 for a more recent acs study of m4 ) . the white dwarf cooling sequences of nearby globular clusters have been analyzed to independently ( e.g. , of main - sequence stars and turnoff morphology ) derive the distances and ages of the clusters @xcite , explore age variations between metal - rich and metal - poor systems using stars whose evolution is almost completely independent of metallicity @xcite , possibly suggest large fractions of helium - core white dwarfs produced through binary interactions @xcite , and hint that mass loss stages leading to the formation of white dwarfs may involve small asymmetric kicks @xcite . for each of these studies , the interpretation of the photometry to yield the results depends on knowledge of the mass of the brightest white dwarfs that are forming today in these old systems . yet , at present , _ no direct measurements of the masses of such population ii white dwarfs exists _ , although several indirect arguments suggest that the value should be @xmath3 @xmath4 0.53 @xmath0 . as discussed in @xcite and @xcite , stellar evolution theory and our understanding of the maximum luminosity of red giant ( rg ) and asymptotic giant branch ( agb ) stars , as well as the luminosity of the horizontal branch ( hb ) , predict that white dwarfs forming in globular clusters today should have 0.51 @xmath5 0.55 @xmath0 . the best observational verification of this theoretical prediction in the case of cluster white dwarfs was provided recently by @xcite , who reported the masses of white dwarfs in ngc 6752 to be 0.53 @xmath0 . however , their spectroscopic measurements were only of sufficient quality to yield the temperatures of the white dwarfs , and therefore the photometry from the cmd was leveraged to derive masses . recently , a study of the masses of field white dwarfs was also presented from the spy project ( eso sn ia progenitor survey ) . pauli et al . ( 2006 ) find several hundred thin disk and thick disk white dwarfs , and seven candidate halo objects based on kinematic analysis . the masses of six of these white dwarfs range from 0.44 0.51 @xmath0 , with an additional lower mass star at 0.35 @xmath0 . in this paper we present the first results from a new study aimed at _ directly _ ( i.e. , purely from spectroscopic absorption line fitting ) measuring the masses of known population ii white dwarfs in different environments . our goal is to target the brightest white dwarfs in the nearest globular star clusters ( such as m4 , ngc 6397 , and 47 tuc ) and in the halo field , with ground - based spectroscopic instruments on 810 m telescopes . the clusters listed above span an appreciable range in metallicity ( factor of 20 ) and are located at distances such that the brightest white dwarfs with @xmath6 = 10.5 have apparent magnitudes of @xmath7 = 22 24 . the faintness of these stars , crowding issues in the cluster , and the need to obtain high signal - to - noise ( s / n ) spectra of the higher order balmer lines at @xmath8 4000 @xmath9 make this a difficult , but possible , project with current telescopes . below , we discuss the results from our first study , in which we target two dozen white dwarfs in m4 with the gemini / gmos and keck i / lris multiobject spectrographs over several observing runs . the spectra for six of these stars , that are confirmed as cluster members and that have sufficient s / n to yield accurate mass measurements , are consistent with @xmath3 = 0.53 @xmath1 0.01 @xmath0 . the impact of this measurement for bettering our knowledge of stellar structure and evolution , and for dating halo globular clusters through white dwarf cooling theory , are discussed . an analysis of the remaining m4 spectra that did not yield masses , e.g. , to define the spectral types of population ii white dwarfs , is presented in the companion paper by @xcite . the deep hst imaging data of m4 illustrated in figure [ fig : globswds ] ( left ) were collected with the wide field planetary camera 2 ( wfpc2 ) camera in 2001 as a part of go-8679 @xcite . the observations spanned a total of 123 orbits , in the broadband @xmath10 and @xmath11 filters , and were centered on a field located at ( @xmath12,@xmath13 ) = ( @xmath14 , @xmath15 ) , 2.4 pc ( 4.8@xmath16 ) s - e of the center of the cluster . although the hst photometry is very accurate , and there is no ambiguity on the membership of the white dwarfs , the bulk of the stars are too faint to be useful for spectroscopic follow up with modern day facilities . additionally , the wfpc2 field of view is 5.3 square arcminutes , about six times smaller than our ground based instruments , and therefore the spectroscopic program would be highly inefficient with just this input list of targets . below we first discuss supplemental imaging of these data with the gemini south telescope ( [ geminiimaging ] ) , followed by spectroscopic follow up ( [ geminispect ] ) . in [ keckspect1 ] and [ keckspect ] , we discuss additional spectroscopic observations of m4 s white dwarf population with the keck i telescope . we reobserved m4 with the gemini south telescope and gemini multiobject spectrograph ( gmos ) , anchoring the deep hst field near the south - eastern edge of the ground based image and placing shallower hst pointings near the western side . the gmos photometry , over a 5.5@xmath17 5.5@xmath16 area , has the added advantage that it extends continuous imaging into the denser region towards the center of m4 and therefore provides a higher density of targets down to a given magnitude limit . of course , crowding in the ground - based data prevents the targeting of white dwarfs near bright stars , and this increases in the direction of the cluster center . a footprint of the hst and gmos images is provided in figure [ fig : foot ] . = 20 and 24 , and colors less than @xmath18 0.5 ) are shown as dots . the objects targeted for gmos spectroscopy are shown as diamonds and those for lris spectroscopy are represented with pluses . the center of m4 is located at ( @xmath12 , @xmath13 ) = ( 254.898@xmath19,@xmath2026.526 ) and is indicated with an asterisk . [ fig : foot],width=315 ] as this map illustrates , less than half of the gemini / gmos field of view overlaps the hst / wfpc2 observations . in order to effectively select white dwarf candidates over the entire gmos field , we obtained imaging in both the @xmath21@xmath22 and @xmath23@xmath22 filters , down to a depth of @xmath425th mag . the images were corrected for bias and flat fielding by the gemini iraf pipeline , version 1.4 . the processed images were then reduced with the standard daophot / allstar photometry package to yield psf photometry @xcite . further details on the steps involved in our application are provided in @xcite for a similar study . the matched catalog of wfpc2 and gmos sources contains @xmath24 stars . we calibrate the @xmath21@xmath22 and @xmath23@xmath22 magnitudes to hst @xmath25 and @xmath11 ( hereafter called @xmath7 and @xmath26 ) using derived transformations for the common stars in the space- and ground - based systems ( see davis et al . 2009 for details on the transformations ) . these offsets are then applied to the entire gmos catalog . the positions of all stars on the gmos image were carefully measured with respect to both hst positions and standard usno guide stars to yield precise astrometry . with an approximate color of @xmath27 = 0.3 ( one brighter object is also present ) . this cmd is contaminated by a more distant field population , located blueward of the main sequence . there is also contamination in the white dwarf region of the cmd from field white dwarfs and possibly blue compact galaxies , which we address in [ membership ] . the open circles encircling the dots represent those objects which were targeted by keck / lris for spectroscopy ( see [ keckspect ] ) . the red open circles are additional objects that were targeted by gemini / gmos , and that are not in the keck sample due to slit conflicts ( see [ geminispect ] and [ keckspect ] ) . the open circles lacking small dots are hst detections for which we do not have ground based photometry ( two of these stars are off the plot ) . these latter objects were selected for the sole purpose of filling the mask in relatively empty regions , and therefore no pre - selection was used . [ fig : cmd3],width=321 ] the gemini / gmos cmd of m4 is presented in figure [ fig : cmd3 ] . both the main - sequence and white dwarf cooling sequence are well measured , the latter containing about two dozen candidates brighter than @xmath7 = 24 . the location of these white dwarf candidates are illustrated on figure [ fig : foot ] as smaller points . the selection here includes stars with 20 @xmath28 @xmath7 @xmath28 24.5 and colors @xmath27 @xmath28 0.5 . note the paucity of ground based candidates in the two inner wfpc2 fields ( top - left ) . there are in fact many white dwarfs at these positions , however , the crowding limits their detection from the ground . we further subdivide the sample of white dwarf candidates into two bins , with priority 1 stars being our strongest white dwarf candidates in the dominant part of the cooling sequence , and priority 2 objects being somewhat more uncertain objects that are located off the cooling sequence ( e.g. , redder or bluer ) and/or near a bright star . these latter objects are only included in the spectroscopic mask if they do not overlap a priority 1 star ( hence , we get a spectrum of the potential white dwarf for free ) . any of the confirmed members of m4 , based on hst proper motions in the region of overlap , were also targeted as priority 1 objects . finally , we include several redder stars that are candidate cvs in the priority 2 sample ; most of these are hst proper motion members as well . in total , 21 white dwarf candidates were targeted with gemini / gmos for follow up spectroscopy . note , a few of the very red or faint objects shown in the cmd were included without pre - selection for the purpose of filling the spectroscopic mask ( these objects were not detected from the ground ) . an example of the gmos and wfpc2 photometry , astrometry , and a postage stamp image of the white dwarf candidate wd 09 is shown in figure [ fig : l09 ] . similar figures for all other white dwarf candidates are provided in the appendix . , width=302 ] in total , we obtained approximately 14.5 and 9 hours of queue scheduled spectroscopic science exposures with gemini / gmos in june / august 2005 and august / september 2006 , respectively . the observations were taken with the b600 grating , covering a broad wavelength region of @xmath42700 @xmath9 centered at 4680 @xmath9 . the data were binned by a factor of two in the spectral direction to improve the s / n . all science exposures were obtained in sub - arcsecond seeing conditions and photometric skies . the raw data frames were downloaded from the canadian astronomy data center ( cadc ) in multi - extension fits ( mef ) format and reduced with the gemini iraf package , version 1.4 . the reductions are described in detail in @xcite . of the 21 targeted white dwarfs , a spectrum was recovered for all but three . remarkably , all 18 of the recovered spectra show pressure - broadened absorption lines indicative of white dwarfs . the spectra for these 18 stars , as well as an analysis of their spectral types is also presented in davis et al . unfortunately , in order to determine reliable spectroscopic masses , characterization of h@xmath29 is crucial , and even h@xmath30 ( 3889 @xmath9 ) is very useful . the lack of flux at these wavelengths in our gemini data limited the utility of the gmos spectra for the central goals of this paper . although we could derive mass measurements from the spectra ( see [ modelfit ] ) , the uncertainties are several tenths of a solar mass and therefore we decided to supplement these data with higher s / n observations , especially at bluer wavelengths . with a sample of confirmed white dwarfs established from the gmos observations , we applied for keck i telescope time to reobserve the targets using the lris multiobject spectrograph . not only does keck provide a larger aperture than gemini , but the spectrograph is very efficient in the blue and has an atmospheric dispersion corrector to minimize light loss at large airmasses ( this is important for this study since keck is in the north and m4 is in the south ) . given the similarities in the field of view ( lris has a 5@xmath31 7@xmath16 footprint ) , we simply used as many of the _ confirmed _ sample of white dwarfs from gemini as our input , as well as several additional redder stars in the same brightness range . the location of the 31 lris targets in the ground - based cmd are shown in figure [ fig : cmd3 ] as larger open circles . the objects marked in red circles are those that were targeted with gemini , but not keck . the lris footprint is also shown in figure [ fig : foot ] . keck / lris spectroscopic observations ( r. m. rich , pi ) of the single m4 mask proved challenging . half nights , required due to the low declination of m4 , were scheduled in 2005 june , 2007 april / july , and 2008 april , with the 2007 and 2008 observations benefiting from the lris atmospheric dispersion correction system . the multiple allocations were required due to poor weather that resulted in the loss of a significant number of scheduled hours . altogether , we obtained 10.6 hours of useful exposure time . even these data are of highly variable quality , with seeing measurements ranging from 08 to over 20 . in these crowded fields , our final spectra are dominated by a few of the contributions from the best nights . lris @xcite has a dichroic that splits the spectrum into two channels at @xmath45500 @xmath9 . on the blue side , we use the 400/3400 grism ( dispersion = 1.09 @xmath9/pixel ) which covers a wavelength region from the atmospheric cutoff to the dichroic . for the red side , we use the 600/7500 grating ( dispersion = 1.28 @xmath9/pixel ) , centered at 6600 @xmath9 , which covers a wavelength baseline of 2620 @xmath9 . at our resolution and central wavelength , only h@xmath12 , which is a rather poor mass indicator , landed on the red side . hence , we chose not to reduce the red - side data for this paper . the blue - side lris spectra were reduced using standard iraf tasks . the trace , sky subtraction , and extraction were all performed with the apall task . as expected given the better blue sensitivity of lris compared to gmos , the trace at blue wavelengths was far more certain . given the high levels of crowding and background light from bright m4 stars in the field , our ability to obtain a reliable trace , and perform accurate sky subtraction is dependent on the particulars of each individual slitlet ( see [ culling ] ) . the wavelength calibration was calculated in iraf from spectra of three lamps containing hg , zn , and cd . the typical final dispersion of the wavelength solution was well constrained , with uncertainties of @xmath40.1 @xmath9 . the reduced , wavelength calibrated lris spectra were flux calibrated using the spectrophotometric standard hz44 . finally , the individual spectra were combined using scombine , with weights assigned according to the individual s / n ratios . we summarize the identification , positions , and brightnesses of all white dwarf candidates in table [ targtab.tab ] . the four stars that have slit conflicts , and therefore were not observed with keck , are listed at the bottom of the table . altogether , 31 objects were targeted with keck / lris . note , the identification of these stars as `` wd '' at this stage does not confirm their nature , although we will find later that most are in fact white dwarfs . we use this naming scheme for convenience as we will refer back to this table in future sections . lcccc wd 00 & 245.9579 & @xmath2026.5589 & 0.32 & 23.32 + wd 01 & 245.9256 & @xmath2026.5540 & 1.18 & 22.77 + wd 02 & 245.9406 & @xmath2026.5531 & 0.30 & 21.32 + wd 03@xmath32 & 245.9710 & @xmath2026.5511 & 0.93 & 25.21 + wd 04 & 245.9638 & @xmath2026.5511 & 0.27 & 22.69 + wd 05 & 245.9224 & @xmath2026.5480 & 0.27 & 22.71 + wd 06 & 245.9262 & @xmath2026.5442 & 0.28 & 22.65 + wd 07 & 245.9844 & @xmath2026.5414 & 0.82 & 21.99 + wd 08 & 245.9763 & @xmath2026.5396 & 0.67 & 22.65 + wd 09 & 245.9419 & @xmath2026.5328 & 0.26 & 22.50 + wd 10@xmath32 & 245.9120 & @xmath2026.5281 & 2.03 & 22.28 + wd 11@xmath32 & 245.9867 & @xmath2026.5300 & 1.31 & 26.96 + wd 12 & 245.9427 & @xmath2026.5241 & 0.22 & 23.52 + wd 13@xmath32 & 245.9764 & @xmath2026.5240 & 2.54 & 24.31 + wd 14@xmath32 & 245.9890 & @xmath2026.5214 & 0.85 & 24.99 + wd 15 & 245.9625 & @xmath2026.5190 & 0.26 & 22.73 + wd 16@xmath32 & 245.9276 & @xmath2026.5175 & 0.28 & 23.53 + wd 17 & 245.9515 & @xmath2026.5147 & 0.32 & 23.20 + wd 18 & 245.9521 & @xmath2026.5125 & 0.25 & 23.64 + wd 19 & 245.9281 & @xmath2026.5092 & 0.43 & 23.33 + wd 20 & 245.9436 & @xmath2026.5090 & 0.32 & 23.01 + wd 21 & 245.9263 & @xmath2026.5067 & 1.22 & 22.15 + wd 22 & 245.9521 & @xmath2026.5037 & 0.19 & 23.41 + wd 23 & 245.9364 & @xmath2026.4994 & 0.51 & 22.54 + wd 24 & 245.9216 & @xmath2026.4984 & 0.26 & 22.72 + wd 25 & 245.9442 & @xmath2026.4941 & 0.39 & 23.21 + wd 26 & 245.9168 & @xmath2026.4860 & 1.89 & 24.53 + wd 27 & 245.9006 & @xmath2026.4856 & 0.58 & 23.44 + wd 28 & 245.9179 & @xmath2026.4840 & 0.30 & 22.81 + wd 29 & 245.9674 & @xmath2026.4789 & 0.30 & 22.94 + wd 30 & 245.9353 & @xmath2026.4719 & 1.07 & 22.86 + gemwd 05 & 245.9195 & @xmath2026.4860 & 1.33 & 22.48 + gemwd 11 & 245.9633 & @xmath2026.5174 & 0.40 & 23.39 + gemwd 15 & 245.9349 & @xmath2026.5426 & 0.25 & 22.75 + gemwd 17 & 245.9504 & @xmath2026.5344 & 1.10 & 21.31 + the ground - based cmd of our targets in figure [ fig : cmd3 ] shows that the dominant , blue white dwarf cooling sequence of m4 extends from @xmath7 = 22.5 to 23.5 , and contains eleven stars that were spectroscopically observed with lris ( note , the clump at @xmath33 22.7 , @xmath27 @xmath4 0.3 contains five lris targets ) . an additional target , wd 02 , is almost one magnitude brighter than the family of fainter points and has @xmath7 = 21.3 , another four stars are found fainter and bluer than the cooling sequence at @xmath7 @xmath4 23.5 , @xmath27 @xmath4 0.25 ( including the one object without ground based photometry ) , and two stars are at the faint end and redder than the cooling sequence at @xmath7 = 23.2 and @xmath27 = 0.4 . most of the stars that are redder than this sample of 18 objects ( i.e. , @xmath27 @xmath2 0.5 ) yielded spectra that are not interesting for the goals of this paper and will be commented upon further in @xcite . in table 1 , these objects are wd 01 , 03 , 07 , 08 , 10 , 11 , 13 , 14 , 21 , 23 , 26 , 27 , and 30 . the one exception to this is possibly the relatively bright star just beyond the @xmath27 = 0.5 color cut at @xmath7 = 22.54 , @xmath27 = 0.51 ( wd 23 ) . this star is a proper motion member of m4 but , as shown in the appendix in figure [ fig : wd18_26 ] , is located very close to a bright neighbor . likely , the color ( and spectrum ) of this star is contaminated from the neighboring star and we therefore eliminate it from our analysis . [ cols="<,^,^,^,^,^,^,^,^ " , ] [ tab : wdresults ] the family of 18 white dwarf candidates in m4 s pseudo cooling sequence are all confirmed to be white dwarfs ( see davis et al . 2009 ) . for our primary goal of measuring accurate masses , we want to first restrict this sample to only include well measured , isolated white dwarfs with robust s / n ( see [ modelfit ] ) . the stars along the cooling sequence in our sample vary in brightness by about a factor of three , and therefore the fainter stars will intrinsically have much less signal than their brighter counterparts . added to this , we note that the final quality of the spectrum for any star will depend on several factors and therefore not always correlate with the brightness . for example , the spatial length of the given slit effects the quality of the sky subtraction , the position of the star relative to m4 s center effects the sky background level , and any astrometric mis - centering of the star within the slit causes light loss . we first eliminate three ( fainter ) stars that are located within @xmath3430 pixels ( @xmath42@xmath35 ) of a bright neighbor . these stars , wd 16 , 19 , and 22 , all exhibit a noisy spectrum that is likely contaminated with light from the nearby , brighter star ( see figures [ fig : wd09_17 ] & [ fig : wd18_26 ] in the appendix ) . we also eliminate five white dwarfs , four of which are isolated , for which our extractions resulted in s / n @xmath28 10 spectra , wd 12 , 17 , 18 , 25 , and 28 . the extracted spectra for these white dwarfs suffer from unusable line profiles blueward of h@xmath13 , but with improved s / n , these particular sources could also be analyzed as described below . our final sample has ten white dwarfs , all of which are close to the m4 cooling sequence and have @xmath36 23.4 ( seven of which have @xmath36 23 ) . as we demonstrate in [ membership ] , four of these ten white dwarfs , wd 00 , 09 , 20 , and 24 , are hst proper motion members of m4 . we present the spectra for these ten white dwarfs in figure [ fig : wdspectram4 ] ( left ) . all of the spectra clearly show pressure - broadened balmer lines extending from h@xmath37 to h8 . these are therefore all da ( hydrogen ) spectral type white dwarfs . we report the spectral types of all of our white dwarfs , including stars that are too faint for mass constraints , in @xcite . the s / n per resolution element in our ten stars is generally 40 50 , and at best 60 for one star ( wd 02).4030 @xmath9 as discussed in [ syntheticspectra ] . ] we also present a detailed view of the white dwarf cooling sequence in the cmd , and indicate the positions of these ten white dwarfs in the bottom - right panel of figure [ fig : wdspectram4 ] ( dotted circles ) . the remaining open circles ( without dots ) in this diagram are those stars that were spectroscopically targeted in this narrow region of cmd space . we note that the magnitudes and colors of wd 15 and 24 are almost identical and so the two dots can not be discerned . we also illustrate a theoretical white dwarf cooling model for @xmath38 = 0.5 @xmath0 @xcite , which cuts through our sample nicely . from this plot , it is clear that wd 02 is well removed from the dominant sequence of points and the cooling model . although it is possible that the photometry for this star is in gross error , it is more likely that the star is a field white dwarf along m4 s sightline ( see [ membership ] for more information ) . the technique of measuring masses of white dwarfs from balmer line fitting has been widely employed for field stars , such as in the palomar - green ( pg ) sample @xcite and sloan digital sky survey ( sdss ) @xcite . the technique has also been cross - checked against independent measures of the mass of a white dwarf such as astrometry for binary systems ( e.g. , sirius b , barstow et al . 2005 ) , fits to the mass - radius relation for stars with trigonometric parallaxes @xcite , the gravitational redshift method @xcite , and pulsation mode analysis @xcite . for temperatures hotter than @xmath412,000 k , these methods are in excellent agreement with the spectroscopic line fitting technique @xcite . in this section , we first briefly describe our method to calculate masses for the ten white dwarfs discussed above . we then verify the uncertainties in the masses using synthetic spectra of the same quality as the observations , and fit using the same methods described below . finally , we remove field white dwarfs from the sample and analyze the mass of the confirmed m4 members . fitting the balmer lines of a white dwarf to model atmospheres involves reproducing multiple line profiles from h@xmath37 to higher order transitions @xcite . typically , four or five well measured lines are enough to constrain the temperature ( @xmath39 ) and surface gravity ( log @xmath21 ) of the star @xcite . the fifth balmer line in this sequence , h@xmath30 , is the weakest and bluest ( at @xmath40 = 3889 @xmath9 ) , and therefore a spectrograph with high throughput in the blue is needed to characterize the feature accurately in faint stars . general details of the fitting technique used to derive @xmath39 and log @xmath21 are provided in @xcite , and recent refinements to this method are given in @xcite . additional information on the fitting of white dwarf spectra similar to these observations , and taken with our instrumental setup , are presented in 6 of @xcite . specifically , the improved method to normalize the line profiles described in @xcite is used in our work . simultaneous fits of model atmospheres to all of the balmer lines , in each of the ten white dwarfs discussed above , are presented in figure [ fig : wdspectram4 ] ( top - right ) . within each panel , the absorption lines of a given white dwarf are arranged with h@xmath37 at the bottom , and successive higher order transitions on top . the model fits are shown as smooth red curves and are excellent in all cases . we also confirmed that the two anomalous features , small spikes caused by bad sky subtraction in the h@xmath37 line of wd 00 and 29 , do not affect the results of the fit . this line is least sensitive to changes in the gravity of the star . the output from the spectroscopic fits are the @xmath39 and log @xmath21 for each star , which we summarize in table 2 . the masses of these ten stars ( @xmath3 ) are simply calculated by combining the measured surface gravity of the stars with the well established mass - radius relation for white dwarfs . in general , most of the white dwarf masses are @xmath40.5 @xmath0 with an uncertainty of @xmath3410% . we also estimate both the cooling ages , @xmath41 , and the theoretical luminosity of each star by interpolating the @xmath39 and log @xmath21 within evolutionary models similar to @xcite , but assuming a 50/50 carbon - oxygen core mix . our default calculations assume thick hydrogen layers with @xmath42 = @xmath43 = 10@xmath44 and helium layers of @xmath45 = 10@xmath46 , and are summarized in table 2 and discussed further below . we also tested the sensitivity of our results to models ( e.g. , wood 1995 ) with pure carbon cores and to models with very thin hydrogen layers ( e.g. , @xmath42 = 10@xmath47 ) . the results from these comparisons , for both the final remnant masses and the theoretical luminosities , are identical to our default values within the measured uncertainties in these properties ( which are small , see table 2 ) . we note , however , that the cooling ages of these bright , young white dwarfs are systematically larger in the pure c core models by 20 30% . at the time of writing this paper , tremblay & bergeron ( 2009 ) have just presented the first results from an improved calculation of the stark broadening of hydrogen lines in dense plasmas that are representative of white dwarf atmospheres . their new results , including non - ideal effects , suggest that the masses measured through the spectroscopic technique require a slight upward revision of 0.034 @xmath0 . as the temperatures of the stars using the new calculations are also slightly hotter ( e.g. , 1,000 2,000 k ) , the luminosities are unaffected . the full implications of this study have not yet been investigated and the dependency of this offset on mass is not well constrained . in the results that follow , we have adjusted our default masses by @xmath480.034 @xmath0 to be consistent with these new calculations ( including the reported masses in table 2 ) . we have not corrected the individual @xmath39 and log @xmath21 values by a common offset ( which would be approximately 1,500 k and 0.1 dex , respectively ) . the s / n of the ten white dwarfs in our sample is listed in column 8 of table 2 . this is calculated from the pseudo - continuum between the h@xmath13 and h@xmath29 balmer lines , at @xmath44030 @xmath9 . the pixel - to - pixel rms scatter is converted to a s / n per resolution element by multiplying by the square root of the number of pixels per fwhm ( six , in our case ) . the bulk of the stars have s / n = 40 50 , although a few are as low as @xmath430 and one object is above s / n = 60 . note , these s / n values should only be used as a relative ranking of the quality of each spectra , and are not indicative of the global s / n . we can see that , due to the throughput , the s / n decreases blueward of the wavelength at which we made this measurement . in figure [ fig : synthetic ] ( top ) , we produce simulated white dwarf spectra over a range of s / n . these spectra have been convolved with a gaussian profile with 6 @xmath9 fwhm . directly comparing the quality of our data in figure [ fig : wdspectram4 ] to these simulated spectra , we can conclude that most of our objects are better than the s / n = 30 case , and similar , if not better , than the s / n = 40 case . for the latter , we generate a monte - carlo with 500 trials and fit each of these spectra in our fitting code . the prescription used in the fitting is identical to that for our actual white dwarfs , namely , the same models are used in the fit with five balmer lines . the results from this test are shown in the middle panel of figure [ fig : synthetic ] . the input stellar parameters for the synthetic white dwarf are @xmath39 = 25,000 k , log @xmath21 = 7.80 , and @xmath38 = 0.536 @xmath0 . from the distribution of recovered masses , we find @xmath49@xmath38@xmath50 = 0.54 @xmath0 with a dispersion of 0.06 @xmath0 . therefore , we expect to measure the masses of our white dwarfs to roughly 10% precision based on this test , which agrees nicely with our reported uncertainties in mass ( see table 2 , column 5 ) . in the bottom panel of figure [ fig : synthetic ] , we extend the test above for other s / n values and plot the associated error from this fitting . thus far , we have established masses for ten white dwarfs in the m4 direction . as we have reported in previous spectroscopic studies of white dwarfs in star clusters , some fraction of the stars along these lines of sight are actually members of the field population and therefore need to be removed before discussing the mass distribution of bonafide members . the best method to establish such membership is through proper motion analysis , and we have such data from the hst overlap of a portion of our lris field ( see figure [ fig : foot ] ) . out of these ten white dwarfs , we can confirm that wd 00 , 09 , 20 , and 24 are proper motion members ( as are wd 16 and 23 ; these are rejected from our sample based on the poor quality of the recovered spectra and/or contamination from a nearby neighbor see [ culling ] ) . this is not a surprising result since the proper motions were folded into the selection process to target white dwarfs in the region of overlap with our hst field . in addition to these five confirmed m4 white dwarfs , we note that the remaining objects in our sample closely follow the 0.5 @xmath0 cooling sequence , and are consistent in position in the cmd with the confirmed members . although we do not have proper motions for these stars , some of them are likely members of m4 . to distinguish members from non - members , we can use the theoretical luminosity of each star to calculate its distance modulus . any stars that have significantly discrepant values from the known distance of m4 , ( @xmath51)@xmath52 = 12.51 @xmath1 0.09 @xcite , should be eliminated from our sample as they likely represent field white dwarfs along the line of sight , binary white dwarfs , or possibly objects for which the spectroscopic fit is incorrect due to some unknown artifact in the spectrum . we compare the calculated distance modulus of each of our ten white dwarfs with the known m4 distance in figure [ fig : mags ] . the 1@xmath53 uncertainty in the theoretical magnitude ( i.e. , from spectral fitting ) is plotted as a horizontal bar for each star , and the mean distance of m4 and @xmath12@xmath53 uncertainties are indicated with the solid and dashed lines . we can use the proper motion members of the cluster as a rough guide to establish the expected scatter of known members in this diagnostic . the calculated distances of three of the four proper motion members of the cluster , wd 00 , 20 , and 24 , are close to , but slightly larger than the distance of m4 , whereas the fourth star ( wd 09 ) appears to be 0.7 mags closer than the cluster . this object could therefore be a non - member that shares the proper motion of the cluster . alternatively , if the object is a member , its overluminous nature may suggest an unresolved double degenerate in the cluster . we note that the mass derived from the spectrum of this star yields the unexpectedly high value of @xmath38 = 0.79 @xmath0 ( see table 2 ) . three additional white dwarfs are clearly members of the cluster , wd 04 , 06 , and 15 , whereas three other stars wd 02 , 05 , and 29 are possibly field white dwarfs . to summarize , our final secure sample of singly evolved , m4 member white dwarfs includes three objects that are proper motion members ( wd 00 , 20 , and 24 ) and three objects which are on the cooling sequence and have the correct distance from spectral fitting ( wd 04 , 06 , and 15 ) . theory predicts that the masses of white dwarfs forming in old , population ii systems such as globular clusters are expected to be in the range 0.51 @xmath54 0.55 @xmath0 @xcite . as table 2 indicates , the six single white dwarfs that we have confirmed as members of m4 have masses that range from 0.50 0.59 @xmath0 . the uncertainties in our mass measurements are all roughly the same , under 10% , and therefore the un - weighted mean mass of the sample of six white dwarfs is measured to be @xmath3 = 0.53 @xmath1 0.01 @xmath0 . the present day temperature of these cooling remnants is also well measured from our data . there is a group of four stars in our sample with @xmath7 @xmath4 22.6 ( @xmath6 = 10.1 ) and all have the same temperature , @xmath55 = 25,000 k. the two fainter stars , wd 00 and 20 , are cooler than this group by about 5,000 k , as expected . the surface gravities , log @xmath21 , of all six stars are in the range log @xmath21 = 7.68 7.95 . as mentioned earlier , the exact @xmath39 and log @xmath21 values have not been adjusted to reflect the small corrections required by the new stark broadening calculations @xcite . our sampling of m4 cluster member white dwarfs in this study puts us in the fortunate situation of having characterized a group of four remnants at the same position on the cmd . @xcite demonstrate that the mass difference between white dwarfs at the tip of the cooling sequence to those five magnitudes fainter at the limit of the hst study is only @xmath40.1 @xmath0 . these four white dwarfs , wd 04 , 06 , 15 , and 24 , differ in luminosity by just 0.07 mag , and therefore should have the same mass , and have evolved from stars with the same progenitor mass . the expected temperature range in this group of four stars should only be @xmath42% . as these differences are much smaller than our uncertainties , we can verify our mean mass measurement of m4 white dwarfs by creating a composite white dwarf spectrum , with very high s / n , from the coaddition of these individual spectra . @xcite have shown that for unresolved double degenerate systems , the atmospheric parameters derived from assuming a single star are in fact an average of the parameters of both components of the system . therefore , spectra with identical gravities should yield the same average gravity in a coaddition as compared to the true individual values . such a coaddition has the advantage that the resulting fit will be minimally influenced by any artifact affecting a given balmer line in one spectrum ( e.g. , presence of a cosmic ray ) . this method has several disadvantages as well , for example , the spectral lines in the composite may be artificially blurred if the wavelength solution is not well determined in each of the individual spectra . we stress that this test is only intended as a verification of our results and should not be attempted in samples where the mass variation between individual white dwarfs is non - negligible ( e.g. , similar studies in younger clusters ) . the composite spectrum from this analysis is shown in figure [ fig : composite ] ( top ) , and exhibits a s / n @xmath4 100 at @xmath40 = @xmath44030 @xmath9 . the balmer lines in this spectrum are very accurately characterized ; the resulting best fit model to the five lines is shown in the lower panel . the spectroscopic mass from the fit is measured to be @xmath38 = 0.54 @xmath1 0.02 @xmath0 , in excellent agreement with our findings above . overall , from our spectroscopic study of white dwarfs in m4 , we can conclude that the mass of population ii remnants forming today is @xmath3 = 0.53 @xmath0 and the temperature of stars presently at the tip of the white dwarf cooling sequence is @xmath39 = 25,000 k. the mass distribution of white dwarfs in nearby open star clusters has been mapped since 1977 when volker weidemann compared theoretical models of mass loss ( e.g. , fusi - pecci & renzini 1976 ) to the observed masses of a few remnants in the hyades and pleiades clusters @xcite . since this pioneering work , similar mass measurements have been made for over 100 stars in a dozen open star clusters . a review of the earlier work is provided in weidemann ( 2000 ) and a compilation of some more recent results is presented in @xcite , excluding very recent studies by @xcite . the mass distribution of the remnants varies as a function of the turnoff mass of their parent cluster , suggesting that more massive progenitors produce more massive white dwarfs ( see salaris et al . 2009 for a detailed analysis of the present relation ) . new measurements by our team in three older star clusters with ages of a few gyrs have recently permitted , for the first time , a purely empirical fit to the data without the need for a theoretical anchor at low masses ( see below ) . with the present study , we can once again extend the initial - final mass relation to new bounds . as the initial mass function in most ( all ? ) stellar populations is bottom heavy , the characterization of mass loss from these low mass stars affects our understanding of stellar evolution for the bulk of all stars today . in figure [ fig : ifmr ] , we present all of the constraints to date at the low mass end of the relation , corrected by @xmath480.034 @xmath0 from our previous reporting to reflect the new stark broadening calculations ( see [ modelfit ] ) . the data includes the two confirmed cluster members in the 1.4 gyr cluster ngc 7789 @xcite , the two white dwarfs in the 2.5 gyr cluster ngc 6819 @xcite , the single carbon - oxygen core white dwarf in the 8.5 gyr cluster ngc 6791 @xcite , and the six white dwarfs in m4 from this study . for the latter , we have fixed the progenitor mass to the expected mass at the turnoff in an old , metal - poor population , @xmath56 = 0.80 @xmath0 , with an adopted error of 0.05 @xmath0 . = @xmath201.1 and @xmath57 = 12.0 gyr , the vandenberg , bergbusch , & dowler ( 2006 ) models yield a turnoff mass of 0.82 @xmath0 . ] the data points for individual white dwarfs in these clusters are shown as open circles and binned averages of the population in each cluster are illustrated with larger , filled circles . for older clusters , all of the white dwarfs at the top of the cooling sequence evolved from progenitors just above the present day turnoff , and therefore we expect a single mass at the cooling sequence in a given cluster . of course , in younger clusters , our spectroscopic measurements can include stars along the entire cooling sequence of the cluster and therefore there can be a large dispersion in white dwarf masses in a single cluster ( e.g. , ngc 2099 , kalirai et al . 2005 and ngc 2168 , williams , bolte , & koester 2009 ) . 0.034 @xmath0 to reflect the new stark broadening calculations presented in @xcite . as discussed in [ masslossmetallicity ] and @xcite , the single white dwarf in ngc 6791 is much more massive than the mean mass of remnants in the cluster , most of which are low mass helium - core white dwarfs ( and not shown here ) . the individual white dwarf progenitor pairs are shown as open circles and the binned averages in each cluster are shown as larger filled circles . the relation shows a linear decline of final remnant mass with decreasing initial progenitor mass , extending down to the extreme low mass end , anchored by our m4 measurement at @xmath56 = 0.80 @xmath1 0.05 @xmath0 and @xmath3 = 0.53 @xmath1 0.01 @xmath0 . the solid line shows the best fit empirical relation to all individual data points excluding the new m4 white dwarfs , and the dashed curve shows the mass of the core at the first thermal pulse ( see [ ifmr ] ) . [ fig : ifmr],width=302 ] the low mass end of the initial - final mass relation in figure [ fig : ifmr ] now contains a total of eleven data points , and shows a roughly linear decrease in white dwarf mass with initial mass , anchored at the extreme left end by the new m4 measurement , @xmath56 = 0.80 @xmath1 0.05 @xmath0 and @xmath3 = 0.53 @xmath1 0.01 @xmath0 . given that the oldest observable stellar populations in the universe are the same age as m4 , this relation is now complete at the lower limit and can not be observationally extended further to lower masses . as discussed earlier , several predictions of stellar evolution theory have long suggested that the masses of population ii white dwarfs forming today should be 0.51 @xmath5 0.55 @xmath0 @xcite , and therefore our first direct measurement is in perfect agreement with these predictions . based on our measurements , the total integrated post main - sequence mass loss of low mass stars in old population ii systems will be @xmath434% . theoretically , there is an expected lower limit to the masses of carbon - oxygen core white dwarfs . this is given by the mass of the stellar core at the first thermal pulse on the asymptotic giant branch , an evolutionary point before the core has had a chance to grow on the branch . this core mass has a very small dependency on the stellar metallicity . in figure [ fig : ifmr ] , we illustrate this expected core mass as a dashed curve @xcite . over the range of initial masses on this plot , the core mass is expected to remain roughly the same at @xmath3 = 0.52 @xmath0 , showing a flat relation for @xmath56 @xmath4 1 2 @xmath0 . at 0.8 @xmath0 , our first observational constraint from a half dozen white dwarfs in this single star cluster suggests the actual remnant masses are just slightly larger than the minimum core mass ( with a slight extrapolation ) . this implies that the progenitors of such low mass stars had just enough mass to experience thermal pulses on the asymptotic giant branch and that the cores of these stars will not grow appreciably during this evolution . as m4 is a typical population ii system in terms of its old age and low metallicity , it is reasonable to assume that these trends may extend to stars that are presently in similar phases of stellar evolution both in the general field halo population and in other globular clusters with similar properties to m4 . in addition to extending the initial - final mass mapping to lower masses , m4 represents the first star cluster with a significantly sub - solar metallicity in which white dwarf masses have been measured . the mean metallicity of the cluster is [ fe / h ] = @xmath201.10 @xcite , 13 times more metal - poor than the bulk of the open clusters used to define the relation and 40 times more metal - poor than the metal - rich system ngc 6791 ( [ fe / h ] = @xmath480.40 , peterson & green 1998 ) . for this cluster , after accounting for the new stark broadening calculations discussed earlier , @xcite actually measured a mean mass for nine white dwarfs of @xmath3 = 0.46 @xmath1 0.06 @xmath0 , much lower than expected . as they show , the bulk of these white dwarfs contain helium - cores , after evolving through a channel involving intense mass loss on the red giant branch , likely driven by the high metallicity of the cluster . the single data point shown in figure [ fig : ifmr ] is based on the only clearly non he - core white dwarf in that sample , and does not affect the discussion above . however , the results from ngc 6791 suggest strong dependencies of mass loss on metallicity , especially at super - solar metallicities . to date , we lack a good theoretical understanding of the dependence of post main - sequence mass loss on metallicity . the general expectation in evolutionary models is for more efficient transport of material outward from the star if the metallicity is higher , and therefore higher rates of mass loss . however , the exact treatment depends on the modeling of winds on the red giant branch and thermal pulses on the asymptotic giant branch ( e.g. , see weidemann 2000 and habing 1996 ) , phases of stellar evolution that are poorly understood . as @xcite summarizes , mass loss prescriptions along the red giant branch alone show very different behaviors with metallicity . for example , over a 1 dex metallicity change from [ fe / h ] = @xmath201.0 to 0.0 , the @xcite and @xcite laws predict that @xmath58@xmath59 should increase by @xmath340.05 @xmath0 whereas the @xcite and @xcite laws suggest @xmath58@xmath38 increases by 0.15 0.25 @xmath0 . to qualitatively compare the dependence of mass loss in post main - sequence evolution with metallicity to these estimates , we can first parameterize the initial - final mass relation . the solid line in figure [ fig : ifmr ] is a linear fit to the entire sample of _ open cluster _ white dwarf main - sequence progenitor mass pairs from @xcite , @xmath60 the zero point of this relation has been adjusted to reflect a global offset of @xmath480.034 @xmath0 in the white dwarf mass measurements , although the slope may in fact also be affected if the offset is found to vary with mass ( see figure 11 and 12 in tremblay & bergeron 2009 ) . the bounds on this relation range from the oldest open cluster white dwarf in ngc 6791 to the youngest white dwarfs in the pleiades , and therefore are @xmath56 = 1.1 6.5 @xmath0 . a small extrapolation of the empirical relation to the mass of the present day turnoff in m4 ( e.g. , @xmath40.8 @xmath0 ) yields @xmath3 = 0.52 @xmath1 0.02 @xmath0 , which is therefore consistent ( at 1@xmath53 ) with our measured value of the masses of m4 s white dwarf population . these data therefore _ hint _ that mass loss rates have a weak or no dependence on metallicity , at least over the range extending from metal - poor populations such as m4 to roughly solar metallicity . interestingly , we know from our study of ngc 6791 that mass loss rates are strongly correlated with metallicity at extremely high [ fe / h ] , and therefore this may lead to an interesting shape for the empirical mass loss metallicity relation from white dwarf spectroscopy . of course , stronger conclusions will benefit greatly from additional observations of clusters spanning a wide range of metallicities , especially at the metal - poor end . for completion , if we add our new data point for m4 to the initial - final mass relation and recalculate the best fit ignoring any possible metallicity - related biasses , we find a slightly shallower relation with @xmath3 = ( 0.101 @xmath1 0.006 ) @xmath56 + 0.463 @xmath1 0.018 @xmath0 . finally , we note that the metallicity of m4 , [ fe / h ] = @xmath201.10 , is only slightly more metal - rich than the global metallicity of both the stellar halo of the milky way ( [ fe / h ] = @xmath201.6 , @xmath53 = 0.6 ; morrison et al . 2003 ) and m31 ( [ fe / h ] = @xmath201.5 , @xmath53 = 0.7 ; kalirai et al . 2006 ) . assuming the dominant population in these galactic components is old ( e.g. , @xmath6110 gyr ) , the same 34% stellar mass loss fraction should be representative of stars evolving in these populations . by directly inputing this mass loss into general stellar evolution models of similar age and metallicity ( e.g. , girardi et al . 2000 ) , we can add an important observational constraint to the modeling of light from the stellar halos of spiral galaxies using population synthesis techniques ( e.g. , bruzual & charlot 2003 ) . in section [ introduction ] we introduced the extensive hst observations that have defined some of the deepest cmds ever measured , with any instrument ( e.g. , the hst cycle 13 observations of ngc 6397 in richer et al . although these data sets have led to a wide range of astrophysical studies , the primary goal of the observations has aimed to accurately establish turnoff - independent white dwarf cooling ages for population ii star clusters . in this respect , the present study serves to establish several of the input ingredients that are needed to make this measurement . for example , knowledge of the masses and spectral types of white dwarfs , and the initial - final mass relation , are all required to fully model an observed white dwarf cooling sequence ( e.g. , hansen et al . 2004 ; 2007 ) . for m4 , two recent studies have modeled the white dwarf cooling sequence using independent data sets , and both have concluded with similar ages ( however , see discussion below ) . first , @xcite used a full 2-d treatment of the white dwarf cooling sequence in the observed hst wfpc2 cmd ( i.e. , luminosity and color information are used ) , and just recently , @xcite have used a more simplified technique of age dating based solely on the luminosity function in newer and deeper acs data . an analysis of these studies illuminates several interesting choices relative to the findings in this paper : * the present day mass of cluster white dwarfs forming today sets the normalization for the masses along the entire cooling sequence . in their modeling , @xcite assume the mass at the tip to be 0.55 @xmath0 and @xcite assume 0.54 @xmath0 . as we have shown , the actual mass of white dwarfs at the bright end of the cooling sequence is only slightly lower at 0.53 @xmath0 , and therefore the mass distribution ( and hence cooling rates ) along the entire cooling sequence are not very different from the assumptions in these studies . * hydrogen ( da ) and helium ( db ) atmosphere white dwarfs have different cooling rates , and so the fraction of stars cooling with these spectral types leads to structure in the luminosity and color function that must be modeled correctly to infer the age of the population from white dwarf cooling theory . for m4 , ngc 6397 , and ngc 6752 , spectroscopy of the white dwarf cooling sequence indicates that 100% of the stars at the bright end have hydrogen atmospheres ( see moehler et al . 2004 and davis et al . @xcite treated the fraction of db white dwarfs as a free parameter and found the best fit to yield an upper limit of 40% ( good fits were obtained as long as more than 60% of the white dwarfs are da ) , whereas @xcite choose an arbitrary value of 30% ( e.g. , the field disk ratio ) . * the initial - final mass relation is required to properly evolve the progenitor stars on to the white dwarf cooling sequence . the normalization and form of the initial - final mass relation adopted in both of these studies ( e.g. , wood 1992 ) , is based on empirical constraints from solar metallicity data points at higher masses , and a theoretical anchor at low masses . this can now be updated to our new result given earlier . we encourage future theoretical modeling efforts to make full use of this new information in their analysis , preferably on the newer acs imaging data of m4 . of course , the white dwarf models themselves should also be updated with new physical insights . for example , the contribution to the opacity of cool white dwarfs from the far red wing of the ly@xmath12 line ( kowalski 2007 ) alone can yield ages of globular clusters ( e.g. , ngc 6397 ) that are higher by @xmath40.5 gyr ( kowalski 2007 ; b. hansen 2009 , private communication ) . we noted above that the @xcite and @xcite studies concluded with similar white dwarf cooling ages for m4 . formally , @xcite reported an age of 12.1 gyr ( 2@xmath53 lower limit of 10.3 gyr ) and @xcite measured an age of 11.6 @xmath1 0.6 gyr ( internal errors ) . a deeper look at these two studies suggests they are , in fact , grossly _ inconsistent _ with one another given differences in white dwarf cooling models and uncertainties in the cluster distance modulus . as our present work is closely linked to obtaining very accurate absolute age measurements of globular clusters ( through the study of white dwarfs ) , we briefly summarize these two studies below and highlight the respective assumptions and differences that are being made . in the @xcite study , the authors compare their data to three sets of models and adopt a distance to m4 based on the well established technique of subdwarf distance fitting . they find @xmath62 = 1.73 @xmath1 0.10 kpc for m4 ( e.g. , richer et al . 1997 ) , which is identical to independent distance measurements for the cluster from the baade - wesselink distance using the m4 rr lyrae stars ( @xmath62 = 1.73 @xmath1 0.01 kpc , liu & janes 1990 ) and from astrometry ( @xmath62 = 1.72 @xmath1 0.14 kpc , peterson et al . 1995 ) . at present , it appears the latter estimate is the most secure . for the subdwarf measurement , the comparison was based on pre - hipparcos parallaxes , and for some other clusters , a calculation with post - hipparcos parallaxes indicates a longer distance by @xmath40.2 mags than initially estimated ( gratton et al . similarly , the baade - wesselink method yields luminosities for m4 s rr lyrae stars that are fainter by @xmath40.2 mags as compared to the @xmath6[fe / h ] relation derived by gratton et al . ( 1997 ) for clusters with low extinction ( using post - hipparcos main - sequence fitting techniques ) . unfortunately , re - deriving the distance to m4 using any photometric technique is problematic given the very high extinction along the line of sight ( @xmath63 @xmath2 1 ) , the variation in @xmath63 across the face of the cluster , and the non - standard shape of the extinction curve along this line of sight ( which passes through the scorpius - ophiucus dark cloud complex vrba , coyne , & tapia 1993 ) . irrespective of the adopted distance ( although see below ) , @xcite find that the white dwarf cooling age they derive for m4 from their own models ( e.g. , hansen 1999 ) is consistent with that measured using the @xcite models , however they rule out the @xcite models which yield ages that are older than the universe . therefore , this suggests that , for the same distance modulus and data set , simply changing the white dwarf model yields an absolute age difference of 2 gyr . in the recent @xcite study , the authors use only the @xcite white dwarf models , and still conclude with an age similar to that derived by @xcite . the difference is actually reflected in the distance to m4 ; @xcite use a much larger ( i.e. , 15% ) value of @xmath62 = 1.98 kpc . this value is formally inconsistent with the three independent studies above . @xcite measure the distance to m4 by simply fitting a single isochrone of approximately the correct metallicity to the turnoff and horizontal branch . in principle , this method can of course yield the distance to any stellar population with a measured turnoff , however it has almost always been avoided for a large number of reasons . first , and most importantly , there are uncertainties in both the microphysics built into the stellar evolution models ( e.g. , opacities , equation of state effects , and nuclear reaction rates ) and in the treatment of processes that do not come from first principles ( e.g. , convection , core - rotation , diffusion , gravitational settling , etc . ) . because of these unknowns , simply adopting an isochrone from another groups models would lead to a different measured distance modulus for the cluster ( e.g. , different models predict different horizontal branch luminosities ) . second , such comparisons of data and models to yield distances require transformation of the model from the theoretical to the observational plane , resulting in another source of error . finally , in this specific case , this comparison does not alleviate any of the aforementioned uncertainties related to our lack of knowledge of the reddening and extinction along this line of sight . in the end , the application of the @xcite models and the longer distance modulus derived by @xcite effectively offset each other , yielding a derived age similar to that of @xcite . summarizing , there are clearly both intrinsic differences in white dwarf models calculated by different groups and there is a lack of consensus on the correct distance to m4 . for the former , increased testing of the white dwarfs models on star clusters with well known fundamental parameters ( e.g. , distance , metallicity , and foreground reddening ) are sure to resolve the correct prescriptions . for the latter , it appears that only the @xcite astrometric measurement is free of the unknown reddening and extinction laws along this line of sight , and therefore should be preferred . until these differences are unambiguously resolved ( e.g. , the space interferometry mission could directly measure the distance to m4 to within a few percent ) , the error budget from these uncertainties will remain large and the new physical constraints on masses , spectral types , and the initial - final mass relation are secondary to the analysis . of course , if the properties of bright white dwarfs in other population ii systems are similar to m4 , the new results can have immediate significance on those studies . this is especially exciting given the remarkable data set that has been presented for ngc 6397 by @xcite and will soon be available for 47 tuc ( h. richer , cycle 17 acs / wfc3 proposal go-11677 ) . the mass measurement of white dwarfs in a globular cluster provides important leverage to the use of these stars as distance calibrators . although the white dwarf cooling sequence is fainter than the bright part of the main - sequence in a star cluster , uncovering the sequence can offer several advantages over canonical techniques involving main - sequence distance fitting . for example , the location of the bulk of the white dwarf cooling sequence in a cmd does not depend on the metallicity or age of the population , or on any of the other systematics that plague main - sequence evolution ( e.g. , convection theory ) . however , the location does depend on mass ( e.g. , wood 1995 ) , with a 0.1 @xmath0 difference translating to a @xmath58(mag ) = 0.25 shift ( i.e. , a 12% distance offset ) . so , first , a simple offset can now be calculated to correct the location of any local calibrating sample of white dwarfs with parallax measurements ( usually 0.60 @xmath0 for field white dwarfs ) to the expected location of 0.53 @xmath0 stars . following this , a straightforward comparison with the apparent luminosity of the globular cluster white dwarf cooling sequence can yield the distance to the population ( e.g. , see renzini et al . unfortunately , it is difficult to actually make this measurement and independently derive the distance of m4 given the aforementioned uncertainties in the slope of the extinction curve along this line of sight . however , the true distances of other clusters with well constrained reddenings and extinctions can be easily calculated from their cooling sequences and the new mass measurements . we have presented the first directly determined spectroscopic mass measurements of white dwarfs that are known to be population ii members . our measurements suggest that the final remnant mass at the tip of the white dwarf cooling sequence in the globular cluster m4 is @xmath3 = 0.53 @xmath1 0.01 @xmath0 , in nice agreement with expectations based on stellar evolution theory . these results suggest that low mass population ii stars will lose @xmath434% of their mass through post main - sequence stellar evolution . if , in fact , all milky way globular clusters have the same age , then this mass loss rate should be applicable to the evolution of stars in all but the most metal - rich galactic globular clusters . we use this measurement to first extend the initial - final mass relation of stars to the lowest possible limit , given the age of the universe . this single observation also hints that mass loss rates are not significantly different for metal poor stars when compared to solar metallicity stars of the same mass . finally , we offer several pieces of new input that are required to measure more accurate white dwarf cooling ages of globular clusters in the galactic halo . imaging observations of globular clusters with ground and space based telescopes have now uncovered thousands of white dwarfs in population ii clusters ( e.g. , see figure [ fig : globswds ] ) . the science that results from these purely photometric surveys is greatly enhanced with spectroscopic follow up , which we have shown can yield the spectral types of the remnants @xcite , the mapping of initial to final mass , and the mass at the tip of the observed cooling sequence ( and therefore a calibration for the masses along the entire cooling sequence ) . all of the nearest globular clusters are in the southern hemisphere , and given their larger distances compared to m4 , an extension of this work to these systems will require superior blue sensitivity and throughput , and much larger allocations of spectroscopic time on 8-meter class telescopes ( e.g. , see moehler et al . future white dwarf spectroscopy in these clusters will benefit greatly from a 30-m telescope with a multiobject spectrograph with the bluest sensitivity . such a telescope would not only allow a direct measurement of the mass loss rates in an expanded set of open and globular clusters , but would permit the first initial - final mass relations to be built from observations of a set of white dwarfs in a _ single _ cluster . deep spectroscopy of white dwarfs over a range in luminosity ( and therefore mass ) can be engaged in a given cluster , minimizing the systematics introduced by metallicity variations in samples of clusters . fortunately , the nearest three globular clusters also sample a wide range in metallicity ( e.g. , a factor of 20 from ngc 6397 at [ fe / h ] = @xmath202.0 to 47 tuc at [ fe / h ] = @xmath200.7 ) , and therefore the dependence of mass loss on metallicity can be probed in detail with such observations . we wish to thank d. reitzel for help with designing the spectroscopic masks in this program . we also wish to thank j. hurley for useful discussions related to stellar evolution , white dwarfs , and binarity . jsk s research is supported in part by a grant from the stsci director s discretionary research fund . the research of hbr is supported by grants from the natural sciences and engineering research council of canada . he also thanks the canada - us fulbright program for the award of a fulbright fellowship . pb is supported in part by the nserc canada , by the fund fcar ( qubec ) , and is a cottrell scholar of research corporation for science advancement . support for mc is provided by proyecto basal pfb-06/2007 , by fondap centro de astrofsica 15010003 , by proyecto fondecyt regular # 1071002 , and by a john simon guggenheim memorial foundation fellowship . rmr acknowledges support from grant ast-0709479 from the national science foundation . barstow , m. a. , bond , h. e. , holberg , j. b. , burleigh , m. r. , hubeny , i. , & koester , d. 2005 , , 362 , 1134 wood , m. a. 1995 , proceedings of the 9th european workshop on white dwarfs held at kiel , germany , 29 august - 1 september 1994 . lecture notes in physics , 443 , edited by detlev koester and klaus werner . springer - verlag , berlin heidelberg new york , 41
low- and intermediate - mass stars ( with masses 0.8 @xmath0 , depending on model details ) play an important role in galactic chemical evolution , thanks to the ejection into the interstellar medium ( ism ) of material containing newly synthesized nuclear products , mainly @xmath1he , @xmath2c , and @xmath3n , and possibly @xmath4o . the interest in the nucleosynthetic history of @xmath2c and @xmath3n , in particular , has recently increased thanks to the observations of high red - shift systems . for instance , it is possible to measure the nitrogen abundance and the n / o ratio in the damped ly-@xmath5 systems ( e.g. pettini et al . 1995 ; lu et al . 1998 ) , whereas carbon is detected in the ly-@xmath5 forest clouds ( e.g. lu 1991 ; crotts et al . 1994 , tytler & fan 1994 ) . the chemical history of both these elements in a galaxy is quite complex , as it depends , besides other factors , on the relative contributions of stars with different masses , hence releasing their yields at different timescales . in the case of @xmath2c the relative role of low- and intermediate - mass stars , or rather of metal - rich , mass - losing massive stars , in the carbon enrichment at late galactic epochs is still a matter of debate ( prantzos et al . 1994 ) . in the case of @xmath3n , large uncertainties affect the theoretical predictions on both the secondary and primary nucleosynthesis of nitrogen in low- and intermediate - mass stars , and the possible contribution of primary nitrogen from massive stars ( woosley & weaver 1982 , maeder 1983 ; see also the review by maeder & meynet 2000 ) . it follows that , in order to trace the history of chemical evolution and of star formation in the universe which the observations of high red - shift systems are intended to infer it is quite important to analyse in detail the nucleosynthesis of these elements taking place in low- and intermediate - mass stars . the contribution of these stars to the chemical enrichment of the ism essentially occurs during the rgb and agb phases , both characterized by the occurrence of mass loss and events of surface chemical pollution ( dredge - up episodes ) . whereas modelling the evolution of a star on the rgb is rather easy by means of modern stellar evolutionary codes , dealing with the most advanced stages characterised by the occurrence of thermal pulses ( the tp - agb phase ) is quite a difficult task , due to both the high complexity of the physics involved , and the remarkable requirement of computing time . then , an alternative theoretical approach is offered by _ synthetic _ models , which summarise the results of _ complete _ stellar calculations through simple and practical analytical relations . this allows quick computing of the models and ready analysis of the results . historically , the first most significant agb synthetic models were those developed by iben & truran ( 1978 ) and renzini & voli ( 1981 ) . these pioneer works focused on the contribution of low- and intermediate - mass stars to the element enrichment of the ism . in particular , the sets of stellar yields presented by renzini & voli ( 1981 ) have been extensively used in chemical evolution models of galaxies , providing almost the only available data source up to recent years . it is important to remark that in renzini & voli ( 1981 ) synthetic agb model the fundamental parameters ( essentially mass - loss and dredge - up ) were specified according to the indications from the current knowledge of the involved physical processes and available complete stellar calculations . in this sense the model was _ uncalibrated_. that approach had the merit of supplying a testing tool for complete stellar models , as it pointed out at some fundamental inadequacies in the predictions ( e.g. too low efficiency of the third dredge - up ) , and assumed prescriptions ( e.g. too low mass - loss rates ) , which led to clear discrepancies between theory and observations ( see , for instance , iben 1981 ; iben & renzini 1983 ; bragaglia et al . 1995 ) . with awareness of that , the later synthetic agb models ( e.g. groenewegen & de jong 1993 ; marigo et al . 1996 , 1999a ) have made the next step ahead , that is update the input prescriptions and _ calibrate _ the model parameters in order to reproduce fundamental observables ( e.g. the carbon star luminosity functions , the initial - final mass relation ) . basing on the results of these model calibrations , various sets of stellar yields from low- and intermediate - mass stars have been presented in recent years as an alternative to renzini & voli ( 1981 ) , namely : marigo et al . ( 1996 , 1998 ) , van de hoek & groenewegen ( 1997 ) ; see also forestini & charbonnel ( 1997 ) , and boothroyd & sackmann ( 1999 ) . to the above reference list we will now add the new homogeneous sets of stellar yields presented in this work . with respect to our previous calculations ( marigo et al . 1996 , 1998 ) , the present yields are derived from stellar models with updated input prescriptions and improved treatment of the relevant processes involved ( i.e. the third dredge - up , see marigo et al . 1999a for all details ) . specifically , we follow the evolution of low- and intermediate - mass stars , coupling the results of complete stellar models ( girardi et al . 2000 ) that cover the evolution from the zero age main sequence ( zams ) up to the onset of the thermally pulsing asymptotic giant branch ( tp - agb ) with synthetic tp - agb models ( marigo 1998ab ; marigo et al . 1999a ) that extend the calculations up to the end of this phase . relevant model prescriptions are briefly recalled in sect . [ models ] . with the aid of these evolutionary calculations , we then derive the stellar yields ( h , he , and main cno elements ; refer to sects . [ yi ] and [ stelyie ] ) for a dense grid of initial stellar masses ( in the range @xmath6 @xmath7 ) and various metallicities ( @xmath80.004 , 0.008 , 0.019 ) . for the most massive stars , experiencing hot - bottom burning ( hereinafter hbb , or envelope burning ) during the tp - agb phase , the corresponding yields are given for three values of the mixing - length parameter , i.e. @xmath9 , @xmath10 , and @xmath11 . this parameter mainly affects the predicted production of @xmath3n and @xmath1he due to hbb ( see sect . [ primy ] ) . the final part of this work ( sect . [ compare_yields ] ) is dedicated to compare our results with the yields calculated by other authors , i.e. renzini & voli ( 1981 ) , and van de hoek & groenewegen ( 1997 ) . in the attempt to single out the causes of the main differences , we analyse the effect of different model prescriptions ( e.g. mass - loss , dredge - up , hbb ) on the predicted yields , testing at the same time the capability of a model to satisfy basic observational constraints ( e.g. initial - final mass relations , white dwarf mass distribution , carbon star luminosity functions , chemical abundances of planetary nebulae ) . let us first define some quantities which will be often used throughout this paper to indicate critical masses for the occurrence of particular physical processes . these quantities are : @xmath12 , @xmath13 , @xmath14 , @xmath15 , and @xmath16 . the first one , @xmath12 , denotes the maximum initial mass for a star to develop a degenerate he - core , hence experience the he - flash at the tip of the red gian branch ( rgb ) , and is comprised between 1.7 2.5 @xmath17 depending on metallicity and model details . the second one , @xmath13 , is defined as the critical stellar mass over which carbon ignition occurs in non - degenerate conditions , marking the boundary between intermediate - mass and massive stars . it is worth recalling that this mass limit is usually comprised within 5 8 @xmath17 , being significantly affected by the adopted treatment of convective boundaries , as discussed in sect . [ compare_yields ] . the third one , @xmath14 , corresponds to the minimum initial mass for a star to undergo hbb during the tp - agb phase . stellar evolution calculations indicate that @xmath18 , depending on metallicity ( see , for instance , marigo 1998b ) . the fourth one , @xmath15 , denotes the minimum initial mass for a star to experience the third dredge - up during the tp - agb phase . observations of carbon stars suggest that @xmath19 , decreasing with the metallicity ( see , for instance , marigo et al . 1999a ) . finally , we recall the basic parameters of the third dredge - up : @xmath20 and @xmath16 . the parameter @xmath20 is intended to measure the efficiency the third dredge - up , being defined as the fraction of the increment of core mass over an inter - pulse period which is dredged - up to the surface at the subsequent thermal pulse . the actual values for @xmath20 are still a matter of debate among theoreticians , and can range from @xmath21 to @xmath22 , depending both on stellar properties ( e.g. mass and metallicity ) and model details ( e.g. treatment of convective boundaries ) . the parameter @xmath16 refers to the minimum core mass for the occurrence of the third dredge - up , and is affected by several factors as well . in general , theoretical models would predict that @xmath16 decreases so that the third dredge - up is favoured at increasing mass and mixing - length parameter , and decreasing metallicity ( see , for instance , wood 1981 ; marigo et al . 1999a ) . in this work we consider low- and intermediate - mass stars , i.e. those with initial masses in the range from about @xmath6 to @xmath13 @xmath23 and and three choices of the original compositions ( i.e. @xmath24 $ ] , @xmath25 $ ] , @xmath26 $ ] ) . their evolution from the zams up to the beginning of the tp - agb phase is taken from the padua stellar models ( girardi et al . 2000 ) , that include moderate overshoot from core and external convection ( chiosi et al . 1992 ; alongi et al . 1993 ) . the reader should refer to girardi et al . ( 2000 ) for more details of the adopted input physics . the models by girardi et al . ( 2000 ) provide the expected changes in the surface abundance of several chemical elements ( h , @xmath27he , @xmath28he , @xmath2c , @xmath29c , @xmath3n , @xmath30n , @xmath4o , @xmath31o,@xmath32o , @xmath33ne , @xmath34ne , @xmath35 mg ) caused by the first and second dredge - up episodes , i.e. prior to the onset of the tp - agb phase . it should be also remarked that these models do not assume any ad - hoc `` extra - mixing mechanism '' , e.g. the so - called cool - bottom process ( wasserburg et al . 1995 ; see also charbonnel 1995 ; boothroyd & sackmann 1999 ; weiss et al . 2000 ) , which is invoked to reconcile discrepant predictions of surface abundances with those measured in field population ii stars , galactic globular clusters , and magellanic clouds clusters ( see , for instance , the results by gratton et al . 2000 in their chemical analysis of field metal - poor stars ) . mass loss by stellar winds suffered by low - mass stars ( with @xmath36 ) on the ascent of rgb , is analytically included applying the classical reimers ( 1975 ) formula to the evolutionary tracks calculated at constant mass by girardi et al . an efficiency parameter @xmath37 is adopted to fulfil the classical observational constraint provided by the morphology of horizontal branches in galactic globular clusters . finally , once the the first significant thermal pulse is singled out in each evolutionary sequence calculated by girardi et al . ( 2000 ) , that point is assumed to define the starting conditions for synthetic calculations of the tp - agb phase ( following the model prescriptions described in marigo 1998ab , and marigo et al . 1999a ) , which are carried on up to the complete ejection of the envelope by stellar winds . mass loss is included according to the semi - empirical formalism developed by vassiliadis & wood ( 1993 ) . nucleosynthesis occurring in the innermost envelope layers of tp - agb stars with hbb is followed adopting the caughlan & fowler ( 1988 ) compilation of reaction rates for cno and p - p reactions . the electron - screening factors are those of graboske et al . following the classical definition by tinsley ( 1980 ) , the _ stellar yield _ , @xmath38 , of a given chemical element @xmath39 , is the mass fraction of a star with initial mass @xmath40 that is converted into the element @xmath39 and returned to the ism during its entire lifetime , @xmath41 . tables a1 a12 give the quantities : @xmath42 expressed in solar masses , for all the chemical elements considered , as function of the initial stellar mass @xmath40 , metallicity @xmath43 , and mixing - length parameter @xmath5 . according to the definition of stellar yield we can write : @xmath44 \,\ , \frac{{\rm d}m}{{\rm d } t } { \rm d}t \label{defy}\ ] ] where @xmath45 is the current mass - loss rate ; @xmath46 and @xmath47 refer to the current and initial surface abundance of the element @xmath39 , respectively . thanks to the fact that the surface chemical composition of stars not suffering hbb during the tp - agb phase ( i.e. with initial masses @xmath48 ) , is altered by the occurrence of discrete and quasi - instantaneous episodes of convective dredge - up , that alternate with periods of continuous mass - loss , the evaluation of stellar yields can be simplified as follows . denoting by @xmath49 , @xmath50 , @xmath51 , the abundance of the species @xmath39 after the first dredge - up , the second dredge - up , and the @xmath52 dredge - up event during the tp - agb phase , respectively , we get : @xmath53 where @xmath54\,\ , \delta m^{\rm ej}_{\rm rgb } \label{rgbml}\ ] ] @xmath55\,\ , \delta m^{\rm ej}_{\rm e - agb } \label{eagbml}\ ] ] @xmath56\,\ , \delta m^{j , { \rm ej}}_{\rm tp - agb } \label{tpagbml}\ ] ] in eq . ( [ rgbml ] ) @xmath57 is the mass of the envelope ejected during the entire rgb phase ( defined only for low - mass stars ) . it is worth noticing that most of @xmath57 is lost close to the tip of the rgb , that is after the occurrence of the first dredge - up , so that eq . ( [ rgbml ] ) is a good approximation of eq . ( [ defy ] ) . as far as mass loss on the agb is concerned , we remark that , with the adopted prescription for @xmath58 ( vassiliadis & wood 1993 ) , the amount mass lost during the e - agb phase is indeed negligible so that we can assume @xmath59 in eq . ( [ eagbml ] ) . the contribution of the tp - agb phase is evaluated with eq . ( [ tpagbml ] ) , that sums all the partial contributions of the pulse cycles , the generic @xmath52 one consisting of a thermal pulse when the @xmath52 dredge - up possibly occurs followed by the inter - pulse period , during which the mass @xmath60 is ejected . it should be noticed that this approximation holds for tp - agb stars which experience only the third dredge - up . for more massive tp - agb stars ( with @xmath61 ) also suffering hbb , the changes in the surface chemical composition and mass loss are concomitant processes , so that the calculation of stellar yields requires the adoption of integration time steps shorter than the inter - pulse periods . in general , _ negative _ @xmath62 correspond to those elemental species which are prevalently _ destroyed _ and diluted in the envelope , so that their abundances in the ejected material are lower with respect to the main sequence values . on the contrary , _ positive _ @xmath62 correspond to those elements which are prevalently _ produced _ so that a net enrichment of their abundances in the ejecta is predicted . figures [ stary1 ] and [ stary2 ] show the quantities @xmath62 for all the chemical elements considered , as a function of @xmath40 and @xmath43 . for stars experiencing hbb results are given for three values of the mixing - length parameter . finally , for the sake of clarity , we remind that the cno cycle does not change the total number of cno nuclei involved as catalysts in the conversion of h into @xmath28he , i.e. @xmath63 . it follows that the first and second dredge - up , though affecting the surface abundances of the cno isotopes , do not alter their _ total abundance by number_. in fact , the material injected into the envelope has experienced the cno cycle involving only isotopes already present in the original composition . on the contrary , the constancy of @xmath64 breaks down as soon as the dredge - up of primary carbon and oxygen , produced by @xmath5-capture reactions at thermal pulses , occurs . however , in all cases the _ total abundance by mass _ of the cno isotopes is somewhat changed because of the conversion of these elements mainly into @xmath3n , so that a small positive cno yield ( in mass fraction ) , is expected , for example , from the rgb phase ( see tables a1 a3 ) . the quantity , @xmath65 , referring to the total net yield of all cno isotopes , is shown in fig . [ stary3 ] . model predictions can be understood more easily considering that the stellar yield of a given element is essentially determined by the _ efficiency _ and _ duration / frequency _ of : * the _ nucleosynthesis / mixing processes _ ( e.g. dredge - up events , hbb ) which alter its abundance in the surface layers ; * the _ mass - loss process _ which ejects the surface layers into the ism according to the physical prescriptions adopted in this work for the tp - agb phase ( see also sect . [ modpre ] ) we can summarise the following points : * the third dredge - up determines the surface enrichment mainly of @xmath1he , @xmath2c , and @xmath4o . the adopted intershell abundances are [ x(@xmath1he)@xmath66 ; x(@xmath2c)@xmath67 ; x(@xmath1he)@xmath68 according to boothroyd & sackmann ( 1988b ) . the process is more efficient ( i.e. higher @xmath20 ) at lower @xmath43 ; * hbb operates via the cno - cycle , hence essentially increasing the surface abundances of @xmath1he , and @xmath3n . the process is more efficient at higher @xmath69 ( provided that @xmath70 ) , lower @xmath43 , and larger @xmath5 . * mass loss is , in general , less efficient ( i.e. lower @xmath58 ) at decreasing @xmath43 and increasing @xmath5 . in fact , both factors tend to produce hotter tracks , and generally @xmath58 anticorrelates with @xmath71 . as a consequence , lower mass - loss rates correspond to longer tp - agb lifetimes , hence greater number of third dredge - up events and a longer duration of hbb . the expected trend of @xmath72 ) for the elements under consideration as a function of the stellar initial mass , metallicity , and mixing - length parameter is shown in figs . [ stary1]-[stary3 ] . we can notice the following : * * h and @xmath1he * + the net yields of these elements have mirror - like trends , being negative for h , and positive for @xmath1he . the maximum of @xmath1he production at around @xmath73 ( depending on metallicity ) is explained considering the effect of the increase of the number of thermal pulses ( hence dredge - up episodes ) with stellar mass for @xmath74 ( see fig . [ fig - npul ] ) . this peak is more pronounced at lower metallicities due to longer tp - agb duration ( fig . [ fig - npul ] ) and larger number of thermal pulses ( fig . [ fig - time ] ) for given stellar mass . the subsequent increase of @xmath1he production towards higher masses , @xmath75 , is caused by the occurrence of hbb in addition to the third dredge - up . the yield of @xmath1he is larger for lower metallicities and higher values of the mixing - length parameter , reflecting the greater efficiency of both the third dredge - up and hbb . * * @xmath76he * + the net yield of this element presents a pronounced peak at very low masses , say at @xmath77 , and decreases at higher masses . this trend is explained considering that the main contribution to the yield is due to the first dredge - up , and that both the related surface enrichment of @xmath76he and the amount of mass lost during the rgb phase are inversely proportional to the stellar mass in the low - mass domain . at higher masses , ( @xmath78 ) , the net yield becomes even slightly negative because of hbb . * * @xmath2c * + the net positive yield increases with the initial mass , up to a maximum located at about @xmath73 , corresponding to largest number of dredge - up episodes suffered during the tp - agb phase , provided that hbb has not operated . + the subsequent decline towards higher masses is initially due to fewer dredge - up events , and then to the prevailing effect of hbb . it follows that no substantial enrichment of @xmath2c is provided from the most massive agb stars . + this general trend is more marked at lower metallicities , because of the longer tp - agb phases , and the greater efficiency of the third dredge - up and hbb . * * @xmath29c * + the first dredge - up causes an increase of the @xmath29c surface abundance in stars of all masses , and the resulting yields are positive . a further contribution is provided by mild hbb , as long as creation of @xmath29c via the reaction @xmath2c@xmath79c prevails over destruction via the reaction @xmath29c@xmath80n . the results also depend on the interplay between the strength of hbb and mass loss . + the most favourable cases correspond to the models [ e.g. ( @xmath81 ) , ( @xmath82 ) ] , in which the efficiency of reactions allows the synthesis of @xmath29c for a long time before the drastic reduction of the envelope causes the extinction of nuclear burning . the spikes of @xmath29c production for these models would suggest that a proper tuning of hbb is required : if nuclear reactions are somewhat too weak @xmath29c is not significantly created , else if somewhat too strong @xmath29c is quickly destroyed in favour of @xmath3n . * * @xmath3n * + it turns out that hbb plays the dominant role for the synthesis of @xmath3n . the positive yield as a function of the stellar mass depends on both the efficiency and duration of nuclear burning . it follows that lower metallicities and higher values of the mixing - length parameter concur to favour nitrogen production . the contribution from low- and intermediate - mass stars to the galactic enrichment of nitrogen may result important , as suggested by chemical evolutionary models of galaxies ( e.g. portinari et al . * * @xmath30n * + the net yield of this element is mostly negative . for stars with initial masses in the range , @xmath83 , the depletion of @xmath30n is due to the effect of the first and second dredge - up . for higher mass stars , the results are affected by hbb , depending on the degree of cno cycling attained in the burning regions . this element has the shortest nuclear lifetime , after that of @xmath32o , so that it quickly attains nuclear equilibrium with @xmath3n . we note that the depletion of @xmath30n is much more efficient at higher metallicities due to the increase of the cno - cycling , which implies a more efficient destruction of this element . the predicted trend of the @xmath30n yield mirrors , to some extent , that of the secondary component of @xmath3n ( cf . [ primy ] ) . * * @xmath4o * + the net yield of this element is positive for stars with @xmath83 thanks to the dredge - up events during the tp - agb phase , whereas it becomes negative at higher masses because of hbb . + the synthesis of fresh @xmath4o occurs via the reaction @xmath2c(@xmath84)@xmath4o during thermal pulses , so that the yield of this element depends , among other factors , on its abundance in the dredged - up material . according to recent tp - agb calculations which include deep overshooting from all convective boundaries herwig et al . ( 1997 ) find that the @xmath4o abundance in the convective intershell is roughly ten times higher , x(@xmath4o)@xmath85 , than previously predicted , i.e. x(@xmath2c)@xmath86 by boothroyd & sackmann 1988b ( standard case ) . then , adopting herwig et al . indications , oxygen production by low- and intermediate - mass stars may be favoured with respect to the standard case , but it is worth recalling that , in general , the final yields are crucially affected by various other factors ( e.g. number and efficiency of dredge - up episodes ; see also sect . [ conclu ] ) . + anyhow , regardless of its abundance in the dredged - up material , the trend of @xmath4o , as a function of @xmath69 and @xmath43 , is expected to be qualitatively similar to that of @xmath2c . moreover , according to the present calculations , a notable dependence on metallicity comes out . the increasing positive trend with decreasing metallicities essentially reflects the longer duration of the tp - agb phase , and hence the greater number of dredge - up episodes , in combination with their larger efficiency . * * @xmath31o * + a certain production of this element is provided by tp - agb stars with hbb , as long as the chain of reactions @xmath4o@xmath87f@xmath88o prevails over nuclear destruction via the reactions @xmath31o@xmath89f and @xmath31o@xmath90n . the behaviour resembles that of @xmath29c , in the sense that under some fine - tuned conditions for particular combinations of the stellar mass and metallicity a production spike of @xmath31o can result . * * @xmath32o * + this element has the shortest nuclear lifetime against proton captures , so that it easily burns even at mild temperatures , quickly attaining the nuclear equilibrium with @xmath31o . the net yield of @xmath32o is negative for all masses , with a trend mirroring that of @xmath31o . moreover , as in the case of @xmath30n , the depletion of @xmath32o is more pronounced at higher metallicities due to the more efficient cno cycling . as far as the cno nuclei are concerned , we can distinguish for each element @xmath39 the secondary , @xmath91 , and primary , @xmath92 , components of the stellar yield @xmath93 calculated with ( see eq . ( [ defy ] ) ) @xmath94 \,\ , \frac{{\rm d}m}{{\rm d}t } { \rm d}t\ ] ] and @xmath95 \,\ , \frac{{\rm d } m}{{\rm d}t } { \rm d}t\ ] ] where @xmath96 and @xmath97 denote the secondary and primary current surface abundances , respectively . moreover , we have @xmath98 and @xmath99 , as follows from the definitions of secondary and primary abundances . we notice that @xmath92 can be only @xmath100 , whereas @xmath91 can be either @xmath100 or @xmath101 , in the respective cases that the mass - averaged secondary abundance of the element in the ejecta is greater , equal or smaller than its original value . a few basic remarks should be made at this point . both the first and second dredge - up affect ( by increasing or decreasing ) only the secondary components of the cno surface abundances . in fact , in these episodes the envelope is polluted by material which has undergone cno - cycling , with a net change in the relative abundances of the cno isotopes synthesized from metal seeds originally present in the star . on the contrary , the third dredge - up enriches the chemical composition of the envelope with @xmath2c and @xmath4o of primary origin ( synthesized by @xmath5-capture reactions ) . finally , hbb affects the abundance distribution of the cno isotopes of both secondary and primary synthesis . keeping in mind these concepts , it turns out that : * stars with initial masses @xmath102 can produce cno yields of secondary origin only ; + * for stars with initial masses @xmath103 the yields of @xmath2c and @xmath4o should include a primary component , as a consequence of the third dredge - up during the tp - agb phase ; and + * for stars with initial masses @xmath104 undergoing hbb during the tp - agb phase , the yields of the cno isotopes should all display primary components , because of the injection of primary @xmath2c and @xmath4o nuclei ( at each dredge - up event ) into regions where the cno cycle is operating . & rv81 & hg97 & m2k ( this work ) + @xmath13 & 8 @xmath17 & 7 @xmath17 & 5 @xmath17 + @xmath105 relation , @xmath106 relation & no metallicity dependence & with metallicity dependence & with metallicity dependence + mass loss & reimers ( 1975 ) @xmath107 & reimers ( 1975 ) @xmath108 & vassiliadis & wood ( 1993 ) + 3 d.up : efficiency @xmath20 & @xmath109 & 0.75 & 0.65 for @xmath110 , any @xmath69 + & function of @xmath111 , any @xmath43 & for any @xmath43 , @xmath69 & 0.55 for @xmath112 , any @xmath69 + 3 d.up : minimum core mass @xmath16 & 0.60 @xmath17 & 0.58 @xmath17 & from envelope integrations + & for any @xmath43 , @xmath69 & for any @xmath43,@xmath69 & function of @xmath69 and @xmath43 + hbb : overluminosity & no & no & yes + hbb : nucleosynthesis & nuclear network & parameterised approx . & nuclear network + [ tab - mod ] comparing the secondary and primary components of the cno yields four cases can be met ( tables a1 a12 ; see also marigo 1998b ) : 1 . @xmath113 and @xmath114 so that @xmath115 2 . @xmath116 and @xmath117 so that @xmath118 3 . @xmath119 and @xmath117 so that @xmath118 4 . @xmath119 and @xmath117 so that @xmath120 the first ( 1 . ) case applies to low - mass stars with @xmath102 , i.e. never experiencing the third dredge - up during the tp - agb phase . no primary component of stellar yields is expected . the second ( 2 . ) case corresponds to both primary and secondary production . in stars with @xmath61 it applies , for instance , to @xmath29c , @xmath31o , and @xmath3n . in general , for these elements a positive secondary contribution may be provided by the first ( and possibly second ) dredge - up and hbb , the latter process being also responsible for the primary synthesis of these elements ( starting from primary @xmath2c and @xmath4o injected by the third dredge - up ) . as far as @xmath29c is concerned ( see also sect . [ stelyie ] ) we notice that a suitable interplay between the strength of hbb and mass loss can occasionally result in very favourable conditions for the production of @xmath29c , giving a peak of the related yields ( for instance , at the model @xmath121 in the case @xmath122 ) . concerning the yields of @xmath3n , it should be remarked that the contribution from intermediate - mass stars may be relevant in view of interpreting , with the aid of chemical evolutionary models of galaxies , the observed trend in the @xmath123(n / o ) vs. @xmath123(o / h ) diagram ( see , for instance , vila - costas & edmunds 1993 ; henry et al . 2000b ) , where the large scatter of data points towards lower metallicities would imply the existence of a significant primary component in the measured nitrogen abundances . the third ( 3 . ) possibility corresponds to a dominant primary production . in stars with initial masses @xmath103 , this case applies to @xmath2c surface abundance , which is first decreased by the negative secondary contribution from the first and second dredge - up , and subsequently increased by the third dredge - up injecting primary nuclei into the envelope . a similar situation occurs for the yield @xmath4o in the same range of stellar masses , as the effect of the third dredge - up prevails over that of the previous mixing episodes ( first and second ) . finally , the fourth ( 4 . ) case corresponds to a dominant secondary depletion . this refers to @xmath30n , @xmath32o for stars of all masses , and to @xmath4o for stars with @xmath104 if the reduction of the original abundance caused by the first and second dredge - up dominates over the injection of primary oxygen via the third dredge - up ( even possibly partially destroyed by hbb ) . under these circumstances , no enrichment of the interstellar medium is expected for these elemental species . tables a10 a12 give the net yield for each element of the cno group ( t entry ) , together with the secondary ( s entry ) and primary ( p entry ) components , for stars with initial masses @xmath124 , for various values of the original metallicity and mixing length parameter . we will compare the stellar yields presented in this work ( hereinafter also m2k ) with those available in two widely used studies , namely the pioneer work by renzini & voli ( 1981 ; rv81 ) , and the more recent one by van de hoek & groenewegen ( 1997 ; hg97 ) . before making a direct comparison between the yields of various elemental species , we consider it useful first to recall the relevant prescriptions adopted in the mentioned agb models , and consequently analyse their effects by showing how the predictions of different models compare with basic observables . table [ tab - mod ] summarises the relevant assumptions adopted in the agb calculations , which the three different sets of stellar yields under consideration are derived from . for further details the reader should refer to the original papers and references therein . first of all , let us consider the quantity @xmath13 ( see also sect . [ somedef ] ) , that corresponds to the maximum initial mass of a star to develop a degenerate c - o core , hence experience the agb phase . this critical mass heavily depends on the previous evolutionary history , mainly the extension of the convective core during the h - burning phase . evolutionary models ( classical models ) that adopt the schwarzschild criterion to define the convective boundaries ( e.g. dominguez et al . 1999 ) predict higher values for @xmath13 than those models that assume some overshoot beyond the last formally stable layer against convection ( e.g. girardi et al . 2000 ) . both rv81 and hg97 yields are based on classical models and have @xmath125 , whereas the pre - agb evolutionary models used in this work ( m2k ) adopt a convective overshoot scheme so that @xmath126 . we remark that models with masses up to @xmath13 cover , by definition , the whole class of low- and intermediate - mass stars . in other words , models with @xmath127 ( for whatever predicted @xmath13 ) would eventually meet the fate of supernova explosion . finally , a cautionary warning should be made in the context of practical application of chemical yields . when stellar yields from stars of different initial masses are to be included in galactic models of chemical evolution , attention should be paid to correctly match sets of yields of different mass intervals ( i.e. low , intermediate , high ) . if the models do not belong to a homogeneous grid of stellar calculations , one should at least care to combine stellar yields of different origin ( i.e. stellar code ) but with the same predicted value for @xmath13 . otherwise , the relative weight of stars belonging to different classes ( i.e. with different nucleosynthetic histories ) to the integrated chemical enrichment may be substantially mistaken ( for instance , by over- or under- estimating the role of supernovae ) . fundamental relations in synthetic agb models are the core mass - luminosity ( @xmath105 ) , and core mass - interpulse period ( @xmath106 ) relations . predictions of stellar yields are significantly influenced by these input prescriptions , i.e. the luminosity affects the mass - loss rates on the tp - agb , and interpulse - periods determine the temporal recurrence of the third dredge - up . with respect to rv81 prescriptions ( iben & truran 1978 ( it78 ) , see fig . [ lmc ] ) , m2k and hg97 models are based on more recent relations ( boothroyd & sackmann 1988a ; wagenhuber & groenewegen 1998 ) in which a notable improvement is the inclusion of a composition dependence . for a given core mass , the quiescent luminosity / interpulse period of a tp - agb star is found to increase / decrease at increasing metallicity . according to evolutionary calculations by boothroyd & sackmann ( 1988a ) , for instance , the quiescent tp - agb luminosity for @xmath128 is about 20 @xmath129 higher than for @xmath130 ( see fig . [ lmc ] ) . the adopted prescription for mass loss on the agb crucially influences the predictions of stellar yields , as it affects the total number of thermal pulses ( hence dredge - up episodes ) suffered by a tp - agb star , hence the growth of its core mass and agb lifetime . in this work we adopt the semi - empirical formalism presented by vassiliadis & wood ( 1993 ) who couple results of pulsation theory with observations of variable agb stars . in rv81 the classical reimers law ( 1975 ) is assumed with the efficiency parameter @xmath131 set equal to 1/3 or 2/3 . hg97 as well use the reimers law , but with @xmath132 , which they find as the best value to fulfil basic observational constraints ( see sects . [ wdmass ] and [ sssec_minmfin ] ) . to this regard the following remark should be made . as a matter of fact , the reimers prescription was originally designed to describe mass loss suffered by low - mass stars climbing up the rgb , and it is usually calibrated in view of reproducing observations of stars in the subsequent horizontal branch phase of quiescent core he - burning ( see sect . [ models ] ) . however , as already shown by rv81 , the straightforward extension of the reimers formula to the agb evolution does not suit important constraints . in fact , the reimers prescription with @xmath133 can not produce the `` super - wind '' mass - loss rates ( @xmath134 @xmath17 yr@xmath135 ) measured in stars close to the agb - tip luminosities , and consequently it can not account for the typical values of masses and radii of planetary nebulae at the observed luminosities . these difficulties have been overcome by later mass - loss prescriptions specifically designed for agb stars ( e.g. bowen 1988 ; fleisher et al . 1992 ; vassiliadis & wood 1993 ; blcker 1995 ) which are all characterised by a more rapid increase of mass - loss rates during the agb evolution , then naturally leading to the development of the superwind regime . the effect of different laws for mass loss is significant with respect to the expected number of thermal pulses experienced on the tp - agb evolution . as an example ( see fig . [ fig - npul ] ) , we can notice that for a ( @xmath136 ) model our calculations yield @xmath137 ( depending on @xmath5 ) , whereas rv81 predict @xmath138 ( 1631 ) with the efficiency parameter @xmath139 ( @xmath140 ) . in general , it turns out that the largest differences in the number of thermal pulses show up for models with higher stellar masses ( i.e. @xmath141 ) , that are expected to experience the super - wind regime according to the the vassiliadis & wood s prescription . moreover , significantly different results are obtained by rv81 and m2k as far as the tp - agb duration is concerned ( see fig . [ fig - time ] ) . first of all , we can notice that our tp - agb lifetimes present a pronounced trend with the stellar mass , showing a maximum at around 2 2.5 @xmath17 ( depending on metallicity ) . in particular , a drastic drop of the tp - agb duration is expected in the highest mass domain , i.e. for stars with hbb . in rv81 the mass - dependence is much less marked , and such a strong reduction of the tp - agb lifetimes of the most massive models is not predicted . this latter point is relevant for the interpretation of the high - luminosity wing of the observed carbon star luminosity functions ( see sect . [ cslf ] ) . we also notice that with the vassiliadis & wood s formalism both the total number of the thermal pulses and the tp - agb lifetimes are notably sensitive to the metallicity , i.e. increase with decreasing @xmath43 . this feature reflects consequently on the predicted yields from stars with the same initial masses but different initial metallicities . finally , it is worth remarking that the efficiency of mass loss on the agb crucially affects the masses of the bare c - o cores left at the end of this phase . it follows that the empirical initial - final mass relation sets important constraints to the theoretical prescriptions for stellar winds ( see sect . [ sssec_minmfin ] ) . it is important also to recall that according to rv81 in stars with @xmath142 the c - o degenerate core grows up to the chandrasekhar limit ( @xmath143 ; see fig . [ minmfin ] ) , leading to explosive carbon ignition ( type i-1/2 supernova event ; see iben & renzini 1983 ) . on the contrary , according to m2k and hg97 this circumstance is always prevented by the earlier removal of the whole stellar envelope . this process is expected to critically affect the actual yields of many elements , mainly @xmath1he , @xmath2c , and @xmath4o . moreover , the reduction of the core mass caused by the third dredge - up concurs to determine the final mass of the remnant left at the end of the agb . the largest differences in the analytical treatment of this process , in the models here considered , can be summarised as follows . in rv81 model the onset of the third dredge - up in low - mass stars possibly occurs later ( higher @xmath16 ) and with a lower efficiency ( lower @xmath20 ) than in the hg97 and m2k calculations , that are carried out with similar values of the parameters . ( see table [ tab - mod ] ) . this can be explained considering the different usage of the quantities @xmath16 , and @xmath20 in the models . as already mentioned in sect . [ intro ] , in rv81 these parameters were derived according to complete stellar models currently available at that epoch . the subsequent comparison between the predictions of synthetic models and observations pointed out at the so - called `` carbon star mystery '' , as denominated by iben ( 1981 ) , i.e. too few faint and too many bright carbon stars expected than observed . differently , the later analyses carried out by hg97 and m2k move from another perspective , that is to consider @xmath16 and @xmath20 as free parameters which should be calibrated in order to reproduce observations of carbon stars . the aim is to provide indications on the average characteristics of the third dredge - up , so as to remove the theoretical discrepancy related to the `` carbon star mystery '' ( see sect . [ cslf ] ) . the most notable effect of this process on stellar yields involves @xmath3n and @xmath1he , which are newly synthesised at the expense of hydrogen and , in general , of the other cno catalysts . it is important to stress that the occurrence of hbb in the most massive agb stars ( @xmath144 ) has not only a direct effect on stellar yields via changing the chemical abundances in the envelope but also determines an indirect action , affecting the energetics of these stars , hence their evolutionary properties . in fact , as a consequence of hbb , the @xmath105 relation breaks down , a feature clearly shown by complete agb stellar calculations ( e.g. blcker & scnberner 1991 ; see also fig . [ hbb7 ] ) . in these massive agb models the luminosity evolution is characterised by a steeper increase with the core mass ( above the @xmath105 relation ) up to a maximum , followed by a decline as soon as the envelope mass is significantly reduced by mass loss . eventually the @xmath105 relation is recovered ( e.g. vassiliadis & wood 1993 ; marigo 1998b ) . this behaviour is exemplified in fig . [ hbb7 ] where we report the results of complete evolutionary calculations performed by blcker ( 1995 ) for a @xmath145 model with solar metallicity ( triangles ) . actually , the effect of the predicted luminosity evolution on stellar yields is at least two - fold . in fact , the stellar luminosity is closely related to the temperature at the base of the envelope , which the nuclear reaction rates crucially depend on . moreover , the overluminosity of agb stars with hbb can trigger high mass - loss rates , thus favouring the onset of the super - wind regime with consequent reduction of the tp - agb lifetimes . such overluminosity effect caused by hbb is included neither in rv81 nor in gdj93 , where the luminosity evolution is assumed to follow the @xmath105 relation by it78 ( with some revision for the composition dependence in the gdj93 work ) . in fig . [ hbb7 ] we show the behaviour of the luminosity for the @xmath145 model ( solid line ) as it would result adopting the it78 relation with the the same values for current @xmath111 and @xmath69 as in the blcker ( 1995 ) model sequence . the discrepancy is notable . to overcome this limitation of synthetic models marigo et al . ( 1998 ; see also marigo 1998ab ) developed a solution scheme based on envelope integrations , so that the overluminosity produced by hbb is taken into account and the results of complete stellar calculations are recovered ( dashed line in fig . [ hbb7 ] ) . in the following we will examine the agb synthetic models under consideration ( rv81 , hg97 , m2k ( this work ) ) in relation to their capability of reproducing important observational constraints , which are closely related to the stellar yields . the initial - final mass relation ( ifmr ) of low- and intermediate - mass stars is intimately linked to the chemical yields , as it determines the total amount of matter ejected by a star during its entire evolution . in a complementary way , it gives information on the reservoir of stellar remnants , irreversibly lost by the star - forming gas . moreover , assessing the upper mass limit for wd progenitors ( @xmath146 ) is an important point , since it affects the expected rate of type ii supernovae . all these aspects are fundamental issues for chemical evolutionary models . figure [ minmfin ] shows a few empirical calibrations of the ifmr for the solar neighbourhood . the first striking point is that the more recent determinations significantly differ from the earlier work by weidemann ( 1987 ) . for instance , the revised relation by herwig ( 1996 ) presents a flatter slope up to hyades location ( @xmath147 , @xmath148 ) , followed by a steeper rise , and a final flattening towards higher initial masses ( @xmath149 ) . the presence of an inflection point at the hyades mean location seems to be confirmed also by reid ( 1996 ) , as discussed by weidemann ( 1997 ) . the second point to be made deals with the critical mass @xmath146 for carbon / oxygen white dwarfs , i.e. not considering the neon / oxygen white dwarfs which would derive from `` super - agb '' stars according to garcia - berro et al . ( 1997 ) evolutionary calculations . ] , that is the maximum initial mass of wd progenitors . at present , this limiting value is still rather uncertain ( most likely in the range 5 8 @xmath17 ) , since it heavily depends on model details . in particular , as already discussed by weidemann ( 1987 ) , the definition of convective boundaries via either the schwarzschild criterion or an overshooting scheme plays a crucial role . it turns out that with the latter choice @xmath146 is lower than assuming the former classical assumption . however , other parameters may affect the predictions for @xmath146 . for instance , the recent metallicity re - determination ( i.e. half - solar ) of the young open cluster ngc 2516 by jeffries ( 1997 ; see fig . [ minmfin ] ) has lead to assign it a younger age . as a consequence , jeffries ( 1997 ) derives @xmath146 around 5 6 @xmath17 , that is considerably lower than @xmath150 as estimated in previous studies ( e.g. weidemann 1987 ; koester & reimers 1996 ) . however , it should be noticed that in a more recent paper jeffries et al . ( 1998 ) still do not exclude that the metallicity of ngc 2516 might be nearly solar . figure [ minmfin ] displays the theoretical imfrs as derived by rv81 , hg97 , and m2k for low- and intermediate mass - models with initial solar metallicity . we can notice that the both hg97 and m2k are satisfactorily consistent with the trend of the most recent observational relations , whereas rv81 is far from reproducing the empirical data . in particular , the rv81 relation shows a quick divergency of the final mass at increasing initial mass , with the most massive stars being able to build c - o cores up to the chandrasekhar limit of 1.4 @xmath17 . the final fate of these stars would correspond to the occurrence of type i-1/2 supernova events , which seems not to be supported by the observations . the white dwarf mass distribution ( wdmd ) is closely related to the ifmr from which it can be theoretically derived , provided that assumptions on the initial mass function ( imf ) and star formation rate ( sfr ) are made to perform the integration over mass and time . the observed wdmd in the solar neighbourhood ( bergeron et al . 1992 ; see fig . [ wdmd ] ) is narrowly peaked in the mass range 0.5 0.6 @xmath17 ( adopting a mass bin of 0.1 @xmath17 ) , which contains more than 45 @xmath129 of the observed objects . we notice that , with a finer bin sampling ( i.e. 0.05 @xmath17 ) , the location of observed peak would fall between 0.5 and 0.55 @xmath17 , which may be difficult to be theoretically reproduced . in fact , according to stellar evolutionary models , these values would be consistent with the minimum remnant mass produced by progenitor stars as old as the age of the galaxy ( @xmath151 gyr corresponding to initial masses @xmath152 ) , provided that their c - o cores do not grow in mass during the tp - agb phase . in fact , these stars are predicted to enter the tp - agb phase with a core mass of already @xmath153 ( girardi et al . then , the location of the observed peak in the range 0.5 0.55 @xmath17 might be explained by theory only if assuming i ) that mass loss suffered by agb low - mass stars is so strong that they leave this phase as soon as they enter it , or ii ) very efficient dredge - up prevents the growth in mass of the core ( herwig et al . 1997 ) . on the other hand , as suggested by bragaglia et al . ( 1995 ) , the origin of such discrepancy between theory and observations would be most likely due to a systematic underestimation of the surface gravities derived from wd models . given this point of uncertainty and considering that wds with @xmath154 are probably helium wds derived from binary evolution , in this work both theoretical and observed wdmds ( see fig . [ wdmd ] ) are derived adopting a mass bin of 0.1 @xmath17 , and normalising them to the observed fraction of wds with @xmath155 . the predicted fraction of wds contained in the @xmath156 mass bin ( from @xmath157 to @xmath158 ) is calculated with : @xmath159 where ( @xmath160 , @xmath161 ) is the corresponding interval of initial stellar mass ; @xmath162 is the imf ( by number ) evaluated at the mean initial mass @xmath163 ; @xmath164 and @xmath165 denote the age of the galaxy , and the lifetime of a star with mass @xmath166 , respectively ; @xmath167 is the sfr evaluated at the birth epoch of the star ; and @xmath168 is the time interval of detectability of the wd . for the sake of simplicity , in our calculations we adopt a constant sfr , the imf given by the salpeter s law ( @xmath169 ) , and suppose that any wd formed in the past is still detectable at the present time . this implies we assume that the wd fading time is always much longer than the wd s age . in this case @xmath170 , i.e. the wd has been detectable since the death of the progenitor . moreover , it is worth noticing the following points . under the assumption of a constant sfr , the wdmd essentially depends on i ) the slope , @xmath171 , of the ifmr , ii ) the imf , and iii ) the lifetimes of the stellar progenitors relative to the age of the galaxy . the first factor favours the population of those mass bins in which the slope of the ifmr is flatter , i.e. where stars with different initial masses build up wds with similar masses . this should be one of the dominant effects which gives rise to the observed narrow peak of the wdmd at around @xmath172 , just where the ifmr is rather flat ( see fig . [ minmfin ] ) . the second and third factors tend to produce opposite effects . the imf preferentially weighs the formation of wds with lower masses , hence evolved from originally less massive stars if the ifmr is a monotonic function . on the contrary , the accumulation factor ( @xmath173 ) favours the contribution of wds of higher masses , evolved from more massive stars , hence with shorter lifetimes . in fig . [ wdmd ] the observed wdmd in the solar neighbourhood is compared with the distributions derived according to eq . ( [ eq_wdmd ] ) assuming different ifmrs , both empirical and theoretical ones . the relations by weidemann ( 1987 ) and herwig ( 1996 ) , though being quite different , yields wdmds both acceptably consistent with the observed one . this can be explained considering that the major differences between the two ifmrs show up for @xmath174 , corresponding to wds that do not contribute to the mass peak . for @xmath175 the relations are quite similar , showing a rather flat trend . as far as the purely theoretical wdmds are concerned ( bottom panels ) , it turns out that a satisfactory reproduction of the observed data is attained by both hg97 and m2k ( this work ) , whereas a notable discrepancy affects the predictions by rv81 . as already anticipated in the discussion on the predicted ifmrs ( sect . [ sssec_minmfin ] ) , in rv81 there is a sizable overproduction of wds more massive than @xmath176 , a feature already pointed out by bragaglia et al . ( 1995 ) . the carbon star luminosity function ( cslf ) is a fundamental observable as it gives indications on , at least , two basic processes occurring in tp - agb stars with different masses , namely : the third dredge - up that determines the increase of the surface carbon abundance , and mass loss by stellar winds , that affects the duration , hence the luminosity excursion during this phase . iben ( 1981 ) first pointed out the so - called `` carbon star mystery '' , corresponding to a long - standing discrepancy between theory and observations , i.e. current stellar models predicted a deficit of faint carbon stars , accompanied by an excess of bright carbon ( in general agb ) stars . this situation is exemplified in fig . [ fig_cslf ] , where we report the predicted cslf for the lmc , according to the calculations performed by iben & renzini ( 1983 ) , with model prescriptions very similar to rv81 . for this particular case , the authors adopt the following set of parameters : efficiency parameter @xmath177 in the reimers ( 1975 ) mass - loss formula ; mixing - length parameter , @xmath178 ; minimum core for the third dredge - up to occur , @xmath179 ; and efficiency of the third dredge - up , @xmath20 , as a function of the core mass . the cslf in the lmc is instead very well fitted ( fig . [ fig_cslf ] ) by the other two agb synthetic models here considered , namely : groenewegen & de jong ( 1993 , gdj93 ; top right - hand side panel ) and marigo et al . ( 1999a , mgb99 ; bottom left - hand side panel ) . we remind again that , differently from rv81 , in these studies the third dredge - up is suitably calibrated in order to reproduce the clsf in the lmc . finally , it should be remarked that marigo et al . ( 1999a ) have extended the analysis to the cslf in the smc ( see bottom right - hand side panel of fig . [ cslf ] ) , so as to include a metallicity - dependent treatment of the third dredge - up in their synthetic agb model . the observed chemical composition of planetary nebulae represents another crucial constraint for stellar models , as it is the record of the nucleosynthetic and mixing processes occurred during the previous stellar evolution . as far as hg97 predictions are concerned , a detailed discussion is given in groenewegen & de jong ( 1994 ) and it will not be repeated here . we restrict here to the results of rv81 and m2k which are compared in fig . [ fig - chempn ] with the measured abundances of he , c , and o in galactic planetary nebulae . predicted pn abundances can be found in tables a13 a15 . since a full analysis is beyond the purpose of this work , we simply consider the most relevant aspects . both rv81 and m2k results shown here are derived from calculations of stellar models with initial solar composition . therefore , they can not reproduce the data points with he / h @xmath180 , since these latter most likely correspond to progenitor stars with initial subsolar metallicity . the expected paths of pn abundances as a function of the initial mass of the progenitor star reflect the efficiency and duration of the involved processes . for instance , in both models the c / o ratio first increases at increasing mass due to the third dredge - up , and then ( for @xmath181 ) it starts to decline because of hbb . an interesting point is the anticorrelation between the n / o ratio and the c / o ratio exhibited by observed pne with the highest helium content ( he / h@xmath182 ; the so - called type i pne according to the classification introduced by peimbert ( 1978 ) ) . this trend is reproduced by m2k synthetic calculations , tracing the signature of hbb in the most massive agb stars ( @xmath183 ) , where carbon in the envelope is quickly converted into nitrogen . note that theoretical results are notably sensitive to the adopted value for the mixing - length parameter , i.e. the larger @xmath5 is , the more efficiently hbb operates , yielding higher n / o and lower c / o ratios . limiting to the observed sample of pne , we might deduce that hbb has operated in the most massive progenitors of solar metallicity , but with a rather mild efficiency , in agreement with the conclusion already mentioned by henry et al . ( 2000a ) . in fact , as we can see from fig . [ fig - chempn ] , predictions for the case ( @xmath184 ) which correspond to strong hbb lead to n / o ratios quite higher than observed . hower , these indications should be considered with some caution , as the considered sample of pne might not cover the whole relevant mass range of the progenitors ( see henry et al . 2000a ) , and predictions of pne abundances are derived under very simple assumptions ( see appendix a ) . a much better approach will be adopted with the aid of a detailed synthetic model of pn evolution , which is being developed ( marigo et al . 2001 , in preparation ; see marigo et al . 1999b for a preliminary presentation ) . finally , we would remark that rv81 results for the most massive stellar models shown in fig . [ fig - chempn ] with dotted lines _ do not _ correspond to pn abundances , but rather to surface abundances just prior the progenitor stars explode as sne i-1/2 . therefore , attention must be paid not to use these data for a comparison with observed pn abundances . we will compare here the stellar yields presented in this work with those calculated by rv81 and hg97 . possible differences are then discussed on the basis of different model prescriptions and observational constraints , already mentioned in the previous sections . we choose not to make a direct comparison between yields produced by models with the same initial mass , because different sets of yields cover different mass - ranges in the domain of low- and intermediate - mass stars ( see sect . [ sssect_mup ] ) . for this reason , it is more meaningful to compare the whole chemical contribution provided by low- and intermediate - mass stars belonging to a given simple ( i.e. coeval ) stellar population . to this aim , we recall that according to the standard definition ( tinsley 1980 ) , the _ yield from a stellar generation _ , @xmath185 , is the mass converted into the chemical element @xmath39 and ejected by all stars per unit mass locked into stars : @xmath186 in the above equation @xmath187 is the stellar yield of the @xmath156 element ( see sect . [ yi ] ) , and @xmath188 is the _ returned fraction _ , expressing the fraction of mass that has formed stars and then been ejected : @xmath189 \phi(m_{\rm i } ) { \rm d}m_{\rm i } } { \int_{m_{\rm l}}^{m_{\rm u } } m_{\rm i } \phi(m_{\rm i } ) { \rm d}m_{\rm i } } \ ] ] where @xmath190 is the imf ( by number ) defined between the lower ( @xmath191 ) and upper ( @xmath192 ) mass limits ; @xmath193 is the remnant mass . in order to weigh the sole contribution from the generation of low- and intermediate - mass stars , let us consider the quantity : @xmath194 that is similar to eq . ( [ ytot ] ) with @xmath195 . the adopted integration extremes are @xmath196 , @xmath197 , and @xmath13 according to the set of stellar yields under consideration ( see table [ tab - mod ] and sect . [ sssect_mup ] ) . the imf is expressed by the classical salpeter s law , i.e. @xmath198 with @xmath199 ( salpeter 1955 ) . the quantities @xmath200 express the relative chemical contribution ( for a given elemental species @xmath39 ) from low- and intermediate - mass stars belonging to a given simple stellar population . they are shown in fig . [ sspy ] as a function of the metallicity for the three sets here considered . it should be remarked that the differences between our results ( m2k ) and those derived by rv81 and hg97 are not only due to the different mass - range covered by low- and intermediate- mass stars , but mainly reflect substantial differences in the adopted physical prescriptions as already illustrated in sect . [ modpre ] . differences essentially show up both in metallicity trends and absolute values of @xmath201 . compared to previous calculations , m2k yields show a pronounced dependence on the metallicity , i.e. positive yields increase with decreasing @xmath43 . conversely , the rv81 and hg97 sets present weak trends with @xmath43 . the metallicity dependence can be explained as follows . on one side , agb lifetimes of low - mass stars increase at decreasing metallicities , as mass - loss rates are expected to be lower . this fact leads to a larger number of dredge - up episodes . moreover , both the onset and the efficiency of the third dredge - up are favoured at lower metallicities . these factors concur to produce a greater enrichment in carbon . on the other side , hbb in more massive agb stars becomes more efficient at lower metallicities , leading to a greater enrichment in nitrogen . the combination of all factors favours higher positive yields of helium at lower @xmath43 . as far as the single elemental species are concerned , we can notice : * m2k yields of @xmath28he are larger than those by hg97 and rv81 towards lower @xmath43 , likely due to the earlier activation and larger efficiency of the third dredge - up in our models . with respect to rv81 and hg97 predictions , our yields of @xmath28he present a significant trend with @xmath43 . * m2k yields of @xmath2c are systematically higher than those of rv81 and hg97 because of the earlier onset ( and average greater efficiency than in rv81 ) of the third dredge - up . * the dominant contribution to the yields of @xmath3n comes from hbb in the most massive agb stars . differences in the results reflect different efficiencies of nuclear reactions and agb lifetimes . in particular , according to m2k the production of @xmath3n , mainly of primary synthesis , is favoured at lower @xmath43 , and is sistematically lower than rv81 and hg97 results . this latter difference can be explained considering the drastic reduction of tp - agb lifetimes for the most massive agb models with hhb ( see fig . [ fig - time ] ) in m2k models with respect to rv81 . in general , our expected dependence of chemical yields on metallicity is far for being linear , and much caution should be used when extrapolating these quantities with respect to @xmath43 in chemical evolutionary models . we can not verify whether such a non - linear relation with metallicity was displayed also in the rv81 models , since just two values of metallicities were considered there . however , the very high number of thermal pulses suffered by stars with hbb regardless of the metallicity , according to the rv81 models , might already explain the apparent lower sensitiveness of their yields to the metallicity . we would like to outline briefly the aims and findings of the present work . in the first part we have presented new homogeneous sets of stellar yields ejected from low- and intermediate - mass stars , in view of providing updated ingredients for modelling the chemical evolution of complex stellar systems . thanks to the updated input physics employed in the calculations , and the improved treatment of both the third dredge - up and hot - bottom burning , the present estimation of the stellar yields from low- and intermediate - mass stars has led to new results and developments . in particular , a pronounced trend of the yields with the metallicity is expected . specifically , at given stellar mass positive yields of @xmath1he , @xmath2c , @xmath3n , and @xmath4o are larger at decreasing metallicity . this feature is the result of concurring factors : at lower @xmath43 both the third dredge - up and hot - bottom burning are more efficient , and tp - agb lifetimes are , on average , longer because of lower mass - loss rates . moreover , it is interesting to notice that low - mass stars may produce positive yields of @xmath4o , which is brought up to the surface by the third dredge - up . the entity of this contribution ( as well as for that of @xmath2c ) depends crucially on the efficiency @xmath20 , number of dredge - up episodes , and chemical composition of the convective intershell ( for this latter , see boothroyd & sackmann 1988b , and herwig 2000 for different model results ) . the possible production and ejection of newly synthesised oxygen from low- and intermediate - mass stars may be an interesting prediction to be tested through its consequences in various possible applications , mostly in relation to the chemical composition of planetary nebulae ( e.g. pquignot et al . 2000 ) , and galactic chemical evolutionary models . the second part of the paper is meant to examine several aspects concerning the stellar models which the chemical yields are derived from . to this aim , we have analysed how the main model prescriptions ( i.e. treatment of convective boundaries , mass loss , analytical relations , third dredge - up parameters , treatment of hot - bottom burning , etc . ) may affect the predictions of stellar yields . finally , the third final part is dedicated to compare our new yields with other available data sets of large usage , in the attempt of explaining the existing differences as the result of particular assumptions . to do this , we have considered basic observational constraints which are closely related to the stellar yields namely : i ) the initial - final mass relation ; ii ) the white dwarf mass distribution ; iii ) the carbon star luminosity function ; and iv ) the chemical composition of planetary nebulae and tested the capability of different models in reproducing them through a direct comparison between predictions and observations . in particular , it has been shown how much the choice of calibrating fundamental efficiency parameters ( e.g. of the third dredge - up and mass loss ) in recent works has changed the predictions of stellar yields compared to earlier studies . to conclude , we wish this work has somehow contributed to clarify a few important general points on theoretical stellar yields , in view of stimulating an aware and critical usage of them . i would like to thank lo girardi , laura portinari , and cesare chiosi for their important professional advice and support , and the referee , dr . henry , for his helpful remarks on this work . this study has been financially supported by the italian ministry of university , scientific research and technology ( murst ) under contract `` formation and evolution of galaxies '' n. 9802192401 . alongi m. , bertelli g. , bressan a. , chiosi c. , fagotto f. , greggio l. , nasi e. , 1993 , a&as 97 , 851 baud b. , habing h.j . , 1983 , a&a 127 , 73 bergeron p. , saffer r.a . , liebert j. , 1992 , apj 394 , 228 blcker t. , 1995 , a&a 297 , 727 ( b95 ) blcker t. , schnberner d. , 1991 , a&a 244 , l43 boothroyd a.i . , sackmann i .- j . , 1988a , apj 328 , 641 boothroyd a.i . , sackmann i .- j . , 1988b , apj 328 , 653 boothroyd a.i . , sackmann i .- j . , 1999 , apj 510 , 232 bowen g.h . , 1988 , apj 329 , 844 bragaglia a. , renzini a. , bergeron p. , 1995 , apj 443 , 735 caughlan g.r . , fowler w.a . , 1988 , atomic data nucl . data tables 40 , 283 charbonnel c. , 1995 , apj 453 , l41-l44 chiosi c. , bertelli g. , bressan a. , 1992 , ara&a 30 , 305 costa e. , frogel j.a . , 1996 , aj 112 , 2607 crotts , a.p.s . , bechtold j. , fang y. , duncan r.c . , 1994 , bull . american astron . 185 , 1807 dominguez i. , chieffi a. , limongi m. , et al . , 1999 , apj 524 , 226 fleischer a.j . , gauger a. , sedlmayr e. , 1992 , a&a 266 , 321 forestini m. , charbonnel c. , 1997 , a&as 123 , 241 garcia - berro e. , ritossa c. , iben i. , 1997 , apj 485 , 765 girardi l. , bressan a. , bertelli g. , chiosi c. , 2000 , a&as 141 , 371 graboske h.c . , de witt h.e . , grossman a.s . , cooper m.s . , 1973 , aj 181 , 457 gratton r.g . , sneden c. , carretta e. , bragaglia a. , 2000 , a&a 354 , 169 groenewegen m.a.t . , de jong t. , 1993 , a&a 267 , 410 groenewegen m.a.t . , de jong t. , 1994 , a&a 282 , 115 henry r.b.c . , kwitter k.b . , bates j.a . , 2000a , apj , 531 , 928 henry r.b.c . , edmunds m.g . , kppen j. , 2000b , apj , 541 , 660 herwig f. , 1996 , in stellar evolution : what should be done , 32nd lige int . astrophys . coll . a. noels et al . , p. 441 herwig f. , blcker t. , schnberner d. , el eid m. , 1997 , a&a 324 , l81 herwig f. , 2000 , a&a 360 , 952 van den hoek l.b . , groenewegen m.a.t . , 1997 , a&as 123 , 305 ( hg97 ) iben i. , 1981 , apj 246 , 278 iben i. , renzini a. , 1983 , ara&a 21 , 27 iben i. , truran j.w . , 1978 , apj 220 , 980 ( it78 ) jeffries r.d . , 1997 , mnras 288 , 585 jeffries r.d . , james d.j . , thurston m.r . , 1998 , mnras 300 , 550 kingsburgh r.l . , barlow m.j . , 1994 , mnras 271 , 257 koester d. , reimers d. , 1996 , a&a 313 , 810 lu , l .. 1991 , apj 379 , 99 lu , l. , sargent w.l.w . , barlow t.a . , 1998 , aj 115 , 55 maeder a. , 1983 , a&a 120 , 113 maeder a. , meynet g. , 2000 , ara&a 38 , in press marigo p. , 1998a , phd thesis , university of padova marigo p. , 1998b , a&a 340 , 463 marigo p. , bressan a. , chiosi c. , 1996 , a&a 313 , 545 marigo p. , bressan a. , chiosi c. , 1998 , a&a 331 , 564 marigo p. , girardi l. , bressan a. , 1999a , a&a 344 , 123 marigo p. , weiss a. , groenewegen m.a.t . , girardi l. , 1999b , in from extrasolar planets to cosmology : the vlt opening symposium , proceedings eso / mpa parallel workshop 2 : star - way to the universe , held in antofagasta ( chile ) , march 1999 , eds . bergeron j. , renzini a. , p. 248 - 251 peimbert m. , 1978 , in iau symposium 76 , planetary nebulae , ed . y. terzian , dordrecht : reidel , p. 215 pquignot d. , walsh j.r . , zijlstra a.a . , dudziak g. , 2000 , a&a 361 , l1 pettini m. , lipman k. , hunstead r.w . , 1995 , apj 451 , 100 portinari l. , chiosi c. , bressan a. , 1998 , a&a 334 , 505 prantzos , n. , vangioni - flam e. , casse m. , 1994 , ras canada j. 88 , 356 rebeirot e. , azzopardi m. , westerlund b.e . , 1993 , a&as 97 , 603 reid , n. , 1996 , aj 111 ( 5 ) , 2000 reimers d. , 1975 , mem . r. sci . lige , ser . 6 , vol . 8 , p. 369 renzini a. , voli m. , 1981 , a&a 94 , 175 ( rv81 ) salpeter e.e . , 1955 , apj 121 , 161 tinsley b.m . , 1980 , of cosmic phys . 5 , 287 tytler d. , fan x .- , 1994 , apj 424 , l87 vassiliadis e. , wood p.r , 1993 , apj 413 , 641 vila - costas m.b . , edmunds m.g . , 1993 , mnras 265 , 199 wagenhuber j. , groenewegen m.a.t . , 1998 , a&a 340 , 183 wasserburg g.j , boothroyd a.i . , sackmann i .- j . , 1995 , apj 447 , l37 weidemann v. , 1987 , a&a 188 , 74 weidemann v. , 1997 , in advances in stellar evolution , proceedings of the workshop stellar ecology , cambridge university press , p. 169 weiss a. , denissenkov p.a . , charbonnel a. , 2000 , a&a 355 , 299 woosley s.e . , weaver t.a . , 1982 , in supernovae : a survey of current research , reidel , dordrecht , eds . rees m. j. , stoneham r. j. , p.79 wood , p.r . , 1981 , in : physical processes in red giants , proc . of the second workshop , erice , italy ( september 3 - 13 , 1980 ) , dordrecht , d. reidel publishing co. , p. 135 tables a1 a12 contain the yields from low- and intermediate - mass stars , in the form @xmath202 ( see eq . ( [ defy1 ] ) ) , as a function of the initial mass @xmath40 and metallicity @xmath43 . all specified masses are expressed in solar units . concerning low - mass stars ( with @xmath203 ) , we give separately the yields ejected during the rgb ( tables a1 a3 ) and agb phases ( tables a4 a6 ) . the quantity , @xmath204 , corresponds to the mass at the onset of the tp - agb phase , which is smaller than @xmath40 by the amount of mass lost during the previous rgb phase , @xmath205 . the mass lost during the agb is denoted with @xmath206 . total yields from stars with @xmath209 are given in tables a10 a12 for three values of the mixing - length parameter @xmath5 , and distinguishing between the secondary ( s entry ) and primary ( p entry ) components in the case of the cno elements . tables a13 a15 present the predicted abundances ratios ( he / h , c / h , n / h , o / h ) in planetary nebulae , as a function of the initial stellar mass ( @xmath40 ) , metallicity @xmath43 , and mixing - length parameter . the pne chemical abundances ( by number , in mole gr@xmath135 ) are calculated by averaging the abundances in the wind ejecta over the last stages ( i.e. a time period @xmath210 yr ) on the agb , weighted by the masses of the ejecta . for the sake of simplicity , we do not consider the question on the actual observability of pne ( which depends on both dynamical and ionisation properties ) , and the fact that evolutionary timescales of pne ( hence the time @xmath211 ) may largely vary according to the mass of the progenitor star . these points deserve a more complex study , which is currently in progress and presented in a preliminary form by marigo et al . ( 1999b ) .
traditionally , people regarded that three - dimensional models are good toy models with which to study the basic nature of pure gravity and quantum gravity @xcite . first , there may be a qualitative relation between a cosmic string in four dimensions and a point source in three dimensions @xcite . second , three - dimensional gravity is simpler than the four - dimensional version , making it useful for obtaining some intuitions about quantum gravity using gauge theoretical correspondences @xcite . however , given that this is too simple , it was doubtful that a black hole solution would be found in a three - dimensional background . the first analytic solution in three - dimensional gravity was found by banados , teitelboim , and zaneli @xcite@xcite . they found the black hole solution in an anti de sitter background . in their solutions , the negative cosmological constant played a crucial role . these solutions can have mass , charge , and angular momentum . moreover , the solutions can have the hawking temperature @xcite . however , they are thermodynamically stable because the specific heat is positive @xcite . the so - called btz solutions are useful in the context of holography . according to the ads / cft correspondence @xcite , there is a relationship between the bulk three - dimensional gravity and the conformal field theory at the asymptotic two - dimensional boundary of such a three - dimensional bulk . that is , we may relate the gravitational perturbation on the bulk to the thermal excitation on the boundary field theory . moreover , it is known that the u - duality transformation maps four or five - dimensional black holes to three - dimensional btz black holes @xcite ; here , these black holes , which are connected by the u - duality , have the same number of microstates . hence , the microscopic derivation of the entropy can be replaced by the study on the corresponding derivation for the btz black hole @xcite . in these senses , three - dimensional black holes are more important than just an artificial toy model . in this paper , we will begin to study the dynamics of @xmath0-dimensional charged black holes , including gravitational collapses and back - reactions , using the semi - classical effects of hawking radiation . we solved time - dependent einstein and field equations using the double - null formalism @xcite@xcite@xcite . of course , it is too ambitious to extend this method to holography or string theory in this paper . however , if we establish a method with which to study dynamical black holes in an anti de sitter background , it will eventually become possible to study the time - dependent correspondences between the bulk and boundary someday . thus far , there is less reference to compare our calculations on the dynamic formation of three - dimensional charged black holes ( e.g. , @xcite ) , whereas there is some literature on the gravitational collapses of neutral matter @xcite , especially in the context of critical collapses @xcite . on the other hand , in four dimensions , there are numerous studies of gravitational collapses and evaporations @xcite@xcite . one important feature in four - dimensional dynamical charged black holes is the existence of mass inflation @xcite . mass inflation is a phenomenon in which the local mass function of an observer near the cauchy horizon increases exponentially . if it is true , then the internal structure of charged black holes can not be described by the reissner - nordstrm solution @xcite and the internal structure should be significantly modified @xcite . this was directly confirmed by numerical simulations , and there are more related works @xcite@xcite . three - dimensional gravity will be similar as regards this point , as it is expected that there is mass inflation in general in charged black holes @xcite . however , there are some expectations that three - dimensional mass inflation is different from four - dimensional cases @xcite . moreover , because black holes are thermodynamically stable in three - dimensions , it will be interesting to determine what the semi - classical back - reactions are in the internal structures of black holes ( there are discussions on the semi - classical effects in the static limit @xcite ) . this paper is organized as follows : in section [ sec : pre ] , we summarize analytic solutions for three - dimensional charged black holes and discuss the mass inflation issue . the paper will then provide the theoretical motivations to study three dimensions . in section [ sec : mod ] , we describe our model of the numerical simulations , the implementation to the double - null formalism , and comment on the initial conditions and dimensional analysis . in section [ sec : dyn ] , we report the results . first , we report results for cases without semi - classical effects . second , we include semi - classical effects and observe how they modify causal structures . finally , in section [ sec : con ] , we conclude and comment on causal structures and cosmic censorship issues . in this section , first we briefly review the analytic and static charged black hole solutions of three dimensions . second , we summarize thermodynamic properties and internal stability issues via mass inflation regarding their solutions . for three dimensions , the first simple black hole solution in the anti de sitter background is known to be the btz solution ( for non - rotating cases ) @xcite . for the einstein - maxwell action @xmath2 the solution becomes @xmath3 where @xmath4 @xmath5 , @xmath6 is electric charge , and @xmath7 is interpreted by the mass . if @xmath8 , then we obtain a neutral black hole solution with @xmath9 . thus although this is in the @xmath10 limit , the center still contains a singularity ( @xmath11 in figure [ fig:3d ] ) . the proper anti de sitter limit is obtained by the @xmath12 limit : @xmath13 if @xmath14 , then there can be two horizons : the outer horizon @xmath15 and the inner horizon @xmath16 . if @xmath17 , then there are always two horizons with @xmath18 and @xmath19 ( case @xmath20 in figure [ fig:3d ] ) . however , if @xmath21 , then the number of horizons depends on @xmath6 : for a small @xmath6 limit , there can be two horizons @xmath22 ( case @xmath20 in figure [ fig:3d ] ) ; for a large @xmath6 limit , there is no horizon ( case @xmath23 in figure [ fig:3d ] ) ; and the extreme limit , @xmath24 and @xmath25 ( case @xmath26 in figure [ fig:3d ] ) . in the neutral limit , the singularity @xmath27 is space - like ( @xmath8 in figure [ fig:3d ] ) ; in the charged limit , the singularity is time - like ( except @xmath11 ) . therefore , it is not difficult to draw penrose diagrams in the static limit @xcite . penrose diagrams of btz solutions for @xmath8 and @xmath28 , @xmath29 , case @xmath20 ( there are two horizons ) , case @xmath26 ( there is one horizon ) , and case @xmath23 ( there is no horizon but a time - like singularity ) . ] the hawking temperature @xmath30 becomes @xmath31 and @xmath30 is positive or zero because @xmath15 is greater than the extreme limit size @xmath32 @xcite@xcite . the entropy @xmath33 becomes @xmath34 however , the specific heat @xmath35 becomes @xmath36 and hence the black hole is in thermal equilibrium and the outer horizon does not shrink @xcite . therefore , when we consider the dynamical formation of a black hole , the outer horizon should be always space - like or null . note that in the double - null coordinates , @xmath37 the einstein equation for the @xmath38 component around the outer apparent horizon @xmath39 is @xmath40 therefore , the outer apparent horizon is space - like or null if and only if @xmath41 . in other words , for btz solutions , even when they are dynamically generated with a hawking temperature , @xmath42 should be positive or @xmath43 around the outer apparent horizon . there are a number of discussions on the instability of cauchy horizons @xmath16 @xcite@xcite@xcite . we can re - write the metric in the following form : @xmath44 here , @xmath45 is the space dimensions ( @xmath26 or @xmath23 ) . now , without loss of generality , we can ignore the angular part to discuss in - going or out - going null geodesics . thus , for an in - falling matter along the in - going null direction ( for coordinates @xmath46 $ ] ) , the energy - momentum tensor components @xmath47 can be represented by @xmath48 where @xmath49 is an arbitrary function that refers to the luminosity function @xcite@xcite . at this point , if there is an observer who moves along the out - going null direction and approaches the cauchy horizon @xmath16 , then for a null geodesic of the observer @xmath50 , @xmath51 and @xmath52 around the @xmath53 limit , where @xmath54 is an affine parameter of the observer and @xmath55 is the surface gravity of the inner horizon . in conclusion , the out - going observer who approaches the cauchy horizon measures the energy density for the energy flow that flows along the cauchy horizon by @xcite@xcite : @xmath56 note that for four dimensions , for a realistic black hole the behavior of the luminosity function should be polynomial to @xmath57 @xcite . although there is no such guidance regarding the luminosity function for three dimensions , if we believe that such a polynomial fall - off is reasonable for three - dimensional cases , then we can conclude that for both three- and four - dimensional cases , any small energy flow along the cauchy horizon is amplified for an observer who approaches to the cauchy horizon . this effect is known as _ mass inflation_. therefore , the inner horizon can not be stable and there should be strong back - reactions due to high energy density @xcite . in many cases , it is believed that the cauchy horizon should be a curvature singularity @xcite . hence , we can not trust the static solutions to study the inside structures of charged black holes . therefore , it is necessary to introduce numerical tools to study these structures . we have two main motivations for this paper . 1 . * theoretical interests : * three - dimension is important . first , it is a toy model of a black string that may be possible in the early universe . second , as we commented in the introduction , in string theory , there is a holographic correspondence between bulk black holes in an anti de sitter space and its boundary conformal field theory . one can relate the gravitational perturbation on the bulk to the thermal excitation on the boundary field theory and one may restrict the perturbation by the spherically symmetric scalar sector . although the effect of the scalar perturbation on the boundary field theory is not yet revealed fully , our study and methodology on dynamical black holes may shed light on the study of the dynamical correspondence between gravity and field theory . * numerical interests : * according to previous discussions , charged black holes in three dimensions will have mass inflation around the inner horizon . this means that the inner horizon and internal structures can not be described by analytical methods . to induce this mass inflation , the black hole should be fully dynamical . then it is inevitable to use numerical methods . therefore , there are various interests to study three - dimensional black holes . to study internal structures , it is inevitable to use the numerical methods . in this paper , we study the dynamical formation of spherical symmetric charged black holes and its back - reactions for three dimensions . we prepare a complex scalar field with the maxwell field ; it is used to form a charged black hole . then , it will make it possible to understand mass inflation and the possible curvature singularities inside of three - dimensional black holes . moreover , we can include hawking radiation by introducing re - normalized energy - momentum tensors with the @xmath1-wave approximation . we can check whether the horizon behavior and the energy condition are consistent with the thermodynamics results , even when we include the re - normalized energy - momentum tensors . we can also determine the back - reaction of the re - normalized energy - momentum tensor to the inside structure of the black hole . in this section , we introduce a model for three - dimensional charged black holes . to use numerical simulations , we introduce the double - null formalism and discuss the details of the initial conditions and free parameters for the simulations . also , we briefly comment on the dimensional analysis of each functions and parameters . let us introduce the action in three dimensions by @xmath58\end{aligned}\ ] ] where @xmath59 is the ricci scalar , @xmath60 is a complex scalar field with a gauge coupling @xmath61 and a gauge field @xmath62 , where @xmath63 . the einstein equation becomes as follows : @xmath64 where the total energy - momentum tensors are @xmath65 the classical part @xmath66 is @xmath67 and @xmath68 is the re - normalized energy - momentum tensor part to include semi - classical effects , where @xmath69 . the field equations for the complex scalar field @xmath60 and the maxwell field @xmath62 are as follows : @xmath70 we use the double - null coordinates @xmath71 assuming spherical symmetry , where @xmath72 is the retarded time , @xmath73 is the advanced time , and @xmath74 is the angular coordinate . we follow the notation of @xcite@xcite@xcite@xcite : the metric function @xmath75 , the radial function @xmath76 , and a scalar field @xmath77 , and define @xmath78 the einstein tensors are then given as follows : @xmath79 via the gauge symmetry of the maxwell field , we can choose the maxwell field @xmath80 by introducing the function @xmath81 . thus , we can obtain the energy - momentum tensors of the classical part with the cosmological constant term by : @xmath82,\\ t^{\mathrm{c}}_{uv } - \frac{\lambda}{8 \pi } g_{uv } & = & \frac{{(a_{,v})}^{2}}{4 \pi \alpha^{2 } } + \frac{\alpha^{2}}{16\pi}\lambda,\\ t^{\mathrm{c}}_{vv } - \frac{\lambda}{8 \pi } g_{vv } & = & \frac{1}{4\pi } z\overline{z},\\ t^{\mathrm{c}}_{\varphi\varphi } - \frac{\lambda}{8 \pi } g_{\varphi\varphi } & = & \frac{r^{2}}{4 \pi \alpha^{2 } } \left [ ( w\overline{z}+z\overline{w } ) + iea(\overline{z}s - z\overline{s } ) + \frac{2 { ( a_{,v})}^{2}}{\alpha^{2 } } \right ] - \frac{r^{2}}{8\pi}\lambda.\end{aligned}\ ] ] note that @xmath83 , where @xmath84 is asymptotically interpreted by the electric current . if we define @xmath85 as the charge function , then @xmath86 and hence corresponds to the concept of the charge . there are several discussions pertaining to re - normalized energy - momentum tensors for three dimensions in the literature @xcite . as expected for two- or four - dimensional cases @xcite@xcite , the re - normalized energy - momentum tensors diverge along the inner horizon for the static metric . however , we know that the internal structures can not be described by the static metric for fully dynamical cases . due to dynamical situations and spherical symmetry , therefore , the following represents the best choice for the re - normalized energy - momentum tensor : use the @xmath87-dimensional results @xcite@xcite divided by @xmath88 and apply the @xmath1-wave approximation by @xmath89 where @xmath90 and @xmath91 is the number of massless scalar fields . after solving these coupled equations , we can write all equations : @xmath92 including the field equations @xmath93 at this point , the equations of @xmath94 , @xmath95 , and @xmath96 parts can be represented by first - order differential equations . we can then implement the same integration scheme used in previous papers @xcite@xcite . we used the second - order runge - kutta method @xcite ; however , due to the differentials in equations ( [ eq : e1 ] ) and ( [ eq : e2 ] ) , our results will converge to the first order if @xmath97 . tests of the convergence are provided in appendix b. one important question when interpreting our results is whether or not these black hole solutions are btz solutions . of course , due to the birkhoff theorem for three - dimensional cases , we can be sure that our solutions will be btz ( for further discussions , see appendix a ) . hence , there will be a natural concept for the mass function . first , we can define the general spherical symmetric metric in @xmath98 dimensions in the static limit by the following coordinates : @xmath99 where @xmath100 in many cases , function @xmath101 vanishes near the horizon ; hence , it will be convenient to redefine the coordinate in the following form : @xmath102 where the coordinate transformation is @xmath103 thus , we can show @xmath104 note that @xmath105 is negative outside of the black hole . therefore , the signature is fine . moreover , we can easily see that the apparent horizons @xmath39 or @xmath106 make @xmath107 , although @xmath75 is not zero . thus , in the four dimensions , we can define the mass function @xcite by @xmath108 also , for three dimensions , we can similarly define the mass aspect by @xmath109 this definition is consistent with the findings in the literature @xcite . integration domain of our simulations . we assign initial conditions along the in - going and out - going null surfaces along the arrows . ] .[table : conditions]summary of the assignments of initial conditions for @xmath110 ( the in - going initial surface ) and @xmath111 ( the out - going initial surface ) . [ cols="^,^,^,^",options="header " , ] therefore , we may conclude that , when @xmath112 , there is surely a central singularity while for the other cases this is not evident . of course , via the singularity theorem , if there is an apparent horizon , there will be a singularity anyway . for the @xmath113 and @xmath114 cases , however , there is no horizon . in these cases , _ these exponents are circumstantial evidence that there may not be a singularity around the center . _ ( this conclusion is consistent with the result of @xcite . ) of course , if the distribution of the charge drastically changes beyond @xmath115 , the central structure will be modified and our extrapolation will be meaningless . therefore , our prospect depends , in all respects , on circumstantial evidence and not on real evidence . top and middle : @xmath116 and @xmath117 for @xmath118 , @xmath119 , and @xmath120 , for the @xmath121 case . bottom : @xmath117 for @xmath122 , @xmath123 , and @xmath124 , for the @xmath121 case.,title="fig : " ] top and middle : @xmath116 and @xmath117 for @xmath118 , @xmath119 , and @xmath120 , for the @xmath121 case . bottom : @xmath117 for @xmath122 , @xmath123 , and @xmath124 , for the @xmath121 case.,title="fig : " ] top and middle : @xmath116 and @xmath117 for @xmath118 , @xmath119 , and @xmath120 , for the @xmath121 case . bottom : @xmath117 for @xmath122 , @xmath123 , and @xmath124 , for the @xmath121 case.,title="fig : " ] there are several studies in the literature that closely investigated the mass inflation in three - dimensional black holes @xcite . one of the interesting predictions was that there is a curvature singularity for a certain curvature component , whereas some of the other curvature components do not diverge as @xmath73 increases . we observed the ricci scalar @xmath59 ( figure [ fig : ricci_p0 ] ) , where @xmath125 @xmath126 for an out - going null observer @xmath127 ( top in figure [ fig : tuu_p0_2 ] ) , @xmath128 for an in - going null observer @xmath129 , for some out - going ( middle in figure [ fig : tuu_p0_2 ] ) and in - going null slices ( bottom in figure [ fig : tuu_p0_2 ] ) ( @xmath118 , @xmath119 , and @xmath120 ; @xmath122 , @xmath123 , and @xmath124 ) . note that we choose ( @xmath130 $ ] ) @xmath131 so that @xmath132 and @xmath133 . thus , any time - like geodesic @xmath134 can be decomposed by @xmath135 with constraint @xmath136 . hence , any time - like observer will see local energy density by @xmath137 therefore , if there is a signature for a case in which @xmath117 or @xmath116 increases exponentially , then it is the evidence of mass inflation . the results are interesting : there is no evidence that the ricci scalar or @xmath116 diverge , while @xmath117 shows exponential growth . thus , this provides clear evidence that there is mass inflation and that a curvature singularity forms around the inner cauchy horizon at the @xmath138 limit . however , this behavior of curvature functions is very different from that noted in four - dimensional cases . for four - dimensional cases , almost all related functions ( e.g. , ricci scalar , kretchmann scalar , misner - sharp mass function ) diverge exponentially , while the ricci scalar and @xmath116 components do not diverge in three dimensions . therefore , this is a typical property of three - dimensional charged black holes . in this section , now we turn to the semi - classical effects and see dynamical back - reactions for the inside structures . we put @xmath139 and vary parameters @xmath61 ( figure [ fig : p ] ) , as was done in the previous section . numerical results for @xmath139 ( with hawking radiation ) while varying @xmath61 . ] for the @xmath140 , @xmath141 , and @xmath142 cases , there is a new horizon where @xmath143 . for the @xmath140 case , there is a space - like singularity . hence , between the @xmath143 horizon and the central singularity , there is a fuzzy region and a cauchy horizon along an ingoing null direction . the @xmath39 and @xmath106 horizons can be separated as in the @xmath121 case , while for small charge cases , they are mostly foliated . because @xmath144 and @xmath145 both change their signs around the foliated region , any observer who moves through the horizon sees an increase of areal radius @xmath76 . therefore , it is a type of a bottleneck in a one - way wormhole . note that the size of the bottleneck is on the order of @xmath146 . for charged black holes in four dimensions , there was a similar effect . however , in those cases , the charge played a crucial role @xcite , while in three dimensions , the charge is not essential to see such a bottleneck , as we can see in the @xmath140 case . rather , it is the effect due to the negative energy via semi - classical effects that concentrated around the center . this bottleneck becomes a curvature singularity as the curvature functions increase exponentially around these regions . here , the charge will make the bottleneck relatively regular . this phenomenon will be discussed in section [ sec : for ] . as we increase the charge , two horizons are separated ( in the @xmath121 case ) and the @xmath106 horizon even disappears ( in the @xmath112 case ) . this suggests that the amount of the negative energy decreases as the charge increases , approaching and exceeding the extreme limit ( @xmath113 and @xmath114 ) . in these cases , the inner apparent horizon can be space - like while the outer horizon remains space - like . this implies that the null energy condition is violated only for inside and that it is not violated around the outer horizon . this will be discussed in section [ sec : ene ] . ricci scalar for @xmath139 along @xmath147 when @xmath148 and @xmath149 ( upper ) and along @xmath150 when @xmath148 and @xmath149 ( lower).,title="fig : " ] ricci scalar for @xmath139 along @xmath147 when @xmath148 and @xmath149 ( upper ) and along @xmath150 when @xmath148 and @xmath149 ( lower).,title="fig : " ] figure [ fig : ricci_2 ] show ricci scalar functions for @xmath139 along @xmath147 and @xmath150 while varying @xmath61 . until @xmath142 , the @xmath39 horizon and the @xmath106 horizon are foliated ; hence , the inner horizon becomes a bottleneck of a wormhole , after which the ricci scalar increases exponentially . unless we assume a large number of scalar fields @xmath91 so that @xmath151 , the internal structure becomes a type of curvature singularity with a non - zero area . note that this is not due to mass inflation , because mass inflation does not make the ricci scalar diverge , as shown in section [ sec : mas ] . also , this effect exists even for the @xmath140 case . therefore , we can clarify that this is a new type of curvature singularity . as was noted in section [ sec : dim ] , the bottleneck occurs via the violation of the null energy condition and the concentration of the negative energy . if the effect of the re - normalized energy - momentum tensor is dominant ( around the order of @xmath152 ) compared to the classical fields , then the region is no longer semi - classical . dvali noticed that such a region is not physically well defined and is hence physically problematic . it is interesting that curvature functions become trans - planckian around such dangerous regions . such a curvature singularity is found by the authors for the first time , and we will term it a _ dvali curvature singularity _ to distinguish it from the mass inflation curvature singularity . both types of curvature singularities are weak , in the sense that the area is non - zero while the curvature becomes trans - planckian @xcite . one can compare the order of magnitude of such curvatures between the dvali curvature singularity and the mass inflation curvature singularity . the former is dominant compared to the latter in terms of the order of magnitude . therefore , the following conclusions are reasonable : * for charged black holes with sufficiently small charge , a space - like singularity partly case . this is due to the fact that the @xmath42 component is positive around the outer horizon via thermodynamic stability , as we shown in the next section . ] changes to a dvali curvature singularity . * for large charged black holes , we can see a mass inflation curvature singularity in the large @xmath73 limit . * between the two limits , both types of curvature singularities can exist . energy - momentum tensor components @xmath42 , @xmath153 , and @xmath42 around the inner horizon , and @xmath154 around the outer horizon , for @xmath121 and @xmath139 . ] figure [ fig : energycondition ] shows the typical distribution of the energy - momentum tensor components of our simulations . the first plot in figure [ fig : energycondition ] is @xmath42 , showing that it is positive during the gravitational collapse ( @xmath155 ) . after the gravitational collapse , the inside structure is governed by the negative energy - momentum tensors @xmath153 and @xmath42 . the second plot in figure [ fig : energycondition ] shows that @xmath153 is negative around the outer part of the @xmath106 horizon and changes its sign around the inner part of the @xmath106 horizon . this is consistent given that @xmath156 and hence , if @xmath145 increases along the ingoing null direction at @xmath106 , @xmath153 should be negative ; if @xmath145 decreases along the ingoing direction at @xmath106 , @xmath153 should be positive . the third plot in figure [ fig : energycondition ] is @xmath42 around the inner horizon . it is negative around the inner apparent horizon . note that around the inner horizon , @xmath157 if @xmath39 horizon is inside of the black hole , to increase @xmath144 from negative to positive , the inner horizon should be space - like , and it is equivalent to @xmath158 . therefore , this is a consistent behavior . finally , the fourth plot in figure [ fig : energycondition ] shows the @xmath154 components . of course , the outer apparent horizon is space - like or time - like if and only if @xmath42 is positive or negative . however , we know that @xmath159 and that competition exists between the two parts . the first classical part is always positive , and hence , to see a time - like horizon , the second quantum part should be negative . moreover , it has to be greater than the first part . however , the fourth plot shows that the quantum part @xmath154 is already positive around the outer horizon . therefore , there is no hope of seeing the evaporation of a black hole in three - dimensional cases . this is consistent with the known thermodynamic stability of three - dimensional black holes , and this calculation confirms the stability in an independent manner @xcite . in this subsection , let us summarize the results of previous sections . in section [ sec : cau ] , we observed causal structures when there is no hawking radiation . in section [ sec : res ] and section [ sec : ene ] , we included hawking radiation and observed some interesting modifications : * space - like singularities are partly changed to a bottleneck of a wormhole ( the dvali curvature singularity ) . * inner horizons at finite @xmath73 can be space - like . ( this depends on the details of the energy - momentum tensors of inside the black hole . ) * a @xmath138 inner horizon remains for sufficiently large charge cases . * the outer apparent horizon is thermodynamically stable . according to section [ sec : mas ] and section [ sec : for ] , we understand that the @xmath138 inner horizons and the bottleneck of the wormhole become curvature singularities . the former is due to mass inflation and the latter is caused by the concentration of negative energy . causal structures with hawking radiation . type @xmath20 , @xmath26 , and @xmath23 in figure [ fig : p0_types ] should be modified to type @xmath160 , @xmath161 , and @xmath162 . ] therefore , for type @xmath20 ( figure [ fig : p0_types ] ) , a part of the space - like singularity will be changed by a dvali curvature singularity ( left of type @xmath160 , figure [ fig : p_types ] ) . we can not trust beyond the dvali curvature singularity , since the curvatures become trans - planckian . however , as was discussed in section [ sec : dim ] , if there are sufficiently large numbers of massless fields that contribute to hawking radiation without changing @xmath163 , beyond the @xmath143 horizon can be trustable in the semi - classical sense , as the curvatures , for example the ricci scalar , will decrease in proportion to @xmath164 . if this is true , our solution implies a branching - off geometry ( right of type @xmath160 , figure [ fig : p_types ] ) . for type @xmath26 ( figure [ fig : p0_types ] ) , again , the space - like singularity should be changed by a dvali curvature singularity and the inner cauchy horizon should be changed by the mass inflation singularity ( type @xmath161 , figure [ fig : p_types ] ) . for type @xmath23 ( figure [ fig : p0_types ] ) , the semi - classical effect can modify the inner horizon to be space - like and there will be a mass inflation singularity in the @xmath138 limit ( type @xmath162 , figure [ fig : p_types ] ) . of course , in that limit , the out - going cauchy horizon will be also a type of mass inflation singularity . in this limit , we may not see the dvali curvature singularity . in this paper , we investigated the dynamical formation and evolution of three - dimensional charged black holes . as a concluding remark , we summarize our new results and point out some interesting issues in terms of two points : classification of causal structures for dynamical black holes and cosmic censorship in three dimensions . we study gravitational collapses of charged matter fields and classified causal structures ( figure [ fig : p0_types ] ) : type @xmath20 for an almost neutral case , type @xmath26 for a small charge , type @xmath23 for a large charge , and type @xmath165-@xmath20 or type @xmath165-@xmath26 for an excessive charge . for near - extreme cases , we show evidence that type @xmath165-@xmath26 will be more probable . when there is an inner cauchy horizon in @xmath138 limit , a local energy density for an in - going null observer increases exponentially , and this confirms mass inflation . this will eventually create a curvature singularity in a large @xmath73 or large @xmath72 limit . however , this was significantly different from the four - dimensional case , as in those cases , other types of curvature functions also diverge . as we include semi - classical effects via hawking radiation , we note that the outer apparent horizon is not affected and is thermodynamically stable . however , the internal structure will be significantly modified . negative energy is concentrated near the center , and this can create a bottleneck in a wormhole . such a bottleneck is not due to a charge but is caused by negative energy . this neck occurs when the semi - classical effects dominate the classical effects . we observed that the region around the bottleneck becomes a curvature singularity ( the dvali curvature singularity ) . moreover , the inner horizon can be space - like via negative energy . therefore , we delineate two types of curvature singularities in three dimensions : the dvali curvature singularity and the mass inflation curvature singularity . therefore , as we summarize these contents , the causal structures type @xmath20 , type @xmath26 , and type @xmath23 should be modified to type @xmath160 , type @xmath161 , and type @xmath162 ( figure [ fig : p_types ] ) . one interesting finding was related to cosmic censorship . there is some controversy about whether or not there is a violation of cosmic censorship in anti de sitter space @xcite@xcite . in terms of btz solutions , there appears to be a time - like or null singularity for @xmath166 or @xmath29 cases . therefore , as long as we rely on the btz solution , a violation of weak cosmic censorship is not odd . however , if we believe that all realistic black holes should be formed by a gravitational collapse , our conclusions seem to say that there is no violation of weak cosmic censorship ( section [ sec : cri ] ) . when there is no horizon and excessive charge , the tendency of the charge distribution shows that the energy density around the center is regular . the absence of a naked singularity has also been observed in four - dimensional gravitational collapses @xcite@xcite . however , in that case , as we give an excessive charge , a near - extreme black hole was formed rather than a disappearance of a black hole . this is an interesting difference between three dimensions and four dimensions . in terms of strong cosmic censorship , there can be two ways to touch the cauchy horizon . first , there is the cauchy horizon along the in - going null direction ( e.g. , the thin red dotted line in type @xmath23 in figure [ fig : p0_types ] ) . second , there is the cauchy horizon along the out - going null direction ( e.g. , thick black dashed line in type @xmath23 in figure [ fig : p0_types ] ) . given that this black hole does not evaporate , for the former cauchy horizon , an ingoing observer requires infinite advanced time @xmath73 ( although this does not imply infinite proper time ) . thus , via mass inflation , it will surely be a curvature singularity . therefore , it is likely not possible to touch in the semi - classical regime . of course , it may be possible for an observer to touch the latter cauchy horizon . however , we know that when there are two horizons ( a space - like outer horizon and a time - like inner horizon ) , it is not evident whether there is a time - like singularity for an ingoing null observer , as we noted in table [ table : fitting ] . if this is not the case , then strong cosmic censorship appears to be safe . the authors would like to thank bum - hoon lee , ewan stewart , and seungjoon hyun for discussions and encouragement . dy and dh were supported by korea research foundation grants ( krf-313 - 2007-c00164 , krf-341 - 2007-c00010 ) funded by the korean government ( moehrd ) and bk21 . dy was supported by the national research foundation of korea(nrf ) grant funded by the korea government(mest ) through the center for quantum spacetime(cquest ) of sogang university with grant number 2005 - 0049409 . hk was supported by the national research foundation of korea(nrf ) grant funded by the korea government(mest ) with the grant number 2009 - 0074518 . in this appendix , we briefly examine the birkhoff theorem for charged btz black holes . there is an argument that the only spherical symmetric solution of @xmath167-dimensional space - time with a negative cosmological constant is a btz black hole or the self - dual coussaert - henneaux space - time @xcite . this was applied only for neutral cases , and hence it is necessary to present an argument about the uniqueness of the solution when the electric charges are involved . in this appendix , we prove that any spherical symmetric charged black holes in a three - dimensional anti de sitter background will approach the charged btz asymptotically . the metric of a spherically symmetric space - time can always be cast in the form @xmath168 involving the two arbitrary functions @xmath169 and @xmath170 . we put the gauge field by @xmath171 as the electric field strength must be of the order @xmath172 asymptotically . we may put the gauge field as an arbitrary function for completeness . however , this restriction of the configuration space is reasonable because the conservation of the energy - momentum tensor , @xmath173 , constrains the form of the gauge field as equation ( [ gauge ] ) . the independent einstein equations are as follows : @xmath174 where @xmath175 and @xmath176 represent the derivatives with respect to @xmath76 and @xmath177 , respectively . from the @xmath178-component and the @xmath179-component of the einstein equations , we have @xmath180 . in addition , a time redefinition can absorb @xmath181 and it will make @xmath170 be the only function to be determined . from the @xmath182-component , we have @xmath183 . finally , from the @xmath179-component of the einstein equations , we obtain @xmath184 where @xmath185 is the integration constant and we can choose by @xmath186 . therefore , this proves our assertion . any spherical symmetric charged black holes in the three - dimensional anti de sitter background will approach the charged btz solution . to sure this expectation , we test the dynamics of scalar field around the outer apparent horizon . for @xmath121 and @xmath187 , we observe the behavior of the scalar field ( figure [ fig : hair ] ) . for convenience , we observed the real part of the scalar field as a function of @xmath73 around the outer apparent horizon ( @xmath188 ) ; the imaginary part is also qualitatively similar . the scalar field falls off approximately @xmath189 with @xmath190 . therefore , in the @xmath138 limit , the scalar field will decay to zero . therefore , this makes sure that our solutions will emerge to a vacuum solution in the anti de sitter . then , our previous proof shows that our results surely emerge to btz solutions . the real part of the scalar field as a function of @xmath73 around the outer apparent horizon . here , we used @xmath121 and @xmath187 . the gradients of each lines are approximately @xmath191 . ] now we test consistency and convergence of this paper . first , to check the consistency , we used one of constraint equations . as an example , we choose the constraint function equation ( [ eq : e2 ] ) @xmath192 to see the constraint . the value should be exactly zero , but due to numerical errors , it may not be zero . the normal scale of the constraint function is on the order of @xmath20 ; hence , if the value is sufficiently less than @xmath20 , it is certain that our simulations give a sound result . top in figure [ fig : convergence ] shows the constraint function for the @xmath139 and @xmath121 case along the @xmath193 , @xmath194 , and @xmath119 slices . the result is sufficiently small and the order is less than @xmath195 . note that when the constraint becomes on the order of @xmath195 , the @xmath196 actually approaches zero and the behavior of the constraint function becomes unstable . however , this is not a significant problem for entire causal structures . second , to check the convergence , we first see @xmath197 and @xmath198 along the @xmath193 , @xmath194 , and @xmath119 slices , where @xmath199 means an @xmath200 finer simulation ( middle in figure [ fig : convergence ] ) . we checked for the @xmath121 and @xmath187 case . for each slice , two curves are almost same ; this shows the second order convergence . numerical errors are less than @xmath201 . next , we test the @xmath121 and @xmath139 case ( bottom in figure [ fig : convergence ] ) and compare @xmath197 and @xmath202 . as we noted , it shows the first order convergence . however , the error is sufficiently small again , and hence it is reasonable to trust our numerical calculations . top : constraint function @xmath203 for @xmath139 and @xmath121 along @xmath193 , @xmath194 , and @xmath119 . middle and bottom : convergence tests @xmath197 and @xmath202 for @xmath187 and @xmath139 with @xmath121 along @xmath193 , @xmath194 , and @xmath119 , where @xmath199 means an @xmath200 finer simulation.,title="fig : " ] top : constraint function @xmath203 for @xmath139 and @xmath121 along @xmath193 , @xmath194 , and @xmath119 . middle and bottom : convergence tests @xmath197 and @xmath202 for @xmath187 and @xmath139 with @xmath121 along @xmath193 , @xmath194 , and @xmath119 , where @xmath199 means an @xmath200 finer simulation.,title="fig : " ] top : constraint function @xmath203 for @xmath139 and @xmath121 along @xmath193 , @xmath194 , and @xmath119 . middle and bottom : convergence tests @xmath197 and @xmath202 for @xmath187 and @xmath139 with @xmath121 along @xmath193 , @xmath194 , and @xmath119 , where @xmath199 means an @xmath200 finer simulation.,title="fig : " ] s. deser , r. jackiw and g. t hooft , annals phys . * 152 * , 220 ( 1984 ) . e. witten , nucl . b * 311 * , 46 ( 1988 ) . m. banados , c. teitelboim and j. zanelli , phys . lett . * 69 * , 1849 ( 1992 ) [ arxiv : hep - th/9204099 ] . m. banados , m. henneaux , c. teitelboim and j. zanelli , phys . d * 48 * , 1506 ( 1993 ) [ arxiv : gr - qc/9302012 ] . s. carlip and c. teitelboim , phys . d * 51 * , 622 ( 1995 ) [ arxiv : gr - qc/9405070 ] . b. reznik , phys . rev . d * 45 * , 2151 ( 1992 ) ; + b. reznik , phys . d * 51 * , 1728 ( 1995 ) [ arxiv : gr - qc/9403027 ] . j. m. maldacena , adv . * 2 * , 231 ( 1998 ) [ int . j. theor . phys . * 38 * , 1113 ( 1999 ) ] [ arxiv : hep - th/9711200 ] ; + e. witten , adv . theor . math . * 2 * , 253 ( 1998 ) [ arxiv : hep - th/9802150 ] ; + s. s. gubser , i. r. klebanov and a. m. polyakov , phys . b * 428 * , 105 ( 1998 ) [ arxiv : hep - th/9802109 ] . s. hyun , j. korean phys . soc . * 33 * , s532 ( 1998 ) [ arxiv : hep - th/9704005 ] ; + k. sfetsos and k. skenderis , nucl . b * 517 * , 179 ( 1998 ) [ arxiv : hep - th/9711138 ] . a. strominger , jhep * 9802 * , 009 ( 1998 ) [ arxiv : hep - th/9712251 ] ; + v. balasubramanian and f. larsen , nucl . b * 528 * , 229 ( 1998 ) [ arxiv : hep - th/9802198 ] ; + s. carlip , phys . d * 51 * , 632 ( 1995 ) [ arxiv : gr - qc/9409052 ] ; + d. birmingham , i. sachs and s. sen , phys . b * 424 * , 275 ( 1998 ) [ arxiv : hep - th/9801019 ] . t. piran and a. strominger , phys . d * 48 * , 4729 ( 1993 ) [ arxiv : hep - th/9304148 ] ; + r. parentani and t. piran , phys . lett * 73 * , 2805 ( 1994 ) [ arxiv : hep - th/9405007 ] ; + t. chiba and j. soda , prog . . phys . * 96 * , 567 ( 1996 ) [ arxiv : gr - qc/9603056 ] ; + s. ayal and t. piran , phys . d * 56 * , 4768 ( 1997 ) [ arxiv : gr - qc/9704027 ] ; + s. hod and t. piran , phys . lett * 81 * , 1554 ( 1998 ) [ arxiv : gr - qc/9803004 ] ; + s. hod and t. piran , gen . grav . * 30 * , 1555 ( 1998 ) [ arxiv : gr - qc/9902008 ] ; + e. sorkin and t. piran , phys . rev . d * 63 * , 084006 ( 2001 ) [ arxiv : gr - qc/0009095 ] ; + e. sorkin and t. piran , phys . d * 63 * , 124024 ( 2001 ) [ arxiv : gr - qc/0103090 ] ; + y. oren and t. piran , phys . d * 68 * , 044013 ( 2003 ) [ arxiv : gr - qc/0306078 ] ; + j. hansen , a. khokhlov and i. novikov , phys . d * 71 * , 064013 ( 2005 ) [ arxiv : gr - qc/0501015 ] ; + a. doroshkevich , j. hansen , i. novikov and a. shatskiy , arxiv:0812.0702 [ gr - qc ] ; + p. p. avelino , a. j. s. hamilton and c. a. r. herdeiro , phys . d * 79 * , 124045 ( 2009 ) [ arxiv:0904.2669 [ gr - qc ] ] ; + a. doroshkevich , j. hansen , d. novikov , i. novikov and a. shatskiy , phys . d * 81 * , 124011 ( 2010 ) [ arxiv:0908.1300 [ gr - qc ] ] ; + a. borkowska , m. rogatko and r. moderski , phys . rev . d * 83 * , 084007 ( 2011 ) [ arxiv:1103.4808 [ hep - th ] ] . j. hansen , d. hwang and d. yeom , jhep * 0911 * , 016 ( 2009 ) [ arxiv:0908.0283 [ gr - qc ] ] ; + d. hwang and d. yeom , class . * 27 * , 205002 ( 2010 ) [ arxiv:1002.4246 [ gr - qc ] ] ; + d. hwang and d. yeom , class . * 28 * , 155003 ( 2011 ) [ arxiv:1010.3834 [ gr - qc ] ] ; + d. hwang , b. h. lee and d. yeom , jcap * 1112 * , 006 ( 2011 ) [ arxiv:1110.0928 [ gr - qc ] ] . j. crisostomo and r. olea , phys . d * 69 * , 104023 ( 2004 ) [ arxiv : hep - th/0311054 ] . s. gutti , class . * 22 * , 3223 ( 2005 ) [ arxiv : gr - qc/0506100 ] ; + s. gutti and t. p. singh , phys . d * 76 * , 064026 ( 2007 ) [ arxiv : gr - qc/0703142 ] ; + c. vaz , s. gutti , c. kiefer and t. p. singh , phys . rev . d * 76 * , 124021 ( 2007 ) [ arxiv:0710.2164 [ gr - qc ] ] ; + r. b. mann and j. j. oh , phys . rev . d * 74 * , 124016 ( 2006 ) [ erratum - ibid . d * 77 * , 129902 ( 2008 ) ] [ arxiv : gr - qc/0609094 ] ; + r. b. mann , j. j. oh and m. i. park , phys . d * 79 * , 064005 ( 2009 ) [ arxiv:0812.2297 [ hep - th ] ] . y. peleg and a. r. steif , phys . d * 51 * , 3992 ( 1995 ) [ arxiv : gr - qc/9412023 ] ; + m. alcubierre , g. allen , b. bruegmann , g. lanfermann , e. seidel , w. m. suen and m. tobias , phys . d * 61 * , 041501 ( 2000 ) [ arxiv : gr - qc/9904013 ] ; + f. pretorius and m. w. choptuik , phys . rev . d * 62 * , 124012 ( 2000 ) [ arxiv : gr - qc/0007008 ] ; + d. garfinkle , phys . rev . d * 63 * , 044007 ( 2001 ) [ arxiv : gr - qc/0008023 ] ; + l. m. burko , phys . d * 62 * , 127503 ( 2000 ) [ arxiv : gr - qc/0008043 ] ; + v. husain and m. olivier , class . grav . * 18 * , l1 ( 2001 ) [ arxiv : gr - qc/0008060 ] . e. poisson and w. israel , phys . lett . * 63 * , 1663 ( 1989 ) ; + e. poisson and w. israel , phys . d * 41 * , 1796 ( 1990 ) . h. reissner , ann . physik . , * 50 * , 106 ( 1916 ) ; + g. nordstrm , proc . kon . ned . wet . , * 20 * , 1238 ( 1918 ) . j. s. f. chan , k. c. k. chan and r. b. mann , phys . d * 54 * , 1535 ( 1996 ) [ arxiv : gr - qc/9406049 ] ; + r. g. cai and d. f. zhao , phys . lett . a * 208 * ( 1995 ) 281 ; + b. wang and r. k. su , phys . rev . d * 53 * , 1950 ( 1996 ) . a. r. steif , phys . d * 49 * , 585 ( 1994 ) [ arxiv : gr - qc/9308032 ] ; + g. lifschytz and m. ortiz , phys . d * 49 * , 1929 ( 1994 ) [ arxiv : gr - qc/9310008 ] . m. simpson and r. penrose , int . j. theor . phys . * 7 * , 183 ( 1973 ) ; + s. chandrasekhar and j. b. hartle , proc . r. soc . a bf 384 , 301 ( 1982 ) ; + w. a. hiscock , phys . d * 15 * , 3054 ( 1977 ) ; + n. d. birrell and p. c. w. davies , nature * 272 * , 35 ( 1978 ) ; + y. gursel , v. d. sandberg , i. d. novikov and a. a. starobinsky , phys . d * 19 * , 413 ( 1979 ) ; + y. gursel , v. d. sandberg , i. d. novikov and a. a. starobinsky , phys . rev . d * 20 * , 1260 ( 1979 ) . g. dvali , fortsch . * 58 * , 528 ( 2010 ) [ arxiv:0706.2050 [ hep - th ] ] ; + g. dvali , m. redi , s. sibiryakov and a. vainshtein , phys . * 101 * , 151603 ( 2008 ) [ arxiv:0804.0769 [ hep - th ] ] ; + r. brustein , g. dvali and g. veneziano , jhep * 0910 * , 085 ( 2009 ) [ arxiv:0907.5516 [ hep - th ] ] . l. susskind , l. thorlacius and j. uglum , phys . d * 48 * , 3743 ( 1993 ) [ arxiv : hep - th/9306069 ] . d. yeom and h. zoe , phys . d * 78 * , 104008 ( 2008 ) [ arxiv:0802.1625 [ gr - qc ] ] ; + s. e. hong , d. hwang , d. yeom and h. zoe , jhep * 0812 * , 080 ( 2008 ) [ arxiv:0809.1480 [ gr - qc ] ] ; + d. yeom and h. zoe , arxiv:0811.1637 [ gr - qc ] ; + d. yeom , int . j. mod . cs * 1 * , 311 ( 2011 ) [ arxiv:0901.1929 [ gr - qc ] ] . t. hertog , g. t. horowitz and k. maeda , phys . lett . * 92 * , 131101 ( 2004 ) [ arxiv : gr - qc/0307102 ] ; + t. hertog , g. t. horowitz and k. maeda , phys . d * 69 * , 105001 ( 2004 ) [ arxiv : hep - th/0310054 ] ; + t. hertog , g. t. horowitz and k. maeda , arxiv : gr - qc/0405050 . m. alcubierre , j. a. gonzalez , m. salgado and d. sudarsky , arxiv : gr - qc/0402045 ; + m. gutperle and p. kraus , jhep * 0404 * , 024 ( 2004 ) [ arxiv : hep - th/0402109 ] ; + m. dafermos , adv . theor . math . * 9 * , 575 ( 2005 ) [ arxiv : gr - qc/0403033 ] ; + d. garfinkle , phys . rev . d * 69 * , 124017 ( 2004 ) [ arxiv : gr - qc/0403078 ] . e. ayon - beato , c. martinez and j. zanelli , phys . d * 70 * , 044027 ( 2004 ) .
study of the conformational transitions of heteropolymers in dilute solutions is important for many applications from the chemical industry to biotechnology . directed more towards the former , there has been a significant amount of theoretical work carried out on concentrated copolymer solutions , mixtures and blends using various types of the density formalism . however , these approaches are not valid at infinitely low dilution where the fundamental interactions of the individual macromolecule determine its conformational state . this situation is more relevant for biochemistry . the problem is even harder to address at non equilibrium conditions typical for biopolymers _ in vivo _ @xcite . for these reasons we have been working for some time on developing an adequate statistical mechanical technique for studying the equilibrium structure and kinetics across phase transitions in heteropolymers @xcite . the main idea was to extend the gaussian self consistent ( gsc ) method , originally proposed for the homopolymer ( see e.g. @xcite and references therein ) , to the case of inhomogeneous monomer interactions . this has been achieved in ref . @xcite where we have derived the complete set of non linear kinetic equations for complex valued equal time correlation functions of the fourier transforms of the monomer coordinates . there we have analysed in some detail the simplest periodic ( ab ) copolymer , but study of more complicated sequences remained out of our practical computational reach . the relation of the kinetic equations to the equilibrium free energy , as well as the expression for the entropy , were also unknown to us at that stage . thus , the phase diagram have not been elucidated and certain other important issues have not been addressed in that work . in the following two works @xcite we have achieved further progress and resolved most of these questions , though at the cost of loosing detailed information about individual sequences . namely , we have performed averaging of the gsc equations over the quenched disorder in the monomer amphiphilic strength . this yielded a closed set of kinetic equations for description of random copolymers with a gaussian distribution of the disorder . such an idealised system is very interesting in itself , particularly since there is a hope that its study could shed light on some features of the extremely complex problem of protein folding @xcite . to render the quenched disorder problem more tractable certain perturbative approximation was necessary @xcite . it became clear in ref . @xcite that this causes some deficiencies in the equilibrium limit of the formalism , and we have alluded as to how these could be alleviated in higher orders of the expansion . therefore , it seems important to revisit the general formalism of ref . @xcite in order to resolve remaining difficulties , maintaining the rich information about particular sequences and avoiding any further approximations . furthermore , there exists , till now , no simple theoretical procedure capable of giving the equilibrium conformational states of a heteropolymer with arbitrary sequence , that can in a consistent manner also give the full kinetic pathway between these states . the work presented here achieves this . to demonstrate the strength of the extended gsc method we consider a number of interesting examples of heteropolymer sequences . different kinds of interactions , such the hydrodynamic interaction , may be straightforwardly incorporated into the scheme . a rich and nontrivial physical picture emerges as a result of this theoretical progress . traditionally , we proceed from the coarse grained description of the polymer chain @xcite with the spatial coordinates @xmath0 ascribed to the @xmath1-th monomer . it is assumed that the long timescale evolution of conformational changes is well represented by the phenomenological langevin equation , which upon neglecting the backflow effect of the solvent may be written as , @xmath2 where @xmath3 is the `` bare '' friction constant per monomer . for the discussion of the hydrodynamic interaction we refer the reader to appendix [ app : hydro ] . the thermal fluctuations are incorporated via the gaussian noise which , according to the einstein law , is characterised by the second momentum , @xmath4 where the greek indices denote the spatial components of 3-d vectors . in the current treatment the solvent is effectively excluded from the consideration @xcite and the resulting monomer interactions are described by the effective free energy functional , @xmath5 where in principle @xmath6 are allowed to have any dependence on the site indices @xmath7 . the first two terms in eq . ( [ eq : h ] ) describe the connectivity and the stiffness of the chain . their coefficients have the following simple meaning : @xmath8 and @xmath9 with @xmath10 and @xmath11 called the statistical segment length and the persistent length @xcite respectively . apart from these interactions , local along the chain , there are also long range volume interactions represented by the virial type expansion @xcite in eq . ( [ eq : h ] ) . the latter reflects the hard core repulsion and weak attraction between monomers , but also the effective interaction mediated by the solvent monomer couplings @xmath12 . the coefficients in this expansion may be calculated as functions of the temperature and the parameters of molecular interaction . however , for our purposes we do not need to know their explicit form here as we shall keep only a few first terms . appropriate coefficients then may be viewed as independent phenomenological parameters which could be directly related to experimentally measurable quantities . in refs . @xcite we have discussed that the case of amphiphilic heteropolymers , for which monomers differ only in the monomer solvent coupling constants , corresponds to the following choice of site dependent second virial coefficients in eq . ( [ eq : h ] ) , @xmath13 the mean second virial coefficient @xmath14 is associated with the quality of the solvent : positive values correspond to the good solvent ( where effective two body repulsion of monomers results in the extended coil conformation ) , and the negative values correspond to the bad solvent condition ( where the effective two body attraction tends to compact the chain ) . the set of couplings @xmath15 expresses the chemical composition , or using the biological terminology , the primary sequence of a heteropolymer . for simplicity , in the consequent sections we shall consider examples for which these constants can be parametrised in a simple way : @xmath16 , where variables @xmath17 can only take three values : @xmath18 or @xmath19 corresponding to ( a ) hydrophobic , ( b ) hydrophilic and ( c ) `` neutral '' monomers respectively . the parameter @xmath20 is called the degree of amphiphilicity of the chain . note that for more complicated than binary sequences there is another relevant dispersion , @xmath21 , and the combination @xmath22 is a more appropriate variable . in work @xcite we have derived a set ( 18 ) of the gsc kinetic equations for the equal time correlation functions of the fourier transforms of monomer coordinates @xmath23 . there , at the end of sec . ii , we have mentioned that a polymer with no periodic structure may be described by choosing the number of blocks @xmath24 . the equations for the correlation functions , @xmath25 contain some redundancies and also are intermixed with the diffusive degree of freedom describing the motion of the centre of mass . obviously , for a single chain the latter can be easily decoupled from the intra molecular degrees of freedom by introducing the mean squared distances between pairs of monomers , @xmath26 it turns out that from eq . ( 18 ) in ref . @xcite after certain algebraic manipulations one can obtain a closed set of equations for the quantities @xmath27 . taking into account the additional bending energy contribution in eq . ( [ eq : h ] ) , the gsc equations may be written down in the most general form as follows , @xmath28 where we have introduced the four point correlation functions , @xmath29 , and the matrix @xmath30 of size @xmath31 with the cofactor @xmath32 , @xmath33 similarly to eq . ( 31 ) in ref . @xcite for the mean energy @xmath34 we have , @xmath35 it is interesting to note that in the case of the homopolymer the right hand side of the kinetic equations ( [ eq : dkineq ] ) may be rewritten via the instantaneous gradients of the variational free energy by introducing the normal modes ( see eq . ( 5 ) in ref . this establishes connection of the stationary limit of our kinetic approach with the equilibrium theory , in which we recover the gibbs bogoliubov variational principle . we also note that if first order phase transitions are involved , one has to possess the expression for the free energy in order to determine the phase boundaries by finding the global free energy minimum . the variational free energy , @xmath36 , based on the gaussian _ ansatz _ for equal time pair correlation functions , contains two terms , @xmath37 . naturally , the mean energy term is given by eq . ( [ eq : e ] ) . the second , entropic contribution , is calculated in appendix [ app : var ] . let us summarise the result of that calculation here . one representation for the entropy that is suitable for numerical analysis is , @xmath38 where we have used the determinant of the @xmath39-dimensional major submatrix of the matrix , @xmath40 and the matrix @xmath41 obviously has the elements equal to @xmath42 . the reason for appearance of the truncated matrix is that we have excluded the zero eigenvalue of @xmath43 related to the translational invariance . here we have also introduced the @xmath39-dimensional orthogonal projector @xmath44 such that , @xmath45 with the matrix elements @xmath46 . this matrix has obviously one zero eigenvalue and @xmath47 degenerate eigenvalues equal to @xmath48 . however , for analytical treatment it is more convenient to obtain a slightly different expression by regularising the zero eigenvalue ( i.e. by imposing the constraint @xmath49 as a `` soft '' condition in eq . ( [ eq : tr ] ) ) , @xmath50 since the matrix @xmath51 is nondegenerate we can easily differentiate eq . ( [ se ] ) so that , @xmath52 where inside the trace @xmath53 over the @xmath39-dimensional subspace projected out by @xmath44 the matrix @xmath43 becomes invertible with the inverse denoted as @xmath54 . this allows us to obtain the combination which appears in the kinetic equations , @xmath55 these preliminaries are sufficient to prove the desired relation . indeed , using eq . ( [ eq : sder ] ) by direct and tedious differentiation of eq . ( [ eq : e ] ) one can show that the kinetic equations ( [ eq : dkineq ] ) may be expressed through the instantaneous gradients @xcite of the variational free energy as , @xmath56}{\partial d_{mm''}(t)}- \frac{\partial { \cal a}[d(t)]}{\partial d_{m'm''}(t ) } \right).\ ] ] this formula , together with eqs . ( [ eq : e],[se ] ) , is the key formal result of the current work . the structure of eq . ( [ eq : kinmain ] ) is sufficiently nontrivial to be guessed from phenomenological arguments and has been derived in a systematic manner proceeding from eq . ( [ eq : dkineq ] ) . we would like to comment here that although , for simplicity , we have presented the explicit formulae above for a ring polymer , our current formalism is general and covariant . in fact , the kinetic equations ( [ eq : kinmain ] ) are valid for any topology of the chain . thus , it is straightforward to consider more complicated topologies such as a star , brush , network , branched chain and so on . for that it is sufficient to modify only the spring and stiffness terms in eq . ( [ eq : h ] ) , and respectively in eq . ( [ eq : e ] ) . for example , to describe an open chain one has simply to suppress the term with @xmath57 in the connectivity contribution , and the terms with @xmath58 in the stiffness contribution to the energy . we undertake a detailed comparison of ring versus open homopolymers in kinetics at the collapse transition of the homopolymer in a separate work @xcite . another interesting possibility is that the general equation ( [ eq : kinmain ] ) is also valid for models with different ways of representing the connectivity and stiffness . for example , the freely rotating model @xcite can be obtained by suppressing the first two terms in eq . ( [ eq : e ] ) and instead keeping fixed the following mean squared distances : @xmath59 and @xmath60 , where @xmath61 and @xmath62 are the bond length and angle . this is easy to prove by adding appropriate `` soft '' constraints to the free energy functional and taking the consequent limit in eq . ( [ eq : kinmain ] ) , so that they become delta functions in the partition function . finally , let us introduce two main observables : the mean squared radius of gyration and the degree of micro phase separation @xcite , @xmath63 the second parameter has a meaning of the dimensionless correlation of the matrices of the relative two body interaction and the mean squared distances . for heteropolymers with two types of monomers it characterises the difference between the mean squared radii of gyration of hydrophilic and hydrophobic species , and for the symmetric composition with their numbers equal we have a simple relation : @xmath64 . the self consistent kinetic equations ( [ eq : kinmain ] ) have been studied numerically using the explicit formulae ( [ eq : c1],[eq : c2],[eq : c4 ] ) for the effective potentials ( see appendix [ app : der ] ) , and the expression given by the first term in the right hand side of eq . ( [ eq : dkineq ] ) for the entropic contribution . we used the fifth order adaptive step runge kutta method @xcite to improve stability of the solution which , for large amphiphilicity parameter , @xmath20 , is characterised by a rather rugged free energy landscape . we note that such kinetic method for finding the equilibrium distributions is more reliable and efficient than the standard methods of free energy minimisation if there are lots of mountains and valleys on its surface . we also refrain from study of the influence of the hydrodynamics in this paper for the analysis and results are complicated enough already . besides , the hydrodynamics in the preaveraged approximation does not affect the equilibrium state , which is recovered by taking the stationary limit in the kinetic equations . we include the volume interactions up to the three body terms only , i.e. @xmath65 for @xmath66 . as can be seen from eqs . ( [ eq : kinmain],[eq : c2 ] ) the computational time per time step scales with the chain length as @xmath67 here . this performance is intermediate between that of the homopolymer @xmath68 @xcite and that of the random copolymer @xmath69 @xcite . the performance of the formalism in ref . @xcite was @xmath70 , where @xmath71 and @xmath72 are the block length and number of blocks respectively . besides , that formalism relied on the use of complex variables , and the unitary transformation to the real basis was not an easily automatised task for complicated sequences . moreover , the treatment of the diffusive mode was nontrivial and sequence dependent . thus , in every respect , the current scheme is most attractive for study of heteropolymer sequences from the computational point of view . it is natural to work here with combinations @xmath73 and @xmath74 as the units of size and time . we choose @xmath75 , @xmath76 and @xmath77 to fix @xmath78 and @xmath79 equal to unity . in addition , we fix the following interaction parameters : the third virial coefficients @xmath80 and the stiffness @xmath81 as we did in ref . @xcite . now , we turn to discussion of concrete results . we have studied ring chains of the lengths @xmath82 . we shall present here the most complete results for @xmath83 and discuss the @xmath84 dependence only briefly . we have examined many different sequences . however , we present our data in this paper only for three particular choices which we found most typical and illustrative of the heteropolymer behaviour . these are chains of : 30 ( ab ) blocks , 10 ( aaabbb ) blocks and 2 ( abacbbcabccbcaaacbcccbbaacbcca ) blocks , which we call the `` short '' blocks , `` long '' blocks and `` random '' sequences respectively ( see the end of section [ sec : model ] for the monomer notations ) . the phase diagrams in terms of the mean second virial coefficient @xmath85 and the amphiphilicity @xmath20 for the above mentioned three sequences are presented in figs . [ fig:1]-[fig:3 ] . for positive values of @xmath14 and comparatively small values of @xmath20 the conformational state of the chain is akin to a homopolymer extended coil ( see fig . [ fig : dmneq]a ) . by decreasing @xmath14 to the negative region the chain is caused to undergo a continuous ( second order like ) collapse transition ( curve ( c ) ) , that is characterised by a rapid fall of the radius of gyration @xcite ( see fig . 1 of ref . @xcite ) and the change of the fractal dimension . proceeding from the collapsed globule at a fixed negative @xmath14 the increase of the amphiphilicity @xmath20 also causes a continuous transition ( curve ( s ) ) to the micro phase separated ( mps ) globule @xcite . during this transition the system size , @xmath86 , monotonically increases ( see fig . [ fig:4 ] ) and the mean energy @xmath87 decreases in agreement with our earlier results in refs . the former change is more pronounced for the `` long '' blocks compared to other sequences , for which the connectivity constraints impede formation of structures with a hydrophilic shell and hydrophobic core . the mps order parameter @xmath88 ( see fig . [ fig:5 ] ) increases almost linearly for small @xmath20 , then after the transition asymptotically saturates . the transition from the coil at large values of @xmath20 to the mps globule turns out to be more complicated , and essentially dependent on the sequence . in case of the `` long '' blocks ( fig . [ fig:2 ] ) the collapse transition to the mps globule becomes discontinuous ( first order like ) . thus , inside the boundaries of metastability , designated by spinodals i and i , there are two competing minima of the free energy : the coil and the mps globule . the former minimum is characterised by a large size , @xmath86 , ( on the right of fig . [ fig:6 ] ) and a small mps order parameter @xmath88 ( fig . [ fig:7 ] ) , while for the latter minimum the situation is reversed ( see the left hand side of figs . [ fig:6],[fig:7 ] ) . the depths of the free energy minima become exactly equal on the transition curve i in fig . [ fig:2 ] . for the `` short '' blocks ( fig . [ fig:1 ] ) , as well as for many random sequences ( such as in fig . [ fig:3 ] ) , the phase diagram is much more complicated . starting from some value of @xmath20 and for intermediate values of @xmath14 there appear additional solutions corresponding to local minima of the free energy . the broad region where this could take place is bounded by the curves i and ii in figs . [ fig:1],[fig:3 ] . with increasing @xmath20 the number of such solutions grows quickly . significantly , in a region of the phase diagram some of these become the main free energy minimum . as the number of such solutions grows roughly exponentially with the chain length , we do not attempt to draw all their boundaries of ( meta)stability . instead , we shall designate them collectively as _ frustrated _ phases , explaining this terminology below . an important point here is that , as our analysis shows , these solutions become dominant in a narrow region of the phase diagram due to a subtle competition between the mean energy and the entropy . the mps globule is entropically unfavourable there because the overall shrinking force is insufficiently strong . the values of @xmath89 and @xmath88 are intermediate for these solutions and lie between those of the coil and mps globule ( see figs . [ fig:8],[fig:9 ] ) . in this sense , we can call them nonfully compacted and misfolded states . in comparison , the mps globule ( see fig . [ fig : dmneq]d ) has a more compact size and better optimised volume interactions , what is manifested in a higher value of @xmath88 . most interesting is the local structure of these additional phases . let us discuss the particular example of the 30(ab ) sequence . the formalism of ref . @xcite has used heavily the assumption of certain symmetries for the mean squared distances @xmath90 due to which these variables can take only @xmath91 independent values , where @xmath72 is the number of blocks . these symmetries are : the block translation invariance , @xmath92 where @xmath71 is the block length , and the more complicated inversion symmetry discussed in detail in ref . these symmetries have a simple meaning a renumbering of monomers does change the average properties over the statistical ensemble the interactions of a ring chain remain the same . however , the maximal possible number of dynamical variables in the gsc method is much larger and equals to @xmath93 . surprisingly , it turns out that the frustrated phases are characterised by spontaneous breaking of these symmetries , and therefore only the current version of the method that takes into account all degrees of freedom can describe them . thus , the property ( [ eq : sym ] ) is no longer valid . this phenomenon describes formation of local frustrated heterogeneities ( see figs . [ fig : dmneq]b , c ) in which pieces of the chain form mps clusters @xcite that are prevented from further coalescing by their hydrophilic shells and high entropic barriers . the role of spontaneous symmetry breaking is well recognised in equilibrium statistical mechanics . what is striking here is that the number of distinct spontaneously broken states becomes huge for large system sizes . this diversity and a special foliating structure of various branches leads in the thermodynamic limit to what is known as a spin glass like frozen phase @xcite of random copolymers ( see e.g. refs . @xcite and numerous references therein ) . we shall return to the issue of spontaneous symmetry breaking in kinetics in subsection [ subsec : kin ] . even though the kinematic symmetries are not present from the outset for arbitrary , or random , sequences , the structure of the phase diagram ( fig . [ fig:3 ] ) and behaviour of main observables ( figs . [ fig:8],[fig:9 ] ) remain very similar . it is the particular structure , number and the shape of boundaries of frustrated phases that are very sensitive to the sequence . the symmetry that may be broken in this case has a subtler , dynamic , meaning and may be expressed in terms of the replica formalism @xcite . in a sense , for a very long periodic chain , blocks may be viewed as identical copies of a smaller block length chain . thus , it is by no means surprising that the replica symmetry breaking in our case for periodic systems takes such an explicit manifestation in the breaking of the block translational symmetry . this important point was completely missed , nor could it be discovered , in our considerations of ref . therefore we have achieved here a new significant insight into the problem by a simple extension of the gsc method . an interesting feature of our phase diagrams is that the region between spinodals i and ii , designating where the frustrated phases can exist , expands dramatically with increasing chain length . for example , at @xmath94 , for the 15(ab ) sequence it lies approximately between @xmath95 , while for the 30(ab ) between @xmath96 , and the curve ii goes nearly vertically downwards for larger @xmath84 . according to the above interpretation , for infinitely long chain somewhere in this broad region and close to its boundaries there are the actual glass transition curves , which should be determined using proper glass order parameters . thus , for short chains the curves i and ii may be viewed as approximate indicators of the freezing transitions . the former distinguishes between the homopolymer like and the random coil , while the latter between the homopolymer like ( liquid ) and the frozen globules . as for the region of stable mps globule , it gets relatively smaller for larger systems . that is not surprising the frustrated phases expand and the phase separation involving larger spatial scales requires stronger interactions . having understood the identifications for the spinodals , we now can recognise in figs . [ fig:1],[fig:3 ] the main features of the phase diagram , though rather distorted , of random copolymer model presented in ref . @xcite . finally , let us comment on the phase diagram of the `` long '' blocks ( fig . [ fig:2 ] ) . the micro phase separation is obviously easier in this case and it dominates for large values of @xmath20 , so that the frustrated phases are suppressed . we found that , qualitatively , in order to form a frustrated phase , in a finite range of @xmath20 values , the number of ( not necessarily identical ) pieces of the chain with competing interactions should be larger than some critical number , in principle , weakly dependent on @xmath84 . here are a few examples conforming to this qualitative criterion : the phase diagram of 10(ab ) behaves roughly as for 10(aaabbb ) , but for 15(ab ) it behaves similarly to 30(ab ) ; while for 15(aabb ) the phase diagram looks like for 15(aaabbb ) at small and moderate values of @xmath20 , it becomes as for 30(ab ) for much larger values of the amphiphilicity . it is reasonable to conjecture therefore that for a large number of ( aaabbb ) blocks , as well as for extremely high values of @xmath20 and just 10 blocks , the frustrated phases may be found again . here we shall consider the time evolution of the conformational state of the system away from its initial equilibrium after it has been subjected to an instantaneous temperature jump that causes the two body interaction parameters @xmath14 and @xmath20 to change . we are interested in quenches from the homopolymer coil , where all monomers are equally hydrophilic ( @xmath97 and @xmath98 ) , to the region of parameters corresponding to the mps globular state , so that the ` a ' species became strongly hydrophobic and the ` b ' species remained weakly hydrophilic ( @xmath99 and @xmath100 ) . the temporal behaviour of the mean squared radius of gyration , @xmath101 , the mps parameter , @xmath102 , and the instantaneous free energy , @xmath103 , in kinetics of folding for different sequences is presented in figs . [ fig:10]-[fig:12 ] . for the homopolymer ( curves ( a ) ) @xmath89 and @xmath36 decrease monotonically to their final equilibrium values , while the mps parameter , @xmath88 , remains identically zero for there is no distinction between different monomer species . these curves agree with the earlier results of ref . @xcite and serve for reference purpose here . now let us discuss the curves ( b ) corresponding to the periodic ( ab ) sequence . in the previous subsection we have mentioned that the current formalism yields results consistent with those of ref . @xcite only beyond the parameter region of the frustrated phases , which are characterised by spontaneous breaking of the block symmetry . in kinetics the situation is somewhat similar , but the consistence with the previous simplified formalism is even more limited . really , the mean squared radius of gyration remains close to that of the `` effective '' homopolymer ( curve ( a ) ) during the first kinetic stage ( see eq . ( 74 ) in ref . @xcite ) . interestingly , this is not quite so for more complicated sequences . the mps parameter , @xmath102 , for the ( ab ) copolymer grows slowly in a way similar to the splitting of the fourier modes @xmath104 for large indices @xmath105 ( see fig . 4 in ref . @xcite ) . however , contrary to the strange conclusion in ref . @xcite that the kinetics of the ( ab ) copolymer proceeds faster than for the homopolymer , we now have a , natural from the point of view of monte carlo simulations @xcite , slowing down of copolymer kinetics . analysis of @xmath27 shows that this effect is entirely due to the spontaneous breaking of the block symmetry in kinetics , something that has not been accounted for in ref . @xcite . indeed , in fig . [ fig:13 ] we exhibit the time dependence of the mean squared distances between two nearest hydrophobic monomers @xmath106 for the ( ab ) copolymer in kinetics after a quench to the mps globule region . for early , as well as for late , times these functions for different @xmath107 are exactly equal to each other . however , there is well defined intermediate period in time where the block symmetry is spectacularly broken ( see fig . [ fig : dmnkin ] ) . we remark that the symmetry breaks and restores in a step like manner ( e.g. @xmath108 and @xmath109 join together at @xmath110 , earlier than with other functions ) and also that this effect is relatively strong . we would like to make a general comment here on the nature of spontaneous symmetry breaking in kinetics . thus , normally in such situations at equilibrium there exist a thermodynamically unstable symmetric free energy minimum and a disjoint set of symmetry broken minima , which may be transformed to each other by the residual subgroup of symmetry transformations . these states may also be obtained kinetically as infinite time limits of the time evolution starting from any , for example , the symmetric initial state , which happens to be the main free energy minimum before the quench . however , the formal structure of the gsc kinetic equations ( [ eq : kinmain ] ) is such that they yield a symmetric solution at any moment in time provided one proceeds from the symmetric initial condition . a question arises then how can one obtain the kinetics that could lead eventually to the multitude of final states with broken symmetry ? the answer is clear in the exact theory there are fluctuations that can transform between different spontaneously broken states in kinetics . the gsc method presents , though an optimised and improved , but still a mean field type theory , where such fluctuations are not properly included . manifestation of the kinetic spontaneous symmetry breaking takes a different form there . namely , at some moment in time the symmetric solution of the kinetic equations becomes unstable with respect to perturbations ( whether of the initial condition , or of the interaction matrix , @xmath111 ) . thus , for example , one can add an infinitesimal symmetry breaking term @xmath112 to the two body interaction matrix and consider the limit of vanishing perturbation in the solution . different choices of the form of @xmath112 in the unperturbed limit would yield all possible types of spontaneously broken kinetic evolution , which are , of course , equivalent to each other . numerically , such a regularisation procedure is not even necessary . there is always an intrinsic perturbation due to computer round off and numerical integration errors . thus , if the symmetry is favourable to be kinetically broken somewhere , numerically one obtains one of the spontaneously broken solutions there , rather than the unstable symmetric solution , unless the symmetry conditions have been imposed by hand . in this situation improvement of the numerical precision would have no profound effect the kinetic process either does not change , or can change only up to an experimentally unobservable residual symmetry transformation . we note also that such a procedure of perturbing the interaction matrix here is analogous to introducing an external magnetic field in the ising model . the spontaneous magnetisation of a macroscopic sample of ferromagnet may be achieved by gradually switching off the external magnetic field . in the absence of the field there would remain domains with long range order , but varying directions of the spontaneous magnetisation canceling each other . unfortunately , in our case it is not evident how to experimentally implement an analogous gradual switching off of @xmath112 . now returning to our results , in fig . [ fig:10 ] one can see that the folding kinetics for the ( ab ) copolymer is about 3 times slower than for the homopolymer of the same length , the effect being even stronger for other sequences we have considered . from fig . [ fig:11 ] and the phase diagrams in figs . [ fig:1]-[fig:3 ] it is evident that the considered quench results in the final state of the mps globule for the `` long '' blocks ( curve ( c ) ) , whereas in the frustrated phases for two other sequences ( curves ( b ) and ( d ) ) . for the latter phases the mps parameter is smaller and the radius is larger than for the mps globule , as we already know from the equilibrium considerations . it is interesting to note that the final relaxation to the frustrated phases may be rather unusual . for example , in figs . [ fig:10],[fig:11 ] for the `` random '' sequence @xmath101 increases and @xmath102 decreases during the last kinetic stage , what is the converse to the behaviour at final relaxations in other cases . another unusual observation is that for some sequences the parameter @xmath88 may even become negative during some time in kinetics , something we never observed at equilibrium . this shows that the structure of non equilibrium conformations can be very complicated . the instantaneous free energy @xmath103 depicted in fig . [ fig:12 ] turns out to be the quantity most sensitive to the conformational structure of the non equilibrium state . from that figure it strikes that , while the homopolymer folding is a single smooth relaxational process , the folding of heteropolymers proceeds through a multistep acceleration deceleration process . the flat regions of a staircase like function correspond to temporary kinetic arrest of the system in transient non equilibrium ( mostly symmetry broken ) conformations . generally speaking , in the gsc method we deal with the time evolution of a statistical ensemble of various initial conformations . the flat regions appear due to transient trappings of various members of the ensemble in their local shallow energy minima . since such minima are encountered at different moments in time for different members of the ensemble , their influence on the overall time evolution of averaged observables is manifested in a smooth characteristic slowing down . qualitatively then , summarising the data for various sequences , most of which we have suppressed here , we can say that the number of local minima affects the number of stair steps , whereas the depth of minima determines the lengths of the steps . this interpretation is highly supported by the strong dependence of the stair case structure on the chain length , the sequence and the interaction parameters , for it is known how the frequency and depths of local potential energy minima depend on these factors . on the one hand , it is known that for a long heteropolymer chain the number of local minima grows exponentially with the chain length @xmath84 , and indeed we see that the number of steps in @xmath103 grows just as quickly . on the other hand , increasing of the amphiphilicity , @xmath20 , leads to higher depths of local minima as well as an increase in their number , and really the steps in @xmath103 become longer and more numerous . let us give another example of such connection . for the `` long '' blocks the interactions are less frustrated than , say , for the ( ab ) blocks , thus there are much fewer of local energy minima , but the barrier between the coil and mps globule is a higher . as a result , the kinetic process ( curve ( c ) in fig . [ fig:12 ] ) proceeds through only one rather long kinetically arrested step . it is worthwhile to make a comment here on the notion of kinetic stages introduced in earlier works @xcite . those we associated with typical structures of the conformation and accompanying kinetic laws . we distinguished at least the following kinetic stages at the collapse of the homopolymer : early time necklace formation , middle time coarsening and a number of final relaxational processes . the multistep character of folding observed in this work affects the middle kinetic stage , resulting in its considerable complication and splitting into multiple substages with respective complex kinetic laws that are determined by the sequence . universality of such kinetic laws is doubtful , but probably it can be recovered by averaging over certain classes of sequences with similar folding properties . apparently , in the gsc method the kinetics is a motion in the space of @xmath93 averaged dynamic variables , @xmath113 , and it is determined by the profile of the free energy . it may be instructive , using figs . [ fig:10 ] and [ fig:12 ] , to present the kinetics via a parametric plot of @xmath36 vs @xmath86 , the latter being the main , though not the only , relevant `` coordinate '' of the system . that would produce a kind of `` bottleneck '' picture that was much discussed by p. wolynes and others @xcite in relation to the protein folding problem . this indicates that our method produces behaviour that permits interpretation in terms of phenomenological energy landscape models . finally , let us utilise once again the connection between the kinetic evolution of the free energy and the ruggedness of the potential energy landscape . this would allow us to shed some light on the general structure of the energy landscape for complex heteropolymers . thus , a typical example of the `` random '' sequence kinetics ( curve ( d ) in fig . [ fig:12 ] ) shows first a fast drop , then appearance of short steps that are growing longer with time , until the last step becomes infinite . this translates to the following equivalent energy landscape versus some collective coordinate : first there is a rapid drop away from the unstable coil state , then the surface flattens and small wrinkles appear on it , they are growing larger in the amplitude gradually becoming high mountains and deep valleys , until there is eventually a very deep ravine corresponding to the `` ground '' state of the system separated by a very high barrier from other minima . in our view , the latter picture bears a remarkable resemblance to a typical mean field energy landscape in spin glass systems @xcite recently discussed by g. parisi @xcite . this observation seems encouraging to us and indicates that the gsc method is capable of describing very complex systems , though it is too detailed and expensive in the complete form for description on a macroscopic scale . nevertheless , by taking the quenched disorder averaging ( similarly to ref . @xcite ) it is possible to achieve an adequate description for heteropolymers that is alternative to the replica formalism . we believe that the underlying connection between both types of approaches has been clarified to some extent by the current work . in the present work we have extended the gaussian self consistent method to permit study of heteropolymers with complicated primary sequences . this has been achieved by transforming the kinetic gsc equations , earlier derived for the fourier modes , to the form containing only the mean squared distances between pairs of monomers and relating these equations to the instantaneous gradients of the variational free energy calculated based on the gibbs bogoliubov principle . the revised gsc formalism possesses important computational advantages , but also it is fundamentally superior to its predecessor in that it can describe phenomena of spontaneous symmetry breaking and formation of structural heterogeneities . we then have applied the extended gsc method to some particular amphiphilic heteropolymers . the equilibrium phase diagrams for these have been obtained in a systematic manner . apart from the coil and two globular states we have discovered that in a wide intermediate region of the phase diagram there may be a large number of frustrated partially misfolded states . despite the fact that such states , the number of which grows exponentially with the chain length , are mostly metastable , some of them become the dominant state in their rather narrow domains . the corresponding potential energy profile of the system has strong resemblance to that of a typical spin glass system . thus , we may conclude that the transition to these states in the thermodynamic limit corresponds to a glassy freezing transition . we note that an observation that for sufficiently long sequences of alternating monomers ( `` short '' blocks ) the micro phase separated state is displaced by the region of glassy frustrated phases has been known for quite some time and well understood in the framework of the density theories @xcite as a result of the density fluctuations which modify the mean field behaviour . the block translational symmetry breaking in our approach is another manifestation of the glassy phenomena earlier extensively studied for melts of random heteropolymers . in addition , we can see how the destruction of the micro phase structures occurs for finite sized systems and how exactly it depends on a particular sequence . the folding kinetics is found to be strongly affected by the presence of these transient frustrated states along the kinetic pathway . this leads to a complicated kinetic process consisting of multiple steps with pronounced slowing down and then acceleration in the folding rate . it is interesting to note that a typical fairly heterogeneous chain with weakly hydrophilic and strongly hydrophobic units folds kinetically first to one of the misfolded states and can undergo a consequent nucleation process to the main thermodynamic state of the micro phase separated globule . these results for equilibrium and kinetics confirm many predictions of the earlier treatment adopted for random heteropolymers in ref . @xcite based on a quasi perturbative averaging of the gsc equations over the quenched disorder . the latter approach describes a large set of random heteropolymer sequences with a gaussian distribution of the monomer amphiphilicity . the kinetics of folding seems to be consistent between the previous and current approaches with respect to main observables such as the radius of gyration , energy and the mps parameter . a smooth slow glass like relaxation in the former approach is produced by averaging over all different sequences with their particular multistep processes of passage through the frustrated states . although here we observe strong dependence of the phase boundaries in the equilibrium phase diagram on the primary sequence , the major conclusion of refs . @xcite that there are three distinct globular states of the homopolymer like ( liquid ) , the frozen ( glassy ) and the micro phase separated ( folded ) globules , and an additional state of the random coil , as well their relative location in the diagram , remain valid . we have previously discussed that the equilibrium limit of the theory in the treatment of ref . @xcite had certain deficiencies due to the additional perturbative approximation , which we attempted to remedy phenomenologically . the current fundamental scheme is free of any such problems and provides a reliable test ground for development of further approximations . we have also discovered that heteropolymer sequences of equal length can be roughly divided into distinct classes possessing similar phase diagrams and kinetic folding properties . for instance , periodic polymers with a few long blocks easily undergo micro phase separation and for them the frustrated states are suppressed at equilibrium and in kinetics . there is a view in the scientific community that complex random sequences also permit more refined classification . the identification of good folding sequences ( see e.g. ref . @xcite ) is believed to be an important prerequisite for unraveling the mysteries of proteins . in the current paper the gsc method has been applied to a single chain problem . however , since the kinetic equations ( [ eq : kinmain ] ) are covariant it would be relatively straightforward to apply it to solutions of many polymers . formation of nontrivial mesoscopic globules in copolymer solutions at low concentrations has been predicted in the framework of the method @xcite and recently observed experimentally @xcite . generally , the gsc method may be viewed as extension to the realm of kinetics of the variational treatment , which is a better theory than standard mean field approaches due to a much larger number of variational parameters . we should admit , however , that in the complete form numerical solution of the gsc equations is computationally expensive for systems of a few hundred of monomers . nevertheless , as we have recently shown @xcite , in some approximation the gsc equations can be reduced to those of a simple mean field theory such as the flory huggins one , and some extra corrections to the latter . importantly , the formalism in terms of the mean squared distances is valid for description of the extended coil as well as of the globular states , whereas the density formalism theories are limited to relatively high densities of the system . the weakest point of the method is the gaussian form of the trial hamiltonian , and thus of the correlation functions , a matter that has been discussed at some length in the past @xcite . however , based on the covariant form of the kinetic equations derived here , we hope , in some not too distant future , to take into account the non gaussian corrections in an analogy to the treatment of the hard core repulsion for van der waals systems . the main strength of the revised gsc approach lies in the ability to describe a given heteropolymer sequence of finite length , rather than an _ infinite ensemble _ of sequences characterised in a certain probabilistic way . this is an inevitable preliminary step in developing adequate techniques for theoretical modelling of complex biopolymers . understanding of the relation between the chemical composition ( primary sequence ) and the 3-d equilibrium ( tetriary ) structure , as well as of the kinetics of folding , in proteins is one of the great challenges of the modern biotechnological science . we hope that methods like the one we have presented here will take their right place in the collection of new tools for bioinformatics . the authors acknowledge interesting discussions with professor c.a . angell , professor k. kawasaki , professor g. parisi , and our colleagues dr a.v . gorelov and a. moskalenko . the langevin equation with account of the hydrodynamic interaction includes the oseen tensor @xcite , @xmath114\ , \phi^{\alpha'}_{n'},\ ] ] where @xmath115 is the right hand side of eq . ( [ eq : lang ] ) and the noise has the second momentum proportional to the inverse oseen tensor . in the preaveraged approximation we have , @xmath116 and the analogs of formulae ( a5,a6 ) of ref . @xcite will be , @xmath117 explicit expressions for a few first terms in the virial expansion of the effective potentials , @xmath118 may be found in appendix [ app : der ] . we introduce the trace as the integration over all monomers coordinates subject to the constraint that the reference point is fixed in the centre - of - mass of the chain , @xmath119 then the partition function is obtained as @xmath120 with @xmath121 given by eq . ( [ eq : h ] ) . the delta function in ( [ eq : tr ] ) removes the trivial divergence in @xmath122 due to the translational invariance . we choose the trial hamiltonian as an arbitrary linear combination , @xmath123 where we have also introduced arbitrary sources @xmath124 . using the delta function one can exclude , say , @xmath125 and derive , @xmath126=(2\pi)^{3(n-1)/2}({\rm det}\tilde{k}^{(n-1)})^{-3/2 } \exp\left(\frac{1}{2}\sum_{mm'}{}'({\bf j}_m-{\bf j}_0 ) ( \tilde{k}^{-1})_{mm'}({\bf j}_{m'}-{\bf j}_0 ) \right),\ ] ] where the @xmath47 dimensional matrix @xmath127 is @xmath128 we can calculate the averages by , @xmath129}{\partial { \bf j}_n \partial { \bf j}_{n'}}=(\tilde{k}^{-1})_{nn'},\ ] ] where again the latter identity holds only for @xmath130 . from these quantities it is straightforward to obtain the mean squared distances using eq . ( [ eq : ddef ] ) which also holds only for @xmath130 . finally , we want to express all independent parameters of the matrix @xmath127 via the quantities @xmath90 . for this we compute the following sums applying the above delta function constraint , @xmath131 substituting @xmath132 from the second formula here to eq . ( [ rtodd ] ) and recalling the relation ( [ d4 ] ) we derive the desired inverse relation ( [ rtodd ] ) for @xmath133 . from the partition function ( [ z0 ] ) we obtain the `` entropic '' term @xmath134 $ ] that yields precisely eq . ( [ sr ] ) . the gibbs bogoliubov variational principle is then based on minimising this variational free energy @xmath135 with respect to the @xmath93 independent variational parameters @xmath90 . for the homopolymer we have the translational invariance along the chain so that @xmath136 , where @xmath137 . then the matrix ( [ rtodd ] ) may be rewritten as , @xmath138 let us apply a unitary transformation to the fourier coordinates generated by the matrix , @xmath139 by a direct evaluation and recalling the relation for the normal modes , @xmath140 we see that the matrix @xmath141 is diagonal with the eigenvalues @xmath142 and @xmath143 . thus , the logarithm of determinant of the @xmath39-dimensional submatrix @xmath144 yields the standard `` entropy '' of the homopolymer ( see eq . ( 3 ) in ref . @xcite ) up to a trivial constant . two first terms in the virial expansion of the derivatives ( [ eq : v ] ) of the mean energy are explicitly given by , @xmath145 where @xmath146 . it is usually assumed that for negative @xmath14 the stability of the system is ensured by the positiveness of @xmath147 . it turns out however that for the whole range of @xmath148 this does not hold and there are additional pathological solutions with singular free energy . the reason for this deficiency of the model ( [ eq : h ] ) is that we have discarded the terms with two coinciding indices in the three body contribution . this standard trick is not satisfactory therefore , but fortunately it could be remedied by using another prescription for these terms : @xmath149 in ref . @xcite we show that the addition of the latter term , which is subdominant in the large @xmath84 limit anyway , does not change the earlier results for the homopolymer , and even quantitatively improves the agreement with known results for the dense globular state @xcite . importantly , for heteropolymers this term removes spurious solutions and makes the theory well defined . the corresponding contributions to the mean energy and the effective potentials are , @xmath150 m. l. huggins , j. chem . * 9 * , 440 ( 1941 ) ; p. j. flory , j. chem . phys . * 9 * , 660 ( 1941 ) ; _ polymer blends _ , edited by d. r. paul and s. neuman ( academic press , 1987 ) ; _ multicomponent polymer systems _ , edited by i .s . miles and s. rostami ( longman scientific and technical , singapore , 1992 ) . d. j. meier , j. polym . * c26 * , 81 ( 1969 ) ; d. j. meier , j. polym . prepr . * 11 * , 400 ( 1970 ) ; * 15 * , 171 ( 1974 ) ; a. n. semenov , sov . jetp * 61 * , 733 ( 1985 ) ; t. ohta and k. kawasaki , macromolecules * 19 * , 2621 ( 1986 ) ; l. leibler , macromolecules * 13 * , 1602 ( 1980 ) ; g. h. fredrickson and e. helfand , j. chem . phys . * 87 * , 697 ( 1987 ) ; t. hashimoto , macromolecules * 20 * , 465 ( 1987 ) . shakhnovich and a.m. gutin , j. phys . ( france ) * 50 * 1843 , ( 1989 ) ; s.v . panyukov and s.i . kuchanov , sov . jetp * 72 ( 2 ) * , 368 ( 1991 ) ; a.v . dobrynin and i.ya . erukhimovich , pisma zh . * 53 ( 11 ) * 545 ( 1991 ) , j. phys . * i 5 * 365 ( 1995 ) . g. wolynes , in _ spin glass ideas in biology _ , edited by d. stein ( world scientific , singapore , 1991 ) ; _ protein folding _ , edited by t. e. creighton ( wiley , new york , 1992 ) ; _ new developments in theoretical studies of proteins _ , edited by r. elber ( world scientific , singapore , 1994 ) . note that one may use the symmetry properties @xmath151 and @xmath152 only after the differentiation of the energy ( [ eq : e ] ) , which is generally speaking expressed through the four point correlation functions . one should use the standard differentiation rule , @xmath153 . the structure of the kinetic equations is such that the symmetry properties are automatically satisfied on the solution of these equations at each moment in time . apparently , for a finite length chain , the collapse transition is characterised by a finite width , and therefore its representation in figs . [ fig:1]-[fig:3 ] is sufficiently schematic with a suitable criterion @xcite adopted to draw a definite curve for it . e.g. timoshenko , yu.a . kuznetsov , and k.a . dawson , physica * a 240 * 432 ( 1997 ) ; proceedings of the international conference on morphology and kinetics of phase separating complex fluids , ed . by f. mallamace ( messina , italy , 1997 ) . a. kuznetsov , e. g. timoshenko , and k. a. dawson , in _ proceedings of the international conference on morphology and kinetics of phase separating complex fluids _ , edited by f. mallamace ( messina , italy , 1997 ) , to be published in il nuovo cimento .
the theoretical description of slow dynamics is a crucial point to elucidate the nature of the glass transition in structural glass - forming liquids . one of the commonly used approaches , mode - coupling theory ( mct ) , was from the very beginning designed for the supercooled simple liquids @xcite , i.e. for the non - disordered models ( as opposed to the models which contain quenched disorder naturally ) . later it has been proven that mc - equations become exact for a number of spin - glass models @xcite as well as for the polymeric manifold in a random media @xcite ( i.e. for the models with quenched disorder ) provided that the number of variables components goes to the infinite . actually applicability of the mc - equations has been substantially extended to the case when the time translation invariance and the fluctuation - dissipation theorem do not hold any more @xcite . this striking similarity between the models with and without quenched disorder suggests that the effective disordered potential ( e.g. , in a supercooled liquid ) is in a sense `` self - induced '' and the difference between such a `` self - induced disorder '' and the quenched disorder might not be crucial @xcite . in order to provide some insight into self - induced disorder we employed in ref.@xcite a feynman variational principle ( vp ) for a set of interacting particles . indeed it was shown that the vp is capable to treat metastable states of the glass - forming system . the main point in ref . @xcite was that the partition function representation in terms of functional integrals is twofold : ( a ) either as an integral over the local density , @xmath3 , or ( b ) over the conjugated to @xmath4 field @xmath5 . it has been _ assumed _ that the _ component average _ free energy , @xmath6 , ( which is only meaningful in the supercooled regime ) is equal to the variational free energy @xmath7 . there are at least four strong reasons in favor of this ( at first sight not obvious ) assumption . * the variational free energy @xmath8 is a upper bound for the _ canonical _ free energy , i.e. , @xmath9 , as it should be since @xmath10 , where the complexity @xmath11 @xcite . * after implementation of vp the initial problem is reduced to a self - consistent random field ginzburg - landau model ( rfglm ) . then , as was shown previously , the corresponding field @xmath12 must be upgraded to a _ replicated _ field @xmath13 , where @xmath14 ( with the final limit @xmath15 ) and the density field @xmath4 plays the role of an `` external '' field . note that here the density @xmath16 is gaussian due to the use of the vp . eventually the correlators of @xmath17 and @xmath18 - fields can be determined self - consistently . * the resulting replicated partition function for rfglm has a typical form which may eventually lead to the replica symmetry breaking ( rsb ) , structural glass transition and the self - induced disorder . * finally , in the case of the long - range interaction the partition function allows the expansion around the saddle point , or mean - field ( mf ) , solution . it is possible to show then that the _ next to the mean - field approximation _ and vp merge and both become exact , i. e. @xmath19 and the glassy phase does not appear . some evidence for this behavior was deduced from the results for the particles on a @xmath20- dimensional hypersphere @xcite at large dimensions , @xmath21 . the aim of this paper is to face the full dynamical problem for a non - disorded model with a long range interaction . using the expansion around the saddle point solution we derive the full equation of motion for the time dependent density - density correlator and show that a `` glassy '' solution does not exist . conversly , if we add a term describing quenched disorder , by random distribution of the strength of the interaction potential , then the resulting equations of motion for two time density correlation and response functions fall in the same class as mc - equations which have been widely discussed @xcite . this means that the `` self - induced '' disorder is not generic for the pure model with the long - range interaction and vice versa on addition of a quenched disorder the phase space becomes very rugged resulting in slow dynamical processes . the paper is organized as follows . in the next section [ ii ] we first introduce the theorotical model without quenched disorder . its dynamics is discussed by using the functional integral technique . the saddle point solution yields the mean field dynamcis . expansions around the saddle point yield one loop corrections . the legendre transformation provides the possibilty of the analysis of the full dynamic correlation matrix . in the section [ iii ] quenched disorder is introduced by a `` random bond model '' and a gaussian disorder . the corresponding generating functional ( gf ) is computed by the self - consistent hartree approximation , which results in a set of coupled langevin equations , that are solved in their asymptotic regimes . more details on the calculations are laid out in the corresponding appendices . we start from a simple model system which consists of interacting particles . to do so , let us consider a set of @xmath22 particles in @xmath23 - dimensional space interacting by a pair potential of the form @xmath24 this is a typical example of a long - range potential with a characteristic length @xmath25 and a coupling constant @xmath26 . the choice of this potential is twofold . it contains a cut off at a @xmath27 and allows thus to control the range of the interaction . moreover at small scales @xmath28 it consists of a typical power law decay with long range character , if @xmath29 . therefore the so chosen potential allows to keep control on range and nature of the interaction , which will become essential below . to ensure extensivity of the total interaction energy we require that the integral @xmath30 , i.e. it does not depend on the number of particles @xmath0 . as a result we have @xmath31 . the intermolecular potential ( [ 1 ] ) has the form of the generalized kac potential @xmath32 which has been used for the rigorous treatment of the van - der - waals theory @xcite . in order to provide conditions for the expansion around a saddle point , carried out later on ( see below ) , we should require that the length @xmath27 must be larger compared to the characteristic size of the system , which scales naturally as @xmath33 ) at @xmath34 . as a consequence we find the limits for the range parameter @xmath35 @xmath36 below we shall restrict our considerations to the case : @xmath37 , and the strength of the interaction @xmath38 ( pure repulsion ) without loss of generalization in the main statements we are going to predict . then the fourier transformation of the potential ( [ 1 ] ) takes especially simple form @xmath39 which allows accurate analytic calculations . in the limit @xmath34 we have thus @xmath40 , but the relevant minimum wave vector is @xmath41 and thus @xmath42 can be actually neglected under the integration over the whole @xmath43 - space . as a result we arrive formally at a _ one - component plasma model _ ( ocp ) @xcite where the electroneutrality is implicitly provided by a neutralizing background . in the following we set up the relevant equations of motion for the model system . we restrict ourselves to the langevin dynamics , which can be comfortably formulated in terms of dynamic functionals which allows the systematic @xmath1 - expansion treatment . the langevin dynamics of @xmath0 particles interacting via the potential ( [ 1 ] ) ( at @xmath37 and @xmath38 ) is described by the equation of motion @xmath44 where @xmath45 and @xmath46 are the mass and the friction coefficient respectively , @xmath47 and @xmath48 . the random force in eq.([4 ] ) is gaussian with @xmath49 and the correlator @xmath50 where from now on we work in units where the boltzmann constant @xmath51 . as was mentioned , it is more convenient to reformulate the langevin problem ( [ 4])-([5 ] ) by using the celebrated martin - siggia - rose generating functional ( gf ) method @xcite . the method was first applied for the @xmath52 - model with the long - range interaction in @xcite and for the polymer melt dynamics in @xcite . despite the fact that the langevin equation ( [ 4 ] ) is of the second order it is possible to show that the jacobian which appear under transformation to the functional variables , still equal to one ( see appendix in @xcite ) . after using this technique for the problem ( [ 4 ] ) - ( [ 5 ] ) , gf takes the form @xmath53 + \int dt \sum_{p=1}^{n}\sum_{m=1}^{n } \frac{\mu}{n } i{\hat r}_{j}^{(p)}(t ) \nabla_{j}^{(p)}v\left({\bf r}^{(p ) } - { \bf r}^{(m)}\right)\right\ } \ ; , \label{6}\end{aligned}\ ] ] where the action of the free system @xmath54 = \int dt \left\ { t\gamma_{0 } \left[i{\hat r}_{j}^{(p)}(t)\right]^{2 } + i{\hat r}_{j}^{(p)}(t)\left[m_{0}\frac{\partial^2}{\partial t^2 } { r}_{j}^{(p)}(t ) + \gamma_{0}\frac{\partial}{\partial t } { r}_{j}^{(p)}(t ) \right]\right\ } \ ; . \label{7}\end{aligned}\ ] ] in the following we are going to transform this functional to collective density variables . by using the transformations to the mass density @xmath55 and the longitudinal projection of the response field density @xmath56 for the gf one gets @xmath57 where the summation over the repeated greek indices is implied . in eq.([10 ] ) we have introduced 2-dimensional field @xmath58 where @xmath59 and @xmath60 . the `` entropy '' of the free system is given as usual by @xmath61\nonumber\\ & \times&\delta\left[\pi({\bf r},t)-\sum_{p=1}^{n}\:i{\hat r}_{j}^{(p)}(t)\nabla_{j}\delta({\bf r}-{\bf r}^{(p)}(t))\right ] \ ; , \label{12}\end{aligned}\ ] ] and @xmath62 is the 2@xmath63 2 - interaction matrix @xmath64 and @xmath65 is a source field . an alternative valuable representation of gf can be obtained through the `` functional fourier transformation '' @xmath66 and its inversion @xmath67 the substitution of eq.([12 ] ) into eq.([14 ] ) leads to the explicit expression for the free - system gf @xmath68\nonumber\\ & - & i\sum_{p=1}^{n}\int dt \psi \left({\bf r}^{(p)}\right ) + \left . i \hat r_{j}^{(p)}(t)\nabla_{j}\phi ( { \bf r})\right\|_{{\bf r}= { \bf r}^{(p)}(t ) } \bigg\}\ ; , \label{16}\end{aligned}\ ] ] where @xmath69 and @xmath70 are components of the column - variable @xmath71 by making use ( [ 15 ] ) in ( [ 10 ] ) and after functional integration over @xmath72 one gets @xmath73\bigl[u^{-1}\bigl]_{\alpha\beta } ( 1,2)\bigl[i\psi_{\beta}(2)+ \chi_{\beta}(2)\bigl ] + \int d1\psi_{\alpha}(1)\lambda_{\alpha}(1)\bigg\}\ ; , \label{18}\end{aligned}\ ] ] where we have also add a source field @xmath74 conjugated to @xmath75 . as a result eqs.([10 ] ) and ( [ 18 ] ) provide two equivalent representations of gf . for the purpose of expansion around the saddle point we use representation ( [ 18 ] ) at @xmath76 which after the transformation @xmath77 , yields @xmath78\bigg\}\ ; , \label{19}\end{aligned}\ ] ] which is appropriate for a saddle point integration , since the particle number @xmath0 is large . the action hereby is given as @xmath79&= & \frac{1}{2}\int dt \int d^3{\bf r}d^3{\bf r'}\psi_{\alpha}({\bf r},t)\left[v^{-1}\right]_{\alpha\beta}({\bf r}- { \bf r ' } ; \kappa ) \psi_{\beta}({\bf r ' } , t ) - \frac{1}{n}\log \int \prod_{p=1}^{n}d{\bf r}^{(p)}(t)d{\bf \hat r}^{(p)}(t)\nonumber\\ & \times&\exp \bigg\{\sum_{p=1}^{n}a_{0}[{\bf r}^{(p)},{\bf \hat r}^{(p ) } ] - i\sum_{p=1}^{n}\int dt r_{\alpha}^{(p)}(t)\left[\psi_{\alpha}\left({\bf r}^{(p)}(t)\right ) + i\chi_{\alpha}\left({\bf r}^{(p)}(t)\right)\right]\bigg\}\ ; , \label{20}\end{aligned}\ ] ] and the interaction matrix @xmath80 recall that the relation @xmath81 is necessary for the validity of the saddle point integration . moreover we have defined the column - vector @xmath82 for convenience . minimization of @xmath83 $ ] with respect to @xmath75 leads to the sp - equations for the mean fields @xmath84 @xmath85 where the average @xmath86 is calculated by using the _ cumulant _ gf @xmath87 -\nonumber\\ i\sum_{p=1}^{n}\int dt r_{\alpha}^{(p)}(t)\left[\overline{\psi_{\alpha}}\left({\bf r}^{(p)}(t)\right ) + i\chi_{\alpha}\left({\bf r}^{(p)}(t)\right)\right]\bigg\}\ ; .\label{24}\end{aligned}\ ] ] the correlation matrix in the random phase approximation ( rpa ) is defined in such a way @xmath88\ ; . \label{25}\end{aligned}\ ] ] after linearization of eq.([24 ] ) with respect to @xmath89 the 2@xmath632 - rpa correlation matrix is easily found to coincide with the well known form @xcite @xmath90^{-1}\right\}_{\alpha\beta}(1 , 2)\ ; , \label{26}\end{aligned}\ ] ] where @xmath91 is the interaction matrix ( [ 21 ] ) and @xmath92 is the correlation matrix for the _ free system _ @xmath93 has the form @xmath94 in eq.([27 ] ) @xmath95 and @xmath96 are response functions whereas @xmath97 stands for the correlation function . the relation between them is given by the fluctuation dissipation theorem ( fdt ) which in @xmath98- representation has the form @xmath99 it is easy to check that in this case the fdt for the rpa type correlation matrix ( [ 26 ] ) also holds @xmath100 where @xmath101 is the inverse temperature . the corresponding elements of the rpa - matrix ( [ 26 ] ) are of an especially simple form in the fourier - @xmath102 - representation , namely @xmath103\left[1+v(k)f_{01}({\bf k},\omega)\right ] } \label{rpares1}\\ s_{01}({\bf k},\omega)&=&\frac{f_{01}({\bf k},\omega)}{1+v(k)f_{01}({\bf k},\omega ) } \label{rpares2}\\ s_{10}({\bf k},\omega)&=&\frac{f_{10}({\bf k},\omega)}{1+v(k)f_{10}({\bf k},\omega)}\ ; . \label{rpares3}\end{aligned}\ ] ] it turns out interesting to recover the wellknown form in the static limit , where we have @xmath104 $ ] and for the correlator @xmath105 one gets @xmath106 where we have used @xmath107 . this expression is completely equivalent to the correlator for the ocp - model ( see eq.(10.1.7 ) in @xcite ) with the direct correlation function @xmath108 and the debye wavenumber @xmath109 . now let us expand the action ( [ 20 ] ) around sp - solution ( [ 23 ] ) up to the second order with respect to the fluctuations @xmath110 . after the functional integration we arrive at the following result for the gf @xmath111\ ; , \label{31}\end{aligned}\ ] ] where @xmath112 is the inverse matrix of the effective interactions @xcite @xmath113_{\alpha\beta}(1 , 2 ) + \frac{1}{n}\left<\delta\rho_{\alpha}(1)\delta\rho_{\beta}(2)\right>_{{\rm sp}}\ ; .\label{32}\end{aligned}\ ] ] in eqs.([31 ] ) - ( [ 32 ] ) we deliberately keep external field @xmath65 nonzero because it to be used in the next subsection for the legendre transformation . the functional legendre transformation is a general way to provide the dyson equation for the _ full correlation matrix _ in doing so the _ irreducible _ gf , @xmath115 , is defined by the identity @xmath116 by doing functional differentiation of ( [ 33 ] ) one gets @xmath117 and @xmath118_{\alpha\beta}(1 , 2 ) = \frac{\delta^{2}\gamma\left\{\left<\rho_{\alpha}(1)\right>\right\}}{\delta\left<\rho_{\alpha}(1)\right > \delta\left<\rho_{\beta}(2)\right > } \ ; .\label{35}\end{aligned}\ ] ] taking into account the result in eq.([31 ] ) we find the following result for gf @xmath119\ ; , \label{35'}\end{aligned}\ ] ] where @xmath120 in eq.([35 ] ) one should consider @xmath65 as a functional of @xmath121 given by eq.([34 ] ) . double differentiation of eq.([35 ] ) leads to an equation of the dyson form @xmath118_{\alpha\beta}(1 , 2 ) = \left[s^{-1}\right]_{\alpha\beta}(1 , 2 ) - \sigma_{\alpha\beta}(1 , 2)\ ; , \label{37 } \end{aligned}\ ] ] where the rpa - correlation matrix , @xmath122 , is defined by eqs.([rpares1 ] ) - ( [ rpares3 ] ) and the `` self - energy '' functional @xmath123 has the form @xmath124 in eq.([38 ] ) the `` trace '' is taken over the variables 3,4 and indices @xmath125 . the explicit differentiation in eq.([38 ] ) leads to the result @xmath126 where @xmath127 is a short hand notation of the matrix @xmath128 and we have took into account that @xmath129 at @xmath130 because the fluctuations are gaussian . further calculation yields @xmath131 where @xmath132 and the full mean field @xmath133 the expression for @xmath134 can be easily found by differentiation of eq.([42 ] ) with respect to @xmath135 . taking into account eqs.([23 ] ) , ( [ 34 ] ) and ( [ 35 ] ) at @xmath136 one gets @xmath137_{\alpha\beta}(1 , 2 ) - \mu \int d4 d3 v_{\alpha\omega}(1 , 4 ) s_{\omega\gamma}(4 , 3)r_{\gamma\beta}(3 , 2)\label{43}\end{aligned}\ ] ] or finally @xmath138^{-1}\right\}_{\alpha\gamma}(1 , 3)\left[\hat g^{-1}\right]_{\gamma\beta}(3 , 2)\ ; , \label{44 } \end{aligned}\ ] ] where the hatted variables stands for the corresponding 2@xmath632 matrices . substitution eqs.([44 ] ) and ( [ 40 ] ) in eq.([39 ] ) yields @xmath139_{\gamma\alpha}(3 , 1)\left[g^{-1}\right]_{\delta\beta}(4 , 2)\;,\label{45}\end{aligned}\ ] ] where 2@xmath632 vertex - matrix has the form @xmath140^{-1}\right\}_{\delta\gamma}(4 , 3 ) s^{(4)}_{\gamma\delta\omega\chi}(3 , 4 , 5 , 6)\nonumber\\ & \times&\left\{\left[\hat 1 + \mu \hat v \hat s \right]^{-1}\right\}_{\omega\alpha}(5 , 1)\left\{\left[\hat 1 + \mu \hat v \hat s \right]^{-1}\right\}_{\chi\beta}(6 , 2)\ ; .\label{46}\end{aligned}\ ] ] in eq.([46 ] ) @xmath141 is the 4 - point ( response ) correlator matrix in rpa @xmath142 the explicit calculation of @xmath143 is implemented in appendix b. the vertex - matrix can be seen as a one - loop diagram ( see fig . + the higher loops contributions which include generally speaking 2 m - point correlators , @xmath144 , can be also considered , however the `` self - energy '' has still the same convolution structure : @xmath145 . here the vertex - matrix @xmath146 is calculated in rpa only . that is why these contributions basically do not change our results . as a result in the @xmath147- representation the dyson equation ( [ 37 ] ) with the `` self - energy '' functional ( [ 45 ] ) reduces to a quadratic one @xmath148_{\gamma\delta}({\bf k } , \omega)g_{\delta\beta}({\bf k},\omega ) - g_{\alpha\beta}({\bf k } , \omega ) + k_{\alpha\beta}({\bf k } , \omega ) = 0\ ; .\label{48}\end{aligned}\ ] ] the coefficients of the eq.([48 ] ) trace the problem back to the _ free system dynamics _ which is embodied in the correlation matrices @xmath149 and @xmath150 . it is not surprising then that a specification of the model for the free system dynamics is necessary , before going to the investigation of eq.([48 ] ) . as we have mentioned the explicit solution of eq.([48 ] ) needs the specification of the free system dynamics . two simple models are most amenable for the theoretical treatment : free diffusion model ( fdm ) and the relaxation time approximation model ( rtam ) @xcite . the latter provide more reasonable dynamical information also for short time intervals , @xmath151 , where the fdm completely failed ( e.g. , the sum rule does not hold ) . it turns out that upon calculation of the trace in eq.([46 ] ) the integral is ultraviolet - divergent for fdm and only rtam leads to the finite result . the matrix elements for rtam has the form @xmath152 where we introduced the diffusion coefficient @xmath153 , the characteristic time scale @xmath154 , and @xmath155 for the overall density . at @xmath156 we return to fdm . in the case of rtam the solution of eq.([48 ] ) for the full correlation matrix reads @xmath157k_{01}({\bf k } , \omega)}}{2 \left[-i\omega\tau_{\rm c } - \omega^{2}\tau_{0 } \tau_{\rm c } + \chi_{st}^{-1}(k)\right]}\label{49'}\\ g_{10}({\bf k},\omega ) & = & g_{01}(-{\bf k},-\omega ) \label{50 } \end{aligned}\ ] ] and @xmath158 k_{01}({\bf k},\omega)}}{- i\omega\tau_{\rm c } - \omega^2 \tau_{0 } \tau_{\rm c } + \chi_{st}^{-1}({\bf k } ) } \right\|^2 - k_{00}({\bf k } , \omega)}{{\rm re } \left\{\sqrt{1 - 4\left [ - i \omega\tau_{\rm c } -\omega^2 \tau_{0 } \tau_{\rm c } + \chi_{st}^{-1}({\bf k } ) \right ] k_{01}({\bf k } , \omega)}\right\}}\ ; . \label{51}\end{aligned}\ ] ] the explicit calculation of the matrix @xmath159 ( see eq.([46 ] ) is given in the appendix c. the overall behavior of the correlation function @xmath160 according eq . ( [ 51 ] ) is shown in fig . 2 ( at @xmath161 ) . it can be seen clearly that there no singularity appears at @xmath162 . + the singularity , however , might be responsible for a glass transition . instead , the low frequency limit of @xmath160 slowly changed with control parameters ( which is not shown in fig.2 ) . that means that for the non - disordered model with a general repulsive long ranged potential ( [ 1 ] ) the glass transition is not generic . this very important conclusion suggest that for the model with a long range interaction the phase space is too smooth to show a glass transition . in order to obtain a glass like transition a competing interactions or a quenched disorder should be added . this leads to glassy dynamics , as we will show in the next section . it is interesting to note , that for the generalized kac potential eq . ( [ kac ] ) , where @xmath163 and its fourier transformation are positive definite functions , mct - memory kernel vanishes at @xmath164@xcite . the corresponding argumentation is relegated to the appendix d. the explanation for this result lies in the fact that the `` cage effect '' , which is a cornerstone of mct , is missing in the mf - limit . the `` glass transition '' which has been studied in ref.@xcite for the particles interacted via the kac potential ( [ kac ] ) has a completely different nature . in ref.@xcite the function @xmath163 has a step form so that its fourier transform @xmath165 is negative at some value of @xmath166 . as a result the system becomes unstable and a nonuniform configuration where the particles are grouped into `` clumps '' shows up . it was found that the slow dynamics of the mf - model is associated with these clumps and does not touch a single particle motion . obviously it is different from the conventional glass transition @xcite . in the previous sections we have shown in detail that in the absence of disorder the dynamical spectrum changed monotonically with a control parameter and no glassy dynamics can be seen . the natural question which arises now is : how will the introduction of competing interactions and / or quenched disorder affect the dynamics of the system discussed above ? to provide an answer to this question we will use already existing models of heteropolymers and their disordered two body interaction @xcite . the use of these models and techniques are here natural , since the behavior of heteropolymers is well discussed in the literature . in principle , two practical possibilities exist . * the strength of the two- body interaction , @xmath167 , in eq.([1 ] ) is now a random function of all pairs of the interacted particles , @xmath168 ( `` random - bond model '' ) . * each particle carry a single `` charge '' @xmath169 , so that @xmath170 and each @xmath169 is randomly distributed ( `` random sequence model '' ) . it turns out to be sufficient for the purpose of this paper to restrict ourself only with the `` random - bond model '' , where @xmath168 does not depend from the choice of pairs and has a gaussian distribution @xmath171 the competing long range interactions frustrates the system of particles and the question is whether a glass transition exist or not . normally frustration and frozen disorder is enough for the existence of glassy phases . here the problem is more complicated , since the long range nature of the interaction may provide opposite effects . the averaging over the quenched disorder in eq.([6 ] ) ( after the substitution @xmath172 can be carried out just in the same way as in ref.@xcite . similarly , typical two - time dependent terms immediately appear . they are also bilinear with respect to the forces of interaction @xmath173 . in order to rationalize these terms it is convenient to introduce ( besides the mass density ( [ 8 ] ) and the response field density ( [ 9 ] ) ) the following collective variables @xmath174 after the introduction of the 4-dimensional column - fields @xmath175\ ; , \end{aligned}\ ] ] where @xmath176and 4@xmath634 - matrix @xmath177\label{55}\end{aligned}\ ] ] the whole expression for gf takes the form @xmath178 where the entropy is given by @xmath179\nonumber\\ & \times&\prod_{{\rm a}=1}^{4}\delta\left[q_{{\rm a}}(1 ; 2)-\sum_{p=1}^{n}\:p_{{\rm a}}^{(p)}(1 ; 2)\delta({\bf r}_{1}-{\bf r}^{(p)}(t_{1}))\delta({\bf r}_{2}-{\bf r}^{(p)}(t_{2 } ) ) \right]\ ; . \label{57}\end{aligned}\ ] ] we had used the column - operators @xmath180 and the external field , @xmath181 , conjugated to @xmath182 , has been introduced also . the two- point collective fields ( [ 53 ] ) have a meaning of the dynamical `` overlaps '' . it is a dynamical generalization of the parisi `` overlaps '' in a replica space @xcite . for example @xmath183 quantify density - density and @xmath184 response - density overlaps respectively between two space - time points . the `` entropy '' ( [ 57 ] ) corresponds to the volume in the dynamical phase space when not only fields @xmath72 but also overlaps @xmath185 are given . in a sense the `` entropy '' ( [ 57 ] ) is again the generalization of the entropy for the heteropolymer spanned in a replica space at the given set of `` overlaps '' @xcite . let us introduce the functional @xmath186 by the functional fourier transformation @xmath187 after substitution in eq.([56 ] ) and integration over @xmath188 and @xmath189 one gets @xmath190_{\alpha\beta}(1,2)\psi_{\beta}(2)\nonumber\\ & - & \frac{n}{\chi_{0}^2}\int d1 d2 d3 d4 \phi_{\rm a}(1;2)\left[\gamma^{-1}\right]_{{\rm ab}}(1 , 2 , 3 , 4 ) \phi_{\rm b}(3;4)\bigg\}\ ; , \label{60}\end{aligned}\ ] ] where @xmath191\nonumber\\ - i\sum_{p=1}^{n}\int dt r_{\alpha}^{(p)}(t)\left[\psi_{\alpha}\left({\bf r}^{(p)}(t)\right ) + i\chi_{\alpha}\left({\bf r}^{(p)}(t)\right)\right]\nonumber\\ - i\sum_{p=1}^{n}\int dt dt ' p^{(p)}_{a}(t;t')\left[\phi_{a}\left[{\bf r}^{(p)}(t);{\bf r}^{(p)}(t')\right ] + ih_{a}\left[{\bf r}^{(p)}(t);{\bf r}^{(p)}(t')\right]\right]\bigg\}\ ; . \label{61}\end{aligned}\ ] ] in order to ensure the extensivity of the whole effective action in eq.([60 ] ) we put the variance @xmath192 ( so that the variance of the whole strength factor in eq.([3 ] ) scaled as @xmath193 akin to ref.@xcite ) . this enable to represent gf in a similar to eq.([19 ] ) form @xmath194\bigg\}\ ; , \label{62}\end{aligned}\ ] ] where @xmath195= \frac{1}{2\xi_{0}}\int dt \int d1 d2\psi_{\alpha}(1)\left[v^{-1}\right]_{\alpha\beta}(1 , 2)\psi_{\beta}(2)\nonumber\\ - \frac{1}{\chi_{0}^2}\int d1 d2 d3 d4 \phi_{\rm a}(1;2)\left[\gamma^{-1}\right]_{{\rm ab}}(1 , 2 , 3 , 4 ) \phi_{\rm b}(3;4 ) -\frac{1}{n}\log \int \prod_{p=1}^{n}d{\bf r}^{(p)}(t)d{\bf \hat r}^{(p)}(t)\nonumber\\ \times\exp \bigg\{\sum_{p=1}^{n}a_{0}[{\bf r}^{(p)},{\bf \hat r}^{(p ) } ] - i\sum_{p=1}^{n}\int dt r_{\alpha}^{(p)}(t)\left[\psi_{\alpha}\left({\bf r}^{(p)}(t)\right ) + i\chi_{\alpha}\left({\bf r}^{(p)}(t)\right)\right ] \nonumber\\ -i\sum_{p=1}^{n}\int dt dt ' p^{(p)}_{a}(t;t')\left[\phi_{a}\left[{\bf r}^{(p)}(t);{\bf r}^{(p)}(t')\right ] + ih_{a}\left[{\bf r}^{(p)}(t);{\bf r}^{(p)}(t')\right]\right ] \bigg\}\ ; .\label{63}\end{aligned}\ ] ] the resulting sp - equation reads @xmath196 where the average @xmath197 is calculated with the gf @xmath198\nonumber\\ & - & i\sum_{p=1}^{n}\int dt r_{\alpha}^{(p)}(t)\left[\overline{\psi_{\alpha}}\left({\bf r}^{(p)}(t)\right ) + i\chi_{\alpha}\left({\bf r}^{(p)}(t)\right)\right]\nonumber\\ & -&i\sum_{p=1}^{n}\int dt dt ' p^{(p)}_{a}(t;t')\left[\overline{\phi_{a}}\left[{\bf r}^{(p)}(t);{\bf r}^{(p)}(t')\right ] + ih_{a}\left[{\bf r}^{(p)}(t);{\bf r}^{(p)}(t')\right]\right]\bigg\}\ ; . \label{65}\end{aligned}\ ] ] thereby we are left with the gf of a free system which experiences the external mean - fields @xmath199 and @xmath200 . in order to calculate gf given by eq.([65 ] ) we will use the self - consistent hartree approximation ( scha ) . for this approximation we replace the real action by an appropriate gaussian one in such a way that all terms which include more then two fields @xmath201 or / and @xmath202 are written in all possible ways as products of pairs of @xmath201 or @xmath202 coupled to self - consistent averages of the remaining fields . the analogy between scha and sp - approximation at @xmath34 for the special case when the non - quadratic terms in the action are only the functions of the mean - squared displacement @xmath203 ^ 2\right>/n$ ] has been proven in ref.@xcite . in our case the action in eq.([65 ] ) has a more general form . in the appendix e we show that the scha and the _ next to the saddle point approximation _ ( nspa ) merge and both become exact , if the gf with an arbitrary action can be treated by a steepest descent approach at @xmath204 . let us make the fourier transformation of the mean - fields @xmath205 and insert it in eq.([65 ] ) . then for eq.([65 ] ) we use the hartree - type action ( see eq.([140 ] ) . by doing so we put for simplicity the expectation value @xmath206 . it is easy to assure oneselves also that the `` response - response overlap '' @xmath207 ( similar to @xmath208 in ref.@xcite ) . in the curse of the derivation we have used sp - equation ( [ sp ] ) and defined the correlator ( or the incoherent scattering function ) @xmath209 as well as the response functions @xmath210 where @xmath211 stands for the averaging with the hartree - type of action . after collection of all terms the final result ( at @xmath212 ) reads then @xmath213 + \int d t d t ' i\hat{\bf r}^{(p)}(t ) { \bf r}^{(p)}(t)\lambda(t , t')\nonumber\\ & - & \int d t d t ' i\hat{\bf r}^{(p)}(t ) { \bf r}^{(p)}(t')\lambda(t , t ' ) + \int d t d t ' i\hat{\bf r}^{(p)}(t)i\hat{\bf r}^{(p)}(t')\eta(t , t ' ) \bigg\}\ ; , \label{71}\end{aligned}\ ] ] where @xmath214 and @xmath215 ^ 2 \ ; .\label{73}\end{aligned}\ ] ] in eqs.([71 ] ) - ( [ 73 ] ) we have restricted ourselves to the homogeneous case @xmath216 for the correlation and response function . the equation of motion for the one particle correlator @xmath217 and the corresponding response function @xmath218 ( which actually does not depend from the particle index @xmath2 ) can be derived from eq.([71 ] ) by using the standard techniques @xcite . the resulting equations read @xmath219{\cal p}(t , t')\nonumber - \int\limits_{-\infty}^{t}d\tau \lambda ( t , \tau){\cal p}(t , t ' ) \nonumber\\ + \int\limits_{-\infty}^{t'}d\tau \ : \eta ( t , \tau){\cal g}(t ' , \tau ) = -2 t \gamma_{0}{\cal g}(t ' , t)\label{76}\end{aligned}\ ] ] and @xmath219{\cal g}(t , t')\nonumber - \int\limits_{-\infty}^{t}d\tau \lambda ( t , \tau){\cal g}(\tau , t ' ) \nonumber\\ = - \delta(t - t')\ ; . \label{77}\end{aligned}\ ] ] eqs.([76 ] ) - ( [ 77 ] ) should be supplemented with the initial conditions @xmath220 and @xmath221 . by making use this condition , equipartition @xmath222 , causality @xmath223 at @xmath224 as well as the condition @xmath225 one find from eq.([76 ] ) the following equation @xmath226{\cal p}(t , t ) - \int\limits_{-\infty}^{t}d\tau \lambda ( t , \tau){\cal p}(\tau , t ' ) + \int\limits_{-\infty}^{t}d\tau \ : \eta ( t , \tau){\cal g}(t , \tau ) = 2 t .\label{76'}\end{aligned}\ ] ] the set of eqs.([76 ] ) - ( [ 77 ] ) have the same structure as the dyson eq.([37 ] ) . after the matrices inversions and going to the time domain eqs.([37 ] ) ( in time - translational invariant case ) take the form @xmath227g_{01}(t , t ' ) & - & \int\limits_{-\infty}^{t}d\tau \sigma_{10}(t - \tau ) g_{01}(\tau - t ' ) = \delta(t - t')\\\label{78 } \left[m_{0}\frac{\partial^{2}}{\partial t^{2 } } + \gamma_{0}\frac{\partial}{\partial t } + \mu(0)\right]g_{00}(t , t ' ) & - & \int\limits_{-\infty}^{t}d\tau \sigma_{10}(t - \tau ) g_{00}(\tau - t ' ) \nonumber\\ & -&\int d\tau\sigma_{11}(t - \tau ) g_{10}(\tau - t ' ) = 2t\gamma_{0 } g_{10}(t - t')\ ; , \label{79}\end{aligned}\ ] ] where @xmath228 and rpa - fourier spectrum @xmath229 eqs.([76 ] ) - ( [ 77 ] ) are turned to the dyson equations ( [ 78 ] ) - ( [ 79 ] ) provided that @xmath230 we can show @xcite that the relation @xmath231 holds , provided that the fdt is satisfied for @xmath232 . we then have in addition @xmath233 bearing eqs.([81 ] ) in mind the eq.([82 ] ) takes in our case the form ( at @xmath234 ) @xmath235 the validity of the relationship ( [ 84 ] ) can be checked by replacing ( [ 72 ] ) and ( [ 73 ] ) in ( [ 84 ] ) . the general eqs.([76 ] ) - ( [ 77 ] ) are equivalent , _ mutatis mutandis _ , to the corresponding equations for the @xmath2 - spin system or a particle in the random potential at the large dimension @xcite . the most important features of these equations are the _ glassy dynamical behavior _ and the universal _ aging regime_. at low temperatures the system tries to minimize the energy and each particle ( with a number @xmath2 ) tends to surround itself with other particles which assure the strength parameter @xmath236 . on the other hand the long - range interaction tries to support other pairs @xmath237 corresponding to @xmath238 . as a result the system becomes `` frustrated '' and many local free energy minima appear . in the spirit of ref.@xcite when @xmath239 we have to discriminate between different cases : ( i ) the _ asymptotic regime _ when @xmath240 and ( ii ) the _ aging regime _ when @xmath241 . the aging regime is much more complicated because the time - translational invariance and fdt are violated . this regime has been extensively investigated both theoretically @xcite and by the computer simulation @xcite . in the following we restrict ourselves only to the asymptotic regime , for the sake of clearness and simplicity , and since the main features will be already visible . this asymptotic regime is characterized by the large time scales , i.e. , @xmath242 but keeping the difference @xmath243 finite . under these circumstances we can define @xmath244 then the equation for the displacement @xmath245 $ ] , response function @xmath246 and the static correlator @xmath247 takes correspondingly the forms @xmath248 { \cal d}_{\rm as}(\tau ) - \int\limits_{0}^{\tau}d\tau ' \lambda_{\rm as } ( \tau - \tau'){\cal d}_{\rm as}(\tau ' ) - \nonumber\\ - \int\limits_{0}^{\infty}d\tau'\left [ \lambda_{\rm as } ( \tau + \tau ' ) - \lambda_{\rm as}(\tau')\right]{\cal d}_{\rm as}(\tau ' ) - 2 \int\limits_{0}^{\infty}d\tau'\left [ \eta_{\rm as } ( \tau + \tau ' ) - \eta_{\rm as}(\tau')\right]{\cal g}_{\rm as}(\tau')= 2 t \label{86}\end{aligned}\ ] ] @xmath248 { \cal g}_{\rm as}(\tau ) - \int\limits_{0}^{\tau}d\tau ' \lambda_{\rm as } ( \tau - \tau'){\cal g}_{\rm as}(\tau ' ) = 0\label{87}\end{aligned}\ ] ] @xmath249 , \label{88 } \end{aligned}\ ] ] where @xmath250 @xmath251 however , it is also convenient to define the `` anomaly '' @xmath252 @xcite . the eqs.([86 ] ) - ( [ 88 ] ) has been analyzed first in the context of polymeric manifold in the random media @xcite and the random - phase sine - gordon model @xcite . the peculiarity of our model is defined by its memory functions @xmath253 and @xmath254 . for example , let as give an explicit expression for @xmath254 . the gaussian form of the correlator , @xmath255 , leads from eq.([73])to the result @xmath256 usually it is assumed that at the high temperature fdt holds , i.e. @xmath257 and @xmath258 in this case eqs.([86 ] ) and ( [ 87 ] ) merge and take a simple form @xmath248 { \cal d}_{\rm as}(\tau ) - \int\limits_{0}^{\tau}d\tau ' \lambda_{\rm as } ( \tau - \tau'){\cal d}_{\rm as}(\tau ' ) = 2 t \ ; .\label{94}\end{aligned}\ ] ] it turns out @xcite that the solution which satisfies the fdt is only stable above a critical temperature @xmath259 . for the stability analysis it is convenient to represent eq.([94 ] ) in the form @xmath260 { \cal d}_{\rm as}(\tau ) -\int\limits_{0}^{\tau}d\tau'\left[\eta_{\rm as } ( \tau - \tau ' ) - \eta_{\rm as}(\tau)\right]\frac{\partial}{\partial\tau'}{\cal d}_{\rm as}(\tau ' ) = 2t\ ; , \label{95}\end{aligned}\ ] ] where @xmath261 for @xmath262 the stability condition which comes out of eq.([95 ] ) reads @xmath263{\cal d}_{\rm as}(\tau ) \le 2 t \quad.\label{97}\end{aligned}\ ] ] then the stationary value of the displacement @xmath264 reads @xmath265 by taking into account eqs.([92 ] ) and ( [ 93 ] ) the stability condition becomes @xmath266 for @xmath267 , where @xmath268 q - \left(\frac{\chi_{0}}{t}\right)^2 \frac{\sqrt{\pi}}{12 } \sqrt{q } + 1 \ ; .\label{100}\end{aligned}\ ] ] the critical values @xmath269 and @xmath259 at which the condition ( [ 99 ] ) first becomes violated is defined by equations @xmath270 consequently , eqs . ( [ 102 ] ) have the simple solution @xmath271 the fig.3 shows the behavior of @xmath272 in the vicinity of the critical point . it can be seen that the minimum , @xmath273 , at which @xmath274 appears continuously , i.e. the instability of the fdt solution shows up as a @xmath275 order phase transition . this is analogous to the dynamics of polymeric manifolds in a medium with the long range correlation in disorder @xcite . in particular , if `` anomaly '' @xmath276 then @xmath277 and @xmath278 , so in this case the fdt solution is unstable for any finite temperature . + let us consider the dynamics at the temperature slightly above the critical point : @xmath279 , where @xmath280 . for large @xmath281 the decomposition @xmath282 is possible , where @xmath283 . the substitution of this decomposition into eq.([95 ] ) and the expansion up to the second order with respect to @xmath284 yields @xmath285 ^ 2 + \frac{1}{2}\int\limits_{0}^{\tau}d\tau'\left[f(\tau - \tau ' ) - f(\tau)\right]\frac{\partial}{\partial\tau'}f(\tau ' ) = 0 \ ; .\label{105}\end{aligned}\ ] ] following ref.@xcite let us make the laplace transformation @xmath286 \equiv { \tilde f}(z)$ ] and introduce the scaling functions , @xmath287 or @xmath288 , in a such way @xmath289 where @xmath290 and @xmath291 . if @xmath292 and @xmath293 then one can write eq.([105 ] ) in the form @xmath294 ( see eq.(2.68b ) from ref.@xcite ) . in the critical regime @xmath295 ( or @xmath296 ) the solution of eq.([108 ] ) has a form @xmath297 . in this limit the first term in eq.([108 ] ) is dropped out and the exponent is defined by the equation @xmath298 the solution of eq . ( [ 109 ] ) gives @xmath299 . in the opposite limit @xmath300 ( or @xmath301 ) the last term in eq.([108 ] ) can be neglected . in this case the solution has the form @xmath302 , where @xmath303 . as a result the overall scaling reads @xmath304 where @xmath305 is some constant . at @xmath306 the fdt is violated for the large time separation @xmath281 and the aging regime is arising . it should be mentioned that the asymptotic regime can not be decoupled from the aging one @xcite . in actual fact , the `` anomaly '' @xmath307 in the asymptotic eqs . ( [ 86 ] ) - ( [ 88 ] ) strictly speaking can be calculated only from the aging regime . because of the distinct aim in this paper we are not going to discuss the aging regime here expecting to return to it in a later publication . in the present paper we have considered the dynamics of two models with the long range repulsive interaction . the interaction potential was designed in a way to enable the saddle point treatment , as well as a fluctuation expansion . for the pure model we have derived eq.([48 ] ) for the full correlation matrix @xmath308 in the one - loop approximation , which has an the explicit solution ( see eqs.([49 ] ) - ( [ 51 ] ) ) . this solution has a `` boring behavior '' at @xmath162 which manifests the absence of the glass dynamics . the physical background of this stems from the fact that the potential is much too soft and the `` cage effect '' is completely missing . this conclusion is in accordance with the interacting particles statistical thermodynamics analysis , which was given in ref.@xcite . it was shown there that for the infinite range interaction potential , which allows a well defined saddle point treatment , the glassy phase is simply suppressed . on the other hand , the same model but with a randomly distributed strength of interaction ( the `` random - bond model '' ) leads to the continuous glass transition . this type of transition is also the case for the polymeric manifolds in the disordered medium with long range correlation @xcite as well as for the @xmath2 - spin interaction spin - glass model at the large external field @xcite . it would be also interesting to investigate the more realistic `` random sequence model '' when each particle carry a random `` charge '' . qualitatively the same glassy behavior have been found in the pure spin models with the deterministic but very rapidly oscillating coupling between variables @xcite . it was assumed that the effective quenched disorder is `` self - induced''@xcite . this means that because of the slow dynamics some degrees of freedom freeze and play the role of the effectively quenched disorder . as a conclusion , the glass transition in the pure systems of the interacting particles , where the disorder is actually `` self - induced '' , goes beyond the mean - field level @xcite . this appears to hard to implement in the present context , because it implies the consideration of the short range interaction potential as well as activated processes . the authors have benefited from discussions with m. fuchs , a. latz , g. migliorini and greatly indebted to the deutsche forschungsgemeinschaft ( dfg ) , the sonderforschungsbereich sfb 262 for financial support of the work . in the full analogy with eq.([25 ] ) the expression for @xmath309 reads @xmath310 \label{113}\end{aligned}\ ] ] the expansion of the @xmath311 up to the @xmath312 order with respect to the mean field @xmath313 can be easily obtained from eq.([24 ] ) @xmath314\nonumber\\ & + & \frac{1}{2 ! } \int d2 d3\left<\delta\rho_{\alpha}(1)\delta\rho_{\beta}(2)\delta\rho_{\gamma}(3 ) \right>_{0 } \left[\chi_{\beta}(2 ) - i\bar{\psi_{\beta}}(2)\right ] \left[\chi_{\gamma}(3 ) - i\bar{\psi_{\gamma}}(3)\right]\nonumber\\ & + & \frac{1}{3!}\int d2 d3 d4\left<\delta\rho_{\alpha}(1)\delta\rho_{\beta}(2)\delta\rho_{\gamma}(3 ) \delta\rho_{\delta}(4 ) \right>_{0 } \nonumber\\ & \times&\left[\chi_{\beta}(2 ) - i\bar{\psi_{\beta}}(2)\right ] \left[\chi_{\gamma}(3 ) - i\bar{\psi_{\gamma}}(3)\right]\left[\chi_{\delta}(4 ) - i\bar{\psi_{\delta}}(4)\right]\quad . \label{114 } \end{aligned}\ ] ] by using the sp - equation ( [ 23 ] ) and after threefold differentiation with respect to @xmath65 ( see eq.([113 ] ) ) we find @xmath315^{-1 } \right\}_{{\bar\alpha}\alpha}({\bar 1 } , 1)\left\{\left[{\hat 1 } + \mu { \hat v } { \hat f}^{(2)}\right]^{-1 } \right\}_{{\bar\beta}\beta}({\bar 2 } , 2)\nonumber\\ & \times&\left\{\left[{\hat 1 } + \mu { \hat v } { \hat f}^{(2)}\right]^{-1 } \right\}_{{\bar\gamma}\gamma}({\bar 3 } , 3)\left\{\left[{\hat 1 } + \mu { \hat v } { \hat f}^{(2)}\right]^{-1 } \right\}_{{\bar\delta}\delta}({\bar 4 } , 4)\quad , \label{115}\end{aligned}\ ] ] where the 4-point free system correlation matrix @xmath316 in eq.([115 ] ) we imply the summation ( integration ) over the barred indices ( barred space - time variables ) . when deriving eq.([115 ] ) we have also kept in mind that @xmath317 and @xmath318 , etc . the fact that the matrix @xmath319 is symmetrical with respect to simultaneous permutations of greek indices and space - time arguments as well as eq.([26 ] ) have been used . it is easy to show that @xmath319 is factorized @xmath320 on the other side it is instructive to check that even in this case @xmath309 can not be factorized . the substitution of the eg.([115 ] ) into eq.([46 ] ) after the straightforward algebra yields @xmath322^{-1 } - \left [ \mu^{-1 } { \hat v}^{-1 } + { \hat f}\right]^{-1}\right\}_{{\bar\beta}{\bar\alpha}}({\bar 2 } , { \bar 1})\nonumber\\ & \times&\left\{f_{{\bar\alpha}{\bar\beta}}({\bar 1},{\bar 2})f_{{\bar\gamma}{\bar\delta}}({\bar 3},{\bar 4 } ) + f_{{\bar\alpha}{\bar\gamma}}({\bar 1},{\bar 3})f_{{\bar\beta}{\bar\delta}}({\bar 2},{\bar 4 } ) + f_{{\bar\alpha}{\bar\delta}}({\bar 1},{\bar 4})f_{{\bar\beta}{\bar\gamma}}({\bar 2},{\bar 3})\right\}\nonumber\\ & \times&\left\{\left[{\bar 1 } + 2\mu { \bar v } { \bar f}\right]^{-1}\right\}_{{\bar\gamma}\alpha}({\bar 3 } , 1)\left\{\left[{\bar 1 } + 2\mu { \hat v } { \hat f}\right]^{-1}\right\}_{{\bar\delta}\beta}({\bar 4 } , 2)\quad,\end{aligned}\ ] ] where as before for the repeated barred indices ( variables ) the summation ( integration ) is implied . for the time - space - translational invariant case the respective fourier transformation leads to the result : @xmath323^{-1 } - \left[\mu^{-1}{\hat v}^{-1 } + { \hat f}\right]^{-1}\right\}_{{\bar\beta}{\bar\alpha}}(-{\bf k } , -\omega ) f_{{\bar\alpha}{\bar\gamma}}(-{\bf k},-\omega)f_{{\bar\beta}{\bar\delta}}({\bf k},\omega)\nonumber\\ & + & \left\{\left[\left(2 \mu \right)^{-1}{\hat v}^{-1 } + { \hat f}\right]^{-1 } - \left[\mu^{-1}{\hat v}^{-1 } + { \hat f}\right]^{-1}\right\}_{{\bar\beta}{\bar\alpha}}({\bf k } , \omega ) f_{{\bar\alpha}{\bar\delta}}({\bf k},\omega)f_{{\bar\beta}{\bar\gamma}}(-{\bf k},-\omega)\bigr\}\nonumber\\ & \times&\left\{\left[{\bar 1 } + 2\mu { \hat v } { \hat f}\right]^{-1}\right\}_{{\bar\gamma}\alpha}(-{\bf k } , -\omega)\left\{\left[{\bar 1 } + 2\mu { \bar v } { \bar f}\right]^{-1}\right\}_{{\bar\delta}\beta}({\bf k } , \omega)\quad,\end{aligned}\ ] ] where the trace @xmath324^{-1 } - \left[{\hat 1 } + \mu^{-1}{\hat f}^{-1}{\hat v}^{-1}\right]^{-1 } \right\}_{{\bar\alpha}{\bar\alpha}}({\bf q } , \omega)\quad.\end{aligned}\ ] ] with the correlation matrix @xmath325 given by eqs.([49 ] ) , by doing integration over @xmath326 one can check that the trace @xmath327 . this gives finally @xmath328 where @xmath329^{-1}\right\}_{\alpha{\bar\gamma}}({\bf k } , \omega)f_{{\bar\gamma}{\bar\beta}}({\bf k},\omega)\nonumber\\ & \times&\left\{\left[{\hat 1 } + \left ( 2 \mu \right)^{-1}{\hat f}^{-1}{\hat v}^{-1}\right]^{-1 } - \left[{\hat 1 } + \mu^{-1}{\hat f}^{-1}{\hat v}^{-1}\right]^{-1 } \right\}_{{\bar\beta}{\bar\delta}}({\bf q } , \omega)\nonumber\\ & \times&\left\{\left[{\bar 1 } + 2\mu { \hat v } { \hat f}\right]^{-1}\right\}_{\alpha{\bar\gamma}}({\bf k } , \omega)\quad .\end{aligned}\ ] ] in this case the direct correlation function @xmath330 and its fourier transformation takes the scaling form @xmath331 let us insert this expression to the mct - memory kernel ( see eq.(3.32 ) in @xcite ) . it is reasonable then to rescale the integration variables in the memory kernel , @xmath332 as well as to put for the external wave vector @xmath333 , where @xmath334 is some reference wave vector . the last scaling means that in the mf - limit an experiment probes a very small wave vector : @xmath335 . the resulting scaling of the memory kernel , @xmath336 , reads @xmath337\right\}^2}{q_{0}^2}\ : \widetilde{s}({\bf k } , t ) \:\widetilde{s}({\bf p } , t)\ ; , \label{kernel}\end{aligned}\ ] ] where we have took into account the scaling form of the correlator : @xmath338 . thus we finally arrive at the scaling @xmath339 and the glass transition dies out . let us prove that scha becomes exact for gf given by ( [ 65 ] ) in the limit @xmath34 . we will consider even a more general gf @xmath340 + \sum_{p=1}^{n}\int d1 x_{\alpha}^{(p)}(1)\chi_{\alpha}(1)\bigg\ } \;,\label{117}\end{aligned}\ ] ] where we have used the shorthand notations @xmath341 and `` 1 '' embraces cartesian indices as well as time variable : @xmath342 . in eq.([117 ] ) @xmath343 $ ] is an arbitrary functional of @xmath344 . instead of the exact action functional in eq.([117 ] ) we consider now the trial one which has a gaussian form @xmath345 & = & \sum_{p=1}^{n}\bigg\ { \frac{1}{2 } \int d1 d2 \:x_{\alpha}^{(p)}(1)a_{\alpha\beta}(1 , 2)x_{\beta}^{(p)}(2)\nonumber\\ & -&\int d1 d2\:x_{\alpha}^{(p)}(1){\gamma}_{\alpha\beta}(1 , 2)x_{\beta}^{(p)}(2 ) - \int d1 l_{\alpha}(1)x_{\alpha}^{(p)}(1)\bigg\ } \ ; .\label{119}\end{aligned}\ ] ] let us look for the `` best '' coefficients @xmath346 and @xmath347 in a sense that the exact `` free energy '' @xmath348 = - \log z\left\{\chi_{\alpha}\right\}$ ] tends to the trial one @xmath349 = - \log \int\prod d x_{\alpha}^{(p)}\exp \ { - s [ x_{\alpha}^{(p)}]\}$ ] , i. e. @xmath350 \longrightarrow f_{0}\left[\chi_{\alpha}\right]\label{120}\end{aligned}\ ] ] and both becomes exact at @xmath34 . we can show that the property ( [ 119 ] ) is satisfied by @xmath351 and @xmath352 which are obtained by extremization of the functional @xmath353\bigg\ } \nonumber\\ & + & \sum_{p=1}^{n}\bigg\{\int d1 d2 \gamma_{\alpha\beta}(1 , 2 ) \left < x_{\alpha}^{(p)}(1)x_{\beta}^{(p)}(2)\right>_{\rm s}\nonumber\\ & + & \int d1 \left[l_{\alpha}(1 ) - \chi_{\alpha } \right ] \left < x_{\alpha}^{(p)}(1)\right>_{\rm s } - \left < w\left[x_{\alpha}^{(p)}(1)\right]\right>_{\rm s } \bigg\}\ ; , \label{121}\end{aligned}\ ] ] where we use the notations @xmath354\bigr\}}{\int\prod_{p=1}^{n}\prod_{\alpha=0 , 1}dx_{\alpha}^{(p)}(1)\:\exp\bigl\ { - s\left [ x_{\alpha}^{(p)}\right]\bigr\ } } \;.\label{122}\end{aligned}\ ] ] the extremization conditions reads @xmath355 the variations in eqs.([122 ] ) can be done directly . during the calculation the generalized wick s theorem @xcite should be also taken into account . namely , because the averaging ( [ 121 ] ) is simply the gaussian integral the wick s theorem yields @xmath356\right>_{\rm s } & = & \left < x_{\alpha}^{(p)}(1)\right>_{\rm s}\left < w\left[x_{\alpha}^{(p)}\right]\right>_{\rm s } \nonumber\\ & + & \int d2 \:\left<\delta x_{\alpha}^{(p)}(1)\delta x_{\beta}^{(p)}(2 ) \right>_{\rm s}\:\left<\frac{\delta}{\delta x_{\beta}^{(p)}(2 ) } \ : w\left[x_{\alpha}^{(p)}\right]\right>_{\rm s}\ ; , \label{124}\end{aligned}\ ] ] where @xmath357 . after the straightforward calculation we find @xmath358\right>_{\rm s}\label{125}\end{aligned}\ ] ] and @xmath359\:\left < x_{\alpha}^{(p)}(2)\right>_{\rm s}\ ; .\label{126}\end{aligned}\ ] ] then equations for the two moments take the form @xmath360_{\alpha\beta]}(1 , 2 ) \left[\left<\frac{\delta}{\delta x_{\beta}^{(p)}(2 ) } w\left[x_{\alpha}^{(p)}\right]\right>_{\rm s } + \chi_{\beta}(2)\right]\label{127}\end{aligned}\ ] ] and @xmath361^{-1}\right\}_{\alpha\beta}(1 , 2 ) \ ; , \label{128}\end{aligned}\ ] ] where @xmath362 and @xmath363 stands for the corresponding @xmath364 - matrices . on the other side the saddle point ( sp ) treatment of eq.([117 ] ) at @xmath365 yields @xmath366 and @xmath367^{-1}\right\}_{\alpha\beta}(1 , 2 ) \ ; , \label{130}\end{aligned}\ ] ] where @xmath368 and @xmath369 stands for the field in sp . in order to show the analogy between eqs.([127])-([128 ] ) and eqs.([129])-([130 ] ) let us make the functional fourier transformation @xmath370\right\ } = \int d x_{\alpha}^{(p)}(1)\exp\left\{w\left[x_{\alpha}^{(p)}(1)\right ] - i\int d1\:x_{\alpha}^{(p)}(1)y_{\alpha}^{(p)}(1)\right\}\label{132}\end{aligned}\ ] ] and its inversion @xmath371\right\ } = \int d y_{\alpha}^{(p)}(1)\exp\left\{k\left[y_{\alpha}^{(p)}(1)\right ] + i\int d1\:x_{\alpha}^{(p)}(1)y_{\alpha}^{(p)}(1)\right\}\ ; .\label{133}\end{aligned}\ ] ] then eqs.([129 ] ) - ( [ 130 ] ) can be written as @xmath372_{\alpha\beta}(1 , 2)\left[i\left < y_{\beta}^{(p)}(2)\right>_{\rm sp } + \chi_{\beta}(2)\right]\label{134}\end{aligned}\ ] ] and @xmath373^{-1}\right\}_{\alpha\beta}(1 , 2)\ ; , \label{135}\end{aligned}\ ] ] where the correlation matrix @xmath374 and @xmath375 + i\int d1\:\bar{x}_{\alpha}^{(p)}(1)y_{\alpha}^{(p)}(1)\right\}}{\int d y_{\alpha}^{(p ) } \exp\left\{k\left[y_{\alpha}^{(p)}\right ] + i\int d1\:\bar{x}_{\alpha}^{(p)}(1)y_{\alpha}^{(p)}(1)\right\}}\ ; .\label{137}\end{aligned}\ ] ] at @xmath376 by making use eq.([133 ] ) one can immediately see that @xmath377\right>_{\rm s } \longrightarrow i\left < y_{\alpha}^{(p)}(1)\right>_{\rm sp}\label{138}\end{aligned}\ ] ] and @xmath378\right>_{\rm s } \longrightarrow - \left<\delta y_{\alpha}^{(p)}(1)\delta y_{\beta}^{(p)}(2 ) \right>_{\rm sp}\ ; , \label{139}\end{aligned}\ ] ] and scha exactly corresponds to nspa . for the case which was treated in sec.iii c @xmath379 and the hartree - type action ( [ 119 ] ) cast the form @xmath345 & = & \sum_{p=1}^{n}\bigg\ { \frac{1}{2 } \int d1 d2 \:x_{\alpha}^{(p)}(1)a_{\alpha\beta}(1 , 2)x_{\beta}^{(p)}(2)\nonumber\\ & -&\frac{1}{2}\int d1 d2\:\left<\frac{\delta^2}{\delta x_{\alpha}^{(p)}(1 ) \delta x_{\beta}^{(p)}(2 ) } \ : w\left[x_{\alpha}^{(p)}\right]\right>_{\rm s } x_{\alpha}^{(p)}(1 ) x_{\beta}^{(p)}(2 ) \bigg\ } \ ; .\label{140}\end{aligned}\ ] ] 99 w.gtze in , _ liquids , freezing and glass transition _ , ed . by j.p.hansen , d.levesque and j.zinn-justin , amsterdam , north - holland , 1991 v.n . prigodin , j. phys . * c 4 * , 785 ( 1992 ) t.r . kirkpatrick and d. thirumalai , phys . rev . * b36 * , 5388 ( 1987 ) t.r . kirkpatrick and d. thirumalai , phys . * b37 * , 5342 ( 1988 ) t.r . kirkpatrick and d. thirumalai , trans . theor . and stat . phys . * 24,927 ( 1995 ) a. crisanti , h. horner and h .- j . sommers , z. phys . 92 * , 257 ( 1993 ) j. - p. bouchaud , l.f . cugliondolo , j. kurchan and m. mzard , physica * a226 * , 243 ( 1996 ) j. - p. bouchaud , l.f . cugliondolo , j. kurchan and m. mzard , in , _ spin glasses and random fields _ , ed . by a.p . young , world scientific , 1997 h. kinzelbach and h. horner j. phys.i ( france ) * 3 * , 1329 ( 1993 ) h. kinzelbach and h. horner j. phys.i ( france ) * 3 * , 1901 ( 1993 ) s. franz and m. mzard , physica * a210 * , 48 ( 1993 ) v.g . rostiashvili and t.a . vilgis , europhys . lett . * 49 * , 162 ( 2000 ) r.g . palmer , adv . phys . * 31 * , 669 ( 1982 ) l.f . cugliondolo , j. kurchan , g. parisi and f. retort , phys . rev . lett.74 1012 ( 1995 ) j.l . lebowitz and o. penrose , j. math . phys.7 , 98 ( 1996);o . penrose and j.l . lebowitz , j. stat . phys.3 , 211 ( 1971 ) j.- p. hansen and i.r . mcdonald , _ theory of simple liquids _ academic press , london , 1986 c. dedominicis and l. pelityphys . rev . * b18,353 ( 1978 ) ; r. bauch , h.k . janssen and h.wagner . z. phys.b24 * , 113 ( 1976 ) v.g . rostiashvili and r. schilling z. phys.b102 , 117 ( 1997 ) v.g . rostiashvili , m. rehkopf and t.a . vilgis , eur . phys . j.7 , 233 ( 1998 ) v.g . rostiashvili , m. rehkopf and t.a . vilgis , j . 110 * , 639 ( 1999 ) v.g . rostiashvili , physicaa231 , 575 ( 1996 ) j. zinn - justin , _ quantum field theory and critical phenomena _ clarendon press , oxford , 1989 d. forster , _ hydrodynamic fluctuations , brocen symmetry and correlation function _ , benjamin , 1975 a. weyersberg and t.a . vilgis , phys . e , * 48 * , 277 , 1993 , t.a . vilgis , a. weyersberg , m.g . brereton phys . e * 49 * , 3031 , 1994 m. fuchs , private communication a.i . melcuk , r.a . ramos , h. gould , w. klein and r.d . mountain phys . rev . lett.75 , 2522 ( 1995 ) ; w. klein , h. gould , r.a . ramos , i. clejan and a.i . melcuk physica * a205 * 738 ( 1994 ) t. garel , h. orland and e. pitardin , in _ spin glasses and random fields _ young , world scientific , 1997 t. garel , h. orland , europhys . lett . * 7 * , 307 ( 1988 ) e.i . shakhnovich and a.m. gutin , j. phys.a22 , 1647 ( 1989 ) c. d. sfatos , a.m . gutin and e.i . shakhnovich phys . * e48,465 ( 1993 ) h. sompolinsky and a. zippelius , phys . b25,6860 ( 1982 ) h.g . schuster , z . phys.b45,99 ( 1981 ) m. mzard , g. parisi and m.a . virazoro , _ spin glass and beyond _ , singapore , world scientific , 1987 l.f.cugliondolo and j. kurchan phys . rev . lett.71 * , 173 ( 1993 ) l.f.cugliondolo and p. le doussal phys . rev.e53 , 1525 ( 1996 ) g. parisi phys . lett . * 79 * , 3660 ( 1997 ) w. kob and j .- l . barrat phys . lett . * 78 * , 4581 ( 1997 ) ; physica * a263 * , 234 ( 1999 ) ; cond - mat/9905248 d. cule and y. shapir phys . rev.e53 , 1553 ( 1996 ) j.p . bouchaud and m. mzard j. phys . i ( france ) * 4 * 1109 ( 1994 ) g. migliorini and f. ritort j. phys . * a27 * 7669 ( 1994 ) w. krauth and m. mzard z. phys.b97 127 ( 1995 ) s. franz and j. hertz phys . 74 * , 2114 ( 1995 )
emergence and manipulation of spins are a major research topic in spintronics . applying a controlled light is one of the promising techniques to manipulate spins . recently , the spin angular momentum of a circularly polarized light has been observed to induce the magnetization in solid - state materials through spin - orbit interactions@xcite . this technique has further been applied to the ultrafast magnetization switching , whose time is much shorter than that by an applied magnetic field@xcite . when a light is irradiated on a surface of a three - dimensional topological insulator ( ti ) , spin is predicted to emerge in the perpendicular direction to the electric field of the light@xcite . here , a ti is an anomalous material with strong spin - orbit interactions . electrons insulate in the bulk , while they conduct on the surface of the ti , where the exotic surface state of the ti is caused by both the spin - orbit interaction and the topological electric structure@xcite . on the surface of the ti , the direction of the spin and that of the momentum are perfectly locked to be perpendicular to each other , which is dubbed spin - momentum locking . because of this spin - momentum locking , the charge current generated along the direction of the electric field causes the spin density in the perpendicular direction@xcite . such a manipulation of spin and charge current using a light may make it possible to develop magneto - optical devices based on tis@xcite . recently , magneto - optical effects and optical excitation using a twisted light beam , whose phase is twisted around the direction of the propagation of light , have been theoretically predicted@xcite and experimentally carried out@xcite . a twisted light has the following two intriguing properties distinct from a plane wave@xcite . first , the phase of the light is twisted around the center of the beam , and hence , has a singularity at the center . as a result , strength of the light becomes zero at the center of the beam . second , because of the twisted phase , the strength of the light strongly depends on the space , whose distributions are manipulated by the angular momentum of the light . the above properties can be well - understood by writing down the electric field of the light . the electric field of the twisted light beam traveling along the @xmath0 axis at @xmath1 , @xmath2 , can be described by@xcite @xmath3,\end{aligned}\ ] ] where @xmath4 is the two - dimensional polar coordinates at @xmath1 , @xmath5 is the time , and @xmath6 and @xmath7 are the momentum and frequency of the twisted light beam , respectively . here , @xmath8 denotes the magnitude of the electric field , which depends on the space and becomes zero at the center @xmath9 for a nonzero @xmath10 due to the phase singularity . @xmath11 and @xmath12 represent the @xmath0 components of the spin and orbital angular momentum of the light , respectively . the former corresponds to the direction of the circular polarization , i.e. , @xmath13 represents a right - handed ( left - handed ) circularly polarized wave , while the latter describes the winding of the electric field in the @xmath1 plane . in fact , the electric field of a twisted light has the topological quantity . we will see later that the winding number of a twisted light given by eq . ( [ eq:1 - 1 ] ) is proportional to @xmath14 and @xmath15[see the discussion below eq . ( [ eq : winding - number ] ) ] . so far , it has been theoretically predicted that in the presence of the spin - orbit interaction , unconventional photo - induced spin excitation and current emerge due to the spatial dependence of the strength of the electric field of the twisted light . however , the orbital angular momentum of the light is not transfered to the spin polarization since the beam waist is much larger than the width of the electron wave function , which is in the order of the lattice constant , and hence each electrons feels locally uniform electric field . actually the experimental investigation of the photo - induced spin polarization@xcite could not detect the orbital angular momentum dependence in the semiconductor with the rashba and dresselhaus type spin orbit interaction as an exception there is a theoretical prediction for cylindrical quantum disks@xcite . in this paper , we theoretically study spin and charge generation due to the electric field of the twisted light beam on a disordered surface of a doped ti by using the green s function technique . we analytically calculate the linear response function of the spin density to a space - time dependent external electric field . we find that the local and nonlocal spin densities are induced by the electric field and the gradient of the electric field , respectively , via the spin - momentum locking . here , the local spin density comes from the charge current that flows along the electric field , whereas the nonlocal one couples to the diffusive charge current due to the impurity scatterings on the disordered surface of the ti . in addition , the gradient of the electric field also induces the charge density and the spin current . applying the obtained results to the electric field of a twisted light beam , we find that various spin distributions appear depending on the orbital as well as spin angular momentum of the light . moreover the spin distributions have topological structures i.e. , magnetic vortex - like textures , characterized with winding numbers , which dependent on both @xmath16 and @xmath10 . the induced spin structure evolves in time but its winding number remains a constant . the manipulation of such a topological spin structure could be applicable for the spintronics related to magnetic vortices and skyrmions . this paper is organized as follows . in sec . [ sec : ii ] , we introduce the model hamiltonian for the disordered surface of the ti in the presence of a space - time dependent electromagnetic field . we also present the green s functions on the disordered surface of the ti . in sec . [ sec : iii ] , we calculate the induced spin density on the surface of the ti within the linear response to the applied electric field . section iv discusses the general properties of the charge density , spin density , charge current , and spin current induced by the electric field on the surface of the ti . section [ sec:4 ] discusses the properties of the twisted - light - induced spin and charge distributions . section [ sec:5 ] summarizes the paper . appendices [ sec : a]-[sec : g ] give the detailed calculations used in sec . iii . we consider a disordered surface of a three - dimensional ti and apply a space - time dependent electromagnetic field to it . we assume that impurities on the disordered surface are nonmagnetic . the hamiltonian we consider is given by @xmath17\psi , \\ \label{eq:2 - 3 } % \ref{eq:2 - 3 } \mathcal{h}_{\textrm{em } } & = - ev_{\textrm{f } } \int d\bm{x } \psi^\dagger ( \hat { \bm{\sigma } } \times \bm{a}_{\textrm{em}})_z \psi , \\ \label{eq:2 - 4 } % \ref{eq:2 - 4 } \mathcal{v}_{\textrm{imp } } & = \int d\bm{x } \ u_\textrm{i } \psi^\dagger \psi , \end{aligned}\ ] ] where @xmath18 and @xmath19 are the creation and annihilation operators of conduction electrons on the surface of the ti , @xmath20 are the pauli matrices , and @xmath21 is the elementary charge of electrons . here , we assume a doped ti , and @xmath22 and @xmath23 are the fermi energy and the fermi velocity , respectively , on the surface of the doped ti . @xmath24 is the gauge coupling between conduction electrons and the electromagnetic field . the vector potential of the electromagnetic field @xmath25 generally depends on the space and time , and the electric field and the magnetic field are respectively given by @xmath26 and @xmath27 . @xmath28 describes the potential due to the nonmagnetic impurity scatterings , where @xmath29 is the potential energy density with @xmath30 being the number of the impurities , @xmath31 a constant , @xmath32 the position of the @xmath33-th impurity on the surface , and @xmath34 the area of the surface . here , the contribution from the impurity potential is treated as the impurity average @xmath35 , where @xmath36 is the fourier transform of @xmath37 . to calculate the spin density and the charge density , we use the green s function method . in the absence of the electromagnetic field , the retarded green s function is given by @xcite @xmath38^{-1}.\end{aligned}\ ] ] here , a variable with a hat denotes a two - by - two matrix . by calculating the self - energy @xmath39 within the self - consistent born approximation @xcite and expanding it with respect @xmath40 up to the linear terms , @xcite eq . ( [ eq:2 - 5 ] ) is rewritten as @xmath41^{-1},\end{aligned}\ ] ] where @xmath42 is the modified fermi velocity due to nonmagnetic impurity scatterings with @xmath43 being a constant depending on the properties of the ti , and the imaginary part of the self - energy @xmath44 defines the transport relaxation time @xmath45 . here , @xmath46 is the concentration of the impurities on the surface and @xmath47 is the density of states of electrons on the surface . since we are considering a metallic state , @xmath48 satisfies @xmath49 . by comparing eqs . ( [ eq:2 - 2 ] ) and ( [ eq:2 - 6 ] ) , we see that the effective hamiltonian for the surface electrons affected by impurities is given by the right - hand side of eq . ( [ eq:2 - 2 ] ) with replacing @xmath50 with @xmath51 . accordingly , @xmath50 in eq . ( [ eq:2 - 3 ] ) is replaced by @xmath51 . [ sec:3 ] we calculate the spin density induced by an applied electric field on a disordered surface of a doped ti by using the keldysh green s function method within the linear response to @xmath52 . the spin density @xmath53 is described by using the lesser component of the keldysh - green s function in the same position and time @xmath54 as @xmath55 \hspace{5mm}(i = x , y , z),\end{aligned}\ ] ] where @xmath56 denotes the trace over the spin indices . then , from the dyson s equation for @xmath57 the induced spin density within the linear response to @xmath52 is given by @xmath58 a_{{\textrm{em } } , u } ( { \bm{q},\omega } ) , \end{aligned}\ ] ] where @xmath59 is the levi - civita symbol , @xmath60 is the spin - spin response function , and @xmath61 and @xmath7 are the momentum and frequency of @xmath62 , respectively . the response function @xmath60 can be decomposed as @xmath63 and @xmath64 is represented by @xmath65^ < & = \sum_{\bm{k } , \omega } [ \hat{g}_{\bm{k}-\frac{q}{2},\omega-\frac{\omega}{2 } } \hat{\lambda}_\nu ( \omega,\bm{q},\omega ) \hat{g}_{\bm{k}+\frac{q}{2},\omega+\frac{\omega}{2 } } ] ^{<}.\end{aligned}\ ] ] here , @xmath66 is the vertex function due to @xmath28 , whose diagram is shown in the fig . [ fig : diagram ] , and is given by . the dashed and wavy lines mean the potential due to the nonmagnetic impurity scatterings and gauge coupling between conduction electrons and the electromagnetic field , respectively . ] @xmath67^n}_{\nu \mu}\hat{\sigma}_\mu,\end{aligned}\ ] ] where @xmath68 is the two - by - two identity matrix and @xmath69 is a @xmath70 matrix defined from the following equation @xmath71 expanding eq . ( [ eq:3 - 3 ] ) with respect to the retarded and advanced green s functions , @xmath72 and @xmath73 , based on the formula@xcite @xmath74 , where @xmath75 is the fermi distribution function , the spin - spin response function can be divided into three terms : @xmath76 here , @xmath77 ( = ) is defined by @xmath78^n}_{\nu \mu}\hat{\sigma}_{\mu } , \\ \label{eq:3 - 5 - 2 } % \ref{eq:3 - 5 - 2 } \hat { \gamma } ^{\textrm{ab}}_\nu(\omega,\bm{q},\omega ) & \equiv n_{\textrm{i } } u_{\textrm{i}}^2 \sum_{\bm{k } } \hat{g}^{\textrm{a}}_{\bm{k}-\frac{q}{2},\omega-\frac{\omega}{2 } } \hat{\sigma}_\nu \hat{g}^{\textrm{b}}_{\bm{k}+\frac{q}{2},\omega+\frac{\omega}{2 } } , \\ & = \sum_{\mu=0,x , y , z}\gamma^{\textrm{ab}}_{\nu \mu}(\omega,\bm{q},\omega)\hat{\sigma}_\mu.\end{aligned}\ ] ] by expanding @xmath79 and @xmath80 with respect to @xmath81 and @xmath7 , we find that they are in the order of @xmath82 and @xmath83 ( @xmath84 ) in @xmath85 ( @xmath86 ) can be approximated with @xmath87 ( see appendices [ sec : a]-[sec : a - c ] for the detailed calculation ) . then , by expanding the fermi distribution function with respect to @xmath7 , the dominant contributions of eqs . ( [ eq:3 - 7 ] ) , ( [ eq:3 - 8 ] ) and ( [ eq:3 - 9 ] ) are written by @xmath88 in addition , @xmath89 and @xmath90 are shown to be much smaller than @xmath91 ( the detailed calculation is given in appendix [ sec : d ] ) . thus , @xmath92 is approximately given by @xmath93 in the low - temperature limit , we approximate the derivative of the fermi distribution function as @xmath94 . then , the integral over @xmath95 in @xmath96 reduces to @xmath97 we further expand @xmath98 as @xmath99 , and rewrite eq . ( [ eq:3 - 13 ] ) as @xmath100 here , we define @xmath101 , which is calculated up to the quadratic terms in @xmath102 and the primary terms in @xmath7 as ( see appendices [ sec : a ] and [ sec : b ] for the detailed derivation ) @xmath103 , \\ \label{eq:3 - 16 } % \ref{eq:3 - 16 } \hat{i}_{\zeta = x , y } & = \frac{\pi\nu_e}{2\eta } \biggl [ \sum_{\nu = x , y } \biggl\ { \frac{1}{2}\biggl(1- i\omega \tau - \frac{3}{4}\ell^2 q^2 \biggr)\delta_{\zeta \nu } + \frac{1}{4}\ell^2 q_\zeta q_\nu \biggr\ } \hat{\sigma}_\nu + \sum_{\alpha = x , y } \frac{i}{2}\ell \hat{\sigma}_0 q_\alpha \epsilon_{\zeta \alpha z } \biggr ] , \\ \label{eq:3 - 16 - 2 } % \ref{eq:3 - 16 - 2 } \hat{i}_{\zeta = z } & = o \left ( \frac{\hbar}{\epsilon_\textrm{f } \tau } \right ) , \end{aligned}\ ] ] where @xmath104 is the mean free path . since @xmath105 is negligibly small as compared with @xmath106 and @xmath107 , we consider only the contributions from @xmath108 . since eqs . ( [ eq:3 - 15 ] ) and ( [ eq:3 - 16 ] ) do not include @xmath109 , they are represented by using the pauli matrices as @xmath110 where @xmath111 is the 3 @xmath112 3 symmetric matrix given by @xmath113 on the other hand , from eq . ( [ eq:3 - 4 - 2 ] ) , the vertex function @xmath114 can be described by @xmath115_{\nu \alpha}\hat{\sigma}_\alpha + \cdots \\ & = \sum_{\alpha=0,x , y}[(1-\gamma^{\textrm{ra}})^{-1}]_{\nu\alpha } \hat{\sigma}_\alpha,\end{aligned}\ ] ] where the second equality holds when @xmath116 is satisfied . by using @xmath117 , @xmath118 , and eq . ( [ eq:3 - 18 ] ) , one can see @xmath119 indeed satisfies @xmath116 . then , the matrix @xmath120 is calculated as @xmath121 from which we obtain the spin - spin response function as @xmath122_{\zeta ' \nu } \hat{\sigma}_{\zeta'},\end{aligned}\ ] ] where we have used the fact that @xmath123 and @xmath124 are symmetric matrices . thus , from eqs . ( [ eq:3 - 2 ] ) and ( [ eq:3 - 20 ] ) , the @xmath125 components of the spin density are given by @xmath126 } % \textrm{tr } [ \hat{\pi}_{\mu\nu } ] a_{\textrm{em},u } & = - \frac{e\tilde{v}_\textrm{f}\nu_e \tau}{2l^2 } \epsilon_{z \nu u } \partial_t\sum_{\bm{q},\omega } e^{i(\omega t -\bm{q}\cdot\bm{x } ) } [ \gamma^{\textrm{ra}}\lambda^{\textrm{ra}}]_{\mu\nu } a_{{\textrm{em } } , u}.\end{aligned}\ ] ] substituting eq . ( [ eq:3 - 19 ] ) in eq . ( [ eq:3 - 20 - 2 ] ) , we obtain @xmath127 } % \ { [ \hat{\gamma}\hat{\lambda}^{\textrm{ra}}]_{xx } a_{{\textrm{em } } , y } % -[\hat{\gamma}\hat{\lambda}^{\textrm{ra}}]_{xy } a_{{\textrm{em } } , x } \ } % \\ \notag & = \frac{1}{2}e\tilde{v}_\textrm{f}\nu_e \tau e_y + \frac{e\tilde{v}_\textrm{f}\nu_e \tau}{2l^2 } \partial_t \sum_{\bm{q},\omega } e^{i(\omega t -\bm{q}\cdot\bm{x } ) } \frac { \ell^2 ( q_y^2 a_{{\textrm{em } } , y } + q_y q_x a_{{\textrm{em } } , x})}{q^2 \ell^2 + i\omega \tau } , \\ \label{eq:3 - 22 } % \ref{eq:3 - 22 } s_y & = -\frac{1}{2}e\tilde{v}_\textrm{f}\nu_e \tau e_x - \frac{e\tilde{v}_\textrm{f}\nu_e \tau}{2l^2 } \partial_t \sum_{\bm{q},\omega } e^{i(\omega t -\bm{q}\cdot\bm{x } ) } \frac { \ell^2 ( q_x^2 a_{{\textrm{em } } , x } + q_y q_x a_{{\textrm{em } } , y})}{q^2 \ell^2 + i\omega \tau}.\end{aligned}\ ] ] the second terms of eqs . ( [ eq:3 - 21 ] ) and ( [ eq:3 - 22 ] ) can be described by using the charge density @xmath128 on the surface . here , @xmath129 a_{{\textrm{em } } , u } ( \bm{q},\omega)$ ] , is obtained from the charge - spin response function @xmath130 as @xmath131_{0\nu } a_{{\textrm{em } } , u } \biggr\ } \\ & = \frac { e^2 \tilde{v}_{\textrm{f}}\nu_e \tau}{l^2 } \ell \partial_t \nabla_\nu \sum_{\bm{q},\omega}e^{i(\omega t-\bm{q}\cdot\bm{x } ) } \frac { 1 } { q^2 \ell^2 + i\omega \tau } a_{{\textrm{em } } , \nu } \label{eq:3 - 23 } % \ref{eq:3 - 23 } \\ & = - 2 e^2 \nu_e d \tau \bm{\nabla } \cdot \langle \bm{e } \rangle_{\textrm{d } } , \label{eq:3 - 24 } % \ref{eq:3 - 24}\end{aligned}\ ] ] where @xmath132 is the diffusion constant . here , @xmath133 is defined by the convolution of @xmath134 and the diffusive propagation function @xmath135 as @xmath136.\end{aligned}\ ] ] the diffusive propagation function @xmath135 is also the green s function satisfying the following differential equation @xmath137 equations ( [ eq:3 - 24 ] ) and ( [ eq:3 - 25 ] ) show that due to the impurities the effect of the applied electric field on the surface electrons is not instantaneous but diffusively propagates on the surface of the ti . equation ( [ eq:3 - 25 ] ) gives the definition of such a nonlocal electric field . suppose that the surface of the ti is isotropic , the gradient of the charge density is given by @xmath138 which is related to the spin density [ eqs . ( [ eq:3 - 21 ] ) and ( [ eq:3 - 22 ] ) ] as @xmath139 on the other hand , the spin density on the surface of the ti is related to the charge current via @xmath140 where we have replaced @xmath23 in eq . ( [ eq:2 - 3 ] ) with @xmath141 so as to take into account the effects of impurities . by substituting eq . ( [ eq:3 - 31 ] ) in eq . ( [ eq:3 - 33 ] ) , we obtain @xmath142 we have confirmed that eqs . ( [ eq:3 - 24 ] ) and ( [ eq:3 - 32 ] ) satisfy the charge conservation law : @xmath143 ( see appendix [ sec : f ] for the detailed calculation ) . using the results in sec . [ sec:3 ] , we discuss the property of the charge , spin , charge current , and spin current densities induced by the electric field applied on the disordered surface of the doped ti . we find that as shown in eq . ( [ eq:3 - 24 ] ) the charge density @xmath128 is induced by the divergence of the nonlocal electric field:@xmath144 . therefore , when we apply a uniform electric field , no charge density is induced . from eqs . ( [ eq:3 - 25])-([eq:3 - 27 ] ) we obtain the diffusion equation for the charge transport : @xmath145 which indicates that the divergence of the applied electric field works as a source of the diffusive propagation of the charge density . we find that from the left side of the equation above , @xmath146 , the diffusion constant is @xmath147 , a twice of that on the surface of a metal@xcite . the factor @xmath148 comes from the difference in the self - energy due to impurity scattering : the self - energy on the surface of an isotropic metal is given by @xmath149 , where @xmath150 is the density of states in the metal , whereas that on the surface of the ti is @xmath151 ; the factor @xmath152 , which originates from the linear dispersion of the surface of the ti , leads to the coefficient @xmath147 . here , the diffusive equation of motion qualitatively agrees with the previous works@xcite . we turn to the discussion on the spin density given by eq . ( [ eq:3 - 31 ] ) . we find that the spin density can be divided into the one induced by the local electric field @xmath134 and that by the nonlocal electric field @xmath133 . we define the local spin density @xmath153 as the first term of eq . ( [ eq:3 - 31 ] ) : @xmath154 the local spin density lies in the @xmath155 plane and is perpendicular to the local electric field @xmath134@xcite , which is the consequence of the spin - momentum locking on the surface of the ti . the spin due to the local electric field is a kind of the edelstein effect@xcite . on the other hand , the nonlocal spin density @xmath156 , the second term of eq . ( [ eq:3 - 31 ] ) , is generated from the spatial gradient of the charge density . the direction of @xmath157 is in the @xmath155 plane and perpendicular to the gradient of the charge density @xmath158 . here , @xmath157 is also written in terms of @xmath159 as @xmath160 , % \\ \label{eq : spin - y10 } % \ref{eq : spin - y10 } % s^{(\textrm{nl})}_y % & = - \frac{e\tilde{v}_\textrm{f } \nu_e \ell^2 } { 2 } [ \nabla_x^2 \langle e_x \rangle_\textrm{d}+ \nabla_x \nabla_y \langle e_y % \rangle_\textrm{d } ] . \label{eq : spin - x10 } % \ref{eq : spin - x10 } \bm{s}^{(\textrm{nl } ) } & = -\frac{e\tilde{v}_\textrm{f } \nu_e\tau \ell^2 } { 2 } \left ( \bm{z } \times \bm{\nabla } \right)(\bm{\nabla } \cdot \langle \bm{e } \rangle_{\textrm{d}}),\end{aligned}\ ] ] which means that the spin density is generated by the second spatial derivative of the nonlocal electric field @xmath159 . the nonlocal spin density diffusively propagates through the impurity scatterings on the surface of the ti . the diffusion propagation of the spin is described by @xmath161 we find that the diffusion propagation of the spin is triggered by an inhomogeneous electric field , @xmath162 . ( this property is also predicted in ref . @xmath163 ) . hence , when we apply a uniform electric field on the surface , the nonlocal spin density is not generated . as in the case of charge density , the diffusion constant for the spin density is @xmath147 . we note that both the local and nonlocal spin densities are proportional to @xmath51 . since @xmath51 s on the top and the bottom sides of the ti have opposite signs , the direction of the induced spin on the top surface of the ti is perfectly opposite to that on the bottom surface of the ti , when we apply the same electric field on both the top and bottom side of the ti . the charge current on the surface of the ti is proportional to the spin density as shown in eq . ( [ eq:3 - 33])@xcite . the origin lies on the spin - momentum locking on the surface of the ti . from eq . ( [ eq:3 - 32 ] ) , we find that the charge current @xmath164 can also be divided into the local and nonlocal parts . the first term of eq . ( [ eq:3 - 32 ] ) gives the local charge current @xmath165 . @xmath166 is the electric current directly induced by the applied electric field . we find from the above result that the longitudinal conductivity is given by @xmath167 , which agrees with the existing work@xcite . the second term of eq . ( [ eq:3 - 32 ] ) corresponds to the nonlocal charge current density @xmath168 . @xmath169 is the diffusion current and is generated by the spatial gradient of the charge density on the disordered surface of the doped ti . the charge current can be rewritten by using the nonlocal electric field @xmath170 as @xmath171 since both the local and nonlocal charge currents are proportional to @xmath172 , their directions are the same for top and bottom surface of the ti . next , we calculate the spin current due to the applied electric field on the disordered surface of the doped ti . the spin current @xmath173 is defined by @xmath174 where the subscript and superscript of @xmath173 denote the direction of the flow and spin , respectively , and @xmath175 represents the spin relaxation . using the hamiltonian in eq . ( [ eq:2 - 1 ] ) and the heisenberg equation for @xmath176 , we obtain @xmath177 note that the direction of the flow and spin is perpendicular each to other . this is the consequence of the spin - momentum locking on the surface of the ti . we also note that the spin current is proportional to the charge density , and from eq . ( [ eq:3 - 24 ] ) , proportional to the divergence of the nonlocal electric field : @xmath178 hence , when we apply a spatially uniform electric field on the surface , no spin current is induced . besides , we find that the spin current is an odd - function of @xmath51 , which means the spin current depends on the chirality on the surface of the ti . namely , the relative direction between flow and spin of @xmath173 on the top side of the ti is opposite to that on the bottom side of the ti . in the conventional spin - orbit coupled systems , the spin current is generated by an applied electric field@xcite , which called the spin hall effect . besides , the generated spin current can be converted into the charge current via the spin - orbit interaction.@xcite these effect can be understand from the coupling between the spin current and the charge current : @xmath179@xcite . on the surface of the ti , on the other hand , we find from eqs . ( [ eq:4 - 3 - 2 ] ) and ( [ eq:4 - 4 - 3 ] ) that the nonlocal charge current is proportional to the gradient of the spin current@xcite : @xmath180 again , this is the consequence of the spin - momentum locking and eq . ( [ eq:4 - 4 - 4 ] ) generally holds for the system of electrons on a surface of tis@xcite . this property in the ti is distinct from that in a conventional metal . the direction of the charge current is parallel to the spatial gradient of @xmath173 . this relation is plausible due to the following reasons . first , the charge density @xmath128 is proportional to the spin current . second , a diffusive particle current generally proportional to a spatial gradient of particles . we note that there is no relation between the spin current and the local charge current @xmath166 . finally , we comment on the property of the spin relaxation torque . the relaxation torque @xmath175 defined in eq . ( [ eq:4 - 4 - 1 ] ) is obtained within the linear response to the electric field as @xmath181 + o [ ( \bm{z } \times \bm{\nabla})_\alpha ( \bm{\nabla } \cdot \langle \dot{\bm{e } } \rangle_{\textrm{d } } ) ] .\end{aligned}\ ] ] here , @xmath182 can be divided into the local and nonlocal terms , which correspond to the first and second terms , respectively , in the first square bracket in the most right - hand side of eq . ( [ eq:4 - 4 - 5 ] ) . the local one is given by the time derivative of the applied electric field , and its direction is perpendicular to both @xmath183 and @xmath184 . the nonlocal one is proportional to the second derivative of the nonlocal electric field @xmath185 . these above results and properties are the same as the spin density on the surface of the ti with magnetism@xcite . using the results obtained in sec . [ sec:4 ] , we discuss the properties of the spin and charge densities due to the electric field of a twisted light beam with various orbital angular momentum . first , we explain the property of the electric field of the twisted light beam with the laguerre - gaussian modes@xcite . we assume that the twisted light beam propagates along the @xmath0 axis and the electric field of the twisted light beam lies in the @xmath155 plane at the top surface of the ti ( @xmath1 ) . the schematic of the system is illustrated in fig . [ fig : fig1 ] . the twisted light beam satisfies the wave equation , @xmath186 , where @xmath187 is the velocity of light in a vacuum . then , the electric field @xmath188 on the surface is written by@xcite @xmath189 where @xmath190 is the distance from the center of the light on the top surface ( @xmath1 ) and @xmath191 is the azimuthal angle . the helicity @xmath192 denotes the right - hand ( left - hand ) circularly polarized light and corresponds to the spin angular momentum of light @xmath193 . the orbital angular momentum of light , @xmath194 , determines the whirling pattern of the electric field on the plane at @xmath1 , which can be manipulated in experiments@xcite . the phase @xmath195 depends on @xmath196 and the distance from the light source , @xmath0 , and is given by@xcite @xmath197 } - \frac{q_z r^2 } { 2r(z_0)},\end{aligned}\ ] ] where the first term denotes guoy phase , @xmath198 $ ] is the radius of the beam curvature with taking the origin of the @xmath0 axis at the beam waist , @xmath199 is the rayleigh range , @xmath200 is the waist size of the @xmath201 mode , @xmath202 is the wavelength , and @xmath6 is the wave vector of the light . the integer @xmath203 is one of the indices that specify the laguerre - gaussian mode and denotes the number of oscillations of the electric field @xmath204 in the radial direction . here , @xmath205 is given by @xmath206 } } ( \sqrt{2}u)^{\|m^z_{\textrm{l}}\| } l_{\|m^z_{\textrm{l}}\|}^{p } ( 2u^2 ) \exp { ( - u^2)},\end{aligned}\ ] ] where @xmath207^{\frac{1}{2 } } ] $ ] , @xmath208 is a constant , and @xmath209 is the laguerre polynomials defined by @xmath210 in the following discussion , we consider only the @xmath211 modes . we also assume that the twisted light beam is focused at the surface of the ti , i.e. , @xmath212 . then , the phase @xmath213 becomes zero , and the magnitude @xmath204 defined in eq . ( [ eq : mathcal(e ) ] ) reduces to @xmath214 figures [ fig : fig2](a)-[fig : fig2](d ) show the snapshots of the electric field for @xmath215 and @xmath216 and @xmath217 . in both cases of @xmath218 and @xmath219 , the amplitude of the electric field exponentially decays with @xmath220 . in addition to this , for the cases of nonzero @xmath15 , the magnitude of the electric field vanishes at @xmath9 because of the phase singularity . this is a characteristic property of the twisted light beam . besides , the direction of the electric field depends on the polar angle around the center of the incident light : while the direction of the electric field is uniform for @xmath221 [ fig . [ fig : fig2](a ) ] , the direction of the electric field at @xmath222 rotates by @xmath223 in the counter - clockwise direction as one goes around the beam center from @xmath224 to @xmath223 [ fig . [ fig : fig2](b ) ] . similarly for the cases of @xmath225 and @xmath226 , the direction of the electric field changes by @xmath227 and @xmath228 , respectively [ figs . [ fig : fig2](c ) and [ fig : fig2](d ) ] . note that the configurations shown in figs . 3(a)3(d ) are snapshots and they evolve in time depending on @xmath14 : @xmath215 means that the electric field at a fixed point rotates in the clockwise direction as time evolves [ fig . [ fig : fig2](e ) ] . the topological properties of the twisted light beam discussed above can be understood by introducing the winding number of the electric field . in general , the winding number of a 2d vector field @xmath229 on a closed loop @xmath230 is defined by @xmath231 \equiv \frac{1}{2\pi}\oint_{c } \delta_{ij}\epsilon^{\mu\nu}dx_i \frac{n_\mu}{\|\bm{n}\| } \frac{\partial}{\partial x_j } \left(\frac{n_\nu}{\|\bm{n}\|}\right),\end{aligned}\ ] ] where @xmath232 is the 2d levi - civita symbol , and @xmath233 is supposed to be nonzero on @xmath230 . the winding number corresponds to the number of times the 2d unit vector @xmath234 rotates about the @xmath0 axis as one traces the contour @xmath230 . for the electric field given by eq . ( [ eq : ef - of - opticalvotex ] ) , the winding number @xmath235 is defined on a contour that encloses @xmath9 . substituting eq . ( [ eq : ef - of - opticalvotex ] ) in eq . ( [ eq : winding - number ] ) , we obtain @xmath236 . for example , for the electric field with @xmath237 , and @xmath226 , we have @xmath238 , and @xmath217 , respectively , which are consistent with the configurations shown in fig . [ fig : fig2 ] . the result @xmath239 is also consistent with the fact that the direction of the electric field with @xmath240 is spatially uniform even for @xmath219 and that the whirling direction for @xmath241 is opposite to that for @xmath215 . , @xmath242 , @xmath243 , and @xmath244 , where @xmath14 and @xmath15 denote the spin and orbital angular momentum of light , respectively . shown are density plots of the magnitude of the electric field , and the black arrows show the direction of the electric field . @xmath200 is the waist size for the @xmath245 mode . ( e ) time evolution of the electric field with @xmath246 ( top ) and @xmath226 ( bottom ) , where @xmath247 with @xmath7 being the angular frequency of the light . ] we consider the charge density due to the electric field of the twisted light beam on the disordered surface of the ti . the setup we consider is schematically described in fig . [ fig : fig1 ] . the induced charge density @xmath128 is given by eq . ( [ eq:3 - 24 ] ) , which can be rewritten as @xmath248 by using eq . ( [ eq : ef - of - opticalvotex ] ) , @xmath249 and @xmath250 the divergence of the electric field for @xmath251 is given by @xmath252}. \label{eq : nabla - e}%\ref{eq : nabla - e}\end{aligned}\ ] ] then , @xmath253 becomes @xmath254 where @xmath255 denotes the total angular momentum of light . we further simplify eqs . ( [ eq : rho_e_laguerre ] ) and ( [ eq : rho_e ] ) . first of all , these equations are valid only for @xmath256 . this condition means that the period of the oscillation of @xmath253 , which is the same as that of the electric field , @xmath247 , is much slower than the electron relaxation time @xmath48 . on the other hand , the length scale of the spatial variation of @xmath253 is in the order of the beam waist @xmath200 [ see eq . ( [ eq : mathcal(e ) ] ) and fig . [ fig : fig2 ] ] , which is comparable to the wavelength @xmath202 of the light . since @xmath202 satisfies @xmath257 , the spatial variation of @xmath253 is much slower than the mean - free path @xmath258 , where @xmath187 is the speed of light and we have used @xmath259 for realistic tis@xcite . then , since @xmath48 and @xmath258 determines the decay time and decay length of the diffusion propagator @xmath260 , respectively , @xmath261 in the integrand of eq . ( [ eq : rho_e_laguerre ] ) can be approximated as @xmath262 , and the convolution can be approximately described by @xmath263 where @xmath264 is a constant coefficient and is estimated by @xmath265 , @xmath266 , @xmath243 , and @xmath267 . the dashed lines in ( a ) , ( b ) , and ( c ) indicate axes of the inversion symmetry . ( e ) time evolution of the charge density with @xmath246 ( top ) and @xmath226 ( bottom ) . in all figures , we use @xmath268 mm , @xmath269 mev , @xmath270 m / s , and @xmath271 s. ] figure [ fig : fig3 ] represents the nonlocal charge density due to the twisted light beam , @xmath128 , for a @xmath272 and @xmath273 and @xmath217 . we find that the distribution of @xmath128 depends on the @xmath0-component of the total angular momentum of light , @xmath274 . for @xmath275 [ fig . [ fig : fig3](b ) ] , the charge density is isotropically induced from the center and the sign of the induced charge changes at @xmath276 . on the other hand , the charge density for nonzero @xmath277 distributes anisotropically . the symmetry of the distribution of @xmath128 with @xmath278 [ figs . [ fig : fig3](a ) and [ fig : fig3](c ) ] and @xmath279 [ fig . [ fig : fig3](d ) ] are the same as that for the electron wave functions with the @xmath280 and @xmath281 orbital , respectively . the dashed lines in figs . [ fig : fig3](a ) , [ fig : fig3](c ) and [ fig : fig3](d ) indicate the axes of the inversion symmetry of the charge density . as time evolves , the distribution of the induced charge density rotates around the beam center ( @xmath282 ) or oscillates around the zero value @xmath283 with the frequency of light @xmath7 [ see fig . [ fig : fig3](e ) ] . the time evolution of the charge density also depends on the total angular momentum of light : when the sign of the total angular momentum is minus ( plus ) , the distribution of the charge density rotates in the clockwise ( counterclockwise ) direction around the phase singularity during the irradiation . when we turn off the incident light , the charge density diffusively propagates on the disordered surface of the ti with obeying the diffusive equation of motion represented in eq . ( [ eq:3 - 3 - 1 ] ) . eventually , the induced charge vanishes . we turn to discuss the spin density induced by the electric field of the twisted light beam in the same setup as that considered in sec . [ sec:5 - 2 ] . as discussed in sec . [ sec:4 - 2 ] , the induced spin density can be divided into the local and nonlocal ones as @xmath284 . the local spin density @xmath153 is described in eq . ( [ eq : local - spin ] ) and its snapshots for several @xmath15 are shown in figs . [ fig : fig4](a1)[fig : fig4](d1 ) . the direction of @xmath153 is perfectly perpendicular to the electric field . we find that the dynamical vortex - like spin structure is generated by the twisted light beam and the winding number of the local spin density , which is defined by eq . ( [ eq : winding - number ] ) with @xmath285 , is identical to that of the electric field , @xmath286=\omega_v[\bm{e}]=-\sigma_{\textrm{l}}^z m_{\textrm{l}}^z\end{aligned}\ ] ] [ see figs . [ fig : fig4](a1)[fig : fig4](d1 ) and figs . [ fig : fig2](a)[fig : fig2](d ) ] . on the other hand , the nonlocal spin density @xmath157 is proportional to the spatial gradient of the charge density [ see the second term of eq . ( [ eq:3 - 31 ] ) ] , and can be estimated by using eq . ( [ eq:5 - 2 - 3 ] ) as @xmath287 . the snapshots of @xmath157 are shown in figs . [ fig : fig4](a2)[fig : fig4](d2 ) . we find that dynamical vortex - like spin structures appear and the spin density becomes zero at the center of the vortex . here , we note that @xmath157 and @xmath288 share the same winding number as they are perfectly perpendicular to each other . since the winding number @xmath289 ) is @xmath290 around the maxima and minima ( the saddle points ) of @xmath128 , the centers of the spin vortices locate at the extrema ( minima , maxima , and saddle points ) of @xmath128 , and therefore , they align on the symmetry axis of the distribution of @xmath128 [ the dotted lines in fig . [ fig : fig3 ] ] . for the cases shown in figs . [ fig : fig4](a2)[fig : fig4](d2 ) , all spin vortices have the winding number @xmath291 except for the one at the center of fig . [ fig : fig4](d2 ) , which corresponds to the saddle point of @xmath128 and has the winding number @xmath217 . since @xmath157 is related to @xmath292 rather than @xmath134 , the configuration of the spin vortices depends on the total angular momentum @xmath293 . note however that with the parameters for a realistic system , @xmath294 is in the order of @xmath295 and @xmath296 is negligibly small as compared with @xmath297 . when we turn off the beam , @xmath296 becomes prominent and diffusively propagates . we expect that the photo - induced spin texture can be observed by pump prove technique with the twisted light beam and circularly polarized light beam@xcite . the profile of the charge current is similar to that of the spin because of the spin momentum locking on the surface of the ti . in fig . [ fig : fig5 ] , we show the snapshots of the charge current for various angular momentum of light , where left ( right ) four panels depict the local ( nonlocal ) components . reflecting the relation @xmath298 , the magnitude of @xmath299 has the same profile as that of @xmath300 , while the direction of @xmath299 is obtained by rotating @xmath301 by @xmath302 about the @xmath0 axis . as in the case of the spin density , the local ( nonlocal ) part of the charge current is related to @xmath303 ( @xmath304 ) and hence its configuration is mainly determined by @xmath305 ( @xmath293 ) . figure [ fig : fig6 ] shows the light - induced spin current . as one can see from eq . ( 55 ) , the magnitude of the spin current is proportional to @xmath306 and the direction of the spin and its current perfectly perpendicular to each other . these properties also come from the spin - momentum locking on the surface of the ti . we find that the spatial profile of the spin current is different from that of the ( spin - polarized ) charge current , which are shown in figs . [ fig : fig6 ] and [ fig : fig5 ] respectively . in fact , they are related to each other via eq . ( [ eq:4 - 4 - 4 ] ) and only the nonlocal charge current couples to the spin current . we study the charge density , the spin density , the charge current density , and the spin current density induced by a twisted light beam shined on a disordered surface of a doped ti by using the keldysh - green s function method . we have discussed the responses of charge and spin to the space - time dependent electric field . the obtained results are summarized in tab . i. the effect of the electric field on the electric charge is twofold . first , it induces a charge current along the direction of the electric field . second , the inhomogeneity of the electric field causes a gradient of the charge density , which then leads to a diffusive charge current . we call the former the local charge current and the latter the nonlocal charge current , based on whether the charge current depends only on the local electric field or is affected by the nonlocal one . since the spin and momentum of electrons on a surface of a ti are locked to be perpendicular to each other , the emergence of the charge current implies that the spin density is induced in the perpendicular direction to the charge current . our calculation based on the linear response theory gives the analytic description for the local and nonlocal spin densities as well as the local and nonlocal charge current densities . we also find that the induced charge density also gives rise to a spin current , which is related to the nonlocal part of the charge current via eq . ( [ eq:4 - 4 - 4 ] ) . by taking into account the spatial and temporal configuration of the electric field associated with a twisted light beam , we have shown that various types of spin vortices arise . since the local spin density is perpendicular to the electric field , their winding numbers are identical and determined by the product of the spin and orbital angular momentum of the twisted light beam . in this paper , we have assumed that the time and length scales of the diffusion of electrons is much faster than those of incident light . in such a situation , we can approximate the nonlocal electric field with the bare electric field , and the nonlocal quantities are described using the divergence of the electric field . thus , the configurations of the nonlocal densities and currents are determined by the total angular momentum of the twisted light beam . [ cols="<,^,^,^,^,^",options="header " , ] the authors would like to thank a. dutt for valuable discussions . this work was supported by grant - in - aid for young scientists ( b ) ( no . 15k17726 ) , by grants - in - aid for scientific research on innovative areas `` topological materials science '' ( no . 15h05853 and no . 15h05855 ) from the ministry of education , culture , sports , science , and technology , japan ( mext ) , by grant - in - aid for challenging exploratory research ( no . 15k13498 ) , and by the core research for evolutional science and technology ( crest ) of the japan science . k.t . acknowledges support from a grant - in - aid for japan society for the promotion of science ( jsps ) fellows . we estimate @xmath307 in eq . ( [ eq:3 - 14 ] ) by expanding with respect to @xmath81 and @xmath7 and by using the green s function @xmath308 + \frac{1}{8 } \sum_{\xi,\xi'=x , y } q_\xi q_{\xi ' } \frac{\partial^2 \hat{g}^\textrm{r}_{\bm{k } } } { \partial k_\xi \partial k_{\xi ' } } + o(q^3 , q\omega , \omega^2 ) , \end{aligned}\ ] ] where we use the short hand notation @xmath309 . @xmath307 is decomposed into four terms as @xmath310 in appendix [ sec : e ] , we list useful formulas for the integral of the green s functions , which are used in the following calculations . first , to calculate @xmath312 in eq . ( [ eq : coefficient pi-1 ] ) , we divide @xmath313 into the even and odd functions of @xmath40 : @xmath314 where @xmath315 , @xmath316^$ ] , @xmath317 , @xmath318 , @xmath319 , and @xmath320 are defined by @xmath321 @xmath315 and @xmath322 are the even functions of @xmath40 . @xmath323 is represented by @xmath324 \\ \label{eq : app-12 } % \ref{eq : app-12 } & \hspace{10 mm } + \hbar \tilde{v}_{\textrm{f } } \sum_{\ell = x , y}\left [ h k_\ell \hat{\sigma}_\mu ( \bm{z } \times \hat{\bm{\sigma}})_\ell + h^ k_\ell ( \bm{z } \times \hat{\bm{\sigma}})_\ell \hat{\sigma}_\mu \right].\end{aligned}\ ] ] the first and second terms are corresponding to the even and odd functions of @xmath40 , respectively . in the following calculation , we simply assume that the surface of the ti is isotropic as a function of @xmath40 : @xmath325 . by taking an average over the direction of @xmath40 , @xmath326 , in eq . ( [ eq : app-12 ] ) and using @xmath327 and @xmath328 , we obtain @xmath329 the second term in the above equation becomes @xmath330 thus , we obtain @xmath331 as @xmath332 the integral of eq . ( [ eq : app-15 ] ) is obtained by @xmath333 [ ( h^)^2 - \hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 ] } % & = \frac{1}{2\pi } \int_0^\infty \frac{kdk } { [ h^2 - \hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 ] [ ( h^)^2 - \hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 ] } % \\ = \frac{\nu_e } { 2\epsilon_{\textrm{f } } } \int_0^\infty \frac{dx } { [ h^2 -x ] [ ( h^)^2 -x ] } \\ = & \frac{\nu_e } { 2\epsilon_{\textrm{f } } [ h^2 - ( h^)^2 ] } \int^\infty_0 dx \biggl [ \frac{1 } { x -h^2 } - \frac{1 } { x - ( h^)^2 } \biggr],\end{aligned}\ ] ] where @xmath334 is the density of states on the surface . here , the above integral is given by @xmath335 = \log { \biggl\| \frac{x - h^2}{x-(h^)^2 } \biggr\|_{x\to0}^{x\to\infty } } = i \biggl [ \arg{(x -h^2 ) } - \arg{(x -(h^)^2 ) } \biggr]_{x\to0}^{x\to\infty}.\end{aligned}\ ] ] in the above equation , we have used @xmath336 , where @xmath337 , @xmath338 . here , @xmath339 is @xmath340 thus , @xmath341 , @xmath342 } \|_{x\to0}^{x\to\infty } = - \pi + o(\hbar/\epsilon_{\textrm{f}}\tau ) ) $ ] , and @xmath343 are satisfied . we have @xmath344 and @xmath345 eq . ( [ eq : app-16 ] ) can be estimated around the fermi energy @xmath346 as @xmath347 to calculate @xmath349 in eq . ( [ eq : coefficient pi-2 ] ) , we divide @xmath350 into the even and odd functions of @xmath40 . here , @xmath351 and @xmath352 becomes @xmath353 + 2 h^ \hbar \tilde{v}_{\textrm{f } } k_\ell ( \bm{z } \times \hat{\bm{\sigma}})_\ell , \\ \notag % % % % % % % \hat{\mathcal{q } } \hat{\sigma}_\mu(\hat{\mathcal{q}}^\dag)^2 = & \biggl [ h \{(h^)^2 + \hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 \ } \hat{\sigma}_\mu + 2 h^ \hbar^2 \tilde{v}_{\textrm{f}}^2 k_\ell k_{\ell ' } ( \bm{z}\times \hat{\bm{\sigma}})_\ell \hat{\sigma}_\mu ( \bm{z}\times \hat{\bm{\sigma}})_{\ell ' } \biggr ] \\ & + \biggl [ 2 \|h\|^2 \hbar \tilde{v}_{\textrm{f } } k_\ell ( \bm{z}\times \hat{\bm{\sigma}})_\ell \hat{\sigma}_\mu \hat{\sigma}_\mu + \ { ( h^)^2 + \hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 \ } \hbar \tilde{v}_{\textrm{f } } k_\ell ( \bm{z}\times \hat{\bm{\sigma}})_\ell \hat{\sigma}_\mu \biggr].\end{aligned}\ ] ] the first and second terms of the above equations are the even and odd functions of @xmath40 . then , we have @xmath354 from which @xmath355 $ ] is obtained by @xmath356 \hat{\sigma}_\mu -\textrm{h.c . } \biggr\ } \\ \notag & = -\frac{\hbar}{2 } \biggl [ \frac{\nu_e}{8\eta^2 } \left ( i\pi - \frac{2\eta^2}{\epsilon_{\textrm{f}}^2 } \right ) \left(1+i\frac{\eta}{\epsilon_{\textrm{f } } } \right ) -\textrm{c.c . } \biggr ] \hat{\sigma}_\mu + o\left(\frac{\hbar}{\epsilon_{\textrm{f}}\tau}\right ) \\ \label{eq : coefficient c-2 no2 } % \ref{eq : coefficient c-2 no2 } & = \frac{-i\hbar \pi \nu_e}{8\eta^2 } \hat{\sigma}_\mu + o\left(\frac{\hbar}{\epsilon_{\textrm{f}}\tau}\right ) , \\\notag \hat{i}^{(1)}_{z } & = -\frac{\hbar}{2 } \biggl\ { \sum_{k } d^\textrm{r } ( d^\textrm{a})^2 \bigl [ h(h^)^2 + ( h - 2h^)\hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 \bigr ] \hat{\sigma}_\mu -\textrm{h.c . } \biggr\ } \\ \notag & = -\frac{\hbar}{2}\left[\frac{-i\pi\nu_e}{4\epsilon_\textrm{f}^2}\left(1+i\frac{\eta}{\epsilon_{\textrm{f } } } \right ) - \textrm{c.c.}\right]\hat{\sigma}_z + o\left(\frac{\hbar}{\epsilon_{\textrm{f}}\tau}\right ) \\ \label{eq : coefficient c-2 no3 } % \ref{eq : coefficient c-2 no3 } & = \frac{i\hbar \pi \nu_e}{4\epsilon_\textrm{f}^2 } \hat{\sigma}_\mu + o\left(\frac{\hbar}{\epsilon_{\textrm{f}}\tau}\right).\end{aligned}\ ] ] @xmath358 is given by @xmath359 + 2\hbar^2 \tilde{v}_{\textrm{f}}^2 k_\xi [ h + \hbar \tilde{v}_{\textrm{f } } \bm{k}\cdot ( \bm{z } \times \bm{\sigma } ) ] } { [ h^2 -\hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 ] ^2 } & = \hbar \tilde{v}_{\textrm{f } } ( \bm{z}\times \hat{\bm{\sigma}})_\xi d^\textrm{r } + 2\hbar^2\tilde{v}_{\textrm{f}}^2 k_\xi \hat{\mathcal{q } } ( d^\textrm{r})^2 . \label{eq : differential1 gr-0 } % \ref{eq : differential1 gr-0}\end{aligned}\ ] ] by using eq . ( [ eq : differential1 gr-0 ] ) , @xmath360 becomes @xmath361 \hat{\sigma}_\mu d^\textrm{a } \hat{\mathcal{q}}^\dag \\ = & \hbar \tilde{v}_{\textrm{f } } ( \bm{z}\times \hat{\bm{\sigma}})_\xi \hat{\sigma}_\mu \langle \hat{\mathcal{q}}^\dag\rangle_k \|d^\textrm{r}\|^2 + 2\hbar^2\tilde{v}_{\textrm{f}}^2 \langle k_\xi \hat{\mathcal{q } } \hat{\sigma}_\mu \hat{\mathcal{q}}^\dag \rangle_k ( d^\textrm{r})^2 d^\textrm{a}. \end{aligned}\ ] ] here , @xmath362 and @xmath363 are given by @xmath364 @xmath365 in eq . ( [ eq : coefficient pi-3 ] ) is given by @xmath366.\end{aligned}\ ] ] here , @xmath367 can be transformed as @xmath368 as a result , @xmath365 is obtained by @xmath369 + o\left ( \frac{\hbar}{\epsilon_\textrm{f } \tau}\right ) \\ \label{eq : a29}%\ref{eq : a29 } & = \frac { i\pi \nu_e } { 8\eta^2 } \hbar \tilde{v}_{\textrm{f } } \epsilon_{\mu\xi z } + o\left ( \frac{\hbar}{\epsilon_\textrm{f } \tau}\right ) , \\ \notag \hat{i}^{(2)}_{z \xi } & = \frac{\hbar \tilde{v}_{\textrm{f } } i\pi\nu_e } { 8 \eta^2 } \epsilon_{\xi z u } \biggl [ \left ( 1 + 2\frac{\eta^2 } { \epsilon_{\textrm{f}}^2 } \right ) \delta_{z u } + \frac{\eta}{\epsilon_{\textrm{f } } } \epsilon_{z u\nu } \hat{\sigma}_\nu \biggr ] + o\left ( \frac{\hbar}{\epsilon_\textrm{f } \tau}\right ) \\ \label{eq : a29-z}%\ref{eq : a29-z } & = \frac { i\pi \nu_e } { 8\epsilon_\textrm{f}\eta } \hbar \tilde{\xi}_{\textrm{f } } + o\left ( \frac{\hbar}{\epsilon_\textrm{f } \tau}\right ) .\end{aligned}\ ] ] @xmath371 in eq . ( [ eq : coefficient pi-4 ] ) is represented by using the partial integral as @xmath372.\end{aligned}\ ] ] in order to consider @xmath373 , we use the following equations : @xmath374 % \\ & = 2\hbar^2 \tilde{v}_{\textrm{f}}^2 \biggl\ { \delta_{\xi\xi ' } \hat{\mathcal{q } } ( d^\textrm{r})^2 + 4\hbar^2 \tilde{v}_{\textrm{f}}^2 k_\xi k_{\xi ' } \hat{\mathcal{q } } ( d^\textrm{r})^3 + \hbar \tilde{v}_{\textrm{f } } [ k_\xi ( \bm{z } \times \hat{\bm{\sigma}})_{\xi ' } + k_{\xi ' } ( \bm{z } \times \hat{\bm{\sigma}})_{\xi } ] ( d^\textrm{r})^2 \biggr\ } , \\ \notag \frac{\partial^2 \hat{g}^\textrm{r}_{\bm{k}}}{\partial k_\xi \partial k_{\xi ' } } \hat{\sigma}_\mu \hat{g}^\textrm{a}_{\bm{k } } & = 2\hbar^2 \tilde{v}_{\textrm{f}}^2 \biggl\ { \delta_{\xi\xi ' } \hat{\mathcal{q } } ( d^\textrm{r})^2\hat{\sigma}_\mu \hat{\mathcal{q}}^\dag d^\textrm{a } + 4\hbar^2 \tilde{v}_{\textrm{f}}^2 k_\xi k_{\xi ' } \hat{\mathcal{q } } ( d^\textrm{r})^3\hat{\sigma}_\mu \hat{\mathcal{q}}^\dag d^\textrm{a } \\&\hspace{20 mm } + \hbar \tilde{v}_{\textrm{f } } [ k_\xi ( \bm{z } \times \hat{\bm{\sigma}})_{\xi ' } + k_{\xi ' } ( \bm{z } \times \hat{\bm{\sigma}})_{\xi } ] ( d^\textrm{r})^2\hat{\sigma}_\mu \hat{\mathcal{q}}^\dag d^\textrm{a } \biggr\}.\end{aligned}\ ] ] the average of @xmath40 in @xmath323 , @xmath375 and @xmath376 become @xmath377\end{aligned}\ ] ] here , we have used @xmath378 . then , @xmath379 is given by @xmath380 \\ \notag & + 2 \|h\|^2 \hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 \delta_{\xi \xi ' } \hat{\sigma}_\mu ( d^\textrm{r})^3 d^\textrm{a } \\ \notag & + \frac{1}{2 } \hbar^4 \tilde{v}_{\textrm{f}}^4 k^4 ( d^\textrm{r})^3 d^\textrm{a } [ ( \bm{z } \times \hat{\bm{\sigma}})_\xi \hat{\sigma}_\mu ( \bm{z } \times \hat{\bm{\sigma}})_{\xi ' } + ( \xi \leftrightarrow \xi ' ) ] \biggr ] \\ & + \textrm{h.c . } \biggl\ } % second \\ \notag = \frac{\hbar^2 \tilde{v}_{\textrm{f}}^2 } { 2 } \sum_{\bm{k } } & \biggl [ \|h\|^2 \delta_{\xi\xi ' } \hat{\sigma}_\mu \biggl ( \|d^\textrm{r}\|^2 ( d^\textrm{r } + d^\textrm{a } ) + 2 \hbar^2\tilde{v}_{\textrm{f}}^2 k^2 \|d^\textrm{r}\|^2 \ { ( d^\textrm{r})^2 + ( d^\textrm{a})^2 \ } \biggr ) \\ \notag & + \biggl [ \hbar^2\tilde{v}_{\textrm{f}}^2 k^2 \|d^\textrm{r}\|^2 ( d^\textrm{r } + d^\textrm{a } ) + \hbar^4 \tilde{v}_{\textrm{f}}^4 k^4 \|d^\textrm{r}\|^2 \ { ( d^\textrm{r})^2 + ( d^\textrm{a})^2 \ } \biggr ] \\ \label{eq : differential-4 grgrgrga 4 } % \ref{eq : differential-4 grgrgrga 4 } & \ \times \bigl [ \epsilon_{\xi z \mu } ( \bm{z}\times \hat{\bm{\sigma}})_{\xi ' } + \epsilon_{\xi ' z \mu } ( \bm{z}\times \hat{\bm{\sigma}})_\xi - \delta_{\xi \xi ' } \hat{\sigma}_\mu \bigr ] \biggr ] .\end{aligned}\ ] ] here , we have used @xmath381.\end{aligned}\ ] ] @xmath382 is given by using equations in appendix [ sec : e ] as @xmath383 \hat{\sigma}_\nu + o\left ( \frac{\hbar}{\epsilon_\textrm{f } \tau } \right ) , \\\label{eq : a38-z}%\ref{eq : a38-z } \hat{i}^{(3)}_{z \xi\xi ' } = & \hbar^2 \tilde{v}_{\textrm{f}}^2 \frac{\pi \nu_e } { 8\epsilon_\textrm{f}^2\eta } + o\left ( \frac{\hbar}{\epsilon_\textrm{f } \tau } \right).\end{aligned}\ ] ] using @xmath384 = q_\alpha q_\beta \bigl [ 3 \delta_{\alpha \beta } \delta_{\nu \mu } - 2 \delta_{\alpha \nu } \delta_{\beta \mu } \bigr]$ ] , we obtain @xmath385 \hat{\sigma}_\nu.\end{aligned}\ ] ] thus , @xmath386 is obtained by @xmath387 \sigma_\nu % \\ \hat{i}_{\mu = x , y } & \simeq \frac{\pi \nu_e } { 4\eta } \biggl [ ( 1 - i \omega \tau - \frac{3}{2}d\tau q^2 ) \hat{\sigma}_\mu + i q_\alpha \ell \epsilon_{\mu \alpha z } + d\tau q_{\mu } q_\nu \hat{\sigma}_\nu \biggr ] , \\ \hat{i}_{\mu = z } & \simeq o\left ( \frac{\hbar}{\epsilon_\textrm{f } \tau } \right),\end{aligned}\ ] ] where @xmath388 and @xmath104 are the diffusion constant and the mean free path of the surface electrons , respectively . we will calculate @xmath389 using the same formalism in the appendix [ sec : a ] . here , @xmath390 can be expanded with respect to @xmath81 and @xmath7 within @xmath391 and @xmath392 as @xmath393 we will calculate @xmath395 in eq . ( [ eq : coefficient pi-1 ] ) . by using @xmath396 and @xmath397 $ ] , @xmath395 becomes @xmath398.\end{aligned}\ ] ] the above equation can be estimated around the fermi energy , @xmath346 as @xmath399 \sum_{k } \|d^\textrm{r}\|^2 = \frac{\pi\nu_e } { 2\eta } .\end{aligned}\ ] ] @xmath400 in eq . ( [ eq : coefficient pi-2 ] ) is calculated by using @xmath401 @xmath402 , and @xmath403 as @xmath404 \\ & = -\frac{i\pi\nu_e } { 4\eta^2}\hbar + o\left(\frac{\hbar}{\epsilon_{\textrm{f}}\tau}\right ) . % \\ % \omega \hat{i}^{(1)}_{0 } & = \frac{\pi\nu_e}{\eta}(-i\omega \tau)\end{aligned}\ ] ] @xmath405 in eq . ( [ eq : coefficient pi-3 ] ) is obtained by @xmath406\|d^\textrm{r}\|^2 % = & \frac{\hbar \tilde{v}_{\textrm{f } } \pi\nu_e } { 2\eta } \epsilon_{\xi z \ell } \hat{\sigma}_\ell \biggl [ \frac{i\eta}{2\epsilon_{\textrm{f}}^2 } + \frac{i } { 4\eta } \biggr ] % \\ \simeq \frac{\pi\nu_e } { 2\eta } \frac{i \tilde{v}_{\textrm{f } } \tau}{2 } \epsilon_{\xi z \ell } \hat{\sigma}_\ell . \end{aligned}\ ] ] here , we have used the following equations @xmath407\|d^\textrm{r}\|^2 % \\ & = \hbar \tilde{v}_{\textrm{f } } \epsilon_{\xi z \ell } \hat{\sigma}_\ell \bigl [ h^ + 2\hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 \epsilon_{\textrm{f } } d^\textrm{r } \bigr]\|d^\textrm{r}\|^2 . % \\ % \langle \hat{g}^\textrm{r}_{\bm{k } } \frac{\partial \hat{g}^\textrm{a}_{\bm{k}}}{\partial k_\xi } \rangle_k % & = \hbar \tilde{v}_{\textrm{f } } \epsilon_{\xi z \ell } \sigma_\ell \bigl [ h + 2\hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 \epsilon_{\textrm{f } } d^\textrm{a } \bigr]\|d^\textrm{r}\|^2 . \end{aligned}\ ] ] @xmath409 becomes @xmath410 the average @xmath411 is reduced to be @xmath412 after we integrate the above equation as a function of @xmath40 , we obtain @xmath413 \|d^\textrm{r}\|^4 ] - ( \hbar \tilde{v}_{\textrm{f } } k)^2 \|d^\textrm{r}\|^2 ( d^\textrm{a } + d^\textrm{r } ) \biggr\}q_\alpha q_\beta \\ % & = \delta_{\alpha\beta } \biggl\ { \frac{\pi\nu_e}{4\eta \epsilon_{\textrm{f}}^2 } % + \frac{\pi\nu_e}{8\eta^3 } % + \frac{\pi\nu_e}{4 \eta \epsilon_\textrm{f}^2 } \biggr\ } & = \frac{\pi\nu_e } { 8\eta^3 } q^2 + o\left(\frac{\hbar}{\epsilon_{\textrm{f}}\tau}\right).\end{aligned}\ ] ] here , we have used @xmath414 % \\ & = - q_\alpha q_\beta \delta_{\alpha \beta } ( d^\textrm{a } + d^\textrm{r}),\end{aligned}\ ] ] thus , @xmath415 becomes @xmath416 we estimate @xmath418 by using the same formalism in appendices [ sec : a ] and [ sec : b ] . here , @xmath419 can be expanded as @xmath420 where coefficients in the above equation are given by @xmath421 = \left\ { \begin{array}{ll } \displaystyle-\frac{n_\textrm{i}u_\textrm{i}^2\nu_e}{2\epsilon_f}\hat{\sigma}_\mu & ( \mu = x , y)\\ \displaystyle- \frac{in_\textrm{i}u_\textrm{i}^2\pi\nu_e}{2\epsilon_f}\arg(-\hbar\omega-\epsilon_\textrm{f}+i\eta)\hat{\sigma}_z & ( \mu = z)\\ \displaystyle-\frac{n_\textrm{i}u_\textrm{i}^2\nu_e}{\epsilon_f } + \frac{in_\textrm{i}u_\textrm{i}^2\pi\nu_e}{2\epsilon_f}\arg(-\hbar\omega-\epsilon_\textrm{f}+i\eta ) & ( \mu=0 ) \end{array}\right . , \\ \hat{c}^{\textrm{rr}(1)}_{\mu}\label{eq : crr1 } = & \frac{n_\textrm{i}u_\textrm{i}^2}{2}\sum _ { k}\left [ \hat{g}^\textrm{r}\hat{\sigma}_\mu\frac{\partial \hat{g}^\textrm{r}}{\partial \omega } -\frac{\partial \hat{g}^\textrm{r}}{\partial \omega}\hat{\sigma}_\mu \hat{g}^\textrm{r } \right ] = 0 , \\ \hat{c}^{\textrm{rr}(2)}_{\mu\xi}\label{eq : crr2 } = & \frac{n_\textrm{i}u_\textrm{i}^2}{2}\sum_{k}\left[\hat{g}^\textrm{r}\hat{\sigma}_\mu\frac{\partial \hat{g}^\textrm{r}}{\partial k_\xi } -\frac{\partial \hat{g}^\textrm{r}}{\partial k_\xi}\hat{\sigma}_\mu \hat{g}^\textrm{r } \right ] = \left\ { \begin{array}{ll } \displaystyle\frac{n_\textrm{i}u_\textrm{i}^2\hbar v_f\nu_e}{2\epsilon_f[(\hbar\omega+\epsilon_f)^2+\eta^2]}[i(\hbar\omega+\epsilon_f)+\eta]\delta_{\xi \mu } \hat{\sigma}_z & ( \mu = x , y ) \\ \displaystyle-\frac{n_\textrm{i}u_\textrm{i}^2\hbar v_f\nu_e}{2\epsilon_f[(\hbar\omega+\epsilon_f)^2+\eta^2]}[i(\hbar\omega+\epsilon_f)+\eta]\hat{\sigma}_\xi & ( \mu = z ) \\ 0 & ( \mu=0 ) \end{array}\right . , \\ \notag \hat{c}^{\textrm{rr}(3)}_{\mu\xi\xi ' } = & \frac{n_\textrm{i}u_\textrm{i}^2}{4}\sum_{k}\left[\hat{g}^\textrm{r}\hat{\sigma}_\mu\frac{\partial^2 \hat{g}^\textrm{r}}{\partial k_\xi\partial k_{\xi ' } } + \frac{\partial^2 \hat{g}^\textrm{r } } { \partial k_\xi\partial k_{\xi'}}\hat{\sigma}_\mu \hat{g}^\textrm{r}\right ] \\\label{eq : crr3 } = & \left\ { \begin{array}{ll } \displaystyle\frac{n_\textrm{i}u_\textrm{i}^2\hbar^2v_f^2\nu_e}{12\epsilon_f[(\hbar\omega+\epsilon_f)^2+\eta^2]^2}\left[(\hbar\omega+\epsilon_{\textrm{f}})^2-\eta^2 - 2i\eta(\hbar\omega+\epsilon_{\textrm{f}})\right]&\\ \displaystyle\hspace{20mm}\times[\epsilon_{\xi z \mu}(\bm{z}\times\hat{\bm{\sigma}})_{\xi'}+\epsilon_{\xi ' z \mu}(\bm{z}\times\hat{\bm{\sigma}})_\xi -2\delta_{\xi\xi'}\hat{\sigma}_\mu]&(\mu = x , y)\\ \displaystyle-\frac{n_\textrm{i}u_\textrm{i}^2\hbar^2v_f^2\nu_e}{4\epsilon_f[(\hbar\omega+\epsilon_f)^2+\eta^2]^2}\left[(\hbar\omega+\epsilon_{\textrm{f}})^2-\eta^2 - 2i\eta(\hbar\omega+\epsilon_{\textrm{f}})\right]\delta_{\xi\xi'}\hat{\sigma}_z&(\mu = z)\\ \displaystyle\frac{n_\textrm{i}u_\textrm{i}^2\hbar^2v_f^2\nu_e}{12\epsilon_f[(\hbar\omega+\epsilon_f)^2+\eta^2]^2}\left[(\hbar\omega+\epsilon_{\textrm{f}})^2-\eta^2 - 2i\eta(\hbar\omega+\epsilon_{\textrm{f}})\right]\delta_{\xi\xi'}\hat{\sigma}_0&(\mu=0 ) \end{array}\right . .\end{aligned}\ ] ] from the eqs . ( [ eq : crr0])-([eq : crr3 ] ) and @xmath422 , the elements of @xmath417 is negligibly small as compared with the ones of @xmath423 , since @xmath82 is satisfied . we estimate the response function composed of only the retarded ( advanced ) green s functions @xmath425 ( @xmath426 ) using the same formalism in appendices [ sec : a ] , [ sec : b ] and [ sec : a - c ] . from eqs . ( [ eq:3 - 11 ] ) and ( [ eq:3 - 11 - 2 ] ) , @xmath427 are written as @xmath428 \\&\hspace{20 mm } + \frac{1}{2 } \omega f'_{\omega } \biggl [ \hat{g}^\textrm{r}_{\bm{k}-\frac{\bm{q}}{2},\omega-\frac{\omega}{2 } } \hat{\sigma}_\nu \hat{g}^\textrm{r}_{\bm{k}+\frac{\bm{q}}{2},\omega+\frac{\omega}{2 } } + \left(\hat{g}^\textrm{r}_{\bm{k}+\frac{\bm{q}}{2},\omega+\frac{\omega}{2 } } \hat{\sigma}_\nu \hat{g}^\textrm{r}_{\bm{k}-\frac{\bm{q}}{2},\omega-\frac{\omega}{2}}\right)^\dag \biggr]\biggr\}. \label{eq : pirr+aa}%\ref{eq : pirr+aa}\end{aligned}\ ] ] the magnitude of second term of the above equation is smaller than that of @xmath91 [ see eqs . ( [ eq:3 - 10 ] ) and appendices [ sec : a ] , [ sec : b ] and [ sec : a - c ] ] . the first term is expanded as @xmath429 \\ = & \hat{d}^{(0)}_{\nu}+\omega \hat{d}^{(1)}_{\nu}+\sum_{\xi = x , y}q_\xi \hat{d}^{(2)}_{\nu\xi}+o(q^2 , q\omega , \omega^2),\end{aligned}\ ] ] where @xmath430 , @xmath431 and @xmath432 in the above equation are given by @xmath433 = 0 , \\\label{eq : d1nu}%\ref{eq : d1nu } \hat{d}^{(1)}_{\nu } = & -\frac{1}{2}\sum_{\bm{k } , \omega}\left [ \left(\hat{g}_{\bm{k},\omega}^\textrm{r}\hat{\sigma}_\mu\frac{\partial \hat{g}_{\bm{k},\omega}^\textrm{r}}{\partial \omega } -\frac{\partial \hat{g}_{\bm{k},\omega}^\textrm{r}}{\partial \omega}\hat{\sigma}_\mu \hat{g}_{\bm{k},\omega}^\textrm{r}\right ) + \textrm{h.c . } \right ] = 0 , \\\label{eq : d2nu}%\ref{eq : d2nu } \hat{d}^{(2)}_{\nu\xi } = & -\frac{1}{2}\sum_{\bm{k } , \omega}f_{\omega}\left[\left(\hat{g}_{\bm{k},\omega}^\textrm{r}\hat{\sigma}_\mu\frac{\partial \hat{g}_{\bm{k},\omega}^\textrm{r}}{\partial k_\xi } -\frac{\partial \hat{g}_{\bm{k},\omega}^\textrm{r}}{\partial k_\xi}\hat{\sigma}_\mu \hat{g}_{\bm{k},\omega}^\textrm{r}\right ) + \textrm{h.c . } \right ] = -\frac{\pi \tilde{v}_{\textrm{f}}\nu_{\textrm{e}}}{2\epsilon_{\textrm{f}}}\hat{\sigma}_z + o\left(\frac{\hbar}{\epsilon_{\textrm{f}}\tau}\right).\end{aligned}\ ] ] here , we have used @xmath434 in the above equation , where @xmath435 is a step function . from eqs . ( [ eq:3 - 2 ] ) , ( [ eq:3 - 6 ] ) , ( [ eq:3 - 33 ] ) and ( [ eq : d2nu ] ) , there are spin density and charge current induced by the magnetic field @xmath436 . since we consider only the electric field , we ignore the densities induced by the magnetic field . we find that the order of @xmath437 are @xmath438 and @xmath439 are negligibly small as compared with @xmath91 . we will show the following integrals as a functions of @xmath40 . the integrals are obtained by @xmath440 d^\textrm{r } ( d^\textrm{a})^2 & \simeq \frac{\nu_e}{8\eta^2 \epsilon_{\textrm{f } } } \biggl [ i\pi - \frac{2\eta^2}{\epsilon_{\textrm{f}}^2 } \biggr ] , \\\label{eq : d5}%\ref{eq : d5 } \sum_{k } [ h(h^)^2 + ( h - 2h^)\hbar^2 \tilde{v}_{\textrm{f}}^2 k^2 ] d^\textrm{r } ( d^\textrm{a})^2 & \simeq \frac{-i\pi\nu_e}{4\epsilon_\textrm{f}^2}\left(1+i\frac{\eta}{\epsilon_{\textrm{f } } } \right ) , \\ \sum_k \hbar^2\tilde{v}_{\textrm{f}}^2 k^2 \|d^\textrm{r}\|^2 ( d^\textrm{a } - d^\textrm{r } ) & \simeq \frac { i\pi \nu_e } { 8\eta^2 \epsilon_{\textrm{f } } } , \\ \sum_k \|d^\textrm{r}\|^2 & \simeq \frac{\pi\nu_e}{4\eta \epsilon_{\textrm{f}}^2 } , \\ % 1 \sum_k \bigl [ ( d^\textrm{r})^2 d^\textrm{a } + ( d^\textrm{a})^2 d^\textrm{r } \bigr ] & \simeq -\frac{\nu_e}{2 \epsilon_{\textrm{f}}^5 } , \\ \sum_k ( \hbar \tilde{v}_{\textrm{f } } k ) ^2 \bigl [ ( d^\textrm{r})^2 d^\textrm{a } + ( d^\textrm{a})^2 d^\textrm{r } \bigr ] & \simeq - \frac { \pi \nu_e } { 4 \eta \epsilon_{\textrm{f}}^2 } , \\ \sum_k ( \hbar \tilde{v}_{\textrm{f } } k ) ^2 ( d^\textrm{r})^3 d^\textrm{a } & \simeq -\frac{\pi \nu_e}{64 \epsilon_{\textrm{f}}^2 \eta^3 } \left ( 1- i\frac{2\eta}{\epsilon_{\textrm{f } } } \right ) , \\ % 4 \sum_k ( \hbar \tilde{v}_{\textrm{f } } k ) ^2 \ { ( d^\textrm{r})^3 d^\textrm{a } + ( d^\textrm{a})^3 d^\textrm{r}\ } & \simeq -\frac{\pi \nu_e}{32 \epsilon_{\textrm{f}}^2 \eta^3 } , \\ \sum_k ( \hbar \tilde{v}_{\textrm{f } } k ) ^4 ( d^\textrm{r})^3 d^\textrm{a } & \simeq -\frac{\pi \nu_e } { 32 \eta^3 } , \\ \sum_k(\hbar v_fk)^2(d^\textrm{r})^{n } & \simeq-\frac{1}{n-1}\sum_k(d^\textrm{r})^{n-1 } \ \ \ \ \ ( n\geq3 ) , \\ \sum_k(\hbar v_fk)^4(d^\textrm{r})^{n } & = \frac{2}{(n-1)(n-2)}\sum_k(d^\textrm{r})^{n-2 } \ \ \ \ \ ( n\geq4 ) , \\ \sum_k(\hbar v_fk)^2(d^\textrm{r})^{3 } & \simeq-\frac{1}{2}\sum_k(d^\textrm{r})^{2 } , \\ \sum_k(\hbar v_fk)^2(d^\textrm{r})^{4 } & \simeq -\frac{1}{3}\sum_k(d^\textrm{r})^{3}= \frac{\pi\nu_e } { 32\epsilon_{\textrm{f}}^2 \eta^3 } , \\ \sum_k(\hbar v_fk)^4(d^\textrm{r})^{4 } & \simeq \frac{1}{3}\sum_k(d^\textrm{r})^{2 } = \frac{\pi\nu_e } { 32 \eta^3}.\end{aligned}\ ] ] where @xmath441 is defined by @xmath442 here , we have used @xmath443 in the above equation . to check validity of our results , we substitute the charge current and charge density in eqs . ( [ eq:3 - 32 ] ) and ( [ eq:3 - 24 ] ) into the charge conservation law @xmath143 . from eq . ( [ eq:3 - 23 ] ) , @xmath444 becomes @xmath445 } \frac { i \omega^2 q_\nu } { q^2 \ell^2 + i\omega \tau } a_{\textrm{em},\nu } . % \\ % & = \frac{\hbar e \nu_e j_{ex } } { \eta v } \sum_{\bm{q},\omega}e^{i[\omega t-\bm{q}\cdot\bm{x } ] } % \biggl [ % \frac { i \ell \omega^2}{q^2 \ell^2 + i\omega \tau } \epsilon_{\zeta \nu z } q_\zeta s^\nu_{\bm{q},\omega } % \biggr]\end{aligned}\ ] ] from eq . ( [ eq:3 - 21 ] ) , ( [ eq:3 - 22 ] ) and @xmath446 , @xmath447 becomes @xmath448 } \biggl [ - \omega q_x a_{{\textrm{em } } , x } + \biggl\ { \frac { \omega q_x^2 q_y \ell^2 } { q^2 \ell^2 + i\omega \tau } a_{{\textrm{em } } , y } + \frac { \omega q_x^3 \ell^2 } { q^2 \ell^2 + i\omega \tau } a_{{\textrm{em } } , x } \biggr\ } \biggr ] , \\ \nabla_y j_{y } & = \frac{e^2 \tilde{v}_{\textrm{f}}^2 \nu_e \tau } { l^2 } \sum_{\bm{q},\omega}e^{i[\omega t-\bm{q}\cdot\bm{x } ] } \biggl [ -\omega q_y a_{{\textrm{em } } , y } + \biggl\ { \frac { \omega q_x q_y^2 \ell^2 } { q^2 \ell^2 + i\omega \tau } a_{{\textrm{em } } , x } + \frac{\omega q_y^3 \ell^2 } { q^2 \ell^2 + i\omega \tau } a_{{\textrm{em } } , y } \biggr\ } \biggr ] , \\ % % % % % % % \nabla_x j_x + \nabla_y j_y & = -\frac { e^2 \tilde{v}_{\textrm{f}}^2 \ell \nu_e } { l^2 } \sum_{\bm{q},\omega}e^{i[\omega t-\bm{q}\cdot\bm{x } ] } \frac { i \omega^2 q_\nu } { q^2 \ell^2 + i\omega \tau } a_{{\textrm{em } } , \nu}. % \\ % & = \frac { \hbar e \nu_e j_{ex } } { \eta v } \sum_{\bm{q},\omega}e^{i[\omega t-\bm{q}\cdot\bm{x } ] } \biggl [ \frac { i\omega^2 \tilde{v}_{\textrm{f } } \tau } { q^2 \ell^2 + i\omega \tau } \epsilon_{\zeta \nu z } q_\zeta s^\nu_{\bm{q},\omega } \biggr].\end{aligned}\ ] ] therefore , @xmath128 and @xmath164 follow the charge conservation law , @xmath449 . diffusive green s function on the disordered surface of the ti can be integrated as follows : @xmath450 } , \\ \notag \sum_{q_x } e^{-(2d t q_x^2 + i q_x x ) } & \sim \frac{1}{2\pi}\int_{-\infty}^\infty dq_x e^{-(2d t q_x^2 + i q_x x ) } % \\ \notag % & = \frac{e^{- \frac{2x^2}{dt}}}{2\pi}\int_{-\infty}^\infty dq_x e^{-2d t \ { [ q_x + i \frac{x}{dt}]^2 \ } } \\ & = \frac{\sqrt{\pi}}{2\pi \sqrt{2dt } } \exp{\left(- \frac{2x^2}{dt}\right)}.\end{aligned}\ ] ] thus , from the above equations , @xmath135 in the coordinates space is obtained by @xmath451}.\end{aligned}\ ] ] a. v. kimel , a. kirilyuk , p. a. usachev , r. v. pisarev , a. m. balbashov , and th . rasing : nature ( london ) * 435 * , 655 ( 2005 ) . f. hansteen , a. kimel , a. kirilyuk , and th . rasing : phys . lett . * 95 * , 047402 ( 2005 ) : phys . b * 73 * , 014421 ( 2006 ) . r. iida , t. satoh , t. shimura , k. kuroda , b. a. ivanov , y. tokunaga , and y. tokura : phys . b * 84 * , 064402 ( 2011 ) . a. kirilyuk , a. v. kimel , and t. rasing : rev . phys . * 82 * , 2731 ( 2010 ) . c. d. stanciu , f. hansteen , a. v. kimel , a. kirilyuk , a. tsukamoto , a. itoh , and t. rasing , phys . 99 * , 047601 ( 2007 ) . k. vahaplar _ et al_. , phys . lett . * 103 * , 117201(2009 ) . s. raghu , s. b. chung , x .- qi , and s .- c . zhang , phys . lett . * 104 * , 116401 ( 2010 ) . t. misawa , t. yokoyama , and s. murakami : phys . b * 84 * , ( 2011 ) 165407 . x. liu and j. sinova , phys . lett . * 111 * , 166801 ( 2013 ) . g. f. quinteiro and p. i. tamborenea , phys . b * 79 * , 155450 ( 2009 ) . g. f. quinteiro and p. i. tamborenea , phys . rev . b * 82 * , 125207 ( 2010 ) . g. f. quinteiro , p. i. tamborenea , and j. berakdar , opt . express * 19 * 26733 ( 2011 ) . j. wtzel , a. s. moskalenko , and j. berakdar , opt . express * 20 * 27792 ( 2012 ) . m. b. faras , g. f. quinteiro , and p. i. tamborenea , eur . j. b * 86 * , 432 ( 2013 ) . m. cygorek , p. i. tamborenea , and v. m. axt , phys . b * 92 * , 115301 ( 2015 ) . y. ueno , y. toda , s. adachi , r. morita , and t. tawara , opt . express * 17 * , 20567 ( 2009 ) . n. b. clayburn , j. l. mccarter , j. m. dreiling , m. poelker , and t. j. gay , phys . b * 87 * , 035204 ( 2013 ) . l. marrucci , c. manzo , and d. paparo , phys . lett . * 96 * , 163905 ( 2006 ) . l. allen , m. w. beijersbergen , r. j. c. spreeuw , and j. p. woerdman , phys . a * 45 * , 8185 ( 1992 ) . g. molina - terriza , j. p. torres , and l. torner , nature phys . * 3 * , 305 - 310 ( 2007 ) . m. padgett , j. courtial , and l. allen , phys . today 57(5 ) , 35 ( 2004 ) . j. t. mendonca , b. thide , and h. then , phys . * 102 * , 185005 ( 2009 ) . v. e. lembessis , m. babiker , and d. l. andrews , phys . a * 79 * , 011806(r ) ( 2009 ) . g. d. mahan , _ many - particle physics _ ( springer , 2000 ) , p.514 . j. fujimoto , a. sakai , h. kohno , phys . b * 87 * , 085437 ( 2013 ) . a. sakai and h. kohno , phys . b * 89 * , 165307 ( 2014 ) . k. taguchi , k. shintani , y. tanaka , phys . b * 92 * , 035425 ( 2015 ) . a. a. burkov and d. g. hawthorn , phys * 105 * , 066802 ( 2010 ) . p. schwab , r. raimondi , and c. gorini , epl ( europhysics letters ) , * 93 * , 67004 ( 2011 ) . k. taguchi , t. yokoyama and y. tanaka , phys . b * 89 * , 085407 ( 2014 ) . s. murakami , n. nagaosa , and s. c. zhang , science 301 , 1348 - 1351 ( 2003 ) . j. sinova , d. culcer , q. niu , n. a. sinitsyn , phys . lett . * 92 * , 126603 ( 2004 ) . e. l. rashuba , phys . b * 68 * , 241315 ( 2003 ) . t. kimura , y. otani , k. tsukagoshi , and y. aoyagi , j. magn . 272 - 276 , e1333 ( 2004 ) . y. k. kato , r. c. myers , a. c. gossard , and d. d. awschalom , science 306 , 1910 , e1333 ( 2004 ) . e. saitoh , m. ueda , h. miyajima , and g. tatara , appl . . lett . * 88 * , 182509 ( 2006 ) . a. a. taskin , s. sasaki , k. segawa , and y. ando , phys . lett . * 109 * , 066803 ( 2012 ) . using @xmath452 , @xmath453 , and @xmath454 , we can obtain the relation : @xmath455 . because of @xmath456 , @xmath457 becomes negligible small as @xmath458 . streakily speaking , there is no problem to take @xmath459 .